Cornerstones of Attachment Research by ROBBIE DUSCHINSKY

Downloaded from https://academic.oup.
com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024
Cornerstones of Attachment Research
Cornerstones of
Attachment Research
R O B B I E D U S C H I N SK Y

Senior University Lecturer, Primary Care Unit, University of Cambridge
Fellow and Director of Studies, Sidney Sussex College
1
3
Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© Robbie Duschinsky 2020
The moral rights of the authors have been asserted
First Edition published in 2020

Impression: 1
Some rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
This is an open access publication, available online and distributed under

the terms of the Creative Commons Attribution – Non Commercial – No Derivatives
4.0 International licence (CC BY-NC-ND 4.0), a copy of which is available at
http://creativecommons.org/licenses/by-nc-nd/4.0/.
For any use not expressly allowed in the CC BY-NC-ND licence
terms, please contact the publisher
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2020933057
ISBN 978–0–19–884206–4
Printed and bound by
CPI Group (UK) Ltd, Croydon, CR0 4YY
Oxford University Press makes no representation, express or implied, that the
drug dosages in this book are correct. Readers must therefore always check
the product information and clinical procedures with the most up-to-date
published product information and data sheets provided by the manufacturers
and the most recent codes of conduct and safety regulations. The authors and
the publishers do not accept responsibility or legal liability for any errors in the
text or for the misuse or misapplication of material in this work. Except where
otherwise stated, drug dosages and recommendations are for the non-pregnant
adult who is not breast-feeding
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
Contents
Introduction vii
1. John Bowlby and the Tavistock Separation Research Unit 1

2. Mary Ainsworth and the Strange Situation Procedure 109

3. Mary Main, Erik Hesse, and the Berkeley Social Development Study 211
4. Alan Sroufe, Byron Egeland, and the Minnesota Longitudinal Study
of Risk and Adaptation 337
5. Phillip Shaver, Mario Mikulincer, and the Experiences in Close
Relationships Scale 427
6. Conclusion 537
Index 573
Introduction
Attachment theory is among the most popular theories of human socioemotional devel-
opment, with a global research community and widespread interest from clinicians, child
welfare professionals, educationalists, and parents. It has been considered ‘one of the most
generative contemporary ideas’ about family life in modern society.1 It is one of the last of the
grand theories of human development that still retains an active research tradition. Indeed,

Simpson and Howland have observed that ‘perhaps no single theory in the psychological
sciences has generated more empirical research during the past 30 years than attachment
theory’.2 Attachment theory and research speak to fundamental questions about human
emotions, relationships, and development. They do so in terms that feel experience-near,
with a remarkable combination of intuitive ideas and counter-intuitive assessments and con-
clusions. Over time, attachment theory seems to have become more, rather than less, ap-
pealing and popular, in part perhaps due to alignment with current concern with the lifetime
implications of early brain development.3 Emerging reports on the economic costs of inse-
cure attachment may make a further contribution to this appeal over the coming years.4 In
a 2018 survey conducted by the British government of organizations working with children
in need of help and protection, attachment theory was, by a large margin, cited as the most
frequently used underpinning perspective.5 Attachment ideas have been used to support
1 Pittman, J.F. (2012) Attachment orientations: a boon to family theory and research. Journal of Family Theory &
Review, 4(4), 306–10.

2 Simpson, J.A. & Howland, M. (2012) Bringing the partner into attachment theory and research. Journal of
Family Theory & Review, 4(4), 282–9, p.282.

3 Wastell, D. & White, S. (2017) Blinded by Science: The Social Implications of Epigenetics and Neuroscience.
Cambridge: Policy Press. Discourses of ‘interpersonal neurobiology’, and the work of Allan Schore in particular,
have been important for the take-up of appeals to attachment within popular and policy discourses emphasising
the importance of child development for the brain. See, for example, Schore, A.N. (2001) Effects of a secure attach-
ment relationship on right brain development, affect regulation, and infant mental health. Infant Mental Health
Journal, 22(1–2), 7–66; and building from Schore’s work, Gerhardt, S. (2014) Why Love Matters: How Affection
Shapes a Baby’s Brain, 2nd edn. London: Routledge. Schore’s work has been much less influential within the attach-
ment research community. An exception is discussed in Chapter 3.
4 Bachmann, C.J., Beecham, J., O’Connor, T.G., Scott, A., Briskman, J., & Scott, S. (2019) The cost of love: finan-
cial consequences of insecure attachment in antisocial youth. Journal of Child Psychology & Psychiatry, 60(12),
1343-50.
5 In response to the question ‘What theories or research do you rely on to inform a plan of how to support a
child?’ attachment theory was mentioned by 11% of respondents. The next most cited responses were general
areas rather than specific theories or research paradigms; and a long way behind, ‘mental health’ at 5% and ‘child
development’ at 4%. The next most cited specific theory was social learning theory at 1%. Mention of ‘attachment
disorder’ appeared in a further 2.5% of responses. Survey responses to a government inquiry offer little sure know-
ledge about what these organizations actually do and why. However, the leading position of attachment theory spe-
cifically, even compared to ‘mental health’ and ‘child development’ as general areas, suggests the position of appeal
to attachment as a dominant and apparently authorized discourse within the justification and conceptualization of
child welfare practice. Department for Education (2018) Children in need of help and protection: call for evidence.
https://www.gov.uk/government/consultations/children-in-need-of-help-and-protection-call-for-evidence. On
the contradictions and diversity of uses of attachment discourses in child welfare contexts, see McLean, S., Riggs,
D., Kettler, L., & Delfabbro, P. (2013) Challenging behaviour in out‐of‐home care: use of attachment ideas in prac-
tice. Child & Family Social Work, 18(3), 243–52; Smith, M., Cameron, C., & Reimer, D. (2017) From attachment to
recognition for children in care. British Journal of Social Work, 47(6), 1606–23.
viii Introduction
recognition of the importance of stable, trusting relationships for children’s socioemotional

development, with the credibility of links to an established empirical research paradigm.6
Attachment also provides a framework for interpreting the underlying logic or meaning of
the behaviour of children and young people following relational adversities.7 These qualities
have contributed to the popularity of attachment among social workers,8 clinicians,9 and
health visitors,10 in training and support provided to foster carers and adoptive parents,11
in parenting education courses and materials given to parents by health and care profes-
sionals,12 in forensic contexts,13 and in professional development courses for teachers.14
Yet the most well-known account of attachment is in many regards based on certain early
claims by John Bowlby and Mary Ainsworth, the originators of attachment theory, at the ex-
pense of their own and others’ later conclusions and qualifications. Ainsworth herself com-
plained of a tendency to describe psychological theories in terms of early findings and ideas;

these enter into circulation, ricochet and rebound among domains of practice, and get re-
peated and repeated. Later developments, even important ones, become difficult to access
6 The particular utility of attachment ideas for clinicians and child welfare practitioners, increasing the cred-
ibility of practice through association with the evidence-base of attachment research, is praised directly in Bennett,
C.S. & Nelson, J.K. (2008) Closing thoughts: special issue on attachment. Clinical Social Work Journal, 36, 109–
11. Concern about many ‘attachment-based’ therapies as pseudoscience has been raised by Mercer, J. (2019)
Conventional and unconventional perspectives on attachment and attachment problems: comparisons and impli-
cations, 2006–2016. Child and Adolescent Social Work Journal, 36(2), 81–95.
7 E.g. Haight, W.L., Kagle, J.D., & Black, J.E. (2003) Understanding and supporting parent–child relationships
during foster care visits: attachment theory and research. Social Work, 48(2), 195–207; Farnfield, S. & Holmes, P.
(eds) (2014) The Routledge Handbook of Attachment: Assessment. London: Routledge.
8 An important contribution to the popularity of attachment theory among social workers, especially in the UK,
was made by Howe, D., Brandon, M., Hinings, D., & Schofield, G. (1999) Attachment Theory, Child Maltreatment
and Family Support: A Practice and Assessment Model. London: Palgrave.
9 For a useful review see Slade, A. & Holmes, J. (2017) Attachment in Therapeutic Practice. London: SAGE.
10 Hogg, S. (2019) Rare Jewels: Specialised Parent–Infant Relationship Teams in the UK. London: Parent–Infant
Partnership UK. https://www.pipuk.org.uk/sites/default/files/PIPUK%20Rare%20Jewels%20FINAL.pdf. In the

UK, National Institute for Health and Care Excellence (NICE) Guidelines for postnatal care state that health vis-
itors should promote the emerging attachment relationship, and that health visitors should assess potential obs-
tacles or problems for this relationship. NICE (2015) Postnatal care up to 8 weeks after birth. London: NICE.
https://www.nice.org.uk/guidance/cg37/resources/postnatal-care-up-to-8-weeks-after-birth-pdf-975391596997.
11 Laybourne, G., Andersen, J., & Sands, J. (2008) Fostering attachments in looked after children: further insight
into the group-based programme for foster carers. Adoption and Fostering, 32(4), 64–76; Benesh, A.S. & Cui, M.
(2017) Foster parent training programmes for foster youth: a content review. Child & Family Social Work, 22(1),
548–59.
12 Wall, G. (2018) ‘Love builds brains’: representations of attachment and children’s brain development in par-
enting education material. Sociology of Health & Illness, 40(3), 395–409. Attachment is a module in the Ready
Steady Baby book, given to all new parents in Scotland. https://www.nhsinform.scot/ready-steady-baby/preg-
nancy/relationships-and-wellbeing-in-pregnancy/attachment-and-bonding-during-pregnancy.
13 Ministry of Justice (2011) Working with Personality Disordered Offenders: A Practitioner’s Guide.
London: HMSO; Baim, C. & Morrison, T. (2011) Attachment-Based Practice with Adults: Understanding Strategies
and Promoting Positive Change. Hove: Pavilion Publishing; Brown, R. & Ward, H. (2012) Decision-Making within
a Child’s Timeframe: An Overview of Current Research Evidence for Family Justice Professionals Concerning Child
Development and the Impact of Maltreatment. London: Childhood Wellbeing Research Centre; Crittenden, P.M.,
Farnfield, S., Landini, A., & Grey, B. (2013) Assessing attachment for family court decision making. Journal of
Forensic Practice, 15(4), 237–48.
14 Geddes, H. (2006) Attachment in the Classroom: A Practical Guide for Schools. London: Worth Publishing;
Beckh, K. & Becker-Stoll, F. (2016) Formations of attachment relationships towards teachers lead to conclusions
for public child care. International Journal of Developmental Science, 10(3–4), 103–10; Rose, J., McGuire-Snieckus,
R., Gilbert, L., & McInnes, K. (2019) Attachment Aware Schools: the impact of a targeted and collaborative inter-
vention. Pastoral Care in Education, 37(2), 162–84. In the UK, NICE has mandated that ‘Schools and other edu-
cation providers should ensure that all staff who may come into contact with children and young people with
attachment difficulties receive appropriate training on attachment difficulties’ (Recommendations 1.2.1). NICE
(2016) Children’s attachment: attachment in children and young people who are adopted from care, in care or at
high risk of going into care. London: NICE. https://www.ncbi.nlm.nih.gov/pubmed/26741018.
Introduction ix
and incorporate, with textbooks and summaries sustaining an outdated caricature.15 Already
in 1968, Ainsworth wrote to Bowlby with concern: ‘attachment has become a bandwagon’.16
In the helping professions, the idea of attachment theory is well known, and even forms
part of the mandatory curriculum for some professions. At the same time, knowledge of
developments in attachment theory and research may not be strong. Qualitative research
by Furnivall and colleagues found that ‘there was a sense that professionals knew the word
but not the underlying theory . . . although there was strong support for the importance
of the fundamental concept’.17 Likewise Morison and colleagues found that staff working
in residential childcare generally stated in interview that their practice was informed by
attachment theory, but struggled to say exactly how.18 Bennett and Blome have observed
that welfare agencies give lip-service to attachment in providing a support for the cred-
ibility of their work, but may provide a protocol-focused organizational culture that ultim-

ately discourages practitioners from gaining expertise regarding attachment research and
its implications.19
Elizabeth Meins, one of the UK’s leading attachment researchers through the 1990s and
2000s, has recently turned her back on the paradigm. She has argued that regardless of the
scientific advances made by attachment research, the benefits arising from these have been
outstripped by the problems caused by public misunderstandings. Meins’ position is un-
usual. However, it sets out clearly the stakes in the gap between attachment research today
and how it is widely understood:
Somewhere along the line, the idea that early attachment is the best predictor of all
aspects of later development has gained credence. We need to get out of our ivory towers
and unite in calling out this caricature of our research. I stand by my claim that laying so
much emphasis on attachment isn’t helpful. Being made to worry about whether you have
a secure attachment with your baby won’t make you a better parent; healthcare profes-
sionals who are provided with oversimplified hype about the predictive power of attach-
ment won’t give families good advice; and letting non-experts who think they know the
attachment literature loose in the political arena won’t result in good policies for children
and families.20
15 Ainsworth, M. (1969) CPA oral history of psychology in Canada interview. Unpublished. http://www.fem-
inistvoices.com/assets/Women-Past/Ainsworth/Mary-Ainsworth-CPA-Oral-History.pdf: ‘I think it was just the

way it is so often with textbooks. The things that get into the textbooks are the early publications and they don’t
ever get around to putting in the later publications, and people write textbooks on the basis of other people’s text-
books.’ Illustrating the caricature and outright mistakes about attachment research available from textbooks, see
Parke, R.D. & Clarke-Stewart, A. (2011) Social Development. New York: Wiley.
16 Ainsworth, M. (1968) Letter to John Bowlby, 27 April 1968. PP/Bow/K.4/12: ‘Attachment has become a band-
wagon. There are so many people now interested in research in this area, and so many approaches, both theoretical
and methodological. I am afraid that people will leap in in a half-baked way, that findings will be equivocal or con-
flicting, and that perhaps interest will move away from “attachment” dismissing it as one more area that did not
“pan out”.’
17 Furnivall, J., McKenna, M., McFarlane, S., & Grant, E. (2012) Attachment matters for all: an attachment
mapping exercise for children’s services in Scotland. Glasgow: Centre for Excellence for Looked after Children in
Scotland (CELCIS). www.celcis.org/knowledge-bank/search-bank/attachment-matters-all/.
18 Morison, A., Taylor, E., & Gervais, M. (2019) How a sample of residential childcare staff conceptualize and use
attachment theory in practice. Child & Youth Services, DOI: 10.1080/0145935X.2019.1583100.

19 Bennett, S. & Blome, W.W. (2013) Implementing attachment theory in the child welfare system: clinical im-
plications and organizational considerations. In J.E. Bettmann & D.D. Friedman (eds) Attachment-Based Clinical
Work with Children and Adolescents (pp.259–83). New York: Springer.
20 Meins, E. (2017) Reply. The Psychologist, 30, 6–9. https://thepsychologist.bps.org.uk/volume-30/march-2017/
attachment-public-and-scientific-discourse, p.9.
x Introduction
Meins’ remarks suggest the value in taking stock of qualifications, innovations, and amend-
ments made by later researchers in relation to Bowlby’s early claims. One important early
attempt at such taking stock was Becoming Attached by Robert Karen, published in 1994.21
Karen described the emergence of the attachment paradigm in the work of Bowlby and
Ainsworth, and its subsequent elaboration by younger researchers such as Sroufe, Main,
and Shaver. Karen interviewed all these researchers, as well as conducting extensive study of
their published works until 1992. He documented how attachment theory was introduced by
John Bowlby in the 1950s and 1960s. He traced how Bowlby sought to revise psychoanalytic
theory in order to create a scientific model that nonetheless retained the strengths of psy-
choanalysis in relevance to clinical work. The theory was not well received at the time by the
psychoanalytic community. However, Karen showed, through the work of Mary Ainsworth,
how attachment theory entered into American developmental psychology, where it took

firm root, first among Ainsworth’s immediate collaborators and then across the subdisci-
pline. Central to the establishment of attachment theory within developmental psychology,
in Karen’s account, was Ainsworth’s introduction of the Strange Situation procedure, an ob-
servational assessment of infant–caregiver attachment relationships using separations and
reunions to examine infants’ expectations about their caregiver. Ainsworth found that in-
fants’ behaviour in the Strange Situation was associated with observations of the care they
received at home. Infants who could confidently explore in the Strange Situation and retreat
to their caregiver for comfort when distressed were those whose caregiver had been attentive
and responsive to their signals over the first year of life. Ainsworth therefore termed this pat-
tern of behaviour ‘secure attachment’.
Karen also documented how longitudinal studies by Ainsworth’s collaborators and stu-
dents had shown the value of attachment theory as an approach within developmental
psychology, and validated the Strange Situation as a predictive measure. Ainsworth’s stu-
dent Mary Main demonstrated that patterns of behaviour towards each parent in the Strange
Situation are largely independent, confirming Ainsworth’s interpretation that the behaviours
were less individual traits than reflections of infant expectations about particular relation-
ships. Qualities from autobiographical interviews with parents were also found to have pre-
dictable associations with their children’s behaviour in the Strange Situation. This formed
the basis of the introduction of the Adult Attachment Interview. Karen also explored the
findings of the Minnesota Longitudinal Study of Risk and Adaptation, led by Alan Sroufe
and Byron Egeland. These researchers followed up a large high-risk sample since the 1970s.
This longitudinal study has been of particular importance in both supporting and qualifying
claims by Bowlby, Ainsworth, and Main about the developmental implications of attach-
ment. Karen also reported another key development: the growth of research on attachment
within social psychology, initiated by Phillip Shaver and colleagues, and drawing on self-
report assessments of adult attachment.
Karen’s book came out at an important moment for attachment research. Bowlby and
Ainsworth were no longer available to act as leaders. Main’s methodological innovations had
been introduced but were still in the process of being validated in other laboratories. Karen
21 Karen, R. (1994) Becoming Attached. New York: Warner Books. Karen’s book developed an earlier art-
icle: Karen, R. (1990) Becoming attached. The Atlantic, February. Karen’s stock-taking not only was influential for
the public reception of attachment theory, but also influenced subsequent attachment research, such as providing a
prompt for the development of the ‘Circle of Security’ intervention: Powell, B., Cooper, G., Hoffman, K., & Marvin,
B. (2016) The Circle of Security Intervention. New York: Guilford, p.9.
Introduction xi
could discuss the Minnesota group’s follow-up of their sample from infancy to preschool, but
the data from later childhood were still being analysed. And Shaver and colleagues had intro-
duced their early ‘love quiz’ self-report measure of adult attachment, but the properties of this
assessment were subject to significant criticism. The relationship between the developmental
psychologists and the social psychologists was relatively hostile, and it was wholly unclear
how the ideas and measures of the two traditions would relate to one another. Yet alongside
sorrow at the loss of Bowlby and Ainsworth, and tensions over the future direction of attach-
ment research and theory, it was a time of great excitement for the field. The standing of at-
tachment research as a scientific paradigm had been established by the 1980s, and support
from research funders had led to a rapid growth in the size of the field by the mid-1990s.22
Since Karen’s book, there has been substantial academic scholarship exploring the early
years of attachment research from a historical perspective. Inge Bretherton, a student and

colleague of Ainsworth, and Jeremy Holmes, a clinician and colleague of Bowlby, also pub-
lished influential celebratory reviews in the early 1990s.23 Work to document the emergence
of Bowlby and Ainsworth’s research by academic historians began in the late 1990s, and re-
mains thriving today.24 There has also been a rich tradition of critical discussion of the re-
lationship between Bowlby’s ideas and post-war gender and parenting cultures.25 However,
the important developments in the field of the past 30 years have not been examined by
historians, a startling gap in light of the revolutions in theory and method that have taken
place in these decades. Peter Fonagy and Chloe Campbell, significant figures in the British
attachment research community, have criticised historians of attachment research for fo-
cusing too exclusively on Bowlby and Ainsworth, neglecting attention to the ways in which
the paradigm has changed over time.26 These changes are also of great interest for the history
of science, illustrating dynamics in the relationship between theory and method in psycho-
logical science, debates about the function of categorization, problems in the conceptual-
ization of emotional development, changing appeals to evolutionary theory and ontologies
of human nature, and shifts in the relationship between developmental science and its pub-
lics. The images of attachment research offered by commentators outside the field are gen-
erally outdated, hackneyed, and too often inaccurate.27 This loses critical psychology access
22 The boom in funding for attachment research in the 1990s is discussed in White, K. & Schwartz, J. (2007)
Attachment here and now: an interview with Peter Fonagy. Attachment: New Directions in Relational Psychoanalysis
and Psychotherapy, 1(1), 57–61.
23 Bretherton, I. (1992) The origins of attachment theory: John Bowlby and Mary Ainsworth. Developmental
Psychology, 28(5), 759–75; Holmes, J. (1993) John Bowlby and Attachment Theory. London: Routledge.
24 E.g. Van Dijken, S. (1998) John Bowlby: His Early Life: A Biographical Journey into the Roots of Attachment
Theory. London: Free Association Books; Mayhew, B. (2006) Between love and aggression: the politics of John
Bowlby. History of the Human Sciences, 19(4), 19–35; van der Horst, F. (2011) John Bowlby—From Psychoanalysis to
Ethology: Unravelling the Roots of Attachment Theory. Oxford: Blackwell.
25 E.g. Birns, B. (1999) I. Attachment theory revisited: challenging conceptual and methodological sacred
cows. Feminism & Psychology, 9(1), 10–21; Vicedo, M. (2013) The Nature and Nurture of Love: From Imprinting to
Attachment in Cold War America. Chicago: University of Chicago Press. As Ruck observes, the popularity of the
development of attachment theory for historians of science resides at least in part in the fact that ‘the theory offers
a looking glass into the social foundations and effects of science; the function and logic of scientific controversies
and disciplinary hierarchies; and the interrelation of descriptive and prescriptive scientific theories, scientific and
popular discourse, and science and ideology all at once’ : Ruck, N. (2014) Review: Marga Vicedo. The nature and
nurture of love. Journal of the History of the Behavioral Sciences, 50(4), 410–11, p.410.
26 Fonagy, P. & Campbell, C. (2016) Attachment theory and mentalization. In A. Elliott & J. Prager (eds) The
Routledge Handbook of Psychoanalysis in the Social Sciences and Humanities (pp.115–31). London: Routledge, p.123.
27 For an example of a work in critical psychology that does little more than repeat stock criticisms with little
relevance to contemporary attachment research, see Walsh, R.T.G., Teo, T., & Baydala, A. (2014) A Critical History
and Philosophy of Psychology. Cambridge: Cambridge University Press.
xii Introduction
to an influential and superbly rich case—one relevant to major current concerns such as the
history of emotion in the human sciences, debates about psychological categorisation, and
ways of imagining human relationships. In turn, attachment research loses effective critical
interlocutors.
This book
Cornerstones re-examines the background and current approaches of key laboratories that
have contributed to attachment research as it exists today. In this way the book traces the
development in a single scientific paradigm through parallel albeit separate lines of inquiry.
The laboratories in focus, those examined by Karen, exemplify particular advances and di-

lemmas the field has faced. Cornerstones seeks to use a focus on five research groups as a lens
on wider themes and challenges faced by the contemporary field as it has emerged. In doing
so, the book uses certain landmarks that suggest some of the fundamental logic, infrastruc-
ture, and points of orientation in attachment research as a terrain.
The book in no way aims to be a comprehensive account of attachment research.28 This
scholarship is diverse. It does not form a single totality, but rather a region with points of
density and intensity. Both the density and intensity of work within attachment research are
structured by research groups, which are shaped by and shape the field. Chapters aim to re-
main with researchers long enough to offer a sense of their characteristic ways of thinking
and tone, to feel them as figures keeping us company—if sometimes quarrelling, sometimes
pulling in unison—within the broader history of developments in the field. Readers will note
that the book does not contain chapters focusing on the research groups of influential direct
students of Ainsworth such as Jude Cassidy, Patricia Crittenden, Roger Kobak, and Bob
Marvin, and other research leaders such as Kim Bartholomew, Jay Belsky, Martha Cox, Peter
Fonagy, Karlen Lyons-Ruth, Avi Sagi-Schwartz, Sue Spieker, and Marinus van IJzendoorn.
A younger cohort of research leaders would also need to be considered in a characteriza-
tion of contemporary research groups, including—but by no means limited to—Marian
Bakermans-Kranenburg, Gurit Birnbaum, Mary Dozier, Robin Edelstein, Pasco Fearon,
Chris Fraley, Patrick Luyten, Sheri Madigan, Carlo Schuengel, Jeff Simpson, Gottfried
Spangler, and Glenn Roisman. All these figures will feature in the present book, however.
And a further book is already underway to examine key research groups that gained prom-
inence only after Karen’s Becoming Attached. This will include attention to major preoccupa-
tions of the 1990s and 2000s, such as studies of attachment and the developing brain.
The concept of a ‘generation’ can be used pragmatically to characterise members of a co-
hort who, facilitated by structural factors that suggest commonalities, regard themselves as
facing a bundle of common challenges, including delimitation and appraisal of the legacy
of an earlier generation.29 In an important sense, just as Karen was writing at a point of
28 One expression of the sheer scale of the historical background to contemporary attachment research is the
six-volume edited work of Slade, A. & Holmes, J. (eds) (2014) Attachment Theory. London: Sage. The editors aimed
to collect 60 essential papers; however, 119 papers ultimately were judged indispensable. On the diversity of fac-
tors involved in canon formation, and above all the importance of subsequent resonance, see Fishelov, D. (2010)
Dialogues with/and Great Books: The Dynamics of Canon Formation. Eastbourne: Sussex Academic Press.
29 The idea of generations in attachment research is heuristic rather than intended as a simple statement of fact.
Certainly, there are figures who do not fall easily within one generation or the other in terms of age and attitudes;
Jude Cassidy, Jay Belsky, Marinus van IJzendoorn, and Gottfried Spangler are all clear examples. And the present
book is centrally concerned with changes over time regarding theory, method, and research priorities that do not
Introduction xiii
transition from the first to the second generation of attachment researchers, this book has
been written during a transition from the second to the third generation of attachment
researchers. The leaders of the research groups considered in this book have, with the ex-
ception of Mikulincer, now retired.30 Consideration of their work is intended to offer an op-
portunity to examine the strengths and the limitations, and clarify some of the debates, that
have characterized the second generation of attachment researchers and which have formed
the context in which a new generation of leaders are inheriting the field of attachment re-
search. In a letter to Mary Main, Bowlby wrote that ‘there is no need for the old to learn from
the young in order for the population to benefit from youthful innovation. The supersession
of an older generation by a younger is sufficient.’31 Now it is Main’s own generation who are
putting down their tools, and a new set of research leaders who must take stock of what they
have learned, and of what hold this learning has on them.

There are several excellent books that have taken a thematic approach in offering the
reader a guide and introduction to contemporary attachment research, most notably the
Handbook of Attachment edited by Cassidy and Shaver, Understanding Attachment and
Attachment Disorders by Prior and Glaser, and Adult Attachment by Gillath, Karantzas, and
Fraley.32 Thompson, Simpson, and Berlin’s Attachment: The Fundamental Questions also will
offer a systematic stock-taking of the present state of attachment research when it is pub-
lished next year.33 However, a thematic approach to synthesis can risk making an area of
research appear seamless and without edges. In particular, it can lose track of the social dy-
namics, debates, and diverging use of the same terminology that organize a field of inquiry
and the relationships between research groups. Thompson has argued that this heterogen-
eity must be captured by any attempt to understand attachment research today, even if it
makes for a more intricate story.34 The chapters of Cornerstones draw on a complete analysis
of published scholarly and popular works by each research group, as well as unpublished
doctoral theses published in English where these were available through inter-library loan.
This signals an important limitation. Much of the imagination, passion, and artistry of
research, much of its process and messy creation out of different elements, much of its in-
fluence by and influence on social interactions, is hidden in textual records, especially those
that go into print.35 Many of the most vital social dynamics of the field of attachment re-
search do not feature in the textual record. For instance, the available texts offer little vantage
on interactions between attachment researchers and clinical and social welfare professionals.
divide by generation. On the concept of ‘generations’ see Aboim, S. & Vasconcelos, P. (2014) From political to
social generations: a critical reappraisal of Mannheim’s classical approach. European Journal of Social Theory, 17,
165–83.
30 Ainsworth’s first doctoral students at Johns Hopkins graduated in 1972, among them Mary Main; her final
doctoral students were Jude Cassidy in 1986 and Carolyn Eichberg in 1987. This means that most of Ainsworth’s
doctoral students have moved into retirement over the past decade.
31 Bowlby, J. (1970) Letter to Mary Main, 18 November 1970. PP/Bow/J.4/1.
32 Cassidy, J. & Shaver, P. (eds) (2016) Handbook of Attachment: Theory, Research, and Clinical Applications,
3rd edn. New York: Guildford; Prior, V. & Glaser, D. (2006) Understanding Attachment and Attachment
Disorders: Theory, Evidence and Practice. London: Jessica Kingsley Press; Gillath, O., Karantzas, G.C., & Fraley,
R.C. (2016) Adult Attachment: A Concise Introduction to Theory and Research. London: Academic Press. See also
Holmes, P. & Farnfield, S. (2014) The Routledge Handbook of Attachment. London: Routledge.
33 Thompson, R.A., Simpson, J.A., & Berlin, L. (eds) (2020) Attachment: The Fundamental Questions.
New York: Guildford.

34 Thompson, R.A. (2017) Twenty-first century attachment theory. In H. Keller & K. Bard (eds) The Cultural
Nature of Attachment: Contextualizing Relationships and Development (pp.301–19). Cambridge, MA: MIT
Press, p.303.
35 Schickore, J. (2008) Doing science, writing science. Philosophy of Science, 75(3), 323–43.
xiv Introduction
Some commentators have described attachment theory and research as little more than
an ideology for the coercive evaluation, classification, and discipline of families by profes-
sionals.36 In response to such accusations, apologists have countered that attachment theory
and research are no different than any other form of knowledge of children and families,
and that contemporary attachment research, adequately understood, offers no support for
oppression of families.37 Both claims are likely too flat, masking the diversity within attach-
ment discourses and their changes over time and between contexts. Neither the accusations
nor the apologetics are based on empirical research, or on textual evidence. In fact, very little
is readily available in the public domain about the circulation of ideas between research and
practice.38 Colleagues and I currently have research on these questions underway, drawing
on interviews, focus groups, ethnography, and analysis of a large archive of clinical case
records.39

Yet even if subject to systematic and important limitations, the written record available
for a study of important research groups in the history of attachment is extensive. And I have
been grateful to have access to some texts in this area beyond the published record. Chapters
draw extensively on materials from the John Bowlby Archive at the Wellcome Collection,
the Mary Ainsworth Archive at the Nicholas and Dorothy Cummings Center for the History
of Psychology, and the Mary Main and Erik Hesse personal archive (currently being cata-
logued for the Wellcome Collection).40 These archives provide access to a treasure trove of
unpublished lectures and seminars, correspondence, notes, and speculations, and drafts of
published works and coding manuals. For Bowlby, I have been able to access the surviving
part of his library at Human Development Scotland (the location of the rest of the Bowlby li-
brary is unknown, even to his family!). This has permitted study of relevant marginalia—for
instance the annotations on his personal copy of the works of Freud. Mary Main and Erik
Hesse have also made available the manuscripts of two major unpublished books from 1986
and 1995, describing the methods and ideas of their Berkeley group. Use of unpublished
materials, such as correspondence, has helped this book attend to the lines of continuity and
discontinuity over time and between research groups.
For readers less familiar with attachment research, it is hoped that Cornerstones can pro-
vide a thorough introduction to theories and methods that form the basis for contemporary
36 For example, Smeeton puts this polemically, alleging that ‘we watch as the next generation of social workers
suffer the consequences of intellectual inbreeding, fumbling through practice with webbed theories and six-
fingered methodologies that give up on families unable to reach the optimal state of a “secure pattern” attachment
with their child’. Smeeton, J. (2017) From Aristotle to Arendt: a phenomenological exploration of forms of know-
ledge and practice in the context of child protection social work in the UK. Qualitative Social Work, 16(1), 14–28,
p.16. See also Garrett, M.P. (2017) Wired: early intervention and the ‘neuromolecular gaze’. British Journal of Social
Work, 48(3), 656–74. Such criticisms are not based on empirical work on how attachment research is conducted,
transmitted, or applied.
37 For example, Ross Thompson’s remarks during discussion in Keller, H. & Thompson, R. (2018) Attachment
theory: past, present & future, recorded at the 2nd ‘Wilhelm Wundt Dialogue’, 28 November 2018, Leipzig
University, hosted by the Leipzig Research Center for Early Child Development (LFE). https://www.youtube.com/
watch?v=_nG5SelEj28.
38 One of the few studies found that the more training professionals had in attachment theory, the less likely they
were to make judgemental comments about parents’ caregiving behaviours. McMahon, C., Huber, A., Kohlhoff, J.,
& Camberis, A.L. (2017) Does training in the Circle of Security framework increase relational understanding in
infant/child and family workers? Infant Mental Health Journal, 38(5), 658–68.
39 Some early findings are presented in Reijman, S., Foster, S., & Duschinsky, R. (2018) The infant disorganised
attachment classification: ‘Patterning within the disturbance of coherence’. Social Science & Medicine, 200, 52–58.
40 Occasional further manuscripts have been made available by other attachment researchers including Chris
Fraley, Klaus and Karin Grossmann, and Alan Sroufe.

Introduction xv
attachment research. Attention to a number of research groups offers an ‘arsenal of exem-

plars’ for how key questions have been approached, as well as clarifying the stakes in debates
between researchers and clarifying the meaning of terminology.41 At the same time, for read-
ers more familiar with attachment research, the book seeks to surprise and defamiliarize.
Historical inquiry offers a point of access for considering how an area of scientific practice
decides its objects, priorities, and tools, and in doing so can provide a way of refreshing a
reader’s perspective on the present and its concerns.42 The interested reader is also advised to
review the book’s detailed footnotes, many of which explore some of the peculiar catacombs
and other structures underneath the more well-known landscapes of attachment research.
Attachment and historical time

Cornerstones is oriented by the perception that, despite their very substantial differences,
historical research and attachment research have points of overlap in how they regard time.43
Several attachment researchers have, in fact, offered substantial—if scattered—commentary
on the idea of history, informed by reading in the history and philosophy of science. From
early in his career, Bowlby was wholly convinced of the importance of patients’ history for
making sense of their trajectory through life, including the capacity of the past to shape or
influence behaviours without the awareness of the individual themselves. As a clinician, his-
tory taking was second nature to him.44 Beyond this, however, Bowlby was an avid, lifelong
reader of social and political history in his spare time. He had a strong belief in the value of
slow, in-depth historical research, which reached its culmination in his decision to dedicate
his final years to a grand study of the life of Charles Darwin.45 In Bowlby’s time, historians of
science tended to shy away from evaluating the ideas of an earlier scientist in light of later de-
velopments. This practice was motivated by an effort to truly understand scientific practices
in their own context.46 Bowlby was impatient with this view. Reflecting on his reading of
philosophers of science such as Kuhn, Lakatos, and Popper, he had a different image of his-
torical analysis.47 His book on Darwin treats it as obvious that later scientific developments
can help historians understanding what an earlier scientist was attempting to feel out and
explore, the constraints they faced, and limitations or tensions within their understanding
and terminology.
41 Bowlby, J. (1974) Marginalia on Kuhn, second thoughts on paradigms. PP/Bow/H.98: Heavily under-
lined: ‘Acquiring an arsenal of exemplars, just as much as learning symbolic generalisations, is integral to the pro-
cess by which a student gains access to the cognitive achievements of his disciplinary group’ (p.471).
42 This function of the history of psychology is discussed well in Capshew, J. (2014) History of psychology since
1945: a North American review. In R. Backhouse & P. Fontaine (eds) A Historiography of the Modern Social Sciences
(pp.144–82). Cambridge: Cambridge University Press.
43 The links between historical research and attachment research in this regard relate especially to researchers in
the developmental tradition. The social psychological tradition of attachment research has offered fewer relevant
reflections on the idea of history, reflecting a predominant tendency in the broader discipline of social psychology.
However, see Billig, M. (2018) Those who only know of social psychology know not social psychology: a tribute to
Gustav Jahoda’s historical approach. Culture & Psychology, 24(3), 282–93.
44 Bowlby, J. (c.1932–33) History taking; methods of examining. PP/BOW/D.2/13.
45 Bowlby, J. (1990) Charles Darwin: A Life. New York: Norton.
46 Wilson, A. & Ashplant, T.G. (1988) Whig history and present- centred history. The Historical Journal,
31(1), 1–16.
47 See Bowlby, J. (1982) A case of mistaken identity. Higher Education Quarterly 36(4): 328–32; Bowlby, J. (1962)
Notes on Feyerabend. PP/BOW/H.99; Bowlby, J. (1974) Marginalia on Kuhn, Second Thoughts on Paradigms. PP/
Bow/H.98.
xvi Introduction
In the Darwin biography, unpublished materials are treated as different but not necessarily
inferior sources of information. Both are asked to play their part in filling out the develop-
ment of ideas and scientific practices over time. Bowlby felt that history can, and at times
should, help ‘exhume’ the ‘archaeological remnant’ of ideas that have been lost or thrust into
the background over time.48 He warned that ‘so long as our history is hidden from us, so long
as we hide our history from ourselves, we are very likely to see the present and future in the
terms of the past’.49 In his view, the history of a research paradigm holds open the possibility
of greater critical awareness of its ideas and methods, including a sense of what avenues have
been or might be more or less fruitful. This can contribute to greater flexibility and freedom
of action in facing contemporary dilemmas. Another potential benefit of historical inquiry,
Bowlby held, is that such research can directly contribute to ‘the formulation of specific hy-
potheses and theories’, even if this is not its primary purpose.50

Bowlby emphasized the ‘appalling complexity’ of history, whether this is the history
of societies, persons, or ideas, since it has to capture ‘highly specific interacting events’.51
Scientific research is partly shaped by the phenomena under investigation. However, Bowlby
stated that historians of science, and their emphasis on the social relationships and cultural
contexts that underpin research, have had a ‘profound influence on my whole conception
of what science is and how scientists operate’.52 In finding a path through this complexity,
Bowlby urged that the historian’s priority must be on attempting to discern what problems a
scientist or a group of scientists were trying to solve.53 If close attention is paid to the prob-
lems that were faced in a particular period, comparison between earlier and later develop-
ments need not result in anachronism. Cautions are required when pursuing such a project.
We should take care not to collapse the problems faced by different periods and how those
problems were understood; we must not assume that words always meant the same thing
over time; and we must not assume that later developments were inevitable or necessarily
the best path that could have been taken. Nonetheless, Bowlby’s book on Darwin strongly
evidences a perspective which has gained ground in recent years within the history of sci-
ence: that earlier and later scientific developments shed light on one another when examined
together.54
Yet, more than this, Bowlby wondered ‘whether something living which has developed
historically can ever be restructured without reference to its historical origins as a social
institution’.55 As such, in Bowlby’s view, historical awareness may not just be helpful but ac-
tually may be a necessary ingredient for the continued vitality of an area of research. This
48 Bowlby, J. (1976) In Dr Martin Bax. Are Mothers Necessary? Radio 3, October 1976. PP/Bow/F.5/7.
49 Bowlby, J. (1989) Attachment and Loss: Continuing Education Seminars. Film produced by David Scott May
and Marion Solomon. Distributed by Insight Media.
50 Bowlby, J. & Dahrendorf, R. (1958) Summary of discussions and topics for final session. Seminar delivered to
members of the Stanford Conflict Seminar, February 1958. PP/Bow/H.67. See also Chang, H. (2017) Who cares
about the history of science? Notes and Records, 71(1), 91–107.
51 Bowlby, J. (1982) A case of mistaken identity. Higher Education Quarterly, 36(4), 328–32.
52 Bowlby, J. (1979) The ten books which have most influenced my thought, 24 October 1979. PP/Bow/A.1/8.
53 Bowlby, J. (1981) Jean Piaget: some reminiscences. The Tavistock Gazette, 5, 3–4, p.4.
54 Tosh, N. (2003) Anachronism and retrospective explanation: in defense of a present-centred history of sci-
ence. Studies in History and Philosophy of Science, 34, 647–59; Oreskes, N. (2013) Why I Am a Presentist. Science in
Context, 26(4), 595–609; Loison, L. (2016) Forms of presentism in the history of science: rethinking the project of
historical epistemology. Studies in History and Philosophy of Science Part A, 60, 29–37.
55 Bowlby, J. (c.1950) Marginalia on Bronfenbrenner’s ‘Toward an integrated theory of personality’. PP/B ow/
J.9/37.
Introduction xvii
argument would also be put forward some years later by both historians and developmental
psychologists who would describe the use of historical methodology in the critical examin-
ation of psychological paradigms as a ‘necessary supplement’ to the hypothesis-testing trad-
ition of academic psychological research.56
Alan Sroufe and the Minnesota group also offered reflections on what it means to know
the past, as part of a deep and abiding concern with the nature of continuities and discon-
tinuities in development over time.57 Like Bowlby, Sroufe was respectful of history and felt
that ‘it is important to bring forward the lessons of the past and at the same time redraw
them with an eye on current problems and current understanding’.58 For Sroufe, the es-
sential commonality between history and developmental psychology lies in the fact that
both acknowledge that early events do not determine later ones. Early events shape what
is taken forward from the past in ways that then frame subsequent interactions between

individuals, groups, or societies and their wider environments. In this account, the past is
not used up but continues to inflect the present, perhaps resourcing and supporting, per-
haps depleting or obstructing what is now possible: ‘the emerging complexity is not speci-
fied by prior features, yet it is founded on them’.59 Furthermore, in Sroufe’s interpretation
of the concept of ‘development’, the present is not the sum of the past. There may well be
ways in which earlier forms possessed strengths for particular purposes that have not been
passed on to later forms. There remains the potential, in Sroufe’s words, for ‘lessons from
the past’.60
Historical entities like attachment research and the structure of a personality can be im-
agined as a maze of little streets and squares, with houses from various periods nonethe-
less situated by earlier structures. It is these structures, which continue to both constrain
and enable what is built today, that come into view when science or a human personality is
considered in historical terms. Both history and developmental science are oriented by an
amazing and strange aspect of the human condition: our pasts are both discontinuous with
our present and, disconcertingly, still with us. The two disciplines agree that we make our
homes on top of and within the standing structures or ruins of our pasts: ‘though we may
be done with the past, the past is by no means done with us’.61 This perspective on the past
suggests a changed attitude to bereavement, to the extent that aspects of the past remain with
56 Van IJzendoorn, M. & van der Veer, R. (1984) Main Currents of Critical Psychology, p.233, trans. M. Schoen.
New York: Irvington Publishers. See also Klempe, S.H. & Smith, R. (eds) (2017) Centrality of History for Theory
Construction in Psychology. New York: Springer.
57 The concept of ‘development’ of course has its own long history. See, for example, Wertheimer, M. (1985)
The evolution of the concept of development in the history of psychology. In G. Eckardt, W.G. Bringmann, &
L. Sprung (eds) Contributions to a History of Developmental Psychology (pp.13–25). Berlin: Mouton; Valsiner, J.
(1994) Irreversibility of time and the construction of historical developmental psychology. Mind, Culture, and
Activity, 1(1–2), 25–42.
58 Sroufe, L.A. (1996) Emotional Development, p.xii. Cambridge: Cambridge University Press.
59 Sroufe, L.A. (2007) The place of development in developmental psychopathology. In A. Masten (ed.)
Multilevel Dynamics in Developmental Psychopathology: Pathways to the Future: The Minnesota Symposia on Child
Psychology, Vol. 34 (pp.285–99). Mahwah, NJ: Lawrence Erlbaum, p.290.
60 Sroufe, L.A. (1996) Emotional Development, p.xii. Cambridge: Cambridge University Press.
61 Roisman, G.I., Madsen, S.D., Hennighausen, K.H., Sroufe, L.A., & Collins, A. (2001) The coherence of dyadic
behavior across parent–child and romantic relationships as mediated by the internalized representation of experi-
ence. Attachment & Human Development, 3(2), 156–72, p.169. An example of attachment researchers ‘building in
the ruins’ is the way that the term ‘internal working model’ has been used by later attachment researchers to show
that Bowlby was attentive to change, since these models were ‘working’, i.e. open to development. However, this
was never Bowlby’s intention with the term: ‘working’ just meant that they were applied (Chapter 1). Nonetheless,
the word ‘working’ has made available this subsequent interpretation.
xviii Introduction
us. Bowlby held that we can even retain the dead as secure attachment figures at a symbolic/
representational level, if we can accept the loss whilst taking courage and reassurance from
memories and other aspects of the person’s legacy.62
Both history and developmental science agree that the past shapes what we can build,
where, and with what stability. Both disciplines recognize that important aspects of our lives
are often best regarded as by-products of the past, rather than immediately functional and
well judged in the present. Yet both perceive that this by-product can be used or adapted re-
sponsively, that contingency is material and runs deep. In making sense of such contingency,
history and developmental science have significant respective commitments that emphasize
the social basis of the self, and the effects of this for the knowing subject. As such, Sroufe and
colleagues expressed concern that when the legacy of the past is ‘unnoticed, disallowed, un-
acknowledged or forgotten’, present-day social practices will likely not be responsive, well

judged, or especially resilient to challenges.63
In agreement with Bowlby’s image of history, Cornerstones takes a stance in proactively
evaluating aspects of attachment as a research paradigm. Particular attention is paid to
aspects of the history of attachment research that have structured or shaped the present, es-
pecially those that have become taken-for-granted over time. Ideas are considered for their
cogency, terminology for its clarity, and empirical claims are appraised against the available
evidence. Appeals by researchers to earlier or contemporary theory or research for authority
or support are evaluated both for the accuracy of the commentary and for the function the
citation appears designed to serve. This includes analysis of the ways in which interpret-
ations of Bowlby and Ainsworth have served as sites for alignment or struggle between later
researchers. Each chapter identifies the strengths and particular insights associated with the
work of the research group under discussion, and changes that have occurred over time in
methodology and theory, and a section at the close of each chapter considers some potential
limitations.
Much of what goes on within a research group occurs behind the scenes. Where the
textual record makes this possible, which is not in every case, the biographical contributories
to the research priorities of principal investigators are identified in the introduction to each
chapter. However, science is a collective work and has a collective legacy. So chapters attempt
to consider the perspectives and efforts of the principal investigators within the context of
their work with collaborators and as embedded within a wider context. Each chapter seeks to
identify the opportunities, debates, and challenges faced by the field of attachment research,
and how these were shaped by and shaped the priorities and concerns of particular research
groups, leading to the making and remaking of methodology, knowledge, and authority over
time. Chapters are intended to be readable as standalones; none necessarily requires know-
ledge of the others. However, the cumulative work of the book as a whole will permit com-
parison and evaluation of the positions of different research groups when confronted with
related concerns. The book as a whole is also intended to facilitate translation, since differ-
ences in method or terminology have often obscured the relationship between the claims of
different groups of researchers.
62 Bowlby, J. (1980) Loss. London: Pimlico: ‘for many widows and widowers it is precisely because they are
willing for their feelings of attachment to the dead spouse to persist that their sense of identity is preserved and
they become able to reorganize their lives along lines they find meaningful’ (p.98).
63 Carlson, E.A., Egeland, B., & Sroufe, L.A. (2009) A prospective investigation of the development of borderline
personality symptoms. Development & Psychopathology, 21(4), 1311–34, p.1315.

Introduction xix
One of the primary forms of attachment-based intervention with families is video-

feedback. Researchers found that showing caregivers exemplars of ‘ideal’ parenting on film
was counterproductive. It did not serve as a useful model, and instead lowered the feelings
of self-worth and self-efficacy of caregivers. However, for a friendly individual to show care-
givers a film of their own behaviour with their child, and watch together, noticing inter-
actions in the film and what stemmed from these moments, had a different effect. This
technique has been repeatedly found to have a meaningful effect on adults’ caregiving be-
haviour (Chapter 6).64 Cornerstones is written with the analogous hope that looking together
at the recent past, with joint attention to how things occurred and what then ensued, may
form a basis for clarifying how things stand and whether there might be other ways of act-
ing in the present and future. This will likely not always be comfortable reading for attach-
ment researchers. At the same time, Cornerstones is written with affection for the field and its

genuine insights into the strange, drunk-dialling human heart.65
Both attachment research and its reception have had enough polemics already. Yet
achieving measured and sincere evaluation is a complex task. There are structural pressures
on historians of science and social historians to take a flatly critical stance towards scientific
claims about family life, treating this as an inappropriate incursion of science as ideology. As
Latour and Bourdieu have observed, the conditions of academic production, which separate
critics in a variety of ways from the practices they are describing, obscure internal differences
in the object of study and may contribute to a wish to ‘debunk’ the scientific work.66 There
has indeed been a tendency in history and sociology to adopt a stance in which psycho-
logical knowledge is regarded as a vast smooth power, without artistry or contingency in its
formation.67 Nonetheless, this tendency for the external observer to be somewhat heedless
of the demands practice makes on insiders may, in certain regards, be part of what history
has to offer. For instance, Cornerstones closely examines matters—for instance the items of
scales for measuring attachment—that researchers themselves have generally simply taken
for granted as workable for practical purposes at a local level. Yet judgements about what is
workable by individual researchers can have huge cumulative unintended consequences as
the years go by. Part of the specific relevance of historical analysis for research psychology is
in the identification and description of such consequences.68
Ainsworth highlighted that empirical research always entails compromises. In her view,
heterogeneity among research groups can be to the benefit of psychological science as a
whole, since the compromises may well be in different places.69 Similarly, reflecting on the
64 Juffer, F., Bakermans-Kranenburg, M.J., & van IJzendoorn, M.H. (eds) Promoting Positive Parenting: An
Attachment-Based Intervention. New York: Psychology Press.
65 Cf. Mykhalovskiy, E., Frohlich, K.L., Poland, B., Di Ruggiero, E., Rock, M.J., & Comer, L. (2018) Critical social
science with public health: agonism, critique and engagement. Critical Public Health, 29(5).
66 Bourdieu, P. (2000) Pascalian Meditations. Stanford: Stanford University Press; Latour, B. (2013) An Inquiry
into Modes of Existence. Cambridge, MA: Harvard University Press. Illustrative of the ‘debunking’ narrative is
Gaskins, S. (2013) The puzzle of attachment. In N. Quinn & J.M. Mageo (eds) Attachment Reconsidered: Cultural
Perspectives on a Western Theory (pp.33–66). London: Palgrave.
67 Sedgwick, E.K. (2003) Touching Feeling: Affect, Pedagogy, Performativity. Durham, NC: Duke University Press.
68 Duschinsky, R. (2019) Attachment and the archive: barriers and facilitators to the use of historical sociology
as complementary developmental science. Science in Context, 32(3), 309–26.

69 Ainsworth, M. (1972) Attachment and dependency: a comparison. In J. Gewirtz (ed.) Attachment and
Dependency (pp.97–137). Washington, DC: Winston: ‘In terms of his problem, theoretical orientation, resources,
opportunities, and personal style, each investigator chooses his own set of compromises. The interests of science
seem likely to be best served in this context by a multiplicity of studies, each with its own compromises, which yet
may in aggregate answer the questions’ (p.126).
xx Introduction
nature of psychological theory, Sroufe has argued that ‘embracing a particular model of dis-
turbance is analogous to putting on lenses which may bring some issues or questions into
focus while distorting others in ways that may not be obvious to the observer’.70 As sug-
gested by Ainsworth and Sroufe’s reflections, though all contributing to the study of attach-
ment in some sense, the research groups considered in Cornerstones have varied strengths
and primary concerns. Treating them together, and with attention to their commentary on
and elaborations of one another, helps reveal these differences and their wider stakes.71 It
also helps in understanding the priorities, methodological choices, and terminology of each
group, which at all times were, in part, structured toward those communicated by or antici-
pated from other research groups as well as the wider discipline. Essentially, Cornerstones
aims to acknowledge and understand the point of view of particular researchers and research
groups, without assuming that this point of view is the only or best one available in the field,

even on their own original ideas and results.72
Ordinary and scientific language
One of the recurrent themes of this book is the way in which communication between re-
search groups, and communication with wider publics, has been hindered by confusion
about the meaning of concepts. Part of the appeal of attachment research lies in its central
reference to experience-near metaphors and terms such as ‘attachment’, ‘mother’, ‘security’,
‘sensitivity’, ‘disorganization’, ‘coherence’, ‘anxiety’, ‘dissociation’, and ‘trauma’. Yet, equally,
part of the difficulty with understanding attachment research is that not one of these terms
is used by attachment researchers in line with ordinary language, and rarely with the same
meaning between research groups.
It is not unusual for terms to take on a life of their own in shaping human perceptions and
actions, both in scientific and ordinary language, a life in turn conditioned by the structures
and conditions within which the language occurs.73 The independent life of language has
been an especially common issue for psychological discourse, as previous historians have
observed.74 The researchers discussed in this book from Bowlby onwards have themselves
been aware of this issue. For instance, Shaver and Brennan have observed that in ordinary
language ‘depression’ can encompass alienation, low self-esteem, helplessness, and dissat-
isfaction with life. However, psychological researchers may well want to distinguish these
states, and investigate their respective contribution to clinical symptoms.75 Ordinary use of
70 Sroufe, L.A. (1997) Psychopathology as an outcome of development. Development & Psychopathology, 9(2),
251–68, p.251.
71 Danziger, K. (1994) Does the history of psychology have a future? Theory & Psychology, 4(4), 467–84: It is
‘when the professional community is divided in some profound way that a critical disciplinary history has a signifi-
cant contribution to make’ (p.478).
72 Hacking, I. (2002) Historical Ontology. Cambridge, MA: Harvard University Press; Collins, H. & Evans, R.
(2014) Actor and analyst: a response to Coopmans and Button. Social Studies of Science, 44(5), 786–92.
73 Cavell, S. (1994) In Quest of the Ordinary: Lines of Skepticism and Romanticism. Chicago: University of
Chicago Press.
74 E.g. Smith, R. (2013) Between Mind and Nature: A History of Psychology. New York: Reaktion Books.
75 Shaver, P.R. & Brennan, K.B. (1991) Measures of depression and loneliness. In J.P. Robinson, P.R. Shaver,
& L.S. Wrightsman (eds) Measures of Social Psychological Attitudes, Vol. 1 (pp.195–289). San Diego: Academic
Press: ‘In addition to being parts of ordinary language, “depression” and “loneliness” are technical terms within
psychiatry and clinical psychology . . . . When ordinary concepts are used technically, definitional confusion may
arise . . . these emotions are closely related to other states discussed in this book: alienation, low self-esteem, ex-
ternal locus of control (helplessness), and dissatisfaction with life. In ordinary language, this is as it should be; in
Introduction xxi
the word ‘depression’ thus diverges from the clinical use of the term—and perhaps both in
turn diverge from the term as used in research contexts. Shaver and colleagues have argued
that the issue expresses the broader predicament of academic and clinical psychology, which
attempt to characterize and support change within the sphere of human everyday life, and
therefore begin with the terms and problems of everyday language.76 And just like everyday
life, ordinary language is unruly, multiply invested, and occasionally nutty or treacherous.
However, even if potential confusions with ordinary language is a broader problem for
psychology, attachment research has been unusually vulnerable from the start. It is helpful
to see that Bowlby was pulled in two directions.77 On the one hand, he was keen to make use
of the advantages of ordinary language. Ordinary language is excellent for doing less precise
work, for making evocative claims, and for communicating with diverse audiences in ways
that resonate with everyday concerns.78 It is intrinsically historical and pitted with depth, a

reserve of images and connotations. Bowlby wanted to communicate simply and evocatively,
making ideas available to a wide public and clinical audiences. He held no tenured academic
post, and especially in the 1950s and 1960s, was intent on speaking beyond the academic
community to support changes in policy and the lives of children and families. Bowlby also
appreciated the flexibility of ordinary language, and the advantages that this could bring
to science. He felt that it can be beneficial in the context of working out ideas, to draw on
ordinary language without strict definitions ‘for, once a definition is laid down, it tends to
straitjacket thought and to control what the worker permits himself to observe’.79
However, there is also a danger in reliance on ordinary language, which Bowlby came to
recognize increasingly over the span of his long career. Use of everyday terms, rich in existing
connotations, ‘makes it extremely difficult to tie any specialised meaning to any particular
word’.80 Above all, the connotations of a term from ordinary language can inadvertently ac-
company the word into scientific language. And no amount of scrubbing and qualifying will
ever fully hold back these connotations from influencing discussions between scientists,
let alone attempts by scientists to communicate with their publics.81 In a chapter drafted
for his final book, Bowlby expressed regret that Darwin’s intellectual legacy has been dam-
aged by assuming the ordinary language connotations of terms that he in fact used in a tech-
nical sense: ‘it has been unfortunate that in his own expositions of his scientific procedures,
Darwin uses words and phrases that have misled readers and have resulted in misconceived
criticism.’ For instance, ‘where we would say we built an explanatory model, Darwin refers to
professional social science it is problematic . . . . Another problem is that the terms “loneliness” and “depression”
harbour implicit causal theories’ (p.195).
76 Shaver, P.R., Morgan, H.J., & Wu, S. (1996) Is love a ‘basic’ emotion? Personal Relationships, 3, 81–96, p.83. See
also Derksen, M. (1997) Are we not experimenting then? The rhetorical demarcation of psychology and common
sense. Theory & Psychology, 7(4), 435–56.
77 Bowlby’s predicament can be seen within the wider context of psychoanalysis discourse in the period, which
both wanted and repudiated the advantages of ordinary language. This issue is considered well in Abram, J. (2007)
The Language of Winnicott: A Dictionary of Winnicott’s Use of Words. London: Karnac.
78 Geertz describes common sense as having five experiential properties: it is felt in use as ‘natural’, ‘practical’,
‘thin’/’simple’, ‘immethodical’, and ‘accessible’. Geertz, C. (1983) Common sense as a cultural system. In Local
Knowledge (pp.73–93). New York: Basic Books.
79 Bowlby, J. (1980) Loss, p.17. London: Pimlico.
80 Bowlby, J. (1972) Notes towards Separation. PP/Bow/K.5./17.
81 Bowlby, J. (1973) Separation, p.118. New York: Basic Books. As a mature scholar, Bowlby regularly warned
his students regarding the use of language in their theorizing. E.g. Issroff, J. (2005) Donald Winnicott and John
Bowlby: Personal and Professional Perspectives. London: Karnac: ‘He concentrated on ensuring that language used
was not loose, and on keeping speculation to a minimum’ (p.26). ‘Often he held forth about the importance of lan-
guage used for conceptualising’ (p.27).
xxii Introduction
himself as ‘speculating’, giving the false impression that he lacked rigour or reason as he put
forward an explanation.82
The most basic example of such problems can be seen with the term ‘attachment’ itself,
which Benjafield situates as one of the most characteristic examples of linguistic polysemy in
all the history of psychology.83 There is a gulf between the ordinary connotations of the term
and how it is used by attachment researchers. And there is a further gap between narrower
and broader uses of the term by Bowlby, and then by subsequent attachment researchers.
In ordinary language, the word means to bind something to something else, physically or
emotionally. In Bowlby’s narrower usage, the word meant a specific set of behaviours and
states that facilitate care-seeking. In Bowlby’s broader usage, the word meant all and any in-
timate relationships. Such multiple investments in the term have had a powerful legacy. They
have contributed to the intuitive appeal of attachment theory, making it seem user-friendly

to diverse publics. And it has contributed to ceaseless miscommunication by and among
researchers. ‘Attachment’ is a fuzzy term, in both the sweet and worst senses. By the 1980s
Bowlby admitted ruefully in correspondence that he kept using the word to describe chil-
dren’s care-seeking behaviours ‘for purely historical reasons’.84 This was despite the fact that
it necessitated work, time and again, to clarify the distinction between the technical usage
and the various connotations of the term, including adjudication between his own earlier
multiple uses.
One of the leaders of the second generation of attachment research, Everett Waters, ac-
knowledged that attachment theory is worse than most areas of psychology for muddling
ordinary and technical scientific language. In Waters’ view, scientists do and likely should
use ordinary language regularly, calling on common metaphors and ideas in order to com-
municate technical notions. This can be a generative process, contributing helpfully to new
theoretical and methodological developments, as well as the circulation of forms of know-
ledge.85 However, Waters has argued, when scientific and ordinary language are mistaken for
one another, the results can be problematic ways that are difficult to notice and redress:
In psychology, and more so, attachment theory, the words we use to label ideas often get
in the way. They misdirect us in what we think we should do next. Many implications that
people draw from their knowledge of attachment theory are probably not rigorously de-
rived from the logic of the underlying theory. Take this example: you ask a college class,
what kinds of developmental problems might arise from being insecure in your attach-
ment to your mother? They start thinking that insecure sounds like afraid, fearful, anxious,
shy, uncomfortable, maybe incompetent, and the reasoning goes on to a conclusion that
82 Bowlby, J. (1987–90) Darwin’s Scientific Achievement. Cambridge University Library, MS Add. 8884.
83 Benjafield, J.G. (2016) The digital history of the anglophone vocabulary of psychology: an exploration using
Zipfian methods. History of Psychology, 19(2), 125–40, p.127.
84 E.g. Bowlby, J. (1983) Letter to Helen Block Lewis, 12 January 1983. PP/Bow/J.9/123: ‘As I expect you know,
some difficulties have arisen over the best use of the term attachment. For purely historical reasons it seems best
now to confine it to protection and comfort-seeking behaviour as seen most obviously in childhood.’
85 On the costs and gains of metaphor for Bowlby’s reception, see Duniec, E. & Raz, M. (2011) Vitamins for the
soul: John Bowlby’s thesis of maternal deprivation, biomedical metaphors and the deficiency model of disease.
History of Psychiatry, 22(1), 93–107. The issue is likewise discussed in Fonagy, P. (2003) Some complexities in
the relationship of psychoanalytic theory to technique. The Psychoanalytic Quarterly, 72(1), 13–47: ‘Science regu-
larly employs metaphor in the absence of detailed knowledge of the underlying process. Provided that metaphor
is not confused with a full understanding—or, to use Freud’s expression, the scaffolding is not mistaken for the
building—heuristic considerations might outweigh any disadvantages of such employment’ (p.36).
Introduction xxiii
insecure is therefore a bad thing. This is not being deduced from some mechanism that is
spelled out in attachment theory. It is merely associative.86
Unless we can be sure that others know just what we mean by ‘attachment’, Waters argued,
severe cautions are needed. In fact, ‘the less often we use the word “attachment” in this dis-
cussion, and the more often we refer specifically to what you are asking about, the better off
we’ll all be’.87 However, Waters’ warning has gone generally unheeded. Attachment research
has had comparatively strong platforms for reporting and synthesizing empirical findings
but weak platforms for the critical discussion of concepts and terminology—besides, to an
extent, the journal Attachment & Human Development and the Handbook of Attachment.88
Luyten has argued that ‘much of the language of . . . attachment theory may have had its
time. There is an unmistakable tendency to reify.’89 Yet any attempt to replace or bypass un-

clear terminology related to attachment will be difficult, perhaps even counter-productive,
unless the various meanings of concepts within the scientific community are understood.
There can be a variety of ways of discerning such meanings, including interviews, focus
groups, and Q-sort tasks.90 However, historical analysis of written material has particular
advantages for tracing lines of continuity and discontinuity in the uses of terms. The five re-
search groups considered in this book are cornerstones in the development of the research
paradigm as it exists today, and have in many regards set the terms of discussion. They are
also, perhaps with the exception of the Minnesota group, the prime originators of the most
serious confusions in the use of attachment language. As a result, historical study of the con-
tributions of these research groups and the debates between them offers both an introduc-
tion to and a clarification of the central concepts and terminology of attachment research.
Summary of chapters
Chapter 1 focuses on the work of John Bowlby. It describes the lines of agreement and dis-
agreement between Bowlby and the psychoanalytic theory of his day, and the extent of his
debt to Robert Hinde and ethology. The chapter clarifies ways in which incompatibilities be-
tween psychoanalysis and ethology have contributed to tensions within Bowlby’s work and
subsequent attachment theory. Access to Bowlby’s unpublished correspondence and notes
provides the basis for a new interpretation of several of Bowlby’s key concepts, including
86 Waters, E. & McIntosh, J. (2011) Are we asking the right questions about attachment? Family Court Review,
49(3), 474–82, p.474.

87 Ibid.
88 Foucault refers to a ‘field of stabilization’ for the concepts and methods of a discipline that make recognition
of equivalence possible. Attachment research has benefited from allowing many an expansive field of stabilization
that allows many phenomena to be recognized as pertaining to ideas relating to attachment. On the concept of
‘field of stabilization’ see Foucault, M. (1969, 1972) The Archaeology of Knowledge and the Discourse on Language,
p.103, trans. A.M. Sheridan Smith. New York: Pantheon.
89 Luyten, P. (2015) Unholy questions about five central tenets of psychoanalysis that need to be empirically veri-
fied. Psychoanalytic Inquiry, 35(1), 5–23.

90 For instance, the term ‘coherence’ has a central place in the conceptualization and coding of the Adult
Attachment Interview (Chapter 3). Beijersbergen, Bakermans-Kranenburg, and van IJzendoorn conducted
a detailed empirical study in 2006 to see whether attachment researchers used the term ‘coherent’ in the same
way as ordinary language or other academic specialisms. The answer was a resounding ‘no’. Beijersbergen, M.D.,
Bakermans-Kranenburg, M.J., & van IJzendoorn, M.H. (2006) The concept of coherence in attachment inter-
views: comparing attachment experts, linguists, and non-experts. Attachment & Human Development, 8(4),
353–69.
xxiv Introduction
monotropy and aggression. The chapter discusses Bowlby’s unpublished book written with
Jimmy Robertson in the 1950s and 1960s on the effects of major separations experienced by
young children. And the chapter presents previously unavailable ideas from Bowlby’s un-
published book on defence mechanisms from the 1960s, which sheds light on his later in-
formation processing model. The chapter also pieces together the full story of Bowlby’s work
with a patient, Mrs Q., the account of which is scattered across a dozen of Bowlby’s writ-
ings. The chapter closes by discussing some ways in which limitations in Bowlby’s work have
proven obstacles for later attachment researchers.
Though the terms ‘attachment theory’ and ‘attachment research’ are sometimes used
interchangeably, attachment as an empirical research paradigm may be regarded as having
fully commenced only with Ainsworth’s work. Chapter 2 begins by introducing the bio-
graphical context of Ainsworth’s work, including her early work at Toronto University.

Ainsworth’s concept of ‘security’ and her attempt to develop self-report measures of security
are reappraised, placing Ainsworth’s work in the context of her debt to her teacher Blatz.
The chapter then draws on Ainsworth’s published and unpublished writings to consider the
strengths and limitations of her Uganda ethnography and Baltimore longitudinal study. This
helps clarify Ainsworth’s goals in her development of a scale to measure sensitivity and in
developing the Strange Situation procedure. A central concern of the chapter is close exam-
ination of the theoretical commitments contained in Ainsworth’s choices in the design of
her coding protocols, including the justifications she provided for characterizing individual
differences in infant attachment as three categories, and how she handled discrepancies.
The chapter also considers the work of Ainsworth’s collaborator and student Everett Waters.
Waters played a critical role in the validation of Ainsworth’s measure, and stimulated an in-
fluential debate about the stability of attachment over time. However, he also acknowledged
the limitations of the Strange Situation procedure, and developed other measures for as-
sessing attachment in childhood and adulthood based on Ainsworth’s ideas. In addition,
the chapter addresses other concerns that have been raised regarding Ainsworth’s work,
including the extent of its cross-cultural validity.
Chapter 3 explores the contributions of Mary Main, Erik Hesse and the Berkeley longitu-
dinal study. The Berkeley group generated the dominant approach to method and theory for
the second generation of attachment research, and helped establish the priorities and values
of the field over recent decades. Drawing on archival materials, the introduction offers a new
interpretation of the development of Main’s work. This highlights the fundamental role she
gave to attentional processes, and leads to a new account of how Main conceptualized min-
imizing and maximizing attachment strategies. Recognition of the centrality of attention to
Main’s theory also helps makes sense of her introduction of the disorganized attachment
classification and her development of the Adult Attachment Interview. The chapter draws
on two unpublished books by Main to describe her methodological innovations, and how
they were achieved, and also to clarify misunderstandings of her goals. This includes dis-
cussion of Main and colleagues’ use of the concepts of ‘disorganization’, ‘fear’, and ‘internal
working model’, and how these related to earlier ideas by Bowlby and Hinde. A particular
focus of the chapter is on the six-year systems for assessing attachment developed by Main
and colleagues. Relatively little information about these coding systems is in print, and yet
close consideration of these methods offers a powerful window into Main’s thinking about
attachment and development, as well as into her more well-known assessments of infant and
adult attachment. The chapter also draws on an examination of the development of the Adult
Attachment Interview coding system from the 1980s to the 2000s to offer clarifications re-
garding Main and Hesse’s ideas regarding ‘lack of resolution’ of loss and trauma. The precise
Introduction xxv
relationship between trauma and dissociation in their thinking—which has often been con-
fused by subsequent attachment researchers—is described, drawing on a major theoretical
work by Main and Hesse published only in Italian.
Chapter 4 considers the work of Alan Sroufe, Byron Egeland, and the Minnesota
Longitudinal Study of Risk and Adaptation. The Minnesota group has served as a fun-
damental source of stability and support for the developmental tradition of attachment
research. The chapter begins by describing the origins of the Minnesota study in the con-
text of growing policy and academic interest in the consequences of child maltreatment.
The chapter presents the first sustained commentary on Sroufe’s ideas about emotion, at-
tachment, and development. These ideas were vital to the selection of measures and in-
terpretation of results in the longitudinal study. Headline concepts like ‘felt security’ were
influential for subsequent attachment theory. However, other ideas such as affects as social

currency, and intrusive intimacy, are interesting but less well known. The chapter exam-
ines the antecedents and sequalae of attachment in the Minnesota study. It then con-
siders the contribution made by the study of attachment at Minnesota to the emergence
of developmental psychopathology as a movement within developmental science. This
includes consideration of Sroufe, Egeland, and colleagues’ distinctive approach to con-
ceptualizing risk and resilience. Two case studies from the Minnesota study are used to
illustrate how the multiple, rich assessments conducted over decades offered the research
group an encompassing picture of human lives. And the legacy of Sroufe and Egeland is dis-
cussed for two former students who have subsequently returned to take leadership roles at
Minnesota: Dante Cicchetti and Glenn Roisman. The chapter also discusses ways in which
Sroufe and Egeland’s theoretical commitment to holism has contributed to both strengths
and limitations in their work.
Chapter 5 discusses Phillip Shaver, Mario Mikulincer and the Experiences in Close
Relationships scale, the most widely used self-report measure of adult attachment. The
chapter begins by revisiting Ainsworth’s reasons for abandoning self-report measures of se-
curity, and Shaver and Hazan’s reason for reigniting this approach. Shaver and Hazen’s devel-
opment of the ‘love quiz’ and early work on adult attachment is discussed, considering ways
in which their ideas converged and diverged from earlier attachment theory. The chapter
then explores the creation of the Experiences in Close Relationships scale, which has pro-
vided a methodological and theoretical basis for the social psychological tradition of at-
tachment research over subsequent decades. The chapter clarifies Shaver and Mikulincer’s
approach to conceptualizing and measuring attachment, and secure base use, and minim-
izing and maximizing strategies. It is anticipated that this will help translation between the
social psychological and developmental traditions of attachment research. The chapter also
considers original contributions made by Shaver and Mikulincer and colleagues through
their inquiries into the relationship of adult attachment styles with sexuality and with reli-
gious practices. The chapter closes with examination of the items of the Experiences in Close
Relationships scale, and the ways in which the mechanics of the measure have limitations for
capturing the implications of both security and trauma for adult attachment.
Acknowledgements
Work on this book has taken five years, during which time I have accumulated an absurd
number of debts. A more detailed account of the experience and nature of these accumu-
lated debts can be found in Duschinsky, R. (2019) Attachment and the archive: barriers
xxvi Introduction
and facilitators to the use of historical sociology as complementary developmental science.

Science in Context, 32(3), 309–26.
First off, my apologies are due to my father for failing, as yet, to turn Chapter 3 into a film.
I agree that it should be. The reader is welcome to be in touch with advice for casting. More
generally, I would be delighted to hear from readers and to learn from their thoughts about
attachment research, past and present.
I am grateful to my mother for first encouraging me to read Bowlby as a teenager. Having
ideas from attachment theory as a reference point through adolescence and adult develop-
ment was a true resource. I have also taken courage from witnessing her indomitability in
the face of health challenges, as well as her thirst for adventure. Conversations with my wife
have likewise contributed directly to this book. Cornerstones benefited from her good judge-
ment in appraising forks in the road, her insights as a clinician, her strength in the face of our

losses, and from the countless ways in which our relationship makes me happy.
The basis of Cornerstones has been access to the John Bowlby, Mary Ainsworth, and Mary
Main/Erik Hesse Archives. My first thanks therefore must go to the Wellcome Collection
for hosting the Bowlby Archive (reference: PP/Bow/), the Nicholas and Dorothy Cummings
Center for hosting the Ainsworth Archive, and to Mary Main, Erik Hesse, and Naomi
Gribneau-Bahm at Berkeley for their efforts to make materials available for study (soon to
arrive at the Wellcome Collections). These are inexhaustible, miraculous collections, and
I feel deeply lucky to have had access to them. This archival research has been enriched by the
kindness Mary Main and Erik Hesse have shown me, in the course of years of correspond-
ence and multiple visits. This has been a transformative gift. I have especially appreciated
our wider conversations about poetry and family, as well as their patience with my endless
questions and criticisms. And—on a personal level—I have appreciated the chance to see,
and hopefully learn something from, a marriage so pervaded in every part by affection and
security.
Special thanks are due to the Wellcome Trust for making it possible for me to work on this
book (Grant WT103343MA), and for encouraging the menagerie of spin-off projects. I have
appreciated Dan O’Connor, Tom Bray, Lauren Couch, Jack Harrington, Jenny Haynes, and
Ross MacFarlane, among others, for their faith in me. Chapter 3 was enhanced by Mary Sue
Moore’s generosity in sharing the detailed notes she took during the 1987 Adult Attachment
Institute in London.
Among colleagues, I am grateful to Sarah Foster, an ally and friend. Sarah’s creativity, me-
ticulousness, and insight have fed this project from the very start, and helped it grow. Thanks
are similarly due to Sophie Reijman, who has also been an ally and friend on this journey. It
has been a joy and a privilege to work with her. Cornerstones is dedicated to Sophie, and to
León, her new little one. Marinus van IJzendoorn and Marian Bakermans-Kranenburg have
been unstinting in their generosity and care, not least in encouraging Sophie Reijman to
come to join me at Cambridge in the first place. I have learnt so much from discussions with
them across the themes of this book, and had great fun spending time together.
Affectionate thanks are likewise due to Judith Solomon for having me to stay during her
Fulbright Visiting Professorship at Vienna University, a visit that has subsequently formed
the basis for many varied and fun conversations. A reader who would like to see ‘proceed-
ings’ from these conversations might take a look at Duschinsky, R., Greco, M., & Solomon,
J. (2015) The politics of attachment: lines of flight with Bowlby, Deleuze and Guattari. Theory,
Culture & Society, 32(7–8), 173–95; or Solomon, J., Duschinsky, R., Bakkum, L., & Schuengel,
C. (2017) Toward an architecture of attachment disorganization: John Bowlby’s published
and unpublished reflections. Clinical Child Psychology and Psychiatry, 22(4), 539–60.
Introduction xxvii
My grateful thanks to Martin Baum, Charlotte Holloway, Janine Fisher, Julie Musk and
Lucía Pérez at Oxford University Press for their support for this book.
Chapters of this book have benefited greatly from feedback from Marian Bakermans-
Kranenburg, Sasha Ban, Kazuko Behrens, Richard Bowlby, Jean-François Bureau, Betty
Carlson, Patricia Crittenden, Tsachi Ein-Dor, Jo Faulkner, Pasco Fearon, Chris Fraley,
Lydia Fransham, Pehr Granqvist, Philip Heslop, Erik Hesse, Jeremy Holmes, Juliet Hopkins,
Michael Lamb, Mary Main, Karin Maraney, Bob Marvin, Mario Mikulincer, Mary Sue
Moore, Mikhael Reuven, Anne Rifkin-Graboi, Glenn Roisman, Avi Sagi-Schwartz, Jessica
Saffer, Carlo Schuengel, Judith Solomon, Alan Sroufe, Paul Stenner, Alessandro Talia,
Anne Tharner, Ross Thompson, Marinus van IJzendoorn, Marije Verhage, Mary Jo Ward,
Everett Waters, and Judy Keiner. I am also grateful to other researchers who have offered
encouragement for this work, including but by no means limited to Byron Egeland, Kelly

Brennan-Jones, Steve Farnfield, Peter Fonagy, Deborah Jacobvitz, Kasia Kozlowska, Mirjam
Oosterman, David Shemmings, Gottfried Spangler, Ruan Spies, Phil Shaver, Miriam and
Howard Steele, Frank van der Horst, Sue White, David Wilkins, and Matt Woolgar.
I am grateful to Tommie Forslund and Kate White for all that I have learnt from them in
the course of editing the Attachment Reader for Wiley and Trauma and Loss: Key Texts from
the John Bowlby Archive for Routledge. The Bowlby family have been remarkable in their
wholehearted support of these endeavours, and their helpful feedback.
At Cambridge, I am grateful to Jonathan Mant and the Primary Care Unit, who provide
my academic home and secure base. The last stage of the book’s composition has been bewil-
dering and painful on various fronts in terms of losses and family health. Jonathan’s kindness
has helped make it possible to continue. I have similarly felt fortunate to be part of Sidney
Sussex College, which has been a supportive community at every turn. Particular thanks are
due to Max Beber, Richard Penty, Brett Gray, and Gary Gerstle for their availability through
difficult times. I have benefited from the mentorship of Mary Dixon-Woods, Claire Hughes,
and Susan Golombok, who have been role models and ever-thoughtful friends. Finally,
I feel deeply fortunate for the intellectual companionship and camaraderie of my imme-
diate research group: Lianne Bakkum, Helen Beckwith, Barry Coughlan, Sarah Foster, Julia
Mannes, Sophie Reijman, Sam Reisz, Guy Skinner, and Melody Turner. I am grateful to them
for conversations about the ideas contained in this book as they have pursued aligned in-
quiries with other methodologies. And I am thankful for their unwavering support through
the challenges of the last year.
1
John Bowlby and the Tavistock
Separation Research Unit
Biographical sketch
John Bowlby reported that his interest in psychological issues was kindled in 1929 whilst

working at a school for troubled children, following studies in natural science at Cambridge.
There he had ‘known an adolescent boy who had been thrown out of a public school for re-
peated stealing. Although socially conforming, he made no friends and seemed emotionally
isolated from adults and peers alike. Those in charge attributed his condition to his having
never been cared for during his early years by any one motherly person, a result of his illegit-
imate birth.’1 The previous year Bowlby had read Sigmund Freud’s Introductory Lectures on
Psychoanalysis; in light of his experiences at the school, Freud’s ideas ‘came alive for me’, and
Bowlby decided to train as a psychoanalyst in London.2 In 1948 he founded the Separation
Research Unit at the Tavistock Clinic and appointed James Robertson as an assistant for a
study of the effects on young children of hospitalisation with no or minimal visitation from
their parents. In 1950 he expanded the research group, appointing Mary Ainsworth as a
clinical postdoctoral researcher (Chapter 2). Reflecting on the observations and ideas of
this research group, Bowlby developed his novel theory of the nature of the parent–child
relationship, of the role of inhibition as a defence against the expectation of rejection, and
the form and nature of grief. This theory found its expression first in a series of articles in
the International Journal of Psychoanalysis, and then in a trilogy of books: Attachment,
Separation, and Loss. Bowlby’s discipline-spanning research made a critical contribution to
thinking about human development, offering a model that mitigated important limitations
of both behaviourism and psychoanalytic theory. This vital contribution to developmental
science, together with his passion for public engagement, has made him one of the most im-
portant and influential psychologists of the twentieth century. It has also made him at once
one of the most accessible to understand superficially, and one of the most difficult to under-
stand in depth.
Introduction
In an unpublished article from 1969, Bowlby wrote that ‘an individual holding an idiosyn-
cratic model of the world or of himself is likely to find himself facing the world alone’.3 This
expectation that individuality and novel ideas will be met with isolation and rejection had a
1 Bowlby, J. (1981) Perspective: a contribution by John Bowlby. Bulletin of the Royal College of Psychiatrists, 5(1),
2–4, p.2.
2 Bowlby, J. (1979) The ten books which have most influenced my thought, 24 October 1979. PP/Bow/A.1/8.
3 Bowlby, J. (1969) Anxiety, Stress and Homeostasis. Unpublished manuscript, April 1969. PP/Bow/H10.
2 John Bowlby
number of sources.4 Bowlby himself attributed his distrust in others in his professional life
in part to the ambush and misunderstanding he received from peers when he first began to
present his ideas at the Institute of Psychoanalysis.5 The expectation of misunderstanding
and rejection led to sharp distinctions throughout his life in what Bowlby would reveal to
different audiences. The result of this was, in a sense, three John Bowlbys, each with some-
what different motivations, and different tones to their writing, reflecting three anticipated
audiences.6 The human self is, of course, always a multiple thing, even if there is a certain
amount of hierarchical organisation. Nonetheless, consideration of three relatively distinct
Bowlbys helps make sense of breakdowns of communication between his writings across
different forums and eras, and between the audiences of these texts.
The most well known to the public is the Bowlby available in works written for a general
audience during the 1950s. These include the famous 1953 Child Care and the Growth of

Love,7 but also his articles in popular women’s magazines8 and presentations to professional
organisations.9 Here a tone of sobriety and daring is expressed as authority, even when the
evidence was—by his own later admission—‘sketchy’ or ‘inadequate’.10 In these texts from
the 1950s, Bowlby argued that a young child needs their mother ‘as an ever-present com-
panion’, providing ‘the provision of constant attention night and day, seven days a week, and
365 days in the year’.11 What he hoped to get across, above all, was that young children should
have someone they feel confident turning to when alarmed. Bowlby had been clearly in-
formed by his wife, based on her own experience, that ‘constant attention’ to a child was both
an impossible and unhelpful aspiration for mothers.12 And late in life, he acknowledged that
he regretted this statement and the implied demand for ever-present care.13 In the 1950s,
however, the language of total presence had the advantage for Bowlby of appealing to the
obviousness and authority of popular British stereotypes about women and children and
nature.14 As has often been remarked, these attitudes were shaped by the context of post-war
4 Van Dijken, S. (1998) John Bowlby: His Early Life: A Biographical Journey into the Roots of Attachment Theory.
London: Free Association Books; Bowlby, U. (1992) A memoir of John. PP/Bow/P.6/3: John believed that ‘his own
childhood had been sufficiently unhappy to want him to investigate—but not so unhappy that he had obliterated
the subject’.
5 Bowlby, J. (1984) Letter to Phyllis Grosskurth, amending discussions of Bowlby in Grosskurth’s biography of
Klein, 10 January 1984. PP/Bow/A.5/7.

6 Bakhtin, M.M. (1981) The Dialogic Imagination: Four Essays by M.M. Bakhtin. Austin: University of Texas
Press; Showers, C.J. & Zeigler-Hill, V. (2003) Organization of self-knowledge: features, functions, and flexibility. In
M.R. Leary & J.P. Tangney (eds) Handbook of Self and Identity (pp.47–67). New York: Guilford.
7 Bowlby, J. (1953) Child Care and the Growth of Love. Harmondsworth: Pelican.
8 See e.g. Bowlby, J. (1954) Should a baby be left to cry? Parents, March 1954, pp.32–35; Bowlby, J. (1958) Should
mothers of young children work? Ladies Home Journal (November) 75, 58–59, 158–61.
9 E.g. Association for Psychiatric Social Workers (1955) Presentation at the Annual General Meeting 1955: Dr
John Bowlby on preventative activities. Modern Records Centre Warwick University. MSS.378/APSW/P/16/6/
19-20.
10 Bowlby, J. (1986) Interview with the BBC. PP/ Bow/F.5/8: ‘I published this report for the World Health
Organisation in 1951, Maternal Care and Mental Health . . . all the evidence was still sketchy, it was inadequate.’
11 Bowlby, J. (1953) Child Care and the Growth of Love. Harmondsworth: Pelican, p.53, 76.
12 Ross, L.R. (2014) Reading Ursula Bowlby’s letters (1939– 1940). Journal of the Motherhood Initiative for
Research & Community Involvement, 5(1), 67–82.
13 Doyle, C. (1987) A continuing case for keeping children at home. Daily Telegraph, 23 June 1987: ‘He insists
that he has never intended to imply that a continuous relationship should mean every minute of the day, and now
adds “intermittent” to “warm and continuous” ’ as the qualities of care he wished children to receive from their fa-
miliar caregivers.
14 Bowlby less fell into the naturalistic fallacy than emblazoned it on his shield. Bowlby, J. (1990, 2011) John
Bowlby: interview by Leonardo Tondo. Clinical Neuropsychiatry, 8(2), 159–71: Tondo—‘Do you agree with the
rather simplistic view that the mother may be more important than the father? Bowlby—‘That view, I think, is well
attested by the information we have . . . This is the way all societies operate. So my concern is always with human
Introduction 3
Britain.15 Bowlby was knowing and explicit that he was drawing on ‘pure prejudice’,16 that in
his popular writings he ‘exaggerates everything’.17 Such a strategy helped him get some core
ideas heard, even if these were mostly the rind of the views he actually held.
Subsequent attachment researchers have engaged in some whack-a-mole efforts to correct
fallacies arising from Bowlby’s populist claims about attachment, for instance his overstrong
claims about the influence of early experiences or his polemical claims about the respon-
sibilities of mothers. However, this has been at most partially successful in shaping public
perceptions of the implications of attachment theory. Some things are irreversible once they
are put out into the world. But additionally, even if they could have been reversed, the second
generation of attachment researchers generally did little, especially compared to Bowlby, to
speak to a wider public.18 This left Bowlby’s early statements in popular works unqualified,
contributing to misalignment between the technical positions of attachment theorists and

criticisms levelled at popular representations of the paradigm.
Among commentators who have read his work carefully, there has been growing recog-
nition that there is little evidence that Bowlby believed in the positions with which his name
became associated in the public imagination.19 At the same time, it should be recognised that
he did little to correct them in his writing for wider publics. In 1954, Donald Winnicott wrote
to Bowlby warning of such consequences, and asking Bowlby to issue a public statement
nature, about which I am most confident about Western culture. When I teach my students, I say, “Look, the first
thing to remember is that Western society is not a human norm.” We behave in a way that human societies have
never behaved in the past. If you take human societies over the past hundred thousand years so far as we know and
around the world, Western societies are peculiar. We do things in funny ways which may be alright, and it may not.
Do not think they are normal. They are not the normal way human beings are meant to behave” . . . You either go
along with human nature or you fight it. If you fight it you get problems. If you don’t fight it life is much more com-
fortable.’ (164–5)
15 E.g. Franzblau, S.H. (1999) Historicizing attachment theory: binding the ties that bind. Feminism &
Psychology, 9(1), 22–31; Vicedo, M. (2011) The social nature of the mother’s tie to her child: John Bowlby’s theory
of attachment in post-war America. British Journal for the History of Science, 44(3), 401–26. Van der Horst has also
situated Bowlby’s work in the wider context of post-war Europe, especially in the context of his travels in the 1950s.
Van der Horst, F.C.P. (2011) John Bowlby—From Psychoanalysis to Ethology. Unravelling the Roots of Attachment
Theory. Oxford: Blackwell.
16 Bowlby, J. (1958) Should mothers of young children work? Ladies Home Journal (November) 75, 58–59,
158–61, p.158. LeVine has claimed that Bowlby had little awareness of the extent to which he was engaging with
contemporary ideologies of family life. Careful examination of the full range of Bowlby’s public and private
writings indicate that the ideological aspects of Bowlby’s popular writings were not simply the result of a lack
of self-awareness, but had a decidedly strategic component. LeVine, R.A. (2014) Attachment theory as cultural
ideology. In H. Otto & H. Keller (eds) Different Faces of Attachment: Cultural Variations on a Universal Human
Need (pp.50–65). Cambridge: Cambridge University Press. In sociological perspective, Bowlby’s appeal to existing
concerns and prejudices of his day may be considered in light of Bourdieu’s remark that when attempting to reach
a mass market, the ‘more directly and completely’ must cultural producers direct their goods to ‘a pre-existing
demand, i.e. to pre-existent interests in established forms’. Bourdieu, P. (1993) The Field of Cultural Production.
Cambridge: Polity, p.97.
17 Bowlby, J. (1987) Baby love. Hampstead and Highgate Express, 3 April 1987.
18 For example, Alan Sroufe (Chapter 4) wrote a popular textbook and a handful of newspaper articles primarily
focused on the overprescription of ADHD medication. But he did not attempt in a sustained way to speak to wider
publics in the manner of Bowlby. In Germany, Klaus and Karin Grossmann were more active in attempting to
speak to wider publics and increase public understanding of attachment. See e.g. Grossmann, K. & Grossmann,
K. (2011) Das Geflecht des Lebens. DVD, Auditorium Netzwerk: Freigegeben ohne Altersbeschränkung. Perhaps
the closest to Bowlby in direct engagement with policy-makers and publics has been Peter Fonagy, e.g. Fonagy,
P. (2018) Evidence submitted to the Evidence-Based Early-Years Intervention Inquiry. Science and Technology
Committee (Commons). http://data.parliament.uk/writtenevidence/committeeevidence.svc/evidencedocument/
science-and-technology-committee/evidencebased-early-years-intervention/written/77644.pdf.
19 E.g. Riley, D. (1983) War in the Nursery: Theories of the Child and Mother. London: Virago Press; Thomson, M.
(2013) Lost Freedom: The Landscape of the Child and the British Post-War Settlement. Oxford: Oxford University
Press, p.87.
4 John Bowlby
revealing his belief that there was, in fact, a shortage of nursery care.20 There is no evidence
that Bowlby replied to this letter, and no public statement was issued. Winnicott later re-
flected on both the advantages and disadvantages of Bowlby’s strategy:
His propaganda for the avoidance of unnecessary breaks in the infant–mother relationship
had gone round the world, though I do also feel that the propaganda element necessarily
led to a fashion in child care and to the inevitable reactions which follow propaganda.21
A good part of the poor relationship between Bowlby and his critics stemmed from this
predicament: in the expectation of misunderstanding and rejection, he had simplified and
excluded qualifications from his position in writing populist propaganda in the 1950s for
the sake of conveying a message that could travel easily between contexts and down through

time; as a consequence, he elicited rejection, and was not able to subsequently make full use
of feedback in further amending his views. For example, he made harsh and mocking re-
marks about his feminist critics, unable or unwilling to take on their concerns, for instance
regarding his often underspecified claims about the harms of ‘separations’.22
When asked in 1976 about the discrepancy between statements in his academic writ-
ings and those in which he used his position as clinician to speak with authority to the
public, Bowlby replied: ‘I would defend this; in different roles one is entitled to speak in
different voices.’23 However, at times it was more than a single person speaking in different
voices: there seemed to be some cases where communication between the different personae
failed or became distorted. For instance, in 1971 he wrote to Michael Rutter ‘I think you mis-
represent me and mislead the reader’ when Rutter attributes to Bowlby the oversimplified
positions—such as an undifferentiated use of the concept of ‘separation’, or a particular focus
on the figure of the mother—that are readily evident from Bowlby’s early popular writings.24
Bowlby appeared genuinely baffled as to why he was being misrepresented, apparently
having forgotten his earlier statements. For instance, in an article from 1981, Rutter wrote
‘Bowlby’s argument is that the child’s relationship with mother differs from other relation-
ships specifically with respect to its attachment qualities’, a passage annotated by Bowlby in
his private copy of the article ‘Where did he get this idea from?’25 A thoroughly different man
is known to those who read Bowlby’s scientific and clinical writings compared to his popular
writings.26 This is surely a good part of why his critics and his successors deal essentially
20 Winnicott, D. (1954, 1987) Letter to Dr J. Bowlby. In F. Robert Rodman (ed.) The Spontaneous Gesture: Selected
Letters of D.W. Winnicott. London: Karnac, pp.65–66. See also Lewis, J. (2013) The failure to expand childcare pro-
vision and to develop a comprehensive childcare policy in Britain during the 1960s and 1970s. Twentieth Century
British History, 24(2), 249–74.
21 Winnicott, D. (1953, 1989) John Bowlby. In C. Winnicott, R. Shepherd, & M. Davis (eds) Psycho-analytic
Explorations (pp.423–32). London: Karnac, p.427.

22 E.g. Ringold, E.S. (1965) Bringing up baby in Britain. New York Times, 13 June 1965: “Dr Bowlby, however,
sticks to his guns. Interviewed in his office at the Tavistock clinic he said: “Whenever I hear the issue of maternal
deprivation being discussed, I find two groups with a vested interest in shooting down the theory. The Communists
are one, for the obvious reason that they need their women at work and thus their children must be cared for by
others. The professional women are the second group. They have, in fact, neglected their families. But it’s the last
thing they want to admit.’
23 Bowlby, J. (1976) Bowlby on latch-key kids: interviews with Dr Nicholas Tucker. Psychology Today, Autumn
1976, 37–41, p.38.

24 Bowlby, J. (1971) Letter to Michael Rutter, 6 October 1971. PP/Bow/J.9/161.
25 Annotations by Bowlby (PP/Bow/J.9/162) on Rutter, M. (1981) Social-emotional consequences of day care for
preschool children. American Journal of Orthopsychiatry, 51(1), 4–28.

26 Bowlby’s scientific and clinical writings are placed together since the tone and arguments are quite well inte-
grated; there are fewer disparities than with Bowlby’s popular writings or his private reflections. Bowlby’s routine
Introduction 5
with two irreconcilable individuals with the same name, and often seem perplexed by one
another. In these scholarly writings, Bowlby presented his ‘idiosyncratic model’ of the world,
laying out its logic. Bowlby’s scientific and clinical writings were just as daring as the popular
texts, but authority is tempered with sobriety. He was concerned with the particularity of
the things he discussed, not the familiar idea of them, even if this demanded dense writing
and some loss of ready readability. In general, there is quite a clean division between the
lively metaphors and appeal to common sense in Bowlby’s popular writings27 and the some-
what dry prose of his writing for medical audiences and the psychoanalytic community, and
in books such as Personality and Mental Illness from 1940 and Attachment, Volume 1 from
1969.28 True, the division between the populariser and the scholar is not complete. For in-
stance, though mild in comparison with the outright homophobia that can be identified in
other psychoanalytic texts in the 1960s, Bowlby leans on metaphors of deficit and futility to

characterise gay and lesbian sexualities in explaining his new account of human motivation
in passages deep inside Attachment, Volume 1.29 However, the fact that even his most hostile
critics have never mentioned these passages evidences the division between his popular and
scholarly readers, corresponding to his popular and scholarly texts.30
The very content of Bowlby’s claims and his use of familiar words was different between
his popular works and his scientific and clinical writings—as was the period of their com-
position: the vast majority of the popular works were published prior to Attachment, Volume
1 in 1969; the works most read and discussed by subsequent attachment research were pub-
lished after 1969. In Attachment, Volume 1, Bowlby was absolutely explicit that ‘almost from
the first many children have more than one figure towards whom they direct attachment
behaviour; these figures are not treated alike; the role of a child’s principal attachment-figure
can be filled by others than the natural mother’.31 In dialogue with academic colleagues,
Bowlby stated clearly that, for children, ‘provided he sees plenty of his principle figure’, who-
ever this is, ‘it is an advantage for him to have others as well’.32 He specified that he intended
the child’s primary caregiver when he used the term ‘mother’, which was standard practice in
the psychoanalytic discourse of his day.33 In his scholarly writings, Bowlby prided himself on
a measured and scientific tone largely without feeling, even as he wrote about its causes. He
largely avoided polemics, and in his academic writings the characterisation of his London
colleagues in the psychoanalytic community was cordial and respectful. Part of Bowlby’s
discipline-spanning courage came from his deep acknowledgement of the limited status of
work as a clinician, and clinical notes, likely represent a further distinct Bowlby. However, there are few available
textual traces, except insofar as this work contributed to Bowlby’s scholarship or his private reflections. Most of
Bowlby’s clinical notes are closed still for several decades in the Wellcome Collections.
27 Duniec, E. & Raz, M. (2011) Vitamins for the soul: John Bowlby’s thesis of maternal deprivation, biomedical
metaphors and the deficiency model of disease. History of Psychiatry, 22(1) 93–107.
28 Bowlby, J. (1940) Personality and Mental Illness. London: Kegan Paul; Bowlby, J. (1957) An ethological ap-
proach to research on child development. British Journal of Medical Psychology, 30(4), 230–40; Bowlby, J. (1969,
1982) Attachment, Volume 1. London: Penguin.
29 Ibid. pp.130–31.
30 A partial exception is Vicedo, M. (2011) The social nature of the mother’s tie to her child: John Bowlby’s
theory of attachment in post-war America. British Journal for the History of Science, 44(3), pp.401–26, where there
is valuable analysis of links between Bowlby’s popular and academic writings of the 1950s.
31 Bowlby, J. (1969, 1982) Attachment, Volume 1. London: Penguin, pp.303–4.
32 Bowlby, J. (1971) Letter to Michael Rutter, 6 October 1971. PP/Bow/J.9/161.
33 Melanie Klein had set a trend in which the word ‘mother’ (or ‘breast’) was deployed as a synecdoche for the
infant’s experience of the caregiving environment in general, rather than referring specifically to the biological
mother. See Hinshelwood, R.D. (1989) A Dictionary of Kleinion Thought. London: Free Association Books.
6 John Bowlby
theory, as a conjecture to then be appraised against the available evidence.34 In his notes on
the philosopher of science Karl Popper, Bowlby wrote: ‘Intelligibility requires the model to
be cast in terms mainly of some analogous and better understood set of ideas. Plausibility
that it does not affront widely held assumptions.’35 His daring and genre-busting theoretical
works, in this sense, were also set up from the start in expectation of being debated, as part of
the scientific process.
Bowlby as populariser treated readers as needing to be coaxed to even the most crude
points through appeal to their stereotypes and preconceptions. The scientific and clinical
commentator held his audience firmly at bay with a carefully orchestrated remoteness.
The Bowlby Archive is also a reflection of another kind of reserve and distrust, evident in
the hidden array of wayward, profound thoughts that he did not trust to print. Bowlby’s
wife described him as ‘completely inarticulate’ when he tried to talk about feelings.36 Yet

in his private writings and correspondence available in the Archive, Bowlby is visible as a
human: with deep feelings, a personal emotional fragility, frequently thoughtful about his
personal limitations. (A volume of writings that reflect this ‘hidden’ Bowlby has recently
been published by Duschinsky and White, with the encouragement of the Bowlby family,
under the title Trauma and Loss: Key Texts from the John Bowlby Archive.)37 In his notes
Bowlby took an interest in monkeys left alone in zoo cages. When angered by visitors, they
self-mutilate. Indeed, remarked Bowlby, how like cages are our inhibitions.38 The Archive
reveals the engine-room of Bowlby’s thinking behind his inhibitions, in the ordering and
transformation of difficult feelings evoked by the world, and by himself, forming the mater-
ials for imagination, insight, and propositions. It is this engine that motored his thinking, as
he sought sense in the stories he was hearing from veterans from the battlefields of France
and the sorrows and symptoms of children.
In the mid-1950s, Bowlby was part of a series of meetings by the ‘Psychobiology of the
Child’ study group, organised by Ronald Hargreaves at the World Health Organisation. The
discussants included leading researchers from across the world: Jean Piaget, Erik Erikson,
Konrad Lorenz, Margaret Mead, and Ludwig von Bertalanffy. A transcription of the dis-
cussions was published. In a strangely unguarded moment, the transcript reveals Bowlby
reflecting on his own experiences in considering child development and the predicament of
patients seen in his clinical work. ‘It seems to me’, Bowlby said, ‘that the main problem with
which we are all faced in the process of growing up is that of making a tolerable and com-
patible synthesis out of a number of manifestly incompatible components.’ Within us are
all kinds of positions and elements, including tender feelings and callousness, kindness and
cruelty, and ‘these things are literally incompatible’. If an individual wishes to benefit from
34 Bowlby, J. (1982) A Secure Base. London: Routledge: ‘All knowledge is conjectural and . . . science progresses
through new theories coming to replace older ones when it becomes clear that a new theory is able to make sense
of a greater circle of phenomena than are comprehended and explained by an older one and is able to predict new
phenomena more accurately.’ (84)
35 Bowlby, J. (c.1982) Popper’s evolutionary epistemology. PP/Bow/H.98. The passage is heavily underlined in
red pen for emphasis.

36 Ursula Bowlby, cited in Karen, R. (1998) Becoming Attached: First Relationships and How They Shape Our
Capacity to Love. Oxford: Oxford University Press, p.29.

37 Duschinsky, R. & White, K. (eds) (2019) Trauma and Loss: Key Texts from the John Bowlby Archive.
London: Routledge.
38 Bowlby, J. (1933–36) Research notes for uncompleted PhD. PP/Bow/D.1/2/13: ‘Zuckerman describes how
monkeys will mutilate themselves if they are alone in their cages and visitors make them angry . . . Although not in
visitor-cages, humans are often in cages of inhibition.’
Introduction 7
these different parts, the first thing is that he must ‘own to all these different parts of himself ’,
rather than pretend that they do not exist. And secondly, he must ‘gradually relate them in
some self-balancing unity’.39
There is little unity between Bowlby as populariser, as scholar, and as private thinker through
the 1950s and early 1960s. From the late 1960s, the relations between these three personae be-
came more integrated. In this period he came to greater acknowledgement of misunderstand-
ings that had arisen as a result of such divisions, and also to reflect in his theoretical work on
the damage that may be done by inhibiting communication between aspects of the self. By the
1970s, with retirement from clinical work, Bowlby further scaled back his activity as a public
figure to focus more on his scholarship and his family. He almost exclusively ceased to publish
popularizing works after 1969. In discussions with his son, Bowlby explained at the time that
he needed to focus on scholarly rather than popular works as he ‘could not afford to be taken

as a lightweight’.40 In interview he continued to offer polemical views that played on sexist at-
titudes towards women.41 But in the final two decades of Bowlby’s life, his central focus was on
securing the scholarly standing and clinical relevance of attachment theory—and on pursuing
meticulous historical work of the life of Charles Darwin. His desire to influence wider publics
was allowed to slide. He made no attempt to update them on developments and qualifications of
his position between the 1950s and 1980s, even in late interviews, focusing instead on his classic
message of the importance of maternal availability for young children.42
The dominance of the three Bowlbys varied over time, with the populariser quiet after
1969 and the academic scholar in much greater ascendance. However, over the span of
Bowlby’s career, a certain generative interplay can be identified. The private thinker fed the
scholar experiences and ideas about intimacy and pain, though these were very thoroughly
intellectualised by the time they hit the printed page. The private thinker also fed the popu-
lariser his passion and his courage—as well as buffering him from the effects of rejection and
misunderstanding, which were anticipated in advance. The scholar permitted the private
thinker some order and containment, and gave the populariser credibility. And the popular-
iser provided the private thinker both a spur and an outlet, and provided the scholar a wildly
increased audience for key claims. There was, however, a personal price to this self-balancing
unity. In the World Health Organisation discussions in the mid-1950s, Margaret Mead re-
sponded to Bowlby’s remarks by asking a penetrating question: ‘What are the conditions of
disassociation in which you do or do not own to this part of your personality?’ Bowlby’s reply
39 Bowlby, cited in Tanner, J.M. & Inhelder, B. (eds) (1958) Discussions on Child Development: Proceedings of the
WTO Study Group of the Psychobiological Development of the Child, Vol. 3. London: Tavistock, p.207. Use of the
first person plural (‘the main problem with which we are all faced’) is sufficiently unusual to permit a biographical
interpretation, especially given the salience of the theme of integration for Bowlby. It should be acknowledged, of
course, that this could have simply been a manner of speaking.
40 Richard Bowlby, personal communication, February 2019. An exception is Bowlby, J. (1974) A guide to the
perplexed parent. New York Times, 2 March 1974.

41 E.g. Bowlby, J. (1976) Bowlby on latch- key kids: interviews with Dr Nicholas Tucker. Psychology Today,
Autumn 1976, 37–41.
42 E.g. Doyle, C. (1987) A continuing case for keeping children at home. Daily Telegraph, 23 June 1987. Careful
examination of the interview signals shifts in Bowlby’s thinking—even his regret in his use of the term ‘continuous’
care by mothers in his early writings. Yet the dominant narrative remains remarkably similar to the early popular
works, and there is no attempt to explain theoretical developments. Bowlby’s lack of explicit attempt to clarify
revisions to his picture of child–caregiver relationships was likely supported by a lack of interest in such develop-
ments in media forums, where reference to basic stereotypes about mothers and children made for more accessible
reading.
8 John Bowlby
is remarkable: ‘I think one could refer to the notions of forgivability and unforgivability.’43
The ultimate question, Bowlby indicates, is whether the parts of oneself can accept one an-
other, can forgive one another, for what may be irreversible.
Bowlby and psychoanalysis
Of special importance for Bowlby’s popularizing, scholarly, and private writings was his
training and work as a psychoanalytic clinician. His central ideas emerged from within psy-
choanalytic theory and clinical work. Not least, Bowlby’s earliest attention to the term ‘attach-
ment’ was in annotations from 1942 on Young Children in War-Time by Dorothy Burlingham
and Anna Freud,44 where he marked the term wherever it appeared. Bowlby not only under-

lined but also highlighted and starred the following passages: ‘Whenever certain essential
needs are not fulfilled, lasting psychological malformations will be the consequence; these
essential elements are the need for personal attachment, for emotional stability, and for per-
manency of educational influence’ and ‘It is a known fact that children will cling even to
mothers who are continually cross and sometimes cruel to them.’ That there will be some
form of ‘attachment of the small child to his mother’ seemed to Burlingham and Freud ‘to
a large degree independent of her personal qualities’.45 The latter passage was prominently
cited 16 years later in Bowlby’s foundational article on ‘Nature of the child’s tie to his mother’,
published in 1958.46
In a late manuscript circulated only to other psychoanalysts,47 Bowlby explained that, as
a young psychoanalyst, he envied colleagues (Donald Winnicott is implied) their clinical in-
tuition and grace. And he saw other peers doing good clinical work in spite of the theories to
which they subscribed. However, ‘I have not that sort of mind, nor am I strong on intuition.
Instead, I tend to apply such theories as I hold in an effort to understand my patient’s prob-
lems. This works well when the theories are applicable but can be a big handicap when they
are not.’ Without clinical grace or intuition, Bowlby felt that he had no choice but to develop a
new theory. Lay and colleagues succinctly identify that this new theory salvaged five aspects
of psychoanalytic theory: ‘(1) that infants have a complex social and emotional life, (2) that
43 Tanner, J.M. & Inhelder, B. (eds) (1958) Discussions on Child Development: Proceedings of the WTO Study
Group of the Psychobiological Development of the Child, Vol. 3. London: Tavistock, p.208. Some legacy of Klein may
be felt here, in the idea of forgiveness as psychological integration.
44 Burlingham, D. & Freud, A. (1942) Young Children in War-time. Oxford: Allen & Unwin.
45 Annotations by Bowlby dated 1942 on Burlingham, D. & Freud, A. (1942) Young Children in War- time.
Oxford: Allen & Unwin, pp.10, 47. Copy held in the library of Human Development Scotland. Earlier than
Burlingham and Freud, Ian Suttie had written that ‘instead of an armament of instincts—latent or otherwise—the
child is born with a simple attachment-to-mother who is the sole source of food and protection . . . the need for a
mother is primarily presented to the child mind as a need for company and as a discomfort in isolation’. Suttie I.
(1935) The Origins of Love and Hate., London: Free Association Books, p.15. It is possible that the use of the term
came to Bowlby via Suttie and/or conversations through the Tavistock or the wider London psychoanalytic scene.
In any case, no textual record is available of Bowlby having read Suttie until after Anna Freud. On the link between
Bowlby and Suttie see van der Horst, F.C. & van der Veer, R. (2010) The ontogeny of an idea: John Bowlby and con-
temporaries on mother–child separation. History of Psychology, 13(1), 25–45.
46 Bowlby, J. (1958) Nature of the child’s tie to his mother. International Journal of Psycho-Analysis, 39, 350–
73. Bowlby’s first published use of the term ‘attachment’ would appear in Bowlby, J., Ainsworth, M., Boston, M.,
& Rosenbluth, D. (1956) The effects of mother–child separation: a follow-up study. British Journal of Medical
Psychology, 29, 211–47, p.237.
47 Bowlby, J. (1985, 1991) The role of the psychotherapist’s personal resources in the treatment situation. Bulletin
of the British Psychoanalytic Society, 27(11), 26–30. Published as Chapter 12 in Duschinsky, R. & White, K. (eds)
(2019) Trauma and Loss: Key Texts from the John Bowlby Archive. London: Routledge.
Bowlby and psychoanalysis 9
early experiences can have lifelong implications, (3) that mental representations of early ex-
periences mediate effects on later behavior and development, (4) that defensive processes
play a role in affect regulation, and (5) that loss of an attachment figure—at any age—is an
emergency and mourning is a process that serves an adaptive affect-regulation function.’48
Yet despite these continuities, Bowlby perceived three major problems with psychoanalytic
theory and method: a weak recognition of actual family experiences in shaping child psych-
ology; a mistaken account of the causes of incompatible elements within the human mind;
and a conflation of self-preservative actions and sexuality.
Actual experience

The issue of actual experiences was perhaps the earliest of the three central problems Bowlby
found with the psychoanalytic theory of his day. Bowlby’s notes from the 1930s and his own
retrospective accounts both suggest that Bowlby entered psychoanalytic training in the be-
lief that Freud attributed the emotional problems of his patients to adverse experiences, es-
pecially in the context of the family. It was an uncomfortable surprise for him to learn that
the psychoanalytic establishment in London, and the work of Melanie Klein in particular,
had come to downplay past experience as the basis of symptoms and instead emphasise the
role of fantasy.49 In a record of his dreams from the early 1930s, Bowlby reported a dream in
which he was imprisoned by the Nazis. Likely with a certain irony, he interpreted the fantasy
material as reflecting his anxiety about how to think about fantasy, the unconscious, and
sexuality, and about the status of psychoanalytic knowledge in the society around him.50
It was not that Bowlby thought that fantasy was unimportant. But he believed that it had
come to be badly overemphasised, at the expense of attention to the effects of biographical
experiences.
In private conversations or in interview, when Bowlby was asked to explain the origins of
attachment theory, he would frequently refer to clinical cases he saw under the supervision
of Melanie Klein. These cases encapsulated key problems Bowlby experienced as a young
psychoanalyst. From around 1938, he began treatment of a small boy of about three years
old, who was hyperactive, anxious, and aggressive. The boy’s mother was evidently highly
troubled, but Melanie Klein insisted that the mother should just sit outside the room. This
made Bowlby uncomfortable. He had previous experience of delivering therapy to families
within a Child Guidance Clinic, which had convinced him that ‘the problem which as a rule
we need to solve is the tension among all the different members of the family’.51 Three or
four months into the treatment of the little boy, the mother was admitted as an inpatient to a
mental hospital. Bowlby was appalled that Melanie Klein was not interested in the impact of
this on the boy, except insofar as it had the practical effect of interrupting the analysis. World
48 Lay, K.L., Waters, E., Posada, G., & Ridgeway, D. (1995) Attachment security, affect regulation, and defen-
sive responses to mood induction. Monographs of the Society for Research in Child Development, 60(2–3), 179–
96, p.179. On continuities between attachment and psychoanalysis see also Eagle, M. (1995) The developmental
perspectives of attachment and psychoanalytic theories. In S. Goldberg, R. Muir, & J. Kerr (eds) Attachment
Theory: Social, Developmental & Clinical Perspectives (pp.123–50). Hillsdale, NJ: Analytic Press.
49 Bowlby, J. (1933–36) Research notes for MD thesis: ‘Anxiety—Essays’. PP/Bow/D.2/46/6; Bowlby, J. (1981)
Perspective: a contribution by John Bowlby. Bulletin of the Royal College of Psychiatrists, 5(1), 2–4.
50 Bowlby, J. (c.1933) Dreams. PP/Bow/D.2/45/7.
51 Bowlby, J. (1949) The study and reduction of group tensions within the family. Human Relations, 2, 123–28.
10 John Bowlby
War II halted Bowlby’s training as a child analyst: ‘When it was over and I was free to resume
my training I could not face doing so however: the absurdity of treating young children and
neglecting their parents was too much for me.’52
In staking out the importance of actual experiences for children, Bowlby developed new
terminology. Kleinian theorists had used the term ‘internal object’ to refer to the image of the
parent held by a child. Bowlby was dissatisfied with this term. He felt that it left the ‘object’ a
shimmering abstraction rather than anything concrete. Influenced by the growing interest in
cognitive and representational processes of the 1960s,53 Bowlby offered instead the term ‘in-
ternal working model’. The term was used by Bowlby to gesture to the cognitive components
associated with the attachment behavioural system, as a way to ‘broach the large, difficult, and
profound questions of how a child gradually builds up his own “internal world” ’:

Starting, we may suppose, towards the end of his first year, and probably especially actively
during his second and third when he acquires the powerful and extraordinary gift of language,
a child is busy constructing working models of how the physical world may be expected to be-
have, how his mother and other significant persons may be expected to behave, how he himself
may be expected to behave, and how each interacts with all the others. Within the framework
of these working models he evaluates his situation and makes his plans. And within the frame-
work of the working models of his mother and himself he evaluates special aspects of his situ-
ation and makes his attachment plans.54
Bowlby’s proposal was that early attachment relationships shape cognitive representations that
then inform action. This accounts for continuities between children’s experiences of early care
and their later expectations of their caregivers, and subsequently of other people. The idea of the
‘internal working model’ was not, as Bretherton has observed, a fully worked out theory. In part
it was a metaphor, one that allowed Bowlby above all to highlight in a general and encompassing
way that representations, as ‘models’, should be regarded as tolerably accurate encapsulations of
the history of past experiences.55
A clear disadvantage was that the ‘model’ included a huge variety of cognitive content. It
is not really clear what the term means, except that it refers to a representation relevant to
attachment. At one point or another, most second-generation attachment researchers have
stated that Bowlby’s gestural use of the term ‘internal working models’ has made it difficult
to generate specific hypotheses using the idea.56 Indeed, part of the ritualised inheritance
Klein, 10 January 1984. PP/Bow/A.5/7. Bowlby requested that Grosskurth add to her biography of Klein the fol-
lowing passage: ‘It was fortunate for Bowlby that the war then intervened. He tells people that it saved him from
open conflict with Melanie Klein for which he would not then have been ready.’
53 See Gardner, H. (1986) The Mind’s New Science: A History of the Cognitive Revolution. New York: Basic Books.
54 Bowlby, J. (1969, 1982) Attachment, Volume 1. London: Penguin, p.353. Internal working models are ‘none
other than the internal worlds of traditional psychoanalytic theory seen in a new perspective’ (82).
55 Bowlby, J. (1969, 1982) Attachment, Volume 1. London: Penguin, p.82. Bretherton, I. (1998) Internal working
models and communication in attachment relationships. In A. Braconnier & J. Sipos (eds) Le Bébé et les Interactions
Précoces (pp.79–90). Paris: Presses Universitaires de France: ‘I would urge, however, that the various metaphors
Bowlby . . . used as tools to think about internal working models are not to be taken too literally.’ (79)
56 See e.g. Crittenden, P.M. (1990) Internal representational models of attachment relationships. Infant Mental
Health Journal, 11(3), 259–77; Bartholomew, K. (1990) Avoidance of intimacy: an attachment perspective. Journal
of Social and Personal relationships, 7(2), 147–78; Main, M. (1991) Metacognitive knowledge, metacognitive moni-
toring, and singular (coherent) vs. multiple (incoherent) models of attachment: some findings and some directions
for future research. In P. Marris, J. Stevenson-Hinde, & C. Parkes (eds) Attachment Across the Life Cycle (pp.127–
59). London: Routledge; Collins, N.L. & Read, S.J. (1994) Cognitive representations of attachment: the structure
of Bowlby’s ideas by second-generation attachment researchers was criticism and revision

of the concept of the internal working model. It has served as a wide, undefined target onto
which each researcher can project his or her own image of what an attachment theory should
be. Nonetheless, at the time, there were distinct advantages to the metaphor for Bowlby’s
purposes. The idea of the ‘working model’ helped Bowlby to emphasise that expectations
about relationships are in constant use, day in and day out, as individuals respond to pre-
sent demands and forecast future needs.57 It also helped him emphasise that expectations
can lead us to search for confirmatory input. Interpretations by later attachment researchers
of the term ‘working model’ as suggesting ‘provisional (in the sense of “working” drafts—
changeable plans)’58 are a post hoc reconstruction: Bowlby apparently did not intend this
connotation.59 His most important emphasis in using the term was to contrast the psycho-
analytic emphasis on fantasy with his own emphasis on models as tolerably accurate repre-

sentations of what actually happened in childhood.
As Peter Fonagy and Morris Eagle have noted, Bowlby’s opposition between the psycho-
analytic theory of his day and his own focus on ‘internal working models’ that reflected the
role of actual childhood experiences was, for polemical reasons, oversimplified.60 There were
analysts like W.R.D. Fairbairn, who held aligned perspectives, as Bowlby himself acknow-
ledged.61 And Anna Freud chided Bowlby for his depiction of her as a secondary-drive the-
orist, with a position similar to Klein’s; Freud felt that her stance was much closer to Bowlby’s
in emphasizing the importance of actual experiences of caregiving for understanding anx-
ious and aggressive behaviours by children.62 Furthermore, it should be highlighted that
and function of working models. In K. Bartholomew & D. Perlman (eds) Advances in Personal Relationships, Vol.
5 (pp.53–90). London: Jessica Kingsley; Waters, H.S. & Waters, E. (2006) The attachment working models con-
cept: among other things, we build script-like representations of secure base experiences. Attachment & Human
Development, 8(3), 185–97.
57 Bowlby, J. (1988) Developmental psychiatry comes of age. American Journal of Psychiatry, 145, 1–10.
58 Mikulincer, M. & Shaver, P.R. (2003) The attachment behavioral system in adulthood: activation, psycho-
dynamics, and interpersonal processes. In M.P. Zanna (ed.) Advances in Experimental Social Psychology, Vol. 35
(pp.53–152). New York: Academic Press, p.60. The idea of working models as changeable, provisional representa-
tions is likewise a central theme in many more applied works making appeal to Bowlby, e.g. Johnson, S.M. (2019)
Attachment Theory in Practice. New York: Guilford.
59 Alan Sroufe, personal communication, January 2019: ‘Bowlby never used “working model” to mean provi-
sional. I asked him explicitly because I kind of liked that idea. He said no, he didn’t mean that.’
60 Fonagy, P. (1999) Points of contact and divergence between psychoanalytic and attachment theories: is
psychoanalytic theory truly different. Psychoanalytic Inquiry, 19(4), 448–80; Eagle, M. (2013) Attachment and
Psychoanalysis. New York: Guilford.
61 Scharff, D.E. & Fairbairn Birtles, E. (1997) From instinct to self: the evolution and implications of W.R.D.
Fairbairn’s theory of object relations. International Journal of Psycho-Analysis, 78, 1085–103: ‘Bowlby specifically
acknowledged his Fairbairnian orientation in the development of attachment theory and the ethological approach
to infant development (personal communication)’ (1100). Bowlby’s annotations on Fairbairn’s writings offer testa-
ment to his alignment and agreement with the latter’s positions. For instance, he wrote ‘crucial points’ in the mar-
gins when Fairbairn emphasized that even a parent perceived as unkind may still be the person a child will want
to turn to for comfort. Annotations by Bowlby on Fairbairn, W.R.D. (1944) Endopsychic structure considered in
terms of object-relationships. International Journal of Psycho-Analysis, 25, 70–92. Bowlby’s copy of this paper is
held by Richard and Xenia Bowlby.
62 Anna Freud urged recognition that the idea of the ‘pleasure principle’ in psychoanalytic theory was not op-
posed to the idea of attachment; all it posited was that when impulses or responses were activated for a child, a
homeostatic response would be initiated to reduce the feeling of tension or motivation, especially in ways that pro-
vide pleasure or comfort. This was therefore aligned with, rather than contrary to, the idea of attachment as a homeo-
static system. Freud, A. (1958, 1960, 1969) Discussion of John Bowlby’s work on separation, grief and mourning.
In A. Freud & D.T. Burlingham (eds) The Writings of Anna Freud, Vol. 5 (pp.167–86). New York: International
Universities Press. Press. Blum’s statement that ‘Bowlby’s ideas angered just about everyone he knew. Anna Freud
dismissed him outright’ is a thoroughgoing oversimplification, based on selective reading. Blum, D. (2002) Love at
Goon Park: Harry Harlow and the Science of Affection. New York: Basic Books, p.59.
12 John Bowlby
Bowlby did not assume that all adult recollection reflected historical experience in any simple
way. Unpublished clinical cases from the 1930s to the 1960s and published later reflections
show Bowlby’s attentiveness in his clinical work to the role of psychological processes that
may shape, edit, or distort a patient’s account of their actual experiences.63 In particular, he
was interested in the way that early experiences of care could contribute to a tendency to ei-
ther unrealistically denigrate or idealise attachment figures, and the therapist.64
Nonetheless, it is also true that Bowlby’s emphasis on actual experience represented a shift
in clinical technique and in epistemology. In terms of clinical technique, it oriented his dis-
cussions with patients and interpretation of their symptoms towards consideration of past
experiences.65 And in terms of epistemology, Bowlby regarded psychoanalytic knowledge as
compatible with other forms of scientific knowledge and measurement. This distinguished
Bowlby from even sympathetic members of the London Institute at the time.66 It was a pos-

ition that would open the door to both the interdisciplinary synthesis and the focus on con-
crete observational measures that would lie at the heart of attachment theory. It would also
contribute to the backlash of psychoanalysis against attachment theory. In contrast to psy-
choanalysts who enacted their rebellions more implicitly, the focus on actual experiences
at the expense of fantasy has meant that even to this day attachment theory is not formally
taught at the British Psychoanalytic Institute.67 This is highly unfortunate. In a longitudinal
follow-up study of children treated at the Anna Freud Centre, Target and Fonagy described
several cases in which therapists, having been taught that representations of parents repre-
sented fantasy, interpreted children’s reports of abuse as such, rather than recognising them
as reports of actual events. These children had especially poor long-term outcomes.68
There were, then, several important advantages to Bowlby’s introduction of the idea of
‘internal working models’ in negotiating the relationship between his ideas and psychoana-
lytic perspectives of the day. However, it should be noted that Bowlby’s use of the concept
incorporated several of the same flaws as the psychoanalytic terms, such as ‘object’, it was
replacing.69 First, and most significantly, the concept was simply too encompassing, since
63 Bowlby, J. (undated) Maternal behaviour: humans. PP/Bow/H.136; Bowlby, J. (1979) On knowing what you are
not supposed to know and feeling what you are not supposed to feel. Canadian Journal of Psychiatry, 24(5), 403–8.
64 This was the topic of his first published paper: Bowlby, J. (1940) The influence of early environment in the
development of neurosis and neurotic character. International Journal of Psycho-Analysis, 21, 154–78: ‘Every pa-
tient who comes to us has a distorted view of his parents . . . Some patients will project all that they feel to be bad in
themselves on to their parents and blame and hate their parents. Others will project all the good and idolize their
parents’ (176). Such reflections on two classes of distortion in adults’ perceptions of their parents align with the
later position of Mary Main (Chapter 3).
65 Bowlby, J. (1977, 1979) The making and breaking of affectional bonds. In The Making and Breaking of
Affectional Bonds (pp.150–88). London: Routledge.

66 See e.g. Isaacs, S. (1944) Letter to Major Bowlby, 20 October 1944. PP/Bow/J.9/111.
67 Fonagy, P. (2015) Mutual regulation, mentalization, and therapeutic action: a reflection on the contributions
of Ed Tronick to developmental and psychotherapeutic thinking. Psychoanalytic Inquiry, 35(4), 355–69.

68 Target, M. & Fonagy, P. (2003) Attachment theory and long-term psychoanalytic outcome: are insecure at-
tachment narratives less accurate? In M. Leuzinger-Bohleber, A.U. Dreher, & J. Canestri (eds) Pluralism and Unity?
Methods of Research in Psychoanalysis (pp.149–67). London: International Psychoanalytical Association, p.163.
69 Fonagy attempted to spell out clearly some of the elements Bowlby absorbed within the concept of ‘internal
working model’: ‘Four representational systems compose the internal working model (IWM): (1) expectations of
interactive attributes of early caregivers created in the first year of life and subsequently elaborated; (2) event repre-
sentations by which general and specific memories of attachment-related experiences are encoded and retrieved;
(3) autobiographical memories by which specific events are conceptually connected because of their relationship
to a continuing personal narrative and developing self-understanding; and (4) understanding of the psychological
characteristics of other people (inferring and attributing causal motivational mind states, such as desires and
emotions, and epistemic mind states, such as intentions and beliefs) and differentiating these states from those
of the self.’ Fonagy, P. (2001) The human genome and the representational world: the role of early mother–infant
it seemed to refer to any and all cognitive content about how interactions work within rela-
tionships between self and others. For instance, the representational model was situated as
both a processor of experience and a repository for experiences, with use of the term often
bouncing between these two very different meanings. Second, there was a lack of clarity re-
garding whether the model was specific to a relationship or general across relationships on
the model of early experiences.70 Third, though clearly infants and adults both have expect-
ations about the availability of attachment figures, it is not clear that it makes sense to use the
concept of ‘model’ to refer to both the basic non-representational goal-directed expectancies
of infants and the elaborate representations held by adults, given all the differences between
them. This was something Bowlby himself acknowledged and mused on in his unpublished
notes.71 In the 1990s, it became general consensus among attachment researchers that pre-
verbal procedural memories of relationship interactions are qualitatively different to the rep-

resentations informed by memory following verbal development, making an encompassing
concept of ‘model’ wholly misleading.72 Each of these issues would leave its legacy for future
attachment researchers (Chapters 3 and 5).
Causes of psychological conflict
Bowlby made another important challenge to the dominant psychoanalytic discourses

of his day. This was in his thinking about the causes of incompatible elements within the
human mind. Bowlby praised Freud for drawing ‘attention to the fact that human beings
are organisms which at times are driven by forces within themselves which they cannot
easily control. We fall in love, we lose our tempers, we panic, we are possessed by forces
which seem alien to ourselves.’73 However, it was especially in his account of the nature of
these forces that Bowlby departed from Freud. In the psychoanalytic theory of the 1930s
and 1940s, ambivalence was a central theoretical concept, considered an inevitable con-
sequence of the incompatibility of human drives. Most critically, psychoanalytic theory
suggested that children feel both love and resentment for their parents. The way that this
predicament—the Oedipus complex—is resolved was thought to be of the utmost import-
ance for a child’s later development. During his analytic training, this was a position that
Bowlby initially accepted.74
interaction in creating an interpersonal interpretive mechanism. Bulletin of the Menninger Clinic, 65(3), 427–48,
p.436. See also Collins, N.L. & Allard, L.M. (2004) Cognitive representations of attachment: the content and func-
tion of working models. In M.B. Brewer & M. Hewstone (eds) Perspectives on Social Psychology: Social Cognition
(pp. 75–101). Oxford: Blackwell.
70 Kobak, R., Rosenthal, N., & Serwik, A. (2005) The attachment hierarchy in middle childhood: conceptual
and methodological issues. In K.A. Kerns & R.A. Richardson (eds) Attachment in Middle Childhood (pp.71–88).
New York: Guilford.
71 E.g. Bowlby, J. (not dated, c. 1955) Thought, conceptualization, language, psycho-analysis. PP/BOW/H.115.
72 E.g. Bretherton, I. (1985) Attachment theory: retrospect and prospect. Monographs of the Society for Research
in Child Development, 50, 3–35; Crittenden, P.M. (1990) Internal representational models of attachment relation-
ships. Infant Mental Health Journal, 11(3), 259–77; Lyons-Ruth, K. (1999) The two-person unconscious: intersub-
jective dialogue, enactive relational representation, and the emergence of new forms of relational organization.
Psychoanalytic Inquiry, 19(4), 576–617.
73 Bowlby, J., cited in J.M. Tanner & B. Inhelder (eds) (1956) Discussions on Child Development: Proceedings of
the WTO Study Group of the Psychobiological Development of the Child, Vol 1. London: Tavistock, pp.182–3.
74 Bowlby, J. (1938, 1950) An examination of the psychological and anthropological evidence. In E.F.M. Durbin
& John Bowlby (eds) Personal Aggressiveness and War (pp.51–150). New York: Columbia University Press.
14 John Bowlby
By the 1950s, however, Bowlby had become critical of a tendency in Freud and Klein to
presume natural individual differences in ambivalence, rather than examining the social
and caregiving context that could intensify such a state. For instance, in the Wolf Man case
study, Freud wrote of ‘the patient’s own ambivalence, which he possessed in a high degree
of development’.75 Next to this passage, in marginalia from the mid-1950s in his personal
copy of Freud’s text, Bowlby wrote: ‘How does Freud explain its genesis?’76 Bowlby’s mar-
ginalia on the work of Melanie Klein likewise expresses this concern. In Contributions to
Psychoanalysis, Klein had observed that ‘Unpleasant experiences and the lack of enjoyable
ones, in the young child, especially lack of happy and close contact with loved people, in-
crease ambivalence, diminish trust and hope and confirm anxieties.’77 However, elsewhere
she had argued against the ‘common tendency to over-estimate the importance of unsatis-
factory surroundings, in the sense that the internal psychological difficulties, which partly

result from the surroundings, are not sufficiently appreciated. It depends, therefore, on the
degree of the intrapsychical anxiety, whether or not it will avail much merely to improve the
child’s environment.’78 Both passages were underlined and highlighted by Bowlby in his per-
sonal copy. In the margins he expressed frustration that according to the second statement,
‘because hatred is primal, the vicious cycle is unavoidable’.79
Bowlby felt that the role of experiences of care in fanning or calming frustration was only
inconsistently acknowledged by Klein, who showed insufficient ‘appreciation that hatred it-
self and therefore intrapsychic conflict can be intensified by bad surroundings’.80 Bowlby
anticipated that if a child receives gentle care that acknowledges their feelings, then con-
flicts can be relatively easily managed. If a child is led to expect rejection or harshness, or
experiences long-term separation from their familiar caregiver, then inner conflicts are in-
tensified.81 The Oedipal potential for love and hate of primary caregivers was acknowledged
by Bowlby as important, since these emotions would stand in necessary conflict within an
intense and important relationship. However, Bowlby regarded this potential as the basis for
mental ill health only when a child’s experiences of the parent had been troubled ones, for
instance as a result of neglect, a parent’s threats to harm the child, or an inability to keep the
child safe.82
In conceptualizing the causes of ambivalence, Bowlby strongly appreciated the emphasis
Klein placed on the experience of loss, and how consequential this was for development.
However, he was frustrated that Klein predominantly thought about loss as a normal de-
velopmental stage within a child–parent relationship. For Klein, every parent would some-
times satisfy a child’s desires and sometimes disappoint them, contributing to a sense of
75 Freud, S. (1918, 2001) From the history of an infantile neurosis. In The Standard Edition of the Complete
Psychological Works of Sigmund Freud, Vol. 17 (pp.1–124). London: Vintage, p.116.

76 Annotations by Bowlby dated 1956 on The Standard Edition of the Complete Psychological Works of Sigmund
Freud, Vol. 17. Copy held in the library of Human Development Scotland.
77 Klein, M. (1948) Contributions to Psychoanalysis. London: The Hogarth Press and the Institute of
Psycho-Analysis, p.314.
78 Ibid. p.280
79 Annotations by Bowlby dated 1948 on Klein, M. (1948) Contributions to Psychoanalysis. London: The Hogarth
Press and the Institute of Psycho-Analysis. Copy held in the library of Human Development Scotland.
80 Ibid. The debate between Bowlby and Klein on the status of aggression can be situated as one skirmish within
a multiparty controversy running over decades on the status of anger within psychoanalytic theory. See Freud, A.
(1972) Comments on aggression. International Journal of Psychoanalysis, 53, 163–71.
81 Bowlby, J. (1960) Separation anxiety. International Journal of Psycho-Analysis, 41, 89–113.
82 Bowlby, J. (1953) The Roots of Parenthood. London: National Children’s Home; Bowlby, J. (1973) Separation.
London: Pimlico, p.287; Bowlby, J. (1983) Letter to Dr Marco Bacciagaluppi, 6 July 1983. PP/Bow/J.9/11.
ambivalence and a sense of loss. Her emblem for this predicament was that of the infant
wanting to access the mother’s breastmilk, but also feeling envy and destructive feelings
towards the breast. Bowlby regarded this account as obscuring the importance of loss as
an event that may damage or remove the child–caregiver relationship, rather than solely as
taking place inside it. It underplayed the potential causal relationship between experiences
of loss and aggression towards caregivers and others.83
In 1956, Bowlby wrote to Klein directly, expressing appreciation for her ‘insistence on
the central importance of the conflict of ambivalence for the loved object starting early in
life . . . all my own thinking stems from that’. However, he insisted in this letter that Klein
was wrong to regard ambivalence about the maternal breast in particular, and relationships
with parents in general, as innate.84 In 1957, he presented ‘The nature of the child’s tie to his
mother’ to the British Psychoanalytic Society. A few days prior to the meeting, Bowlby had

given a copy of the paper to a colleague. At the meeting, Klein and her followers had clearly
received a copy in advance, and were well prepared with an array of hostile remarks. Above
all, Bowlby felt that they criticised his emphasis on actual experiences, rather than inherent
conflict about feeding, as the origin of feelings of ambivalence within family relationships.85
Bowlby experienced this ambush as incontrovertible evidence for the rest of his career that
his ideas would inevitably be met with rejection. As a by-product, it led to a tendency for
Bowlby to throw scorn on any proposal, even coming from Ainsworth, that early feeding ex-
periences could be important for emotional development (Chapter 2).
Self-preservation
A third central problem for Bowlby with the psychoanalytic theory of innate drives was the
status of self-preservatory behaviours. Bowlby’s first exposure to psychoanalysis was reading
the Introductory Lectures on Psychoanalysis. In these lectures Freud posited a drive for self-
preservation, which would include seeking safety, seeking help, and eating. However, during
the period in which Bowlby began training as a psychoanalyst, Freud changed his stance: he
posited a sexual drive as primary, which secondarily became altered into a concern with self-
preservation through unpleasurable experiences in the world.86 From his earliest writings,
Bowlby was not happy with this shift:
It must be realised that Freud equates pleasurable and sexual . . . Psycho-analytically, the
term sexual should be used only to designate pleasurable and should not be used as iden-
tical with genital. I think however there is some confusion here. In the act of sucking, Freud
distinguishes between the activity of taking nourishment and the pleasure obtained . . . On
83 Bowlby, J. (1960) Separation anxiety. International Journal of Psychoanalysis, 41, 89–113, p.108.
84 Bowlby, J. (1956) Public letter to Melanie Klein, 7 February 1956, following presentation of her paper ‘A study
of envy and gratitude’. PP/Bow/G.1/4.
Klein, 10 January 1984. PP/Bow/A.5/7.

86 Young-Bruehl, E. & Bethelard, F. (1999) The hidden history of the ego instincts. Psychoanalytic Review, 86(6),
823–51. Bowlby, J. (1933–36) Functional approach to super-ego. PP/Bow/D.2/49: ‘Freud uses the term sex to de-
scribe all positive sentiments between two people . . . These needs clearly are not always sexual—may be nutritive
or self-protective e.g. parent–child. The point may be, however, that the sexual impulses are apt to be aroused
in such situations.’ This early work by Bowlby appears as Chapter 5 in Duschinsky, R. & White, K. (eds) (2019)
Trauma and Loss: Key Texts from the John Bowlby Archive. London: Routledge.
16 John Bowlby
the other hand he does not make this distinction in describing the reproductive act . . . It
seems to me that the difficulty has arisen through using the term sexual to equal pleasure in
the first instance and then to refer to its reproductive significance in the second.87
Bowlby perceived that Freud’s ‘discovery that symptoms sometimes represented the patient’s
sexual activity led him into an over-generalisation . . . I should like to suggest that symp-
toms can be excited purely in the interests of avoiding danger-situations.’88 In this, one in-
fluence on Bowlby was likely the concern of evolutionary theory with both survival and
reproduction, which had been significant in his undergraduate training in natural sciences
at Cambridge. Whereas Freud treated self-preservation as an essentially rational tendency
adapted to the perception of experienced reality, in his notes in the 1930s Bowlby set out the
position that the self-preservative response has its own intrinsic predispositions and pref-

erences: ‘Although the baby neither foresees nor reasons about cold being bad for its health,
[this] does not preclude the possibility of its being concerned with an inclination to react
to cold with anxiety and to gentle warmth with pleasure. In such a primitive mechanism,
I should recognise the gears of the tendency to self-preservation.’89
Further reflections on self-preservation were prompted by Bowlby’s own experiences of
fatherhood. John and Ursula Bowlby’s first child, Mary Hamilton, was born in February
1939. Their second, Richard, was born in August 1941; Pia Rose was born in February 1945;
finally, Robert was born in April 1948. A traditional upper-class household in this regard,
with the additional separations caused by Bowlby’s wartime responsibilities, John had no
responsibility for feeding the children. Nonetheless, he recalled observing during the 1940s
that his young children would seek him out both for affectionate interaction and when
alarmed.90 This ran contrary to the idea, common at the time, that a child’s relationship with
their parent developed out of the pleasure that came from feeding (‘cupboard love’). In this
seeking of a familiar caregiver when alarmed, Bowlby saw a tendency to self-preservation,
which he distinguished from the 1950s onwards from feeding activity in strict terms (indeed
overstrict. See Chapter 2).
This position had a number of major practical consequences. One implication was that
affection shown to children would not ‘spoil’ them and make them anxious about separ-
ation, unable to cope without affection. In fact, Bowlby argued, the opposite would be true.
Children who feel confident that they are ‘cherished’ would be less likely to be anxious about
separation.91 Another critical implication was that the signs of depression and grief seen by
children in the context of long-term separations from their caregivers should, indeed, be
recognised as mourning for this relationship.92 From the vantage of the present, this seems
like an obvious point. However, it ran against psychoanalytic orthodoxy. Even in 1970, major
87 Bowlby, J. (1933–36) Research notes for MD thesis. PP/Bow/D.2/44/3.

88 Bowlby, J. (1933–36) Mechanisms—symptom-formation. PP/Bow/D.2/44/1.
89 Bowlby, J. (1933–36) Anxiety: essays. PP/Bow/D.2/46/6.
90 Bowlby, J. (1986) Attachment theory: new directions. ACP-Psychiatric UPDATE, 7(2), panel discussion,
Washington 1986. PP/BOW/A.5/1.

91 Bowlby, J. (1956) The growth of independence in the young child. Royal Society of Health Journal, 76, 587–91.
The concept of ‘cherishing’ was an irregular but important one for Bowlby. He refers to the concept in various
places, e.g. Bowlby, J. & Robertson, J. (1965) Protest, Despair and Detachment. PP/BOW/D.3/38.
92 Bowlby’s attention to the issue of mourning was stimulated by reading Marris, P. (1958) Widows and their
Families. London: Routledge. Marris’s book was unusual for the time in giving consideration to typical as well as
atypical mourning processes. The idea that adult and child mourning represents the same process was one Bowlby
inherited from Melanie Klein.
Following Hinde 17
psychoanalysts were still claiming that ‘mourning as defined by Freud and as observed
in the adult is not possible until the detachment from parental figures has taken place in
adolescence’.93
Bowlby interpreted childhood loss in terms of adult mourning, and adult bereavement
in light of children’s response to separations: ‘Since the evidence makes it clear that at a de-
scriptive level the responses are similar in the two age groups, I believe it to be wiser meth-
odologically to assume that the underlying processes are similar also, and to postulate
differences only when there is clear evidence for them’.94 This transposition between child-
hood and adulthood proved an influential move for later attachment research. It represented
the beginnings of a heuristic, or even a method, within subsequent attachment theory and
the development of attachment assessments in which adulthood and childhood are inter-
preted as on analogy with one another. Above all, adult attachment would be interpreted by

the second generation of attachment researchers through analogy and extrapolation from
individual differences in infant attachment. In the 1980s, this was Mary Main’s approach
in developing the Adult Attachment Interview (Chapter 3) and Hazan and Shaver’s in the
development of the ‘love quiz’ (Chapter 5). However, more recently, the analogy has been
reversed. Researchers such as Fraley and Roisman have deployed ideas about the underlying
dimensions of adult attachment as the basis for reimagining individual differences in infant
attachment (Chapters 2 and 3).
Following Hinde
Activating and terminating conditions
For Bowlby, the mistaken idea that children cannot mourn their relationships was a conse-
quence of the assumption in psychoanalytic theory that the relationship is secondary, be-
cause sexuality is primary. He saw this view as an obstacle to recognition that affectionate
bonds are of various kinds and can serve various functions, at an individual level and for
humans in general at an evolutionary level. It was in conceptualizing human social responses
in the context of research on animal behaviour that Bowlby found strongest support for his
thinking about these different kinds and functions of relational behaviour.
In 1951, Bowlby was introduced by Julian Huxley to the work of Lorenz, which revealed
an exciting development: that in the study of animal behaviour, careful work was taking
place differentiating between behavioural tendencies.95 Niko Tinbergen, Konrad Lorenz,
Karl von Frisch, and colleagues, working under the label of ‘ethology’, had developed a revo-
lutionary approach to the study of behaviour. This approach took as its central premise the
idea that not just biological structures but also sequences of observable behaviour could be
the product of evolution through natural selection, contributing in predictable ways to an
93 Nagera, H. (1970) Children’s reactions to the death of important objects: A developmental approach.
Psychoanalytic Study of the Child, 25(1), 360–400.

94 Bowlby, J. (1963) Pathological mourning and childhood mourning. Journal of the American Psychoanalytic
Association, 11(3), 500–41, p.521.

95 Van der Horst, F.C.P. (2011) John Bowlby— From Psychoanalysis to Ethology. Unravelling the Roots of
Attachment Theory. Oxford: Blackwell.
18 John Bowlby
individual’s chances of survival and reproduction.96 Ethologists asked four questions of be-
havioural sequences they observed: How did it develop in the individual? What causes it?
What is its function? How did it evolve in the species? When these questions are not ad-
equately distinguished, the ethologists warned, researchers will talk past one another about
‘adaptiveness’, a concept that can refer to any of these levels. They were also worried that the
idea of ‘adaptiveness’ can imply the implicit value judgement that a response is warranted
or useful for the individual, when in fact this would need to be demonstrated (Chapter 2).97
Indeed, the ethologists argued that responses that support survival and reproduction in gen-
eral may contribute to highly counterproductive behaviours by an individual, depending on
circumstances. Similarly, they observed that the current deployment of a behavioural pat-
tern need not be equated with the function for which it evolved. In fact, an action pattern
may become active, reach its predictable outcome, and terminate, all without any direct rela-

tionship with the function for which it evolved.
Bowlby was primed to take an interest in ethology by his personal passion for bird-
watching. Growing up, Bowlby had quite a formal, distant relationship with his mother;
however, she was a passionate and knowledgeable naturalist, with a particular delight in
birds. Bowlby learnt that he could retain his mother’s attention by engaging her about his
sightings of birds.98 Over time, birdwatching became a firm hobby. However, it should also
be emphasised that Bowlby was exhilarated by the quality of the research being done in eth-
ology, especially in contrast to the rudimentary observations of human development avail-
able in the 1950s: ‘They were brilliant, first-rate scientists, brilliant observers, and studying
family relationships in other species—relationships which were obviously analogous with
that of human beings—and doing it so frightfully well. We were fumbling around in the
dark; they were already in brilliant sunshine.’99
Critical for Bowlby’s engagement with this literature was his friendship with the ethologist
Robert Hinde, who he met in 1954. For the next ten years, Hinde read and gave comments
on most of Bowlby’s scholarly writings, and attended Bowlby’s weekly research seminars at
the Tavistock.100 Together, Bowlby and Hinde identified several ways in which debates in
ethology could advance psychoanalytic theory. The most important point of intersection
and difference was in the theory of motivation. Both Freud and Lorenz tended to think about
activity as motivated or inhibited by the availability of a somewhat underdefined notion of
‘psychological energy’ or ‘drive’. Hinde addressed this model in his paper on ‘Ethological
models and the concept of “drive” ’, published in 1956.101 Here Hinde acknowledged that
it was possible in general terms for an organism to become exhausted after various forms
of energy expenditure. However, from this he argued that it should not be concluded that
all forms of behaviour draw from the same psychological energy or drive. The concepts of
‘energy’ and ‘drive’, Hinde noted, had become nodes within a taken-for-granted theoretical
96 Burkhardt Jr, R.W. (2014) Tribute to Tinbergen: putting Niko Tinbergen’s ‘Four Questions’ in historical con-
text. Ethology, 120(3), 215–23.

97 Tinbergen, N. (1963) On aims and methods of ethology. Zeitschrift Tierpsychology, 20, 410–33.
98 See Bowlby’s correspondence with his mother, writing from boarding school, 1921–1924. PP/BOW/A.1/17/1.
99 John Bowlby interview with Robert Keren, 14 and 15 January 1989, cited in Karen, R. (1998) Becoming
Attached: First Relationships and how They Shape Our Capacity to Love. Oxford: Oxford University Press, p.94.
100 Bowlby, J. (1977–79) Interview with Alice Smuts and Milton J.E. Senn. PP/BOW/A.5/2. The interview has
been published as Chapter 11 in Duschinsky, R. & White, K. (eds) (2019) Trauma and Loss: Key Texts from the John
Bowlby Archive. London: Routledge.
101 Hinde, R. (1956) Ethological models and the concept of ‘drive’. British Journal for the Philosophy of Science,
6(24), 321–31. See also Hinde, R.A. (1959) Unitary drives. Animal Behaviour, 7(3), 130–41.
Following Hinde 19
framework; it was time to reappraise the actual behavioural sequences that they were being
used to describe and explain. Hinde proposed that, if behaviours are observed closely in
comparative perspective across species, different action patterns could be distinguished,
along with their activating and terminating conditions. In contrast to the idea of a single
reservoir of energy, an advantage of the idea of distinct action patterns was that it was easier
to ask the four key ethological questions, considering the development, causation, function,
and evolution of behavioural sequences.
Looking back on his career, Bowlby situated Hinde’s 1956 paper on drives as one of the
most influential works he ever read. It led directly to the account of motivation at the heart
of attachment theory.102 Hinde’s approach lent itself, much more than the psychoanalytic
notion of drive, to observational and experimental research to identify the activating and
terminating conditions of behavioural responses.103 For instance, Bowlby grudgingly ac-

knowledged, it can happen that babies stop crying when they are exhausted. However, he had
no patience with any model of motivation that implied energy or exhaustion. Personally and
intellectually, experiences of depletion were generally disavowed by Bowlby, at least until his
book on Darwin where they were treated as a symptom of mental illness.104 He emphasised
that babies generally stop crying because the terminating conditions for the crying have
been met, for instance when they are picked up.105 The specific antecedents, processes, and
consequences of the crying response and its termination could be studied empirically much
more easily than when crying was theorised in terms of psychological energy or drive.
A second reason that the work of the ethologists appealed to Bowlby was that it distin-
guished between the infants’ desire for proximity with familiar caregivers and the desire for
nutrition. Lorenz had found that exposure to a familiar moving figure was sufficient to elicit
a following response in baby geese, a process he termed ‘imprinting’.106 Lorenz’s work was
of special interest to Bowlby as it validated his experiences of fatherhood, which suggested
that offspring could seek affection and protection from a caregiver based on familiarity, not
on the basis of the pleasure of nutrition. The birds studied by Lorenz were not dependent on
their parents for food, but could feed themselves by catching insects. Nonetheless, the geese
families remained together for at least 12 months.107 Bowlby discussed these findings with
Hinde in 1956, who concluded that they indicated the critical importance of the following
102 Bowlby, J. (1979) The ten books which have most influenced my thought, 24 October 1979. PP/Bow/A.1/
8: ‘Robert Hinde (1956) Ethological models and the concept of drive. British Journal of the Philosophy of Science.
I first met Hinde in 1954 and in the years following read almost all his papers on publication and often before. It
was this paper and others of this published around the same time that led me to the concepts of instinctive behav-
iour presented in Attachment.’
103 Bowlby, J. (1957) An ethological approach to research on child development. British Journal of Medical
Psychology, 30(4), 230–40.

104 Bowlby’s niece has pointed to Bowlby’s disidentification with his disabled younger brother as important in
shaping this stance. Hopkins, J. (in press) The need to put things right: a response to Bowlby’s chapter ‘Hysteria
in Children’. Attachment. On tiredness as one of the symptoms of Darwin’s illness see Bowlby, J. (1990) Charles
Darwin: A Life. New York: Norton.
105 A few years later, Ainsworth reported from her Baltimore home observation study that picking a baby up
stopped his or her crying 86% of the time: ‘this degree of effectiveness is remarkable when one notes that it oc-
curred irrespective of the conditions that activated crying’. Ainsworth, M., Bell, S., & Stayton, D. (1972) Individual
differences in the development of some attachment behaviors. Merrill-Palmer, 18(2), 123–43, p.132.
106 Lorenz, K. (1937) The companion in the bird’s world. Auk, 54, 245–73.
107 Lorenz, K. (1949, 1952) King Solomon’s Ring. London: Methuen & Co. Bowlby, J. (1986) Interview with the
BBC. PP/Bow/F.5/8: ‘First of all you see this following response had nothing to do with food because young geese
feed themselves on insects . . . It’s very powerful and these geese families, they stay together for at least 12 months.
Very important. So I said well if this is true of some animal species it might be true of humans too.’
20 John Bowlby
response for the safety of offspring, since it emerges even before the sequence of behaviours
that would allow a gosling to flee from a threat.108 This perspective helped Hinde and Bowlby
account for findings by animal researchers such as John Paul Scott and Harry Harlow that
non-human animals would continue to show a following response, and indeed might inten-
sify their following response, to a caregiver who was unkind or maltreated them.109
Whereas Lorenz had depicted ‘imprinting’ as a mechanism that was either on or off,
Hinde’s experiments showed that the following response, though initially elicited by a wide
range of objects and sounds, was elaborated by practice and subsequently organised around
the particular objects that elicited the response. In a passage heavily underlined in Bowlby’s
personal copy, Hinde made a comparison with the human infant, for whom an incipient fol-
lowing response can often be elicited by even a stranger walking away, a stronger response
can be elicited by a sibling, but for whom the full following response will generally occur pri-

marily with a familiar adult caregiver.110 Though the following response might be primed by
the nervous system, it required interaction for its elaboration and specification. For Hinde,
this implied that not only behaviour but the motivational response itself was integrated and
altered on the basis of experience.
An additional source of discussion between Hinde and Bowlby was the observation by
ethologists that behavioural sequences might contain components that are unrelated to the
function of the whole. This shed light for the vexed problem of ‘infantile sexuality’ that had
caused significant consternation and confusion for the psychoanalytic community and its
publics, especially during the 1930s when Bowlby was training as an analyst.111 For Bowlby,
the work of the ethologists provided an elegant explanation: ‘A great tit whilst still a fledg-
ling may, for example, show isolated fragments of reproductive behaviour—snatches of sub-
song, nest-building, and copulatory behaviour—but those fragments appear in contexts
quite divorced from the context in which they appear in the adult.’112 Likewise, many of the
component sequences of sexual behaviour may well be present in humans already from in-
fancy. And human children can sometimes show these fragments before they become coord-
inated as the sequences of adult sexual behaviour. However, Bowlby argued, such behaviours
do not imply a unitary sexual drive present from birth in all humans.113
At a paper read to the Association for the Study of Animal Behaviour in April 1955,
Bowlby praised the ethological research community especially for their combination of close
observation and evolutionary theory, which led to conceptually precise distinctions be-
tween sequences of child to parent safety-seeking responses, parent to child care-providing
108 Bowlby, J. (1956) Sequence in maturation of drives: notes from discussion with Hinde, August 1956. PP/
Bow/H146.
109 E.g. Harlow, H.F. (1958) The nature of love. American Psychologist, 13(12), 673–85; Scott, J.P. (1963) The pro-
cess of primary socialisation. Monographs of the Society for Research in Child Development, 28, 1–47.
110 Annotations by Bowlby (PP/Bow/H.226) on Hinde, R.A. (1961) The establishment of the parent–offspring
relation in birds, with some mammalian analogies. In W.H. Thorpe & O.L. Zangwill (eds) Current Problems in
Animal Behaviour (pp.175–93). Cambridge: Cambridge University Press.
111 E.g. Ferenczi, S. (1933, 1980) Confusion of tongues between adults and the child: the language of tenderness
and passion. In M. Balint (ed.) Final Contributions to the Problems and Methods of Psychoanalysis (pp.156–67).
New York: Brunner.
112 Bowlby, J. (1969, 1982) Attachment, Volume 1. London: Penguin, p.157.
113 This account would, essentially, supplant the psychoanalytic model within the psychoanalytic community
over the subsequent decades. Holmes, J. (1998) The changing aims of psychoanalytic psychotherapy: an integrative
perspective. International Journal of Psychoanalysis, 79, 227–40; Zeuthen, K. & Gammelgaard, J. (2010) Infantile
sexuality—the concept, its history and place in contemporary psychoanalysis. Scandinavian Psychoanalytic
Review, 33(1), 3–12.
Following Hinde 21
responses, and sexuality. Bowlby felt that ‘Freud’s observations that these are apt to become
mixed up with each other, although certainly true, does not necessarily mean that there are
not three main responses.’114 In a later version of this argument, written up for the British
Journal of Medical Psychology in 1957, Bowlby expressed his exhilaration that ethological re-
searchers were finding common aspects to these behaviour patterns across different species,
despite all their vast differences. This raised the prospect that human behaviour, too, could
be conceptualised in this way.115 Behavioural patterns such as child safety-seeking and fol-
lowing responses, care-providing responses, and sexual behaviour might be discerned that,
in human evolutionary history, contributed in predictable ways to individual survival or re-
productive fitness. This did not mean that the behaviour as seen in any individual member
of a species would be serving this function. However, Bowlby felt that this ethological per-
spective nonetheless offered the potential to shed light on both typical and atypical forms of

development in humans.
Discrimination
One quality of following, care provision, and sexual behaviours was that all three seemed
to be directed towards preferred targets, even when suitable alternatives were available. In
November 1955, Bowlby wrote to the classicist D.C.H. Rieu to ask for help in describing
a quality of the following response: ‘What I am seeking is a term to denote this tendency
to restrict these basic social responses to particular individuals.’ He gives the example of
the way an infant discerns their mother from among other mothers. Rieu wrote back: ‘I
think the word you want is “monotropy”.’116 In retrospect there is an important mismatch
here. Bowlby was requesting a word to describe the restriction of following behaviour to
‘particular individuals’—plural. However, Rieu’s term implies the restriction to a single
individual—‘mono’—of the tendency to ‘turn to’ the familiar person. It is rather tragic in the
fact that Bowlby, whose appeal to ordinary language so often led to misunderstandings of his
ideas even as it helped popularise them, unusually sought a new and technical Latinate term
on this occasion, but was handed one by Rieu that differed from his request and contributed
no less to confusion and polemics.
The term ‘monotropy’ was first used and defined by Bowlby in a 1958 paper as ‘the ten-
dency for instinctual responses to be directed towards a particular individual or group of
individuals and not promiscuously towards many’.117 In a footnote, Bowlby clarifies that the
meaning is the same as William James’s concept of ‘the law of inhibition of instincts by hab-
its’, which is the observation that ‘when objects of a certain class elicit from an animal a cer-
tain sort of reaction, it often happens that the animal becomes partial to the first specimen
of the class on which it has reacted’.118 The idea that ‘monotropy’ was intended to mean ‘re-
striction on the individuals or groups towards whom a response is directed on the basis of
114 Bowlby, J. (1955) Paper read to the Association for the Study of Animal Behaviour, April 1955. PP/Bow/
H.146.
Psychology, 30(4), 230–40.

116 Bowlby/Rieu correspondence, November 1955. PP/Bow/H146.
117 Bowlby, J. (1958) The nature of the child’s tie to his mother. International Journal of Psycho-analysis, 39(5),
350–73, p.370.
118 James, W. (1890, 2003) Psychology: The Briefer Course. Toronto: Dover Books, p.266.
22 John Bowlby
experience’ is also evident in Bowlby’s close collaborators, who used the term in this sense.119
With changes in experience, it is possible for the restriction itself to alter, though this be-
comes more difficult with time; there is no implication that attachment is limited to one
person, or that it is fixed regardless of later experiences.120 However, this was certainly not
the impression of Bowlby’s readers. The reasons for this seem clear. In Attachment, Volume
1 from 1969, the term appears only once: Bowlby referred his reader back to this 1958 dis-
cussion for the meaning of this term, but summarised this earlier account of monotropy
briefly—and inaccurately—as ‘the bias of a child to attach himself especially to one figure’.121
It is this latter characterisation that was the public understanding of the term and of
Bowlby’s position, supported by the literal implication of the word ‘monotropy’. This im-
plication became a natural rallying point for Bowlby’s critics, as ‘monotropy’ neatly encap-
sulated within one word the complicity between Bowlby’s dense and scholarly theory and

the polemical claims in his popular writings about mothers’ natural responsibilities. Later
attachment researchers have worked hard to dispel the idea that attachment theory implies
monotropy in this sense,122 but this has been a slow and incomplete process. Some things
are, unfortunately, irretrievable. Apparently unaware of his characterisation of the term in
Attachment, Volume 1, and of the consequences of his appeals to the natural role of mothers
in his popular writings of the 1950s, from the 1970s onwards Bowlby expressed dismay that
critics insisted on making him a ‘straw man’, seeking to discredit him through inaccurate
characterisation of his discussion of monotropy. He disagreed vehemently with critics who
supposed that he introduced the term to imply attachment only to the biological mother, and
expressed bafflement in his private notes as to the origin of this view, which he regarded as
‘nonsense’.123
In fact, ‘monotropy’ was intended by Bowlby as a technical term for the way that experi-
ence leads a response to become oriented towards particular targets. For Bowlby, ‘monot-
ropy’ was intended to mean a relationship to a particular ‘person or place or thing’ that is
personally significant, based on a felt sense of need, and not superficial or interchangeable
with other people, places, or things even if they are somewhat similar.124 We cherish certain
people, places, or things and not others, and this is part of what it is to cherish. This is in con-
trast to forms of ‘liking’ where the targets are interchangeable, without hierarchy, but our
investment in them is also bloodless. In an unpublished text from 1955 ‘Notes on child at-
tachment and monotropy’, Bowlby wrote: ‘Focusing of instinctive responses on individuals.
This is the rule—not the exception and allies to all three basic social responses’—presumably
119 E.g. Hinde (1986) Ethology. New York: Fontana, p.230.

120 See also Hinde, R.A. (1963) The nature of imprinting. In B.M. Foss (ed.) Determinants of Infant Behaviour,
Vol. 2 (pp.227–30). London: Methuen.
122 See e.g. van IJzendoorn, M.H., Sagi, A., & Lambermon, M.W.E. (1992) The multiple caregiver paradox. Some
Dutch and Israeli data. In R.C. Pianta (ed.) New Directions for Child Development, No. 57. Beyond the Parent: The
Role of Other Adults in Children’s Lives (pp.5–25). San Francisco: Jossey-Bass.; Cassidy, J. (1999) The nature of the
child’s ties. In J. Cassidy & P.R. Shaver (eds) Handbook of Attachment: Theory, Research, and Clinical Applications
(pp.3–21). New York: Guilford.
123 Bowlby’s correspondence with Peter K. Smith, PP/Bow/J.9/184, discussing Smith, P.K. (1980) Shared care
of young children: alternative models to monotropism. Merrill-Palmer Quarterly of Behavior and Development,
26(4), 371–89. See also Bowlby’s correspondence with Michael Rutter, PP/Bow/J.9.161-2.
124 Indeed, Bowlby offered the speculative claim that such relationships are the basis for all that constitutes
‘deep feeling’ within human life. Bowlby, J. (1960) Separation anxiety. International Journal of Psycho-Analysis,
41, 89–113: ‘It is because of this marked tendency to monotropy that we are capable of deep feelings; for to have a
deep attachment to a person (or place or thing) is to have taken them as the terminating object of our instinctual
responses.’ (101)
Following Hinde 23
child-to-parent following, parent-to-child care, and sexuality—and ‘it is in the nature of the
instinctive response to focus on an individual, though this may be everything from complete
to very partial’.125 As such, ‘ “monotropy” is only a special case of discrimination becoming
heightened through learning’. The term ‘discrimination’ captures Bowlby’s intention better
than monotropy, in fact—and indeed was preferred by Ainsworth.126
In the discussions of the ‘Psychobiology of the Child’ study group at the World Health
Organisation in the 1950s, and then later in Separation in 1973, Bowlby argued that the dis-
crimination of particular figures occurs in the case of all the basic behavioural responses.
The sexual response, for example, is not evoked indiscriminately in adults by any poten-
tial stimulus, but has been trained by experience to become restricted to certain kinds of
people or situations.127 There is likewise discrimination in the ‘flight’ response. We have par-
ticular people that make us feel safe, who we turn to when worried or scared.128 Bowlby

made an argument for ‘the following response’ as another such basic behavioural response,
where discrimination occurs regarding its target. Though his focus was on people, Bowlby
allowed that objects can also be treated as targets of following and clinging, and sought for
‘the intense sense of reassuring comfort’ that they can, at least in some regards, provide.129
Elsewhere Bowlby added that we also have particular institutions, ideas, and places—such as
our nation or our place of prayer130—that are discriminated by the flight and following re-
sponses as signifying safety, which we might seek when alarmed.
Tinbergen called attachments to things or locations in the physical environment ‘site at-
tachment’, and Bowlby the ‘personal environment to which we are attached’ in Separation.131
‘Home’ as an idea and as a location is a clear example. Bowlby acknowledged that though
his focus was generally on ‘distress felt and expressed when a person, particularly a child,
is separated from his mother figure, distress is felt also on separation from certain familiar
objects of other kinds. Attachment to a particular house and environs as home is usual in
humans (as it is in animals of other species).’132 Gruneau Brulin and Granqvist have argued
125 Bowlby, J. (c.1955) Notes on child attachment and monotropy. PP/Bow/H146.

126 E.g. Ainsworth, M. (1964) Patterns of attachment behavior shown by the infant in interaction with his
mother. Merrill-Palmer, 10(1), 51–58, p.56. More recently, Lyons-Ruth and the Boston Change Process Study
Group have suggested describing these discriminated relationships as ‘charged’. For them, ‘charged’ relationships
are characterized by some underlying valuation, their priority over other relationships, and sufficient continuity
to scaffold trust. Boston Change Process Study Group (2018) Engagement and the emergence of a charged other.
Contemporary Psychoanalysis, 54(3), 540–59.
127 Bowlby cited in Tanner, J.M. & Inhelder, B. (1956) Discussions on Child Development: Proceedings of the WTO
Study Group of the Psychobiological Development of the Child, Vol 1. London: Tavistock, pp.184–5; Bowlby, J. (1973)
Separation. London: Pimlico. See also Bowlby, J. (1969, 1982) Attachment, Volume 1. London: Penguin, pp.137–8.
128 Bowlby, cited in Tanner, J.M. & Inhelder, B. (1956) Discussions on Child Development: Proceedings of the
WTO Study Group of the Psychobiological Development of the Child, Vol 1. London: Tavistock, pp.184–5.
129 Bowlby, J. (1973) Separation: Anxiety and Anger. New York: Basic Books: ‘Laughlin (1956) has proposed a
new term “soteria”, as an obverse of phobia, to denote the intense sense of reassuring comfort that a person may get
from a “love object”, be it a toy . . .’ (148).
130 Bowlby, J. (1956) The growth of independence in the young child. Royal Society of Health Journal, 76,
587–91, p.589.
131 Tinbergen, N. (1956) The functions of territory. Bird Study, 4(1), 14–27; Bowlby, J. (1973) Separation: Anxiety
and Anger. New York: Basic Books: ‘It is still too little realized, perhaps, that the individuals of a species, so far from
roaming at random throughout the whole area of the earth’s surface ecologically suitable to them, usually spend
the whole of their lives within an extremely restricted segment of it, known as the home range . . . each individual
has its own relatively small and very distinctive personal environment to which it is attached’ (177).
132 Bowlby, J. (undated) Distress at loss of home: Chapter 3. c.1969–1971. PP/Bow/H.55. See also Bowlby, J.
(1965) Attachment behaviour: a note after Ciba 1965. Revised Note January 1966. PP/Bow/H.146: ‘Further defin-
ition of attachment: equilibrium point is proximity to a certain type of object . . . Consider also habitat attachment.’
It must be acknowledged, however, that Bowlby was inconsistent on this point. Likely in an attempt to hack away at
the thicket of wider connotations of the word ‘attachment’, he came to emphasize that the object of an attachment
24 John Bowlby
that attachment to ‘home’ is simply a secondary effect of its association with family and at-
tachment figures.133 This claim seems overstated: the very concepts of a secure base and
safe haven (Chapter 2) are metaphors for territorial movement, away from and back to a
base in the context of potential threat.134 Nonetheless, most of Bowlby’s published works
focus narrowly on attachment relationships with parents, allowing other attachments such
as siblings or to home to slide into the background. Such issues would both have to be re-
discovered by later researchers, spurred by the primacy of these concerns in fields such as
social work and in ecological approaches.135 Furthermore, Bowlby was unclear regarding
what processes exactly led to the discrimination of attachment figures. Main and Fonagy
admonished Bowlby for this.136 In the 1990s, they later argued that discrimination and the
basis for selective attachments occur especially when someone, or something, is perceived as
contingently responsive to us.

Following and attachment
In historical perspective, without access to comparisons with animal behaviour, and to

Robert Hinde in particular, it seems likely that there would have been no attachment theory.
This was also Bowlby’s own view.137 Though use of the phrase has dropped away since the
mid-1990s, in the 1970s and 1980s researchers used ‘ethological attachment theory’ as the
official name for the paradigm as a way to ‘commemorate the influence of ethological theory
and research on Bowlby’s early thinking’.138 Key strengths of Bowlby’s thinking can be found
relationship had to be human. Stevenson-Hinde, J. (2007) Attachment theory and John Bowlby: some reflections.
Attachment & Human Development, 9(4), 337–42: ‘During the first conference, I recall John emphatically stating,
“We cannot allow ‘attachment to an umbrella’!” He insisted that “attachment” be used to describe an emotional
bond to someone (i.e., a person) usually perceived as older or wiser (e.g., mother or father). While other kinds
of bonds undoubtedly exist, they should not be called “attachment,” in order to keep some precision in the use of
terms’ (338). This, exclamation, of course, produces an excluded middle: our feelings for our home are neither the
trivial affection we might feel for a favourite umbrella, nor the feelings we have for a parent.
133 Gruneau Brulin, J. & Granqvist, P. (2018) The place of place within the attachment-religion framework: a
commentary on the circle of place spirituality. Research in the Social Scientific Study of Religion, 29, 175–85.
134 The claim that home cannot be a secure base or safe haven would seem to depend on a reification of ‘at-
tachment relationship’ beyond its constituent elements. Gruneau Brulin and Granqvist claim that home is non-
individual, which is implausible. They also claim that it cannot be an object of attachment since it is non-reciprocal.
However, reciprocity was never part of either Bowlby or Ainsworth’s definition of an attachment relationship.
135 On attachment to place see e.g. Scannell, L. & Gifford, R. (2013) Comparing the theories of interpersonal
and place attachment. In L.C. Manzo & P. Devine-Wright (eds) Place Attachment. Advances in Theory, Methods
and Applications (pp.23–36). London: Routledge. On attachment to siblings see e.g. Teti, D.M. & Ablard, K.E.
(1989) Security of attachment and infant–sibling relationships: a laboratory study. Child Development, 1519–28;
Farnfield, S. (2009) A modified strange situation procedure for use in assessing sibling relationships and their at-
tachment to carers. Adoption & Fostering, 33(1), 4–17.
136 Fonagy, P., Steele, M., Steele, H., Higgitt, A., & Target, M. (1994) The Emanuel Miller memorial lecture
1992: the theory and practice of resilience. Journal of Child Psychology and Psychiatry, 35(2), 231–57; Main, M.
(1999) Epilogue. Attachment theory: eighteen points with suggestions for future studies. In J. Cassidy & P. Shaver
(eds) Handbook of Attachment (pp.845–87). New York: Guilford, p.848.
137 E.g. Bowlby, J. (1991) Ethological light on psychoanalytic problems. In P. Bateson (ed.) The Development
and Integration of Behaviour: Essays in Honour of Robert Hinde (pp.301–14). Cambridge: Cambridge University
Press: ‘The typed drafts of major parts of the first edition of his [Hinde’s] Animal Behaviour (1966) which he lent
me in 1965 when I was starting work on my volume on Attachment (1969). Whatever merits my own volume has
owed a tremendous debt to his’ (303).
138 Waters, E., Kondo-Ikemura, K., Posada, G., and Richters, J.E. (1991) Learning to love. In M.R. Gunnar &
L.A. Sroufe (eds) Self Processes and Development (pp.217–55). New York: Psychology Press, p.227. Ainsworth, M.
& Bowlby, J. (1991) 1989 APA award recipient addresses: an ethological approach to personality development.
Following Hinde 25
precisely in the revision of psychoanalytic theory on the basis of ethology: the differentiation
of behavioural responses; the care in thinking about value of behaviour for the individual
and for the species; recognition that the predisposition for certain responses may come pre-
programmed but that their expression requires elaboration in the context of experience and
learning; the privilege given to observational methodology; and the specification of the fol-
lowing response in birds and mammals, and in comparison with humans.139 This is a large
chunk of attachment theory.
Whereas Bowlby made appeal to the idea of ‘love’ throughout the 1950s in both his
popular and scientific works—most importantly in Child Care and the Growth of Love—
from 1961 the term was expunged from his vocabulary.140 Hinde’s contribution had made
Bowlby acutely aware that the term was too absorptive, that it hid within itself profoundly
diverse processes with different causes and consequences. Clinical experience had taught

Bowlby time and again that humans can love fiercely even while we fail one another; we can
seek comfort even in the absence of love; we can experience affection as claustrophobic and
dangerous, as holding us captive.141 It may appear that ‘love’ designates an intense and deter-
minate feeling. In fact, however, the term is more a magnet and placeholder for complex and
diverse intensities. Even if appeal to ‘love’ could help Bowlby with intelligibility and plausi-
bility to his readers, Hinde’s precise conceptualisation of motivation had shown the neces-
sity of going beyond evocative euphemism. Michael Rutter later situated the replacement of
‘love’ with ‘attachment’ as one of the most important landmarks in the history of attachment
research.142
However, the establishment of attachment theory on the foundation of Hinde’s account of
the following response also generated problems. Hinde was discussing a specific behavioural
pattern: the following, greeting, and clinging response shown by offspring to their parents.
And Bowlby and Ainsworth were both insistent that literal following and bodily clinging
were essential referents of the term ‘attachment’.143 However, the concept was fused into the
metapsychology of psychoanalytic theory, with its focus on the emotional and symbolic
American Psychologist, 46(4), 333–41: ‘The distinguishing characteristic of the theory of attachment that we have
jointly developed is that it is an ethological approach’ (333).
139 See e.g. Bowlby, J. (1963) Remarks at the MRC Ethology Meeting organised by Bowlby, 23 May 1963. PP/
Bow/D.6/5: ‘The role of a comparative approach: a) in facilitating observation b) in elucidating the evolution of be-
haviour c) in taxonomy d) in providing a basis for generalisation e) in leading to an understanding of function’ (3).
140 Bowlby, R. (2017) Growing up with attachment theory—a personal view. Psychodynamic Psychiatry, 45(4),
431–9: ‘When my father sat down he told us that he was looking for a new term to replace [a]‌child’s “tie.” He said
the image of a child being tied to mother or mother being tied to a child had become socially unacceptable, and he
was thinking of using the child’s “attachment” instead. We all groaned and said how boring and why couldn’t he use
“love” like he had originally? He explained that “love” was not strictly accurate and anyway he had already decided
he was going to use attachment from then on’ (436).
141 Bowlby, J. (1985) Letter to John Byng-Hall, 12 April 1985. PP/Bow/J.9/45.
142 Rutter, M., Kreppner, J., & Sonuga- Barke, E. (2009) Emanuel Miller lecture: attachment insecurity, dis-
inhibited attachment, and attachment disorders: where do research findings leave the concepts? Journal of Child
Psychology and Psychiatry, 50(5), 529–43. In fact, ‘love’ would always be threatening to return, hammering at
the door of attachment research. This was especially the case for the social psychology tradition (Chapter 5), e.g.
Hazan, C. & Shaver, P. (1987) Romantic love conceptualized as an attachment process. Journal of Personality and
Social Psychology, 52(3), 511–24. However, appeal to ‘love’ is also a feature in some work in the developmental trad-
ition of attachment research. For instance, ‘loving’ would be one of the scales of the Adult Attachment Interview,
and used in the assessment of transcripts as ‘earned secure’ (Chapter 3).
143 E.g. Ainsworth, M. (1967) Infancy in Uganda. Baltimore: The Johns Hopkins Press: ‘Attachment is more than
a discrimination between people and implies something far more active—a literal or figurative seeking out, fas-
tening on’ (440).
26 John Bowlby
meanings of parent–child relationships.144 On the one hand, this gave the ethological con-
cept of ‘following’ a much deeper emotional resonance. On the other hand, it gave a psycho-
analytic model greater behavioural specification. The resulting concept, ‘attachment’, ended
up with both narrow and broad meanings (Chapters 2 and 5).145 Narrowly, attachment could
mean the following response and related actions that serve to monitor and maintain access
to the caregiver; broadly, the same term could mean an emotionally invested relationship, as
a symbolic source of comfort and protection. Bowlby shuttled between these distinct mean-
ings, sometimes intending one, sometimes the other, and sometimes both.
Though this certainly contributed to conceptual muddle, the basis for this movement was
in Bowlby’s attempts to capture, as best he could, the expression of a cross-species phenom-
enon in human beings, a species with particular capacities for symbolisation and shared
meanings.146 In Bowlby’s unpublished writings and correspondence of the 1950s, his dis-

satisfaction is visible with this predicament.147 His basic referent for an expression of the
attachment system was of infant proximity-seeking in the context of distress as a means of
eliciting care. Part of his positioning against psychoanalytic theory was his emphasis on ob-
servable interactions. Yet, for example, he wished to maintain that a widow or widower who
experiences security and comfort in remembering their spouse is likewise benefiting from
a real and ongoing attachment relationship, even if this is now solely at a symbolic level.148
Between these two extremes, he knew that most attachment relationships after toddlerhood
would have elements of the basic behavioural system and of its symbolic elaboration: ‘A
child separated from his mother comes to crave both for her love and for its accompanying
144 The centrality of symptom formation through symbolization in psychoanalysis was established in Breuer,
J. & Freud, S. (1893–95, 2001) Studies on hysteria. In The Standard Edition of the Complete Psychological Works
of Sigmund Freud, Vol. 2 (pp.1–305). London: Vintage. Bowlby was especially interested by Rycroft, C. (1956)
Symbolism and its relationship to the primary and secondary processes. International Journal of Psycho-Analysis,
37, 137–46, and had correspondence with the author (PP/Bow/H.116). Part of the significance of Rycroft’s work on
symbolization for Bowlby was that it showed that ‘the emphasis placed in Kleinian theory on the fact that “psych-
ical reality” and “external reality” are both subjectively real does, I think, tend to obscure the fact that there are
none the less essential differences between them, and that psychical reality is itself divisible into one part which is
developmentally bound to external reality and another which has been formed by idealization’ (141).
145 For an example of Bowlby taking away implications for relationships in general from a discussion with Hinde
of specific aspects of the following response see e.g. Bowlby, J. (1957) Discussion with Hinde, January 1957. PP/
Bow/H.128: ‘Attachment behaviour comprises all those responses which subserve the total task of relating to an-
other human being.’
146 The capacity for symbolization was, in a sense, highlighted by Main and colleagues’ ‘move to the level of rep-
resentation’ (Chapter 3). However, Main and colleagues did not distinguish the cross-species ethological aspects
of attachment from those aspects associated with human symbolic capabilities. Implicit acknowledgement of the
issue also appeared in the 1990s in the writings of researchers focused on intersubjectivity and rhythms of parent–
infant interaction, e.g. Beebe, B., Lachmann, F., & Jaffe, J. (1997) Mother–infant interaction structures and presym-
bolic self-and object representations. Psychoanalytic Dialogues, 7(2), 133–82; Trevarthen, C. (1998) The concept
and foundations of infant intersubjectivity. In S. Braten (ed.) Intersubjective Communication and Emotion in Early
Ontogeny (pp.15–46). Cambridge: Cambridge University Press. However, concepts of intersubjectivity and rhythm
emphasize continuities between presymbolic procedural expectations in relationships and the symbolic capacities
that emerge from them. The distinction and potential disjuncture between presymbolic and symbolic senses of
the concept of ‘attachment’ was not drawn out. In the history of attachment research, this distinction appears
to have first been made focally and clearly in the application of attachment theory to religious life: Kirkpatrick,
L.A. (1999) Attachment and religious representations and behavior. In J. Cassidy & P.R. Shaver (eds) Handbook of
Attachment: Theory, Research, and Clinical Applications (pp.803–22). New York: Guilford.
147 E.g. Bowlby, J. (not dated, c.1955) Thought, conceptualization, language, psycho-analysis. PP/BOW/H.115;
Bowlby, J. (1955) Letter to C.F. Rycroft, 26 April 1955. PP/Bow/H.116.

148 Bowlby, J. (1980) Loss. New York: Basic Books, p.243. The confusion evoked by this stance for later inter-
pretations of Bowlby is detailed in Fraley, C.R. & Shaver, P.R. (2016) Attachment, loss, and grief: Bowlby’s views,
new developments, and current controversies. In J. Cassidy & P.R. Shaver (eds) Handbook of Attachment: Theory,
Research, and Clinical Applications (3rd edn, pp.40–62). New York: Guilford.
Following Hinde 27
symbols.’149 However, the interrelation between cross-species behavioural system and the
symbolic elaboration especially characteristic of humans remained a source of theoretical
and terminological problems for him, firmly tangled up within his use of the word ‘attach-
ment’, as well as his difficulties in rooting the mind in the body.150
Another conceptual issue made a potent contribution to this confusion. In Bowlby’s writ-
ings, the broad notion of attachment was generally used alongside, specifically, a narrower
concept of internal working model—to mean the specific symbolic and affective represen-
tations made by humans about attachment figures and their availability, and the value of
the self to these attachment figures. By contrast, when Bowlby used the concept of attach-
ment narrowly to mean the specific ethological following response, it was accompanied by
a broader concept of internal working model—to mean expectations about the other’s likely
availability in response to attachment behaviour.151 This is how infants, puppies, and lambs

can have ‘internal working models’ in Attachment, Volume 1 even without the capacity for
much representational thought: the concept was being used in this second sense, not the
first. This is also why more ethologically oriented attachment researchers, such as Kobak, de-
fine internal working models very simply as expectations about attachment figures and how
interaction with them will go,152 whereas less ethologically oriented attachment researchers
require more elaborate interpretations (Chapter 5).
Already in the 1960s, Hinde was very worried by Bowlby’s misleading use of broad and
narrow meanings of the concepts of attachment and internal working model. He felt that
‘behaviour is diverse: an attempt to squeeze it into a system involving only a few explanatory
concepts is liable to lead to one of two results—either facts which do not fit will be ignored,
or the concepts will be stretched until they become valueless’.153 Hinde tried to warn Bowlby,
arguing for greater precision in use of the term ‘attachment’ and what it meant for interpret-
ing observable behaviour. For instance, with early drafts and ideas from Attachment, Volume
1 in hand, Hinde expressed concern with the way Bowlby was conceptualizing the termin-
ating conditions of attachment behaviour, and the idea of internal working models. He wrote
to Bowlby in 1967 directly: ‘I think that you sometimes reify the concepts that you are using
for explanation as though they were mechanisms.’154
149 Bowlby, J. (1944) Forty-four juvenile thieves: their characters and home-life (II). International Journal of
Psycho-Analysis, 25, 107–28, p.121.

150 On this latter point, criticisms of Bowlby are presented by Fonagy, P. & Target, M. (2007) The rooting of the
mind in the body: new links between attachment theory and psychoanalytic thought. Journal of the American
Psychoanalytic Association, 55(2), 411–56. However, Fonagy and Target did not adequately recognize the polysemy
of Bowlby’s concept of attachment, which hindered their own discussion of attachment and mentalization.
151 Note Bowlby’s wording in his very definition of internal working models as ‘starting, we may suppose, to-
wards the end of his first year’, in a procedural form, and as developing into but conceptually distinguishable
from semantically elaborated internal working models ‘during his second and third when [the child] acquires the
powerful and extraordinary gift of language’. Bowlby, J. (1969) Attachment. London: Penguin, p.353.
152 Kobak, R. & Esposito, A. (2004) Levels of processing in parent–child relationships: implications for clin-
ical assessment and treatment. In L. Atkinson & S. Goldberg (eds) Attachment Issues in Psychopathology and
Interventions (pp. 139–66). Mahwah, NJ: Lawrence Erlbaum, p.140.
153 Hinde, R. (1957) Consequences and goals: some issues raised by Dr Kortland’s paper on aspects and pro-
spects of the concept of instinct. British Journal of Animal Behaviour, 5, 116–18, p.116.
154 Hinde, R. (1967) Letter to John Bowlby, 28 June 1967. PP/Bow/K.4/11. Hinde would later make the ob-
servation in print, looking back on Bowlby’s overall contribution to the study of behaviour in Hinde, R. (1991)
Relationships, attachment and culture: a tribute to John Bowlby. Infant Mental Health, 12(3), 154–63; and Hinde,
R. (1991) Commentary. In P. Bateson (ed.) The Development and Integration of Behaviour: Essays in Honour of
Robert Hinde (pp.411–18). Cambridge: Cambridge University Press.
28 John Bowlby
Seeing insufficient change in Bowlby’s stance, in a 1982 chapter Hinde criticised Bowlby
on this matter in print. He argued against the implication in Bowlby’s writing that evolution
had wired human infants to seek proximity as the sole strategy for achieving the set-goal of
the attachment behavioural system. This seemed implausible. Natural selection, Hinde felt,
would likely ‘favour individuals with a range of potential styles from which they select ap-
propriately’.155 Whilst direct proximity-seeking might be regarded as the desirable response
in many circumstances, Hinde emphasised that survival of infants would be more likely if
they could adapt to the conditions of care in which they found themselves. They therefore
needed alternative strategies for other conditions. Hinde therefore anticipated that evolution
would have given humans a repertoire of ‘conditional strategies’ for responding to caregiving
environments where direct proximity-seeking was not possible or effective.156 The avail-
ability of conditional strategies could be anticipated to contribute to survival under such

circumstances.
The emphasis on proximity as the goal of the infant attachment behavioural system also
had discrepancies with Ainsworth’s empirical findings. As Chapter 2 documents, in the
1960s Ainsworth noticed that some infants—those she labelled B1—clearly achieved ter-
mination of the attachment behavioural system without proximity. Ainsworth also found
other infants—who she labelled C—who did not achieve termination of the attachment be-
havioural system even when they did gain proximity, and continued to exhibit behaviours
seemingly aiming to retain the attention of their caregiver. Citing Ainsworth’s observations
as inspiration, across the 1970s and 1980s researchers demonstrated empirically that the car-
egiver’s physical availability facilitated infant exploration in a novel room less effectively than
the caregiver’s emotional and attentional availability.157
Nonetheless, ‘proximity’ remained enshrined in theory as the terminating condition
of the attachment behavioural system; it was only late in his career that Bowlby consist-
ently amended his descriptions of the attachment behavioural system in infancy to specify
that it could be terminated by the ‘availability’, not simply proximity, of the caregiver.158
However, by this time, a generation of attachment researchers had come of age. For instance
155 Hinde, R.A. (1982) Attachment: some conceptual and biological issues. In C.M. Parkes & J. Stevenson-Hinde
(eds) The Place of Attachment in Human Behavior (pp.60–76). London: Tavistock, p.71.
156 Ibid.
157 Carr, S.J., Dabbs Jr, J.M., & Carr, T.S. (1975) Mother–infant attachment: the importance of the mother’s visual
field. Child Development, 46(2), 331–8; Sorce, J.F. & Emde, R.N. (1981) Mother presence is not enough: effect of
emotional availability on infant exploration. Developmental Psychology,17(6), 737–45. See also Joffe, L.S., Vaughn,
B.E., Barglow, P., & Benveniste, R. (1985) Biobehavioral antecedents in the development of infant–mother at-
tachment. In M. Reite & T. Field (eds) The Psychobiology of Attachment and Separation (pp.323–49). Orlando,
FL: Academic Press.
158 A first acknowledgement appears in Bowlby, J. (1973) Separation: Anxiety and Anger. New York: Basic
Books: ‘Accessibility in itself is not enough. Not only must an attachment figure be accessible but he, or she, must
be willing to respond in an appropriate way; in regard to someone who is afraid this means willingness to act as
comforter and protector. Only when an attachment figure is both accessible and potentially responsive can he, or
she, be said to be truly available. In what follows, therefore, the word ‘available’ is to be understood as implying that
an attachment figure is both accessible and responsive’ (234). Throughout the 1970s, Bowlby tended still to refer
to proximity as the set-goal of the attachment system. His final word, however, is in Bowlby, J. (1991) Ethological
light on psychoanalytic problems. In P. Bateson (ed.) The Development and Integration of Behaviour: Essays in
Honour of Robert Hinde (pp.301–14). Cambridge: Cambridge University Press: ‘The goal of attachment behav-
iour is to maintain certain degrees of proximity to, or of communication with, the discriminated attachment
figure(s)’ (306). Some of the tensions in Bowlby’s account of the set-goal of the attachment system had been dis-
cussed already by Bretherton, I. (1980) Young children in stressful situations: the supporting role of attachment
figures and unfamiliar caregivers. In G.V. Coelho & P. Ahmed (eds) Uprooting and Attachment (pp.179–210).
New York: Plenum Press.
Following Hinde 29
‘disorganised/disoriented attachment’ had already been established by Main and Solomon as

behaviour suggesting disruption in an infant’s proximity-seeking (Chapter 3).159 Likewise,
severe conceptual difficulties had been experienced as the field attempted to extend Bowlby’s
model into adulthood (Chapter 5). For instance, in the 1970s Bob Marvin observed pat-
terns of availability-seeking in otherwise secure-seeming three and four year olds in the
Strange Situation that did not rely on proximity. This seemed developmentally expectable.
Nonetheless, because proximity was not sought by the children in the Strange Situation, in
attempting to square his observations with Bowlby’s stated theory, he was improbably forced
to wonder whether these were even attachment relationships.160
As well as his insistence on proximity as the terminating condition for following, a por-
tion of the confusion lay in Bowlby’s use of the term ‘attachment’. As Rutter later observed,
the word ‘attachment’ was used by Bowlby ‘to refer to discrete patterns of behaviour (such

as proximity-seeking), to a dyadic relationship, to a postulated inbuilt predisposition to de-
velop specific attachments to individuals, and to the hypothesised internal controlling mech-
anisms for this predisposition’.161 ‘Attachment’ (inbuilt predisposition) means that children
show ‘attachment’ (discrete behaviours) within ‘attachments’ (dyadic relationships). How
does this happen? Attachment! (a hypothesised internal controlling mechanism). Where the
different meanings of the term are not clearly held in mind and distinguished, the result is
a recipe for clouded and for circular thinking, and a weak basis for identifying theoretical
differences and deviations. Bowlby and subsequent attachment researchers only came to ac-
knowledge the contribution of the sprawling connotations of the term ‘attachment’ to mis-
appropriations of the theoretical tradition in the 1970s. By this time the term, swollen as if
infected, had come to define the research paradigm as a whole, and was inextricable.162
Ambiguities about the meaning of ‘attachment’ as the headline concept for Bowlby’s theory
have had a widespread legacy. Charles and Alexander expressed concern that Bowlby’s varied
and underspecified uses of the term ‘attachment’ have permitted many spurious forms of
therapy to claim a basis in the subsequent attachment research paradigm.163 Likewise, where
reference to ‘attachment’ appears within social policy, it is common to find circular and con-
fused reasoning, with the term—and the associated research paradigm—invoked for various
159 Still by 1990 Main was adamant that physical touch with the caregiver was ultimately the set-goal of the at-
tachment system in infancy. Main, M. (1990) Parental aversion to infant-initiated contact is correlated with the
parent’s own rejection during childhood: the effects of experience on signals of security with respect to attachment.
In T.B. Brazelton & K. Barnard (eds) Touch (pp.461–95). New York: International Universities Press.
160 Marvin, R.S. (1977) An ethological–cognitive model for the attenuation of mother–child attachment be-
havior. In T.M. Alloway, L. Krames, and P. Pliner (eds) Advances in the Study of Communication and Affect, Vol. 3
(pp.25–60). New York: Plenum Press, pp.56–7.
161 Rutter, M. (1995) Clinical implications of attachment concepts: retrospect and prospect. Journal of Child
Psychology and Psychiatry, 36(4), 549–71, p.551. See also Stern, D. (1985) The Interpersonal World of the Infant.
New York: Basic Books: ‘Attachment is a set of infant behaviours, a motivational system, a relationship between
mother and infant, a theoretical construct, and a subjective experience for the infant’ (25).
162 From the 1970s onwards, Bowlby tested out referring to ‘care-seeking’ rather than ‘attachment behaviour’,
e.g. Bowlby, J. (1986) Attachment, life-span and old age. In J. Munnichs & B. Miesen (eds) Attachment, Life-Span
and Old Age. Utrecht: Van Loghum, p.11.
163 Charles, G. & Alexander, C. (2014) Beyond attachment: mattering and the development of meaningful mo-
ments. Relational Child and Youth Care Practice, 27(3), 26–30: ‘Herein lies another problem with “attachment.” It
is one of those terms which we all think we understand the meaning of but when we actually examine it we find
that it has significantly different meanings for different people . . . Today the term is often a loose metaphor for a
relationship-based intervention, and those using the term do not necessarily have an accurate understanding of
the concept. The absence of a precise and universally understood definition has led to a wide variety of interpret-
ations of what is a practical “attachment intervention.” For example, there are a number of controversial “attach-
ment” treatments based on various forms of “therapeutic holding” ’ (27).
30 John Bowlby
purposes so long as they can align with the idea of the importance of child–caregiver rela-
tionships. So, for instance, in the UK, since the 2010s there have been consistent appeals to
Bowlby and the idea of attachment by the political right, who have argued that a policy focus
on the early years justifies cuts to other public services, with attachment security presented
as an alternative to social security.164 Much the same goes for ‘attachment parenting’ dis-
courses that, in fact, lack an evidence-base or anything but the most selective and strategic
relationship with the tradition of attachment research, and that play off cultural stereotypes
about motherhood.165 Given the common use of the ‘attachment’ label and the gulf that sep-
arates attachment research from these ‘attachment parenting’ discourses, Ross Thompson
observed that ‘if you talk to attachment researchers, you will find involuntary wincing—
sometimes followed by groaning—when someone brings up attachment parenting’.166

Separation
Forty-four thieves
Bowlby’s discussions with Hinde and his emerging model of attachment were fed by and
in turn contributed to a concern with the role of long-term separations in childhood for
subsequent development. Bowlby regarded himself as predisposed to an interest in major
separations by his own childhood experiences, such as being sent to boarding school.167
This interest was then intensified by his early experiences as a clinician. From the late 1930s,
Bowlby began work in the London Child Guidance Clinic. Soon after his arrival he saw
two cases, one after the other, where the child had been referred for conduct problems.
164 Duschinsky, R., Greco, M., & Solomon, J. (2015) Wait up! Attachment and sovereign power. International
Journal of Politics, Culture, and Society, 28(3), 223–42. The polyvalence of attachment as a political discourse
should be highlighted, however. For uses of attachment theory for a social agenda more aligned with the political
left see e.g. Kraemer, S. & Roberts, J. (1996) The Politics of Attachment: Towards a Secure Society. London: Free
Association Books. There are also a variety of policy texts invoking attachment without a marked political agenda,
and instead using the term in a general sense to characterize the value of ‘positive’ parent–child relationships. See
e.g. Scottish Government (2012) National Parenting Strategy: Making a Positive Difference to Children and Young
People through Parenting. Edinburgh: Scottish Government.
165 ‘Attachment parenting’ is one of the most powerful discourses of intensive parenting. It was introduced by
Bill and Martha Sears (1993) in The Baby Book: Everything You Need to Know about your Baby (Boston: Little and
Brown). The Sears already had their ideas in place, but initially called them ‘immersion parenting’. Use of the idea
of attachment and appeal to Bowlby’s authority was post-hoc and strategic, made available by Bowlby’s ambiguous
and overgeneral statements about the dangers of separation and the need for mothers to spend time with their
baby: ‘At a talk one time in Pasadena, a grandmother came up to Bill and said she thought the term immersion
mothering was a good one, because some moms find themselves “in over their heads.” When he told me of this,
I realized we needed to change the term to something more positive, so we came up with AP, since the Attachment
Theory literature was so well researched and documented, by John Bowlby and others’ (http://attachedattheheart.
attachmentparenting.org/faq/). On the weak evidence for ‘attachment parenting’ and its lack of a link with at-
tachment research see Chaffin, M. (2006) Report of the APSAC Task Force on Attachment Therapy, Reactive
Attachment Disorder, and Attachment Problems. Child Maltreatment, 11(1), 76–89; Fairclough, C. (2013) The
problem of ‘attachment’. In E. Lee, J. Bristow, C. Faircloth, & J. Macvarish (eds) Parenting Culture Studies (pp.147–
64). London: Palgrave.
166 Keller, H. & Thompson, R. (2018) Attachment theory: past, present & future. Recorded at the 2nd ‘Wilhelm
Wundt Dialogue’, 28 November 2018, Leipzig University, hosted by the Leipzig Research Center for Early Child
Development (LFE). https://www.youtube.com/watch?v=_nG5SelEj28.
167 Bowlby, J. (1990, 2015) John Bowlby: an interview by Virginia Hunter. Attachment, 9, 138–57, p.147. See
also Van Dijken, S. (1998) John Bowlby, his Early Life: A Biographical Journey into the Roots of Attachment Theory.
New York: Free Association Books.
Separation 31
Both had been caught stealing, but more generally were considered rude and disobedient.
Bowlby was curious that both of these children had spent nine months in hospital for fever
when they were toddlers, during which time they were isolated and separated from their
caregivers.
During this period, Bowlby was beginning his training in child psychoanalysis, and
reading Klein’s works carefully. He was questioning her description of ‘loss’ only as a norma-
tive developmental stage, and her inattention to the possibility of actual separation and loss
as consequential experiences for a child. Bowlby found support for his concerns in the fact
that, in both of these cases from the Child Guidance Clinic, the parents reported that their
children were ‘emotionally remote’ when they returned home from the hospital. Bowlby
found it remarkable that ‘these stories were extraordinarily similar. So I generalised from a
sample of the two . . . I found a lot of other cases and the upshot was that I wrote this mono-

graph’, ‘Forty-Four Juvenile Thieves’.168 In this paper, published in 1944, Bowlby examined
the records of 44 consecutive cases of children who had been caught stealing, but where the
relevant authorities had sent the child to the Child Guidance Clinic rather than to court. Half
the referrals had come from schools. These children were compared to unselected control
cases from the Clinic with no history of stealing. Fourteen of the 44 thieves he described
as ‘affectionless’ types; of these, 12 had experiences of early separation, compared to 10%
of the controls. He proposed a general theory that early separation experiences predispose
later conduct problems by disrupting the bases of self-worth and capacity for empathy.169
A more specific, and speculative, theory linked early separations specifically to delinquency
and criminal behaviour.170
The importance of separation was, as commentators often note, likely made especially
salient by Bowlby’s observations of evacuated children during World War II, as part of
his work on the Cambridge evacuation survey.171 In unpublished papers written at the
time, Bowlby documented the mental health symptoms shown by these children, which
168 Bowlby, J. (1986) Interview with the BBC. PP/Bow/F.5/8.

Psycho-Analysis, 25, 107–28. An earlier version of the paper, delivered on 30 November 1937, is available as ‘Some
unconscious motives in habitual pilfering’, PP/BOW/C.3/10/2.
170 The specific theory linking separations to delinquency has not subsequently been well supported. There are,
however, few relevant studies. Ryan and colleagues have reported that group care, as opposed to family or foster
care, increases the likelihood of criminal activity: Ryan, J.P., Marshall, J.M., Herz, D., & Hernandez, P.M. (2008)
Juvenile delinquency in child welfare: investigating group home effects. Children and Youth Services Review, 30(9),
1088–99. In the attachment literature, Allen and colleagues found a concurrent relationship between criminal ac-
tivity and insecurity with the Adult Attachment Interview: Allen, J. P., Hauser, S.T., & Borman-Spurrell, E. (1996)
Attachment theory as a framework for understanding sequelae of severe adolescent psychopathology: an 11-
year follow-up study. Journal of Consulting and Clinical Psychology, 64(2), 254. By contrast, Brennan and Shaver
found no association between self-reported attachment style and criminal activity: Brennan, K.A. & Shaver, P.R.
(1998) Attachment styles and personality disorders: their connections to each other and to parental divorce, par-
ental death, and perceptions of parental caregiving. Journal of Personality, 66(5), 835–78. For a review see also
Schimmenti, A. (2020) The developmental roots of psychopathy: an attachment perspective. In S. Itzkowitz & E.F.
Howell (eds) Psychopathy and Human Evil: Psychoanalytic Explorations. London: Routledge.
171 This interpretation of Bowlby was first offered in Chodorow, N. (1978) The Reproduction of Mothering.
Berkeley: University of California Press, p.75. As well as the clinical cases seen at the Child Guidance Clinic, an-
other experience that may have been relevant to Bowlby’s attention to the pathogenic role of separations was his
training as a child analyst. Whereas at the Child Guidance Clinic Bowlby generally saw children and their primary
caregiver or caregivers together, the technique for child analysis in the 1930s was to meet with the child alone.
Bowlby would later recall to his former student Victoria Hamilton that he had been upset by the distress he would
cause young children, time and time again, as he would separate the children from their caregiver in the waiting
room and take them to the consulting room. Hamilton, V. (2007) The nature of a student’s tie to the teacher: remin-
iscences of training and friendship with John Bowlby. Attachment, 1, 334–47.
32 John Bowlby
included ‘tempers, sullenness, disobedience, stealing, sleeplessness, bed-wetting, tim-

idity, pains of an undefined sort’.172 They also seemed disoriented when they saw their
parents again, which Bowlby suspected was a bad sign.173 Bowlby attributed the mental
health symptoms of the evacuated children to feelings of being abandoned by their par-
ents. However, he noticed that such symptoms were reduced—though still present—
among those children billeted with affectionate foster parents. And he observed that such
symptoms were more common when children passed through multiple carers or foster
homes with very large carer-to-child ratios.174 Bowlby noted, too, that these symptoms
were shown not just by children who missed kind and affectionate parents but also by
those children who had cruel or unkind parents. Ultimately, he reflected, ‘it must be re-
membered that even socially bad homes are nevertheless the child’s only harbours in life.
Without a home a child feels lost.’175

An additional factor, rarely mentioned in commentaries on Bowlby, should also be high-
lighted as contributing to his attention to long-term separations. Examination of the ‘Forty-
Four Juvenile Thieves’ case study reveals that some form of child maltreatment appears in
around half of the case studies, alongside the separations. Maltreatment is not mentioned in
Bowlby’s discussion, however, which focuses solely on separation as the cause of the children’s
conduct problems.176 This was done knowingly. Later, in an interview in 1986, Bowlby ac-
knowledged that working with abusive interactions between parents and their children was
‘the run of the mill of what we were doing clinically’ in the Child Guidance Clinic.177 For in-
stance, in an unpublished case history, likely from the early or mid-1960s, Bowlby described
the case of Martin, who lived with both parents, but whose experiences of physically abusive
punishment from his mother contributed to a tendency to turn to his father for comfort when
he was alarmed.178 However, Bowlby reported in the 1986 interview his perception that abu-
sive caregiving could not be turned into a viable research topic with the tools available in the
1940s: ‘My only reason for focusing so exclusively on separation and loss was the fact that one
could get comparatively reliable evidence about them whereas during the forties and fifties we
had neither manpower nor means of recording less crude variables. As a result of adopting this
research strategy, it often appeared that I was unaware of other adverse family events.’179 In par-
ticular, he recalled that one thing that his ‘Forty-Four Juvenile Thieves’ paper ‘misses terribly’
was the physical abuse many of the children had also suffered, alongside multiple separations
or neglect.180
As well as being difficult to measure with the tools of the time, in the 1940s Bowlby felt
that a report on the prevalence of abuse among children referred to the Clinic for conduct
172 Bowlby, J. & Fairbairn, C.N. (c.1939–42) The billeting of unaccompanied school children. PP/Bow/C.5/4/1.
173 Bowlby, J., Miller, E., & Winnicott, D. (1939) Evacuation of small children. Letter to the Editor of the
British Medical Journal, 16 December 1939. In Winnicott, D. (1984) Deprivation and Delinquency (pp.13–14).
London: Tavistock.
174 Bowlby, J. (c.1939–42) Psychological problems of evacuation. PP/Bow/C.5/4/1. This early work by Bowlby
appears as Chapter 7 in Duschinsky, R. & White, K. (eds) (2019) Trauma and Loss: Key Texts from the John Bowlby
Archive. London: Routledge.
175 Bowlby, J. & Fairbairn, C.N. (c.1939–42) The billeting of unaccompanied school children. PP/Bow/C.5/4/1.
176 Follan, M. & Minnis, H. (2010) Forty-four juvenile thieves revisited: from Bowlby to reactive attachment dis-
order. Child: Care, Health and Development, 36(5), 639–45.

177 Bowlby, J. (1986) An interview with John Bowlby on the origins and reception of his work. Free Associations,
1, 36–64, p.39.
178 Bowlby, J. (undated) Untitled case history beginning ‘Mrs E. consulted the Clinic about her son Martin’, in
the file Maternal Behaviour: Humans. PP/Bow/H.136.

179 Bowlby, J. (1985) Letter to Tirril Harris, 17 September 1985. PP/Bow/J.9/94.
Separation 33
problems would raise scandal, and prove unacceptable to public or clinical opinion.181 The
overarching issue of the importance of actual childhood experiences for later development
would risk getting lost. The initial priority, Bowlby felt, had to be to support the study of child
development as a science.182 Bowlby castigated his fellow psychoanalysts in 1943 for their
hostility towards scientific methodology, indeed any methodology besides clinical observa-
tion. He felt that this stance was rendering psychoanalysis increasingly irrelevant to matters
of policy or professional practice: ‘We find ourselves in a rapidly changing world and yet, as a
Society, we have done nothing, I repeat nothing, to meet these changes, to influence them or
to adapt to them. That is not the reaction of a living organism but of a moribund one. If our
Society died of inertia it would only have met the fate that it has invited.’183
To try to be intelligible and credible in context, Bowlby sacrificed reporting the abuse
experiences of the children in his clinic in favour of a focus only on documentable separ-

ations. This hard decision to focus on loss rather than abuse had to be taken, he believed,
so that later researchers, living in a society more frank about family life and with the tools
to do more rigorous work, could make child abuse and its prevention the object of scien-
tific measurement. Even with such justifications, it was one of the professional decisions that
Bowlby found most difficult to forgive. And in an interview the year he died, Bowlby de-
scribed that he remained ‘appalled’ at himself and colleagues: ‘Although I was plugging real
life events, there were a whole lot of events I didn’t give enough attention to. Sexual abuse is
one . . . Physical abuse is another.’184 In a manuscript circulated only to fellow psychoanalysts
late in his career, Bowlby reported that ‘I often shudder’ when he found himself thinking
about the accounts by patients of abuse experiences that he ignored.185
The focus on documentable separations was successful in drawing clinical, research, and
public attention. However, in certain regards the strategy backfired. In the 1950s Bowlby had
something of a tendency to document separations as present or absent, as a crude measure
with high reliability. Anyone could check the record and agree that there had been a separ-
ation or not. Yet this methodology contributed to a tendency for Bowlby to think and write
about separations as merely present or absent, at least until the early 1970s.186 Kinds of sep-
arations were not distinguished. Instead in his early work they were all grouped under the
label ‘maternal deprivation’, and in his later work discussed as ‘separation’ or ‘lack of con-
tinuity’. The concept of ‘maternal deprivation’ had the problem that it implied that the child
belonged, specifically, with their mother. Ainsworth repeatedly criticised Bowlby for the fact
that at times he used the term ‘maternal deprivation’ when describing children whose pre-
dominant experience had actually been cruelty or abuse from their caregiver, with separation
181 Bowlby, J. (1984) Violence in the family as a disorder of the attachment and caregiving systems. American
Journal of Psychoanalysis, 44, 9–27: ‘It was, indeed, largely because the adverse behavior of parents toward their
children was such a taboo subject in analytic circles when I was starting my professional work that I decided to
focus my research on the effects on children of real-life events of another sort, namely separation’ (10).
182 Bowlby, J. (1990, 2011) John Bowlby: interview by Leonardo Tondo. Clinical Neuropsychiatry, 8(2),
159–71, p.160.
183 Bowlby, J. (1943, 1992) Contribution to business meeting. In P. King & R. Steiner (eds) The Freud–Klein
Controversies 1941–45, 2nd edn. London: Routledge, p.369.

184 Bowlby, J. (1990, 2015) John Bowlby: an interview by Virginia Hunter. Attachment, 9, 138–57, p.156.
185 Bowlby, J. (1985, 1991) The role of the psychotherapist’s personal resources in the treatment situation.
Bulletin of the British Psychoanalytic Society, 27(11), 26–30, p.29–30. Published as Chapter 12 in Duschinsky, R. &
White, K. (eds) (2019) Trauma and Loss: Key Texts from the John Bowlby Archive. London: Routledge.
186 By the 1980s, Bowlby was admonishing colleagues for not being specific enough in their use of the term ‘sep-
aration’, e.g. Bowlby, J. (1985) Letter to John Byng-Hall, 12 April 1985. PP/Bow/J.9/45: ‘I think one needs to be a little
more precise about lengths of separation—words like moderate and prolonged are obscure. I suggest a week or two
instead of “moderate”, and “longer than that” in place of prolonged.’
34 John Bowlby
only a lesser feature.187 However, an additional problem was that ‘maternal deprivation’, ‘sep-
aration’, and ‘lack of continuity’ were all used in an undifferentiated way, and could connote
everything from a child sleeping alone in a room, to use of daycare, to child neglect, to insti-
tutionalisation in an orphanage.188
In unpublished writings, and in correspondence, Bowlby was quite capable of making
these distinctions from the 1940s onwards.189 Writing to Michael Rutter in 1971, he
stated: ‘As regards long-term effects of brief experiences, we [Bowlby’s research group]
have endeavoured to keep an absolutely open mind. The view I have held for some years
is rather like Doll’s view of smoking. Whilst serious effects are found almost always only
by prolonged and heavy smoking, even lighter and less prolonged smoking can have ad-
verse effects in some people. Where one draws the line in practice then becomes a matter
for private judgement.’190 This more qualified position in private is unfortunately gen-

erally absent in Bowlby’s published writings. His tendency to document separation as
a binary variable aligned with populist appeal to a stereotyped image, sun-lit and stock,
of mother and child as a natural whole. In both his popular and his scholarly writings
Bowlby tended to describe separation as a quantitative phenomenon, without qualitative
differentiation. In Separation, for instance, the qualifications are removed from the com-
parison previously offered to Rutter, and the claim is made that more separations of any
kind are simply wore: ‘the effects of separation from mother can be likened to the effects
of smoking or of radiation. Although the effects of small doses appear negligible, they
are cumulative’.191 The crude stance on separations in much of Bowlby’s published work
has been considered mistaken and misleading not only by feminist critics and other re-
searchers, but also by his collaborators, followers, and even Bowlby’s own family.192 It has
had, additionally, an abiding impact on perceptions of Bowlby and of attachment theory
in general.
187 E.g. Ainsworth, M. (1962) The effects of maternal deprivation: a review of findings and controversy in
the context of research strategy. In Deprivation of Maternal Care: A Reassessment of its Effects (pp.87–195).
Geneva: WHO, p.99. See also Yarrow, L. (1961) Maternal deprivation: toward an empirical and conceptual re-
evaluation. Psychological Bulletin, 58, 459–90.
188 Rutter, M. (1972) Maternal Deprivation Reassessed. London: Penguin; Rutter, M. (2002) Maternal depriv-
ation. In M. Bornstein (ed.) Handbook of Parenting, 2nd edn (pp.181–202). Mahwah, NJ: Lawrence Erlbaum.
189 E.g. Bowlby, U. & Bowlby, J. (1940) Difficult children. PP/Bow/H7: ‘The term—broken home covers a multi-
tude of situations—illustrate. Vague, don’t intend to use term but to examine different situations . . . Broken home
does not cause trouble.’
190 E.g. Bowlby, J. (1971) Letter to Michael Rutter, 6 October 1971. PP/Bow/J.9/161.
191 Bowlby, J. (1973) Separation: Anxiety and Anger. New York: Basic Books, p.96. Another contributing factor
appears to have been Bowlby’s initial difficulties in articulating the distinction between observable attachment
behaviour and the invisible attachment behavioural system. Bowlby would often make claims urging parents to
always do their best to follow the cues of a child’s attachment to ensure their child’s wellbeing. To the degree that
this refers to the attachment behavioural system, the claim is clearly overstated, and neglects his friend Robert
Hinde’s criticisms that weaning and other requirements on parents mean that the short-term demands of the
attachment system should not always be given priority in facilitating children’s long-term security. However, to
the degree that Bowlby was implying that parents should follow the dictates of attachment behaviours—such
as distress on separation—the result is an even more extreme position. It would imply that any separation is, in
itself, potentially harmful. See, for instance, Bowlby, J. (1987) Baby love: an interview. Hampstead and Highgate
Express, April 1987. PP/Bow/A.5/19: ‘The more a child’s attachment is respected and responded to, the more he’ll
feel secure.’
192 Ainsworth, M. (1962) The effects of maternal deprivation: a review of findings and controversy in the context
of research strategy. In Deprivation of Maternal Care: A Reassessment of its Effects (pp.87–195). Geneva: WHO,
p.101; Bowlby, R. (2005) Fifty Years of Attachment Theory. London: Karnac; Vicedo, M. (2011) The social nature
of the mother’s tie to her child: John Bowlby’s theory of attachment in post-war America. British Journal for the
History of Science, 44(3), 401–26.
Separation 35
Prospective studies of separations
The treatment of any form of caregiver–child separation as equivalent was unwarranted, and
to an extent polemical. Bowlby had no evidence on which to base such claims. He did have
some limited evidence regarding the consequences of relatively long-term separations in
childhood. The ‘Forty-Four Juvenile Thieves’ paper looked backwards from clinical cases
to find potential pathological causes. However, Bowlby was well aware that this research
strategy had significant methodological flaws, not least the problem of confirmation bias for
pre-existing theoretical ideas.193 Instead, he advocated a longitudinal methodology, which
began by taking children who had experienced a long-term separation and examining its
sequelae. Belief in the promise of slow, empirical, longitudinal study of emotional and family

life would become a hallmark of attachment research.
In 1946 Bowlby took up a position with clinical and research responsibilities at the
Tavistock Clinic in London. There, in 1948, he founded the Separation Research Unit,
and hired James Robertson as a research assistant to study children undergoing hospi-
talisation.194 Mary Boston and Dina Rosenbluth, child psychotherapy trainees at the
Tavistock, also contributed to the research.195 At the time, it was common policy to keep
parents from visiting their children. For infectious diseases this was partly to prevent
the spread of infection. However, it was also general policy for non-infectious diseases
as well, since children tended to become distressed and difficult for the hospital staff to
manage after their parents had visited. Bowlby had long been interested in the predica-
ment of hospitalised children, which he regarded as a natural experiment to test the ef-
fects of separation.196 In the 1930s, he had seen a 15-year-old girl, Joan, in the Child
Guidance Clinic:
Joan was examined because of severe headaches, which had begun when she was ten years
old. Since no organic basis could be found and psychical factors were obvious she was
treated with psychotherapy. After some weeks she described how she suffered from absent
periods which were evidently hysterical dream-states. For periods up to two or three hours
her head would feel funny and she would be unable to remember the recent past. Things
looked different and seemed unreal. If she went to touch a thing she found it was not there.
She herself remarked ‘It’s as if I’m in a dream’. The episode ended suddenly and other girls
at school would tell her she had been staring curiously, ‘looking beyond usual things’. . . The
episodes began with severe headaches and would come on when she was frightened or
anyone was angry. In this case the hysterical headaches and dream-states had begun when
she was ten years old after she and her brother had been in hospital with scarlet-fever and
diphtheria.197
193 Bowlby, J. (1965) Comments on Joffe and Sandler 1965 ‘Notes on Pain, Depression and Individuation’. PP/
Bow/J.9/168-9.
194 Van der Horst, F.C.P. (2011) John Bowlby— From Psychoanalysis to Ethology. Unravelling the Roots of
Attachment Theory. Oxford: Blackwell.
195 Rustin, M. (2007) John Bowlby at the Tavistock. Attachment & Human Development, 9(4), 355–9.
196 Bowlby, J. & Robertson, J. (1965) Protest, Despair and Detachment. PP/BOW/D.3/38: ‘Separation can be lik-
ened to the natural experiments that are exploited by students of nutrition.’

197 Bowlby, J. (1939) Hysteria in children. In H. Milford (ed.) A Survey of Child Psychiatry (pp.80–94).
Oxford: Oxford University Press, p.84.

36 John Bowlby
From 1948 Robertson began making direct observations of a sample of children who had
been hospitalised. The first outputs from Robertson’s research were a film and a 1952 paper
entitled ‘A two-year-old goes to hospital’, which documented the behaviour of a two-and-a-
half-year-old girl, Laura, who was hospitalised for eight days.198 This film helped contribute
to recognition of the sorrow major separations can cause children, and to the important
movement to change hospital visitation regulations in the 1950s.199
Robertson documented that Laura initially showed a great deal of distress and protest.
Her affect then turned towards apparent depression, though accompanied by tic-like stress
movements. Three months after she returned home, her mother was away in hospital herself,
to have a baby. On her mother’s return, Laura seemed avoidant and somewhat disoriented
on reunion:

Laura was very excited and keen to return home. Half an hour later she arrived and the
mother could hear her banging on the outside door and calling “Mummy, Mummy”. But
when her mother opened the door, Laura looked at her blankly and said, “But I want my
Mummy”. For the next two days she did not seem to recognise her mother, and although
quite friendly was completely detached. This naturally upset her mother very much. When
the father came home an hour or two after Laura’s return, Laura was for a few moments
mute toward him but then recovered quickly and was friendly and sure of him.200
Robertson went on to observe 50 cases, although less than half of these were observed in-
tensively, and the context and kind of hospitalisation was highly diverse.201 Reviewing these
cases, Robertson and Bowlby came to the conclusion that because the children’s efforts to
regain their familiar caregivers had failed both chronically and painfully, they responded by
inhibiting their feelings, and especially their yearning for their family. Bowlby termed this
‘detachment’.202 Ainsworth sought, essentially unsuccessfully, to persuade him that the term
was misleading, since Bowlby did not intend to suggest that the child was no longer attached
to their caregiver, but that an inhibition was observable that blocked intense feeling and its
expression.203 Robertson and Bowlby proposed that such inhibition was the cause of the
depressed affect, the tic-like tension movements, and also the avoidant or disoriented behav-
iour on reunion.204 Robertson noted his qualitative impression that children who sustained
198 Bowlby J., Robertson, J., & Rosenbluth, D. (1952) A two-year-old goes to hospital. Psychoanalytic Study of
the Child, 7, 82–94. During the war, Robertson had worked with Anna Freud at the Hampstead Nurseries with
children, many of whom had been evacuated from London or who had no family to care for them. Freud and col-
leagues documented avoidant behaviour by young children to caregivers, including following reunions. This likely
primed Robertson’s interest in the avoidant behaviour shown by Laura and other hospitalized children on reunion.
See Burlingham, D. & Freud, A. (1944) Infants without Families. London: Allen and Unwin, p.63.
199 Van der Horst, F.C. & van der Veer, R. (2009) Why we disagree to disagree: a reply to commentaries by
Robertson and McGilly, and Lindsay. Attachment & Human Development, 11(6), 569–72.
200 Bowlby J., Robertson, J., & Rosenbluth, D. (1952) A two-year-old goes to hospital. Psychoanalytic Study of the
Child, 7, 82–94, p.85.

201 Bowlby, J. & Robertson, J. (1965) Protest, Despair and Detachment. PP/BOW/D.3/38.
202 Bowlby, J. (1960) Separation anxiety. International Journal of Psycho-Analysis, 41, 89–113, p.90. See also
Southgate, J. (1998) Attachment, intimacy, autonomy. British Journal of Psychotherapy, 14, 389–93: ‘During my
supervision with John . . . he argued to keep its original meaning on the grounds that once in an attachment space
or relational field there is always some form of attachment even if it is disorganized and chaotic’ (390).
203 Ainsworth, M. (1962) Letter to John Bowlby, 11 December 1962. Mary Ainsworth papers, Box M3168, Folder
1. For an example of an otherwise careful reader and friend of Bowlby assuming that ‘detachment’ meant the op-
posite of attachment see Birtchnell, J. (1987) Attachment—detachment, directiveness—receptiveness: a system for
classifying interpersonal attitudes and behaviour. British Journal of Medical Psychology, 60(1), 17–27.
Separation 37
this inhibition of feeling longer, and who received less comfort from staff during the hospi-
talisation, were those that showed more psychological disturbance once they returned home.
These children displayed more anxiety and aggression, and less affection or help-seeking to-
wards their parents; and these affects were also more likely to occur at odd moments, without
apparent reason.205
Bowlby’s confident tone in his academic writing about the impact of the hospitalisations,
and crude statements warning about the dangers of separation in his popular writings, led
many readers, even sympathetic ones, to assume that he saw a direct relationship between
early separations and later behaviour.206 Hazen and Shaver called this, in their assessment,
‘one of the most common misconceptions about attachment theory’.207 This impression was
likely reinforced by a strong tendency in Bowlby’s reporting of his own and his analysis or
exposition of others’ quantitative findings to neglect attention to moderators and interaction

effects. One contributing factor may have been Bowlby’s lack of expertise in using and inter-
preting statistical tests for interaction effects in empirical research. In fact Bowlby, in his
theoretical writing already from the early 1950, explicitly acknowledged two additional fac-
tors that need to be included in any causal model, based on Robertson’s observations. The
first is that children will have ‘differential susceptibility of an inherited kind’ to the effects of
separation.208 The other is that the implications of the separation for a child’s development
will be mediated through the consequences of this event and the child’s behaviour for family
interactions: ‘Most mothers find their children’s failure to recognise them and to respond,
or their outright rejection of them, on reunion extremely hard to bear, and later, when feel-
ings have thawed, their intense possessiveness and whining mummyishness tries patience
to the limit. Events such as these set up vicious circles which undoubtedly play an important
part in establishing adverse patterns of behaviour in the personality.’209 Rather than treating
long-term separation itself as solely directly harmful as implied by Bowlby in popular works
and papers reporting empirical findings, in theoretical reflections Bowlby emphasised that
harms could be expected to be partially mediated through the vicious cycles that could be
expected to be established in family relationships.210
In fact, to Bowlby’s disappointment, the quantitative measures from the Robertson
study of the effects of hospitalisation did not supply the evidence he expected of negative
205 Ibid: ‘It is our impression, and that of others, that children who have reached a detached state . . . “blow up”
more easily and more violently event than the ordinary child at home.’ An example was Bobby: ‘Everyone was
punched when necessary—including father—and it was usually possible to detect the immediate reason for it. But
his treatment of mother was exceptional in that she was often punched for no apparent reason. At times he would
approach her with a bland or smiling face that gave no warning of the severe body blow that was to follow.’ See also
Robertson, J. & Robertson, J. (1971) Young children in brief separation: a fresh look. The Psychoanalytic Study of
the Child, 26(1), 264–315.
206 E.g. Eagle, M. (2013) Attachment and Psychoanalysis. New York: Guilford.
207 Hazan, C. & Shaver, P.R. (1994) Deeper into attachment theory. Psychological Inquiry, 5(1), 68–79, p.70.
208 Bowlby, J. (1953) Some pathological processes set in train by early mother– child separation. Journal of
Mental Science, 9, 265–72, p.271. The exact meaning of the term ‘differential susceptibility’ for Bowlby is unclear.
It may or may not necessarily imply that genetic factors may be ‘for better or for worse’, as later for Belsky, van
IJzendoorn, and Bakermans-Kranenburg.
209 Ibid.
210 Bowlby was pleased with later research findings indicating the mediating role of family relationships on
the impact of parental loss in childhood on an individual’s later mental health. See PP/BOW/F.4/1:Box 40 and
PP/BOW/J.9/19:Box 60 for Bowlby’s reflections on and correspondence with Harris, T., Brown, G.W., & Bifulco,
A. (1986) Loss of parent in childhood and adult psychiatric disorder: the role of lack of adequate parental care.
Psychological Medicine, 16(3), 641–59.
38 John Bowlby
consequences of long-term separation.211 The quality of the data was, by his own admission,
unusably poor. Ainsworth later commented that Bowlby had been overconfident in his hy-
pothesis, and so did not take sufficient care in choosing his measures: ‘He had expected this
to be so conspicuous that he used very crude measures of assessment—teachers’ ratings—
and centred in on the IQ. Actually there was nothing in the IQ. The IQs of these children
were not lower. And the teachers’ ratings were not sufficiently sensitive really, to turf up any
very conspicuous differences.’212 These quantitative results were therefore held back from
publication. However, the qualitative descriptions made of the children by Robertson during
the hospitalisation and on their return home were very rich and suggestive in their detail.213
Ainsworth later described Robertson’s qualitative descriptions as ‘entrancing’ and ‘deeply
impressive’.214 In 1968 she remarked to Bowlby that ‘despite all the lapse of time and subse-
quent research’ it is ‘still the best’ and most revealing descriptions made by early research on

attachment.215 In an oral history conducted in the same period, Ainsworth recalled:
I was tremendously impressed with this material. Jimmy was a social worker at the time but
he has since been qualified as an analyst. His observations were the most sensitive direct
observations I had ever encountered. I don’t think I have ever encountered anyone who
was more perceptive.216
One of the most consequential aspects of Robertson’s work was the distinction between ‘am-
bivalent’ anxiety/preoccupation and ‘withdrawn’ avoidance. Following separation, these
were often—though certainly not always217—observed in sequence, with protest at sep-
aration followed by flattened affect over time. Robertson also identified related behaviour
upon and following reunion. Anxious preoccupation with the parent was often shown by
the formerly hospitalised children.218 And avoidant behaviour was sometimes evident at the
moment of reunion with the parents, and could also manifest as withdrawal from the par-
ent in the months after the child returned home.219 In a paper from 1956 reporting on their
follow-up study with the hospitalised children, Bowlby, Ainsworth, and colleagues wrote
211 Bowlby, J. (1976) Bowlby on latch-key kids: interviews with Dr Nicholas Tucker. Psychology Today, Autumn
1976, 37–41: ‘I felt that because we had used such very superficial measures we were in no real position to give
an adequate account of how these children had developed . . . I really don’t think the study has much scientific
value’ (38).
212 Ainsworth, M. (1969) CPA oral history of psychology in Canada interview. Unpublished. http://www.femi-
nistvoices.com/assets/Women-Past/Ainsworth/Mary-Ainsworth-CPA-Oral-History.pdf.
213 Though a proportion of the most important data was second-hand. Bowlby, J. & Robertson, J. (1965) Protest,
Despair and Detachment. PP/BOW/D.3/38: ‘As experience accumulated, it came to be realised that the way a child
greets his mother on the occasion of his return home is of great interest. This, however, was not realised in our
earlier studies and for this reason in all but a few cases our data referring to this event were obtained second-hand.’
214 Ainsworth, M. (1983, 2013) An autobiographical sketch. Attachment & Human Development, 15(5–6),
448–59, p.454.
215 Ainsworth, M. (1968) Letter to John Bowlby, 2 November 1968. Mary Ainsworth papers, Box M3168,
Folder 3.
217 Van der Horst, F.C. & van der Veer, R. (2009) Separation and divergence: the untold story of James Robertson’s
and John Bowlby’s theoretical dispute on mother–child separation. Journal of the History of the Behavioral Sciences,
45(3), 236–52.
218 That the preoccupied response was a reflection of concern about the availability of the caregiver was con-
firmed by Bowlby’s clinical observations of the possessiveness, anger, and jealousy with which toddlers responded
to the birth of a new sibling, especially if the toddler already had a troubled relationship with his or her caregiver.
Bowlby, J. (1955) New baby jealousy. Parents, December 1955, p.42–4.
Separation 39
that ‘the personality patterns of children who have experienced long separation tend to fall
into one or other of these two opposite classes’: either (i) ‘over-dependent’ and ‘ambivalent’
or (ii) ‘mother-rejecting . . . having repressed their need for attachment’.220
Ainsworth found it thought-provoking that the two major classes of behaviour appeared
to be ‘opposites’. She expressed her fascination with the ‘anxious over-dependence on the one
hand, and superficiality and affectionlessness on the other’ in the follow-up study, and the
way that this seemed to correspond to ‘the anxious clinging response following reunion after
relatively brief or mild separations on the one hand, and the detachment and failure to re-
establish affectional relations after long and severe separations on the other’ (Chapter 2).221
Bowlby, Ainsworth, and colleagues discussed this extensively, and they came increasingly to
regard the two classes ‘as the prototypes of responses that, when seen in acute and chronic
form in older individuals and out of family context, are habitually labelled as psychiatric

symptoms. Thus, for example, when the family situation is ignored, it becomes easy to label
despair as depression, detachment as psychopathic lack of affect, protest as hysterical.’222
However, these observations of hospitalised children were dismissed as circumstantial
evidence by Bowlby’s critics. For instance, the behaviour of hospitalised children could
be rejected as unrepresentative of children who were healthy. Experimental research was
needed to support the case, but major separations of human children in experimental re-
search was not morally permissible. Bowlby helped Hinde secure the funds to set up a colony
of rhesus monkeys at Madingley, Cambridge.223 Rhesus monkeys were chosen on the basis
of the intimate relationship between infant and mother during early childhood, which made
attachment phenomena especially visible. One of many important findings from Hinde’s
Madingley colony of rhesus monkeys from the 1960s was observation of both the ‘classes’
of behaviour following separation that had been identified by Bowlby, Ainsworth, and col-
leagues in their 1956 paper. Spencer-Booth and Hinde studied the effects of separating
rhesus monkey infants from their mothers for six days, too long to be treated as an ordinary
separation and so able to offer a parallel to hospitalisation. The researchers observed that ‘the
infants’ behaviour during the mothers’ absence can only be described as depressed. They
sat in the hunched, passive attitude of a subordinate animal.’224 After reunion, they ‘showed
exceptionally intense tantrums when rejected by their mothers, and often flung themselves
220 Bowlby, J., Ainsworth, M., Boston, M., & Rosenbluth, D. (1956) The effects of mother–child separation: a
follow-up study. British Journal of Medical Psychology, 29, 211–47, p.238. On the two classes as ‘opposites’ see
also Bowlby, J. (1968, 1970) Disruption of affectional bonds and its effects on behavior. Journal of Contemporary
Psychotherapy 2(2), 75–8: ‘In the separated children, two forms of disturbance of affectional behaviour were
seen, neither of which were observed in the comparison group of non-separated children. One form is that of
emotional detachment; the other, its apparent opposite, i.e., an unrelenting demand to be close to mother’ (82).
Both responses would be documented again, decades later, by Stovall-McClough and Dozier, observing chil-
dren one to two years old in the first months after joining a foster-family. Stovall-McClough, K.C. & Dozier, M.
(2004) Forming attachments in foster care: infant attachment behaviors during the first 2 months of placement.
Development & Psychopathology, 16(2), 253–71. Stovall-McClough and Dozier observed that these behaviours
may mis-cue the foster-parent about the child’s needs. These findings fed into the construction of the Attachment
and Biobehavioral Catch-up intervention (Chapter 6).
221 Ainsworth, M. (1962) The effects of maternal deprivation: a review of findings and controversy in
the context of research strategy. In Deprivation of Maternal Care: A Reassessment of its Effects (pp.87–195).
Geneva: WHO, p.140.
222 Bowlby, J. (1976) Human personality development in an ethological light. In G. Serban & A. Kling (eds)
Animal Models in Human Psychobiology (pp.27–36). New York: Plenum Press, p.28.
223 Bateson, P., Stevenson-Hinde, J., & Clutton-Brock, T. (2018) Robert Aubrey Hinde CBE. 26 October 1923–23
December 2016. Royal Society Biographical Memoirs, 65.

224 Spencer-Booth, Y. & Hinde, R. (1967) The effects of separating rhesus monkey infants from their mothers for
six days. Journal of Child Psychology & Psychiatry, 7, 179–97, p.187.

40 John Bowlby
violently on to their mothers, or sometimes, when the mother had rejected them, on to
aunts’:
A further interesting feature was the way in which the infants could change from being re-
laxed to being very upset and clinging without apparent cause. Thus Tim on Days 11 and
12 was recorded as coming off his mother in an apparently calm fashion, then suddenly
panicking and going on her geckering. The most dramatic example was that of Linda, who
on Day 16 was playing in a very relaxed fashion for about the first 35 min of the watch,
then went on the mother and slept. When she awoke she seemed very upset and, terrified,
cringed and would hardly leave her mother.225
Spencer-Booth and Hinde also documented that ‘the deprivation experience accentuated

pre-existing individual differences. When the various symptoms were combined to give a
distress index, those infants which had been rejected most and/or played the largest role in
maintaining proximity before separation were most disturbed after. By contrast, the actual
time the infant spent off or at a distance from its mother before separation bore practically
no relationship to the disturbance it showed subsequently.’226 Spencer-Booth and Hinde fur-
ther found that all effects were intensified if they separated the infants from their mothers for
two six-day periods or for thirteen days.227 These experimental studies with rhesus monkeys
therefore provided further support that depression or detachment could be expected during
a major separation from the primary caregiver, and that anxious preoccupation with the
caregiver occurred following reunion, especially for those infants who had received the least
proactive and nurturing caregiving.
Disorientation
However, the ambivalent and avoidant classes of behaviour were not the only behaviours
observed by Robertson that would prove consequential for attachment theory. Robertson
identified other anomalous behaviours, especially during separation, but also sometimes on
reunion. These were described in detail in an unpublished book of Robertson’s observations,
written and re-written with Bowlby over the span of about ten years from the mid-1950s to
mid-1960s.228 The central theme that Robertson perceived in these anomalous behaviours
was disorientation. Bowlby and Robertson regarded the behaviours as suggesting some dis-
ruption of the attachment response. However, what exactly they signified was not clear, and
this may have contributed to the decision by Bowlby to hold back the book from publication,
alongside Robertson’s gradual departure from the Tavistock in the early 1960s. There was
also little academic interest during this period in disoriented behaviour as a mental health
symptom.229 Ainsworth regarded the absence of the Robertson and Bowlby book as a major
225 Ibid. p.190.

226 Hinde, R. & Spencer-Booth, Y. (1970) Individual differences in the responses of rhesus monkeys to a period
of separation from their mothers. Journal of Child Psychology & Psychiatry, 11, 159–76, p.174.
227 Hinde, R. & Spencer-Booth, Y. (1971) Effects of brief separation from mothers on rhesus monkeys. Science
173(3992), 111–18.
229 Ross, C.A. (1996) History, phenomenology, and epidemiology of dissociation. In L.K. Michelson & W.J. Ray
(eds) Handbook of Dissociation: Theoretical, Empirical and Clinical Perspectives (pp.3–25). New York: Springer;
Separation 41
loss to developmental psychology, and a source of personal sadness given the quality of the
observations.230 The behaviours only came back into central focus in the work of Mary Main,
some decades later (Chapter 3).
One sign of disorientation was that behaviour became unmoored from environmental
cues for activation and termination. Robertson noted that some children would swing,
seemingly without external prompt, between the ambivalent and avoidant classes of behav-
iour. For instance: ‘By about the end of her fifth month at home Jacqueline was frequently
seeking “baby cuddles” from her mother not only in the evening but also in the daytime. In
these brief moments she would curl up and revel in the mutual indulgence she and mother
permitted themselves. But between the extremes of being a helpless sensual “baby” and of
being detached and independent there was little behaviour of a moderate and quietly affec-
tionate kind.’231

Robertson also documented many signs of disorientation on reunion. He offers a vivid
description of Laura’s behaviour during a brief visit by her mother to the hospital:
One of the most striking features about Laura is that, despite being only two years and five
months old, she contrived much of the time to control the expression of her grief. Mother
announced “I’m going home now”. Laura’s expression was instantly tense and unhappy.
Mother insisted “Don’t cry” and pointed an admonitory finger; Laura nodded uncertainly.
As her mother left and before she was out of sight Laura turned away with an expression of
the deepest misery on her face. The relief of tears, which would have come to most children
of that age in that situation, was not available to Laura. As she tried to keep her feelings in
check she idly turned the pages of a book, fingered her hair, and both hands fluttered impo-
tently before her face as if she had been momentarily disoriented.232
The withdrawn avoidant class of behaviour was, he thought, especially often accompanied
by disorientation in relation to the caregiver. Robertson interpreted both disorientation and
avoidance as suggesting the ‘repression of attachment behaviour’.233
A different child showed fear on reunion with her mother. Robertson interpreted this as
an effect of disorientation, with the mother misrecognised as a stranger due to repression of
the attachment response:
When they reached home and Mary saw her mother for the first time in six weeks, she
screamed and refused to go near her. This rejection continued for several days during
which she treated her mother so much as though she was a frightening stranger . . . Only
after a week did Mary begin to show a wish to be near her mother. Then her behaviour
moved to the other extreme and she followed mother about continuously as if afraid to let
her out of her sight. She became increasingly aggressive towards her sisters, and made sev-
eral vicious attacks on her new baby brother.234
Van der Hart, O. & Dorahy, M.J. (2009) Dissociation: history of a concept. In P.F. Dell & J. O’Neill (eds) Dissociation
and the Dissociative Disorders: DSM-V and Beyond (pp.3–26). London: Routledge.
230 Ainsworth, M. (1970) Letter to John Bowlby, 28 September 1970. Mary Ainsworth papers, Box M3168, Box 3.
231 Bowlby, J. & Robertson, J. (1965) Protest, Despair and Detachment. PP/BOW/D.3/38. Bowlby’s colleagues
Heinicke and Westheimer also documented such responses following brief separations. Heinicke, C.M. &
Westheimer, I. (1966) Brief Separations. Oxford: International Universities Press.
233 Ibid.
234 Ibid.
42 John Bowlby
Another behaviour that Robertson described as having a disoriented quality was freezing
when alarmed rather than looking for support from a familiar caregiver. For instance, Vicky
showed such behaviour soon after returning from the hospital: ‘On the third day her grand-
mother took her out. Although she had wanted to see the traffic, when she was halfway across
a road she suddenly became petrified by fear, and refused to move, while the cars stopped
and hooted and people stared.’235
Robertson also described forms of disorientation that led to care-seeking behaviours
being directed towards strangers rather than familiar caregivers, what might now be re-
ferred to as ‘disinhibited social engagement’. For example, he described the behaviour of
Jacqueline: ‘unlike ordinary children of this age, when hurt she did not go to her mother
for comfort. Indeed, on an occasion when she bruised a finger quite badly Jacqueline did
not turn to her mother but to a visitor who happened to be in the room.’236 Robertson’s

impression was that a significant minority of the hospitalised children seemed to show
‘a tendency to make shallow relationships with one and all, and thus to be undiscrimin-
ating’ in their attachment behaviour following their hospitalisation.237 He described the
behaviour of Stephanie: ‘Although she had not seen me for 18 months, she was immedi-
ately friendly and asked for a ride in my car: she was driven around the avenues without
showing anxiety, a response mother commented upon. The parents also expressed a re-
lated concern—that she was unable to concentrate on anything for more than a very short
time. They linked this, with considerable alarm, to the fact that she had been wandering
away from home.’
Behavioural systems
The idea of a behavioural system
In the unpublished Robertson and Bowlby book, the authors described many observations
relevant to the disruption of the attachment response: ambivalent behaviour, avoidant
behaviour, and various forms of what appeared to be disorientation. However, the inter-
pretation of these observations kept running into trouble, across the various drafts and
redrafts of the book. ‘Attachment’, ‘detachment’, ‘disorientation’, ‘preoccupation’—all these
terms were used at times to refer to observable behaviour and at times to an inferred pro-
cess at a motivational level. Sometimes they referred to voluntary actions; sometimes to
involuntary responses or predispositions. This led the research group in circles at times, as
Bowlby became increasingly aware as the 1950s progressed.238 Bowlby’s work on what be-
came attachment theory was driven in part by a desire to develop a conceptual apparatus
adequate to the findings of the Separation Research Unit. Indeed, Attachment, Volume 1
began life as a single theoretical chapter for a book reporting on the empirical work with
235 Bowlby, J. & Robertson, J. (undated, 1950–65?) Cases relating to part II and III. PP/BOW/D.3/11-12.
237 Ibid.
238 Bowlby, J. (1960) Separation anxiety. International Journal of Psycho-Analysis, 41, 89–113, p.95.
Behavioural systems 43
hospitalised children.239 The fundamental concept that Bowlby elaborated was that of a
‘behavioural system’.
The idea of a behavioural system was developed within ethology.240 It referred to ‘sys-
tems postulated as controlling a group of behaviour patterns that together serve to achieve
a given biological end’.241 For Bowlby, the concept was essentially a metaphor, ‘conceived
on the analogy of a physiological system organised homeostatically to ensure that a certain
physiological measure, such as body temperature or blood pressure, is held between appro-
priate limits’.242 If these limits are breached, then steps are taken by the individual to alter
the environment or itself to regain them, achieving what Bowlby termed the ‘set-goal’ of
the system and re-estabilishing homeostatic equilibrium.243 Each behavioural system can
recruit various kinds of resources, most visibly behaviours, to respond flexibly to the envir-
onment to achieve the set-goal. The concept of a behavioural system therefore presupposes

a hierarchy between an overall goal and the means of achieving this goal, which must be as-
sembled and coordinated together.244
As such, different behaviours commonly associated with the same system may not cor-
relate, or may even negatively correlate with one another, as they are different paths to the
same goal. For instance, crying and smiling directed towards the caregiver may be alterna-
tive paths to achieving the availability of an attachment figure; freezing and fleeing alterna-
tive paths to avoiding danger for the fear system. The concept of behavioural systems allowed
Bowlby to clearly distinguish between the behavioural system (the motivation) and behav-
iour (the observable actions undertaken in the service of the motivation). Admittedly these
terms are sufficiently close, and Bowlby was insufficiently careful in his use of the concepts,
that this was and has remained an enduring point of confusion for his readers.245 However,
the distinction between invisible motivation and observable behaviour is sharp enough
when Bowlby is being careful, and it was a consequential distinction for subsequent attach-
ment theory and research.
When activated, behavioural systems such as attachment, caregiving, exploration, fear,
and aggression initiate a disposition to try to achieve a set-goal. Where past and present in-
formation suggest that the environment will be receptive, the motivation finds expression
in behaviour. Indeed, in contexts where the environment may demand the intense activa-
tion of a behavioural system, but provides only subtle cues, Bowlby observed in Loss that
normal attentional processes may be sharpened and narrowed, as ‘perceptual vigilance’. Such
vigilance will lower the threshold at which conditions in the environment stir a slumbering
239 Ainsworth, M. (1966) Letter to John Bowlby, 29 October 1966. Mary Ainsworth papers, Box M3168, Folder 2.
240 Sevenster, P. (1961) A causal analysis of a displacement activity (Fanning in Gasterosteus aculeatus L.).
Behaviour, 9, 1–170; Baerends, G.P. (1976) The functional organization of behaviour. Animal Behaviour, 24, 726–
38. The movement of the concept from ethology to attachment theory is discussed further in Grossmann, K. &
Grossmann, K. (2012) Bindungen—das Gefüge psychischer Sicherheit Gebundenes. Stuttgart: Klett-Cotta.
241 Hinde, R. (1983) Ethology and child development. In P.H. Mussen (ed.) Handbook of Child Psychology
(pp.27–93). New York: Wiley, p.57.

242 Bowlby, J. (1982) Attachment and loss: retrospect and prospect. American Journal of Orthopsychiatry, 52(4),
664–78, p.670.
244 Marvin, R.S., Britner, A.A., & Russell, B.A. (2016) Normative development: the ontogeny of attachment in
childhood. In J. Cassidy & P.R. Shaver (eds) Handbook of Attachment: Theory, Research, and Clinical Applications,
3rd edn (pp.273–90). New York: Guilford.
245 An important case was that of contemporary social learning theorists, e.g. Gewirtz, J.L. (ed.) (1972)
Attachment and Dependency. Washington, DC: Winston.

44 John Bowlby
behavioural system to life; it will also increase the intensity of motivation, which will likely be
expressed in more elaborate and intense behaviour to achieve the set-goal of the system. This
model appeared to Bowlby to account for the clingy behaviour documented by Robertson in
children following hospitalisation: the distress of the separation prompted perceptual vigi-
lance regarding proximity to their parents in these children. This led to an intense activation
of the attachment system even to minor cues of caregiver unavailability. In turn, the opposite
process could explain the ‘detached’ behaviour of the children who had been hospitalised
for some time, and on reunion with their caregiver. Where the environment has come to be
perceived as unreceptive and unwelcoming for a behavioural system, the activation and ex-
pression of a behavioural system can be inhibited.
Ethologists had used the concept of behavioural systems solely to refer to behaviour.246
Bowlby expanded the concept, arguing that a behavioural system also recruits cognitive

aspects, most especially expectations based on repeated past experiences and interactions,
and perceptions of the present situations.247 In Attachment, Volume 1 he used the term ‘in-
ternal working models’ to characterise these cognitive components of a behavioural system.
However, too much has often been made of internal working models as an element in
Bowlby’s thinking, especially among commentators in the late 1980s and early 1990s who
took the internal working model as the ultimate ‘content’ of qualitative differences in attach-
ment, in part based on an interpretation of Main’s work (Chapter 3).248 In fact, Bowlby used
the concept relatively sparingly from the 1970s onwards, though he remained keen to em-
phasise that behavioural systems have cognitive components.249
Furthermore, for Bowlby behavioural systems contain not just behavioural and cognitive
components, but also emotional aspects. This aspect of behavioural systems is acknowledged
but not detailed in Attachment, Volume 1, which sought only to secure the argument that
affects alter our perception of the point that a behaviour system requires activation or ter-
mination.250 More than in his books, Bowlby’s key remarks on the emotional components of
behavioural system are generally contained in his publications in psychoanalytic journals—
which have generally been little read since the publication of Bowlby’s books—as well as in
unpublished work. He observed: ‘I conceive overt behaviour to be only one component of
a motivational system within the organism, and fantasies, thoughts and affects, conscious
and unconscious, to be integral to, and other components of, such systems.’251 Furthermore,
some affects, he proposed, accompany the activation of a behavioural system, some to its ter-
mination, and others its lack of assuagement. For instance, the sexual system may be accom-
panied by arousal, its termination by satisfaction and comfort, and its lack of assuagement to
yearning and dissatisfaction. The attachment system is accompanied by distress, its termin-
ation by comfort, and its lack of assuagement to yearning and dissatisfaction, and eventually
despair.252
246 E.g. Baerends, G.P. (1976) The functional organization of behaviour. Animal Behaviour, 24, 726–38.
247 Bowlby, J. (1958) The nature of the child’s tie to his mother. International Journal of Psycho-Analysis, 39, 350–
73, pp.365–6; Bowlby, J. (1980) Loss. London: Pimlico, p.348.
248 E.g. Zeanah, C.H. & Anders, T.F. (1987) Subjectivity in parent–infant relationships: a discussion of internal
working models. Infant of Mental Health Journal, 8(3), 237–50.

249 E.g. Bowlby, J. (1982) Outline of attachment theory. PP/Bow/H.260.
250 Bowlby, J. (1969, 1982) Attachment, Volume 1. London: Penguin: ‘The process of categorising parts of the
environment in terms of fitness to elicit a particular class of behaviour is itself experienced as coloured by the ap-
propriate emotion’ (114).
251 Bowlby, J. (1962) Defences that Follow Loss: Causation and Function. PP/Bow/D.3/78.
Behavioural systems 45
Groundplans of desire
Another important consequence of the concept of behavioural systems was that it allowed
Bowlby to circumvent simplistic notions of nature and nurture. Harkness observed that first
among the concerns raised by critics of attachment research since the 1970s has been the
question ‘How can attachment be both biologically based and determined by context?’.253
The theory of attachment as a behavioural system was Bowlby’s answer to this question,
though it is not an answer that has been widely or well understood. For Bowlby, humans
may be primed to develop behavioural systems along certain lines, but the very assemblage
of a system depends upon experience. As Bowlby put it in his notes in the 1960s, discussing
the attachment behavioural system: ‘What I mean by this is that in an infant’s behavioural

equipment there is a built-in bias, genetically determined, that in the ordinarily expectable
environment leads to the development of object relations. The built-in bias is there from the
first: the actual development of object relations takes time.’254
Attachment behaviour is neither inevitable nor pre-wired, which is why reference by later
researchers to the attachment system as a whole using terms such as ‘innate’255 can be mis-
leading. Behavioural systems are a very specific subset of responses that are neither simply
innate nor learnt, and that are not primarily expressions of emotion without environmental
responsiveness. On the one hand, behavioural systems are neurologically primed at the level
of motivation; they represent behaviour towards particular goals that would increase indi-
vidual survival or reproductive success, at least in the environment within which humans
evolved. However, on the other hand, behavioural systems are assembled out of component
behaviours and experiences in a way that is flexibly responsive to the environment, with
behaviours used interchangeably to achieve the set-goal. Shaver stated that ‘according to
Bowlby and Ainsworth, an infant can be viewed as an assembly of behavioural systems’.256
This is essentially incorrect, or at least suggests a broader use of the concept of ‘behav-
ioural system’ by Shaver than that of Bowlby and Ainsworth (Chapter 5). For Bowlby and
Ainsworth an infant is much more than the sum of behavioural systems: they observed that
infants have some almost entirely innate and pre-given responses, and some entirely learnt
responses.257 Additionally, not all behaviours have a set-goal. Some recognisable and pat-
terned responses appeared to Bowlby and Ainsworth to be expressions of overwhelming
affect, such as rage and despair, but without being environmentally responsive in a way that
recruits component behaviours to achieve a specific end.258
For Bowlby, behavioural systems have specific properties, including that they are neither
simply innate nor simply learnt, and that they have a set-goal for their termination. In one
253 Harkness, S. (2015) The strange situation of attachment research: a review of three books. Reviews in
Anthropology, 44(3), 178–97, p.179.

254 Bowlby, J. (1965) Comments on Joffe and Sandler 1965 ‘Notes on Pain, Depression and Individuation’. PP/
Bow/J.9/168–9.
255 E.g. Mikulincer, M. & Shaver, P. (2008) Contributions of attachment theory and research to motivation sci-
ence. In J.Y. Shah & W.L. Gardner (eds) Handbook of Motivational Science (pp.201–16). New York: Guilford, p.204.
256 Koski, L.R. & Shaver, P.R. (1997) Attachment and relationship satisfaction across the lifespan. In R.J.
Sternberg & M. Hojjat (eds) Satisfaction in Close Relationships (pp.26–55). New York: Guilford, p.27.
257 The correspondence between Bowlby and Rene Spitz on smiling addressed just this issue (see PP/Bow/
H.169). As an example of a largely pre-given response requiring only sufficient maturation and the absence of a
grossly inhibiting environment, Bowlby offered the example of infant sucking. As an example of an essentially
learnt response, Bowlby gave the example of throwing a ball.
258 Bowlby, J. (1973) Separation. New York: Basic Books, p.288.
46 John Bowlby
of Bowlby’s key metaphors: ‘As in the case of a military operation, the master plan gives only
main objective and general strategy; each commander down the hierarchy is then expected
to make more detailed plans and to issue more detailed instructions for the execution of
his part in the master plan.’259 This phrasing might imply that the activating and termin-
ating points of the attachment system are constant, that only the behaviour is shaped by ex-
perience. This was a common misunderstanding of Bowlby’s position among his critics, as
Ainsworth identified.260 In fact, Bowlby’s claim was more radical: that the very parameters
of the motivational system, though predisposed by evolution, are nonetheless also shaped by
our encounters with others, such as experiences of expressing behaviour and how this be-
haviour is received.261 Not just how, but what and when we desire or do not desire, and what
that desire means to us, are inscribed by experience, even if the groundplan of desire may be
available from human evolutionary history.

In Attachment, Volume 1, Bowlby proposed that, in humans, ‘most behavioural systems are
to some extent environmentally labile—namely, the form they take in an adult turns in some
degree on the kind of environment in which that adult is reared. The advantage of this is that
the form ultimately to be taken by the system is in some degree left open so that it can, during
development, become adapted to the particular environment in which the individual finds him-
self.’262 The complex cognitive capabilities of humans can also permit this flexibility to extend to
individual or collective forms of substitution, so that if a desired set-goal is unattainable, some
interim, partial, or symbolic form can nonetheless reduce activation of the behavioural system
and offer some portion of the emotional rewards of achieving the set-goal.263
Nonetheless, experiences cumulatively channel this flexibility. ‘Once a human has had ex-
perience of reaching a consummatory situation the behaviour that leads to it is likely to be-
come reorganised in terms of a set-goal and a plan hierarchy. That is what appears to occur in
sexual behaviour’, where early experiences of relationships in general, and in particular of in-
timacy lit by intensity, serve to channel the later shape and targets of desire.264 Furthermore,
this ontogenetic flexibility is not limitless. A basic quality of behavioural systems is that they
require certain kinds of environments for the homeostatic system to operate successfully.
A ‘feature of control systems that it is important to note is that they can operate effectively
only within certain environmental limits. For example, the system responsible for maintain-
ing body temperature is effective only provided the ambient temperature lies within certain
limits. Outside those limits it fails.’265
259 Bowlby, J. (1969, 1982) Attachment, Volume 1, London: Penguin, p.78.

260 Ainsworth, M., Blehar, M., Waters, E., & Wall, S. (1978, 2015) Patterns of Attachment: A Psychological Study
of the Strange Situation. Bristol: Psychology Press, p.11.
261 In a remarkable statement, Bowlby observed that this gives attachment behaviour a certain developmental
and causal priority over the attachment relationship itself: ‘Attachment behaviour leads to the development of af-
fectional bonds or attachments, initially between child and parent.’ Bowlby, J. (1980) Loss. London: Pimlico, p.39.
262 Bowlby, J. (1969, 1982) Attachment, Volume 1. London: Penguin, p.129. Bowlby later reflected that this flexi-
bility is especially potent for humans: ‘the longer-lived an individual the more necessary is ontogenetic flexibility to
enable it to adapt to changes in the environment’. Bowlby, J. (1982) Evolution theory. PP/Bow/H102.
Psychology, 30(4), 230–40, p.239.

Animal Models in Human Psychobiology (pp.27–36). New York: Plenum Press, p.31. This idea has its origins for
Bowlby not only in control systems theory, but also in his reading of Freud. Annotations by Bowlby dated 1960 on
the ‘Project for a Scientific Psychology’ in Sigmund Freud: The Origins of Psychoanalysis. Copy held in the library
of Human Development Scotland. Bowlby highlighted and starred the passage: ‘Every contrivance of a biological
Five behavioural systems 47
In his popular and early writings, Bowlby tended to imply that the expected environment
for an infant was their primary caregiver, generally the mother. However, over time he came
to acknowledge that a context with multiple familiar caregivers simply could not lie outside
of the bounds of the attachment system’s responsiveness. His last published work explicitly
stated his mature view that the attachment system ‘contributes to the individual’s survival
by keeping him or her in touch with one or more caregivers’,266 and he told colleagues that
‘a baby interacting with and forming trusting (secure) relationships (attachments) with a
larger number of significant persons will as a child and later as an adult walk more securely
in the world’.267 For example, in the environment within which humans evolved it would
have been quite possible that grandparents would have been on hand when a baby required
care.268 By contrast, he supposed that institutional care, as in orphanages, with a rapid turn-
over of paid professional caregivers, was likely outside the limits of what the attachment

system had evolved to handle. The attachment system is not pre-given, and requires envir-
onmental learning and input, wrapping itself around and elaborating itself on the basis of
discriminated relationships. Growing up in an institutional context with continual turnover
of professional caregivers, it would be much harder for a child to experience any caregiver as
familiar enough to entrain the attachment system and its elaboration.
Five behavioural systems
In his scholarly writings, Bowlby gave particular attention to five behavioural systems: the
attachment system; the caregiving system; the exploratory system; the fear system; and ag-
gression. This is not an exclusive list—for instance, Bowlby also described a sexual system;
he suggests sleep may have qualities of a behavioural system; and he discussed an affiliative
system that organises friendly behaviour towards others. However, attachment, caregiving,
exploration, fear, and aggression are the most well-characterised instances in his writings,
and will be discussed in turn (Table 1.1).
The attachment behavioural system
In Bowlby’s account of behavioural systems, a central place is given to the system’s activating
and terminating conditions. Activating conditions trigger a motivation and behaviour to
achieve the set-goal; achievement of the set-goal deactivates the motivation and its behav-
iour. Bowlby theorised that the attachment system in infancy is usually comparatively dor-
mant, tasked primarily with monitoring the caregiver’s whereabouts and checking in from
time to time to ensure that a line of retreat to the caregiver remains open. In this dormant
state, the system is also engaged with gaining relevant information about the caregiver and
nature has limits to its efficiency, beyond which it fails. Such failures exhibit themselves as phenomena bordering
on the pathological’ (368).
266 Bowlby, J. (1991) Ethological light on psychoanalytic problems. In P. Bateson (ed.) The Development and
Integration of Behaviour: Essays in Honour of Robert Hinde (pp.301–14). Cambridge: Cambridge University
Press, p.306.
267 Bowlby 1984 personal communication, cited in Harwood, I. (2003) Creative use of gender while addressing
early attachment, trauma, and cross-cultural issues in a cotherapy group. Psychoanalytic Inquiry, 23, 697–712.
48 John Bowlby
Table 1.1 Behavioural systems and their parameters in infancy in Bowlby’s writing
Behavioural Activating Terminating Key component Species-level

system condition(s) condition behaviour function
Attachment Alarming stimuli The perceived Sucking, clinging, Protection of

or the potential for availability of a crawling, smiling, self from harm
separation from familiar caregiver distress cries (especially
familiar caregivers predation)
Caregiving Perceptions of The perceived Retrieval, care, Protection
infant need, cessation of infant encouragement, and support of
especially for signals of need protection others
retrieval
Exploration Novel and/or Perceived Exploratory Learning
complex stimuli familiarity manipulation

Fear Alarming stimuli Perceived safety Startle, wariness, Escape
or the potential for fleeing, hiding,
separation from freezing
familiar caregivers
Aggression Frustration Satisfaction of Threatening, attack Coercion
(anger of hope) of another the frustrated
behavioural system behavioural system
the environment, especially information related to the caregiver’s responses to the child’s
behaviour and to potential cues for danger. When a child is alarmed, the attachment system
will prompt attempts to gain proximity with the caregiver.
From the primate researchers Harry Harlow and Robert Zimmermann, Bowlby took
the phrase ‘haven of safety’ to refer to the way that an infant’s alarm and motivation to seek
their caregiver was terminated once they achieved proximity with the caregiver.269 However,
Bowlby was keen to make clear that the extent of proximity required by the attachment
system was flexible, not a biological given. The set-goal of the attachment system may, indeed,
abruptly change to specify the degree of proximity more narrowly or more loosely, and this
will bring about attachment behaviours of different forms and intensities. Even the same en-
vironmental cue, such as a loud and sudden noise, may elicit only a look to the caregiver from
an infant when accessibility feels ready and sure, but may elicit swift approach and clinging
when the set-goal of the attachment system has been calibrated at full physical contact.
In Attachment, Volume 1 Bowlby highlighted three conditions as of particular importance
to the calibration of the set-goal. A first was past experience: for children who have come to
expect that their caregiver might not be accessible when the attachment system is activated,
a lower threshold would be set for its activation and a higher threshold for termination. By
contrast, for children confident in the accessibility of their caregiver, physical proximity-
seeking may only occur at a high threshold, and can be more readily terminated.270 A second
factor that would influence these thresholds, Bowlby suggested, was the extent of current
269 Bowlby, J. (1958) Nature of the child’s tie to his mother. International Journal of Psycho-Analysis, 39, 350–
73; Harlow, H.F. & Zimmermann, R.R. (1958) The development of affectional responses in infant monkeys.
Proceedings of the American Philosophical Society, 102(5), 501–9.
270 Bowlby, J. (1973) Separation: Anxiety and Anger. New York: Basic Books, p.228–9.
sources of threat. In environments with alarming stimuli, or cues that an unexpected or

major separation might occur, the parameters of the attachment system might well be al-
tered.271 Finally, a third factor that Bowlby thought would especially impact the attachment
system’s parameters was human cognitive development. With maturation beyond infancy,
we have at our disposal more sophisticated strategies for maintaining contact, including
verbal communication and mental representation.
As Ainsworth’s student Bob Marvin documented in his 1972 doctoral thesis, matur-
ation raises the threshold for the activation of the attachment system, and also the inten-
sity of the attachment behaviours that are likely to be expressed.272 Infants may cry and
seek to cling to their attachment figure following an unexpected separation. By contrast,
preschoolers, with greater tools at their disposal for understanding and achieving the
caregiver’s accessibility, may simply offer a look and smile of greeting. Ultimately the

set-goal of the attachment system is not a particular inner state or set of external con-
ditions, Bowlby argued, but a point at their intersection: ‘a certain sort of relationship
with another specified individual’.273 Until his final writings, Bowlby generally claimed
that this relationship was physical accessibility. Later, taking heed of Ainsworth’s ob-
servations and arguments, he acknowledged that the ‘certain sort of relationship’ is, in
infancy and across childhood, both the physical and psychological ‘availability’ of the
attachment figure.
Though this was a point generally missed by his critics, and could have been better
highlighted, Bowlby regarded few, if any, of the component behavioural sequences of the
attachment system as exclusive to it.274 Behaviours can and often do serve more than one
system. The attachment system takes as key early components behavioural sequences such
as sucking, clinging, crawling, and smiling.275 However, sucking is also important for what
Bowlby called the ‘eating’ or ‘feeding’ behavioural system276 and what Ainsworth termed the
‘food-seeking system’ (Chapter 2). The shared components of the following and the food-
seeking systems were critical for the conflation of the two in psychoanalytic theory, under
the notion of the ‘oral stage’.277 Smiling is not only used for activating caregiver responses, but
is more generally critical for the affiliative system, and can be co-opted by the fear system as
a form of placation. Developmental maturation also allows behaviours usually characteristic
of other systems to be recruited by the attachment system if its usual repertoire for seeking
care has not been successful. As discussed below in the section ‘The caregiving behavioural
271 Bowlby, J. (1979) By ethology out of psychoanalysis: an experiment in interbreeding. (The Niko Tinbergen
Lecture.) Animal Behaviour, 28(3), 649–56, p.651.

272 Marvin, R.S. (1972) Attachment, exploratory and communicative behavior of two, three and four
year old children. Unpublished doctoral dissertation, University of Chicago. Discussed in Bowlby, J. (1973)
Separation: Anxiety and Anger. New York: Basic Books, p.70.
274 The point would need to be clarified repeatedly by early members of the second generation of attachment
researchers. The most important contribution on this score was that of Everett Waters, who also demonstrated the
claim empirically (Chapter 2).
275 Bowlby, J. (1958) The nature of the child’s tie to his mother. International Journal of Psycho-Analysis, 39,
350–73, p.351.
276 Bowlby, J. (1980, 1988) Caring for children. In A Secure Base (pp.1–21). London: Routledge, p.6; Bowlby, J.
(1973) Letter to Scott Henderson, 30 July 1973. PP/Bow/J.9/98.

277 Bowlby, J. (1960) Grief and mourning in infancy and early childhood. Psychoanalytic Study of the Child, 15,
9–52: ‘I believe that the hypothesis now advanced would have been advocated earlier had it not happened that the
phase of attachment to a mother figure was so late in being recognized and had not theory become preoccupied
instead on the one hand with primary narcissism and on the other with orality’ (14).
50 John Bowlby
system’, Bowlby’s particular interest was in the recruitment of the caregiving system in the
service of the attachment system. However, he also noted how sexuality could be used in the
service of the attachment system when other forms of care-seeking have failed, and ‘one can
regard attempted suicide as an aberrant form of care-eliciting behaviour resorted to by only
those who have had very unstable relationships in the past, and who have learned that more
normal types of care-eliciting behaviour fail to work. As an aberrant form of attachment
behaviour, of course, it is not far removed from total despair.’278 Behaviours can also serve
more than one system at the same time: for instance, Bretherton and Ainsworth argued that
coy and submissive behaviour is coordinated by the fear and affiliative systems together.279
Rough-and-tumble play integrates, stylises, and perhaps ritualises elements from both the
exploratory and the anger systems. Close observation of children during such play reveals
how aggression and exploration, specifically, can be used in the service of the other.280 For

Bowlby:
Affect laden behaviour I tend to view in terms of structures built of component bricks.
The bricks are relatively stereotyped behaviour patterns, e.g. bird song or sucking, which,
according to the species, may be built in or learnt or a combination of both. The larger
structure, e.g. courtship or nest building, is less stereotyped and a complex synthesis of
these components. Although in principle any component is available for any synthesis, in
practice each synthesis tends to select a particular group of components. None the less it
is probably usual for certain component items to be utilised in more than one synthesis.281
For instance, Bowlby discussed the close relationship between the attachment response and
the flight response. These may have many common components, since we often flee to those
we turn to with the expectation of protection, though there may be occasions when we do
not do so, for instance when we do not perceive such individuals as available to us. Bowlby
also expected the attachment and fear behavioural systems to share important cognitive
components, such as expectations about the effectiveness of previous attempts to gain safety,
and present perceptions of sources of danger.
Another pair of behavioural systems with significant shared components are attachment
and sexuality. Ainsworth felt that Bowlby had seriously underplayed the importance and
complexity of the sexual system, as part of his attempt to pull away from psychoanalytic
theory.282 This is certainly true of the 1970s and 1980s. However, during the late 1950s, whilst
278 Bowlby, J. (1979) Letter to Professor K.S. Adam, 12 February 1979. PP/Bow/J.9.2.
279 Bretherton, I. & Ainsworth, M.D.S. (1974) Responses of one-year-olds to a stranger in a strange situation.
In M. Lewis & L.A. Rosenblum (eds) The Origin of Fear (pp.131–64). New York: Wiley. This was also discussed
by Tony Ambrose, working within Bowlby’s group: Ambrose, T. (1960) The smiling and related responses in early
human infancy: an experimental and theoretical study of their course and significance. Unpublished PhD thesis,
Birkbeck College, London.
280 See Attili, G. & Hinde, R.A. (1986) Categories of aggression and their motivational heterogeneity. Evolution
and Human Behavior, 7(1), 17–27.

281 Bowlby cited in Tanner, J.M. & Inhelder, B. (1960) Discussions on Child Development: Proceedings of the
WTO Study Group of the Psychobiological Development of the Child, Vol. 4. London: Tavistock, p.40.
282 Ainsworth, M. (1997) Peter L. Rudnytsky—the personal origins of attachment theory: an interview with
Mary Salter Ainsworth. Psychoanalytic Study of the Child, 52, 386–405: ‘PLR: Is there any sense in which your view
and Bowlby’s diverge at all, or is there a complete meeting of the minds? MSA: Well, I think there’s more to the
oedipal situation than he does. Bowlby just doesn’t talk about it’ (399); ‘In the oedipal situation, there’s no question
in my mind that in the parents’ dynamics there is a lot of sexuality’ (402).
Ainsworth was relatively out of contact in Uganda (Chapter 2), comparison of the sexual
response and the following response was key to discussions between Bowlby and Hinde.
Bowlby discussed the sexual system in some detail during the meetings of the WTO Study
Group of the Psychobiological Development of the Child in the late 1950s.283 In these dis-
cussions Bowlby did not consider the activating and terminating conditions of the sexual
system. Instead, his focus was on how the sexual response takes component behaviours
from attachment relationships: ‘certain components in the human sexual response are de-
rived from parent–child response’.284 Such common components might include affective
and cognitive elements such as trust, affection, gentleness, and expectations around what
interactions such as acceptance or rebuff will ultimately mean for a relationship. Common
aspects might include behavioural components such as gazing, kissing, and coming into
close proximity. This account allowed Bowlby to qualify Freud’s claim that early relationship

patterns are organised as a whole into later sexual desire. For Bowlby, it seemed more cau-
tious and more probable that only some components of earlier patterns are integrated into
how adults show affection, enact courtship and sexuality, and provide care to their own off-
spring.285 These components are the thin wires connecting early care to later sexuality; there
is no inevitable or single line of continuity.
Even if components of the attachment behavioural system ultimately get recruited and
repurposed in adulthood, and even if the threshold for activating the system is increased,
Bowlby nonetheless claimed that the attachment system remains available throughout
life. In an emergency, or when hurt or scared, we are prompted to seek those we know and
trust, and feel the more troubled when they are not available.286 As adults, when alarmed we
may seek a spouse, a parent—or, ‘more often than might be supposed’, adults may look to
their own child for reassurance.287 However, the surplus of meanings assigned to the term
283 Ainsworth does not cite the transcript of these discussions, and she may not have owned a copy.
284 Tanner, J.M. & Inhelder, B. (1956) Discussions on Child Development: Proceedings of the WTO Study Group
of the Psychobiological Development of the Child, Vol 1. London: Tavistock, p.184–5. For later considerations of the
relationship between the attachment and sexual behavioural systems see e.g. Crittenden, P.M. (1998) Patterns of
attachment and sexual behavior: risk of dysfunction versus opportunity for creative integration. In L. Atkinson
& K.J. Zucker (eds) Attachment and Psychopathology (pp.47–93). New York: Guilford; Diamond, L.M. (2003)
What does sexual orientation orient? A biobehavioral model distinguishing romantic love and sexual desire.
Psychological Review, 110(1), 173–92.
WTO Study Group of the Psychobiological Development of the Child, Vol 4. London: Tavistock, p.41.
286 Bowlby, J. (1969) Affectional bonds: their nature and origin. In H. Freeman (ed.) Progress in Mental Health
(pp.319–27). London: J. & A. Churchill, p.323.

Affectional Bonds (pp.150–88). London: Routledge, p.157. The use of a child as an attachment figure by adults
appears to have been deliberately neglected by subsequent attachment research, in part to maintain—against
misunderstanding—the characterization of attachment as something a child shows to their adult caregiver. See
e.g. Ainsworth, M.D.S. (1991) Attachments and other affectional bonds across the life cycle. In C.M. Parkes, J.
Stevenson-Hinde, & P. Marris (eds) Attachment Across the Life Cycle (pp. 33–51). London: Routledge: ‘We talk
of the bond of a mother to her child . . . a mother does not normally base her security on her relationship with
her child, however eager she may be to give care and nurturance’ (40). The primary exception is Doherty, N.A. &
Feeney, J.A. (2004) The composition of attachment networks throughout the adult years. Personal Relationships,
11(4), 469–88. Forty percent of participants with children reported using their child as a safe haven, as a secure
base, and experiencing separation anxiety. However, both the attachment and caregiving systems can prompt
separation anxiety and proximity-maintenance. This difficulty disentangling these systems may be a secondary
reason why attachment researchers have neglected the phenomenon of use of children as attachment figures by
adults.
52 John Bowlby
‘attachment’ caused Bowlby to swing between two very different positions.288 On the one
hand, ‘attachment’ was sometimes used broadly to mean early relationships as a whole. With
this meaning in mind, Bowlby made strong claims for the influence of attachment on ‘a per-
son’s whole emotional life’, and as the condition of possibility for all later emotional devel-
opment and mental health.289 In Loss, for instance, he claimed that ‘attachments to other
human beings are the hub around which a person’s life revolves . . . From these intimate at-
tachments a person draws his strength and enjoyment of life.’290 Such claims later contrib-
uted to the impression by audiences of attachment theory that individual differences in early
attachment are fixed for life (Chapter 2).
On the other hand, when ‘attachment’ was understood narrowly as a specific behavioural
system, Bowlby’s stance was rather different. With the narrow ethological meaning in mind,
he proposed that influences on this behavioural system in early life could have a pervasive

influence on later mental health, but nonetheless tended towards more qualified claims.
Human development is very complex, and no single factor is likely to play more than a mod-
erate role. However, for a single behavioural system, Bowlby anticipated that the attachment
system would be quite influential and therefore worthy of particular attention. The period
of early following and help-seeking has its importance because ‘the period when they are
most active is also the period when patterns of control and of regulating conflict are being
laid down’. It therefore has a particular significance for socio-emotional development.291 Yet
even children who have had ‘ghastly experiences’ may ‘nevertheless develop favourably’, with
maybe only 20–30% subsequently showing severe problems. Furthermore, those who show
severe problems are often those who have experienced combinations of adversities, rather
than disruptions of attachment alone.292 Nonetheless, Bowlby anticipated that individual
differences in early attachment experiences would have some implications for later mental
health and wellbeing across the spectrum (see also Chapter 4).
In his final years, Bowlby saw the life of Charles Darwin as offering an especially clear case
illustrating the importance of the disruption of the attachment system in childhood for later
development, and also the emphasis he placed on family interactions for mediating the ef-
fects of separation or loss. Especially after seeing visitors or after some other excitement, as an
adult Darwin experienced numerous symptoms including gastric pains, vertigo, vomiting,
288 This hinge seems to have been primarily the result of conceptual imprecision. It may at times have served as a
‘motte and bailey’ rhetorical strategy, with a poorly defendable but expansive outer area and a more defendable but
narrower inner position. However, the fact that the term’s meaning slides around just as much in Bowlby’s private
notes suggests that such a strategy was not his intent. On the ‘motte and bailey’ rhetorical strategy see Shackel, N.
(2005) The vacuity of postmodernist methodology. Metaphilosophy, 36(3), 295–320.
289 E.g. Bowlby, J. (1984) Violence in the family as a disorder of the attachment and caregiving systems. American
Journal of Psychoanalysis, 44, 9–27, p.11.

290 Bowlby, J. (1980) Loss. London: Pimlico, p.442.
291 Bowlby, J. (1960) Separation anxiety. International Journal of Psycho-Analysis, 41, 89–113, p.105; see also
Steele, H. & Steele, M. (1998) Response to Cassidy, Lyons-Ruth and Bretherton: a return to exploration. Social
292 Bowlby, J. (1986) Interview with the BBC. PP/Bow/F.5/8: ‘I’m often accused of exaggerating and various
people point out that not every child who has had these sorts of experience comes to grief . . . First of all, maybe he’ll
be vulnerable if things go wrong in his life . . . Another point is that, supposing it’s true that some children having
had some pretty ghastly experiences nonetheless develop favourably . . . But then the question arises, what per-
centage of people who contract polio are left with long term paralysis? The answer is less than 1 per cent. The fact
is 99 per cent get by. Now in the case of severe maternal deprivation—first of all I think that much more than 1 per
cent suffer. I think the percentage is way up in the 20s or 30s. And the other thing is that crippling of personality is
much more serious’. See also Bowlby, J. (1971) Letter to Michael Rutter, 6 October 1971. PP/Bow/J.9/161: ‘The more
serious effects of experiences of the type we are discussing are when there are certain special combinations of vari-
ables present.’
tremors, cramps, headaches, eczema, anxiety, and a compulsion to wander or search for
something without knowing what. There remains much speculation about the cause of these
symptoms, and the relative contribution of organic and psychological illness.293 In Charles
Darwin: A Life, published in 1990, Bowlby argued that several other symptoms could essen-
tially be explained as panic attacks, caused by the poor integration of behavioural, affective,
and cognitive components of the attachment system following the loss of Darwin’s mother.
The issue was not this disruption of the attachment relationship alone, however. Bowlby in-
stead highlighted that the family insisted that Darwin was not permitted to talk about his
mother after she died when he was eight years old. Bowlby speculated that this could have
contributed to exciting emotions being misrecognised as threatening ones.294 In turn, he
proposed that this misrecognition may have contributed to Darwin’s tendency to hyperven-
tilate, with his body always on the cusp of physiological overarousal.295 In support of his

claims, Bowlby offers evidence that Darwin’s symptoms were exacerbated whenever as an
adult he felt threatened or experienced a loss.
The caregiving behavioural system
It is sometimes assumed that Bowlby also regarded parental behaviour and feelings as an ex-
pression of attachment. This error is more common among Bowlby’s critics, who generally
privilege his writings of the 1950s and ignore his later qualifications and clarifications.296
However, such usage can sometime be seen in the work of attachment researchers too, albeit
less since the 2000s.297 It follows some confusing and overextended use of the term ‘attach-
ment’ by Bowlby in the early 1950s that did include caregiving behaviour. Though he at-
tempted subsequently to distinguish attachment and caregiving, the early work set up a chain
reaction that shaped both interpretation of his theory and measurement tools built from
it. For instance, the Maternal–Fetal Attachment Scale was introduced by Cranley in 1981
to measure a mother’s ‘attachment’ (feelings of care for) her unborn baby.298 Such institu-
tionalised uses of ‘attachment’ to mean caregiving have caused even attachment researchers
assiduous about terminology, such as van IJzendoorn and Bakermans-Kranenburg, to
293 Finsterer, J. & Hayman, J. (2014) Mitochondrial disorder caused Charles Darwin’s cyclic vomiting syndrome.
International Journal of General Medicine, 7, 59–70.

294 Bowlby, J. (1987– 90) Papers on Charles Darwin. Cambridge University Library. MS Add. 8884, margi-
nalia on Liotti, G. (1986) Structural cognitive therapy. In Dryden, W. & Golden, W. (eds) Cognitive-Behavioural
Approaches to Psychotherapy. Milton Keynes: Open University Press, p.125.
295 Bowlby, J. (1990) Charles Darwin: A Life. New York: Norton, p.6.
296 Two clear cases are Vicedo, M. (2013) The Nature and Nurture of Love: From Imprinting to Attachment
in Cold War America. Chicago: University of Chicago Press, and Gottlieb, A. (2014) Is it time to detach from
attachment theory? Perspectives from the West African rainforest. In H. Otto & H. Keller (eds) Different
Faces of Attachment: Cultural Variations on a Universal Human Need (pp.187–214). Cambridge: Cambridge
University Press.
297 Bretherton, I., Biringen, Z., Ridgeway, D., Maslin, C., & Sherman, M. (1989) Attachment: the parental per-
spective. Infant Mental Health Journal, 10(3), 203–21; Waters, E. & Cummings, E.M. (2000) A secure base from
which to explore close relationships. Child Development, 71(1), 164–72: ‘To examine relations between early at-
tachment experience and both secure base use and secure base support skills later in life and to examine them
across contexts such as marriage, parenting, caring for adult parents, and requesting care from others. As currently
formulated, attachment theory suggests that these are all organized by the same attachment control system’ (171).
298 Cranley, M.S. (1981) Development of a tool for the measurement of maternal attachment during pregnancy.
Nursing Research, 30(5), 281–4.

54 John Bowlby
refer to ‘attachment’ when they mean caregiving.299 Perhaps a source of confusion between
caregiving and attachment for Bowlby in his early writings was that there did seem to be
common elements. As mentioned earlier, in his mature writings Bowlby identified that some
components of the attachment system, like the capacity to communicate about emotion with
others, may be recruited when an adult is required to provide care. However, this does not
imply that caregiving is a reflection or expression of attachment. The two systems evolved in
parallel, and function reciprocally, producing important forms of potential cooperation and
friction.300
In Attachment, Volume 1, Bowlby developed an account of the caregiving behavioural
system modelled on his existing ideas regarding the attachment behavioural system. The
caregiving system was specifically anchored in the retrieval response, as a reciprocal partner
to the infant’s following response for maintaining proximity:

Retrieving can be defined as any behaviour of a parent a predictable outcome of which
is that the young are brought either into the nest or close to mother, or both . . . The re-
trieving behaviour of a primate mother gathers the infant into her arms and holds him
there. Having a similar outcome to the attachment behaviour of young, it is probably best
understood in similar terms—namely, as being mediated by a number of behavioural sys-
tems the predictable outcome of which is maintenance of proximity to the infant . . . espe-
cially when her infant is playing contentedly with other known individuals in the vicinity,
a mother may let things be. Yet her tendency to retrieve him is not wholly dormant: she is
likely to keep a watchful eye on him.301
Yet just as the attachment behavioural system was anchored in the infant’s following re-
sponse, but inserted into a broader image of comfort-seeking and relationship, the care-
giving system had a parallel hinge. Like the attachment system, then, Bowlby’s notion of
the caregiving system sustained an ambiguity between the narrow notion of the retrieval
response (and, presumably, related aspects of holding) and a more expansive notion of the
provision of ‘encouragement, support, help and protection’.302 In notes from 1978, disliking
throughout his career the ambiguity of the word ‘empathy’, he referred to the key emotional
component of the caregiving system as ‘concern for the welfare of others’.303 And in a late
interview he emphasised the importance of attentional processes that serve as architecture
for the caregiving system.304 Yet Bowlby’s written reflections on the set-goal of the caregiving
299 Huffmeijer, R., van IJzendoorn, M.H., & Bakermans-Kranenburg, M.J. (2013) Ageing and oxytocin: a call for
extending human oxytocin research to ageing populations—a mini-review. Gerontology, 59(1), 32–9: ‘In (soon-to-
be) mothers, increases in plasma oxytocin concentrations over the course of pregnancy have been found to predict
greater attachment to the unborn baby’ (33).
300 Bowlby, J. (1986) Attachment, Life-Span and Old Age. Eds J. Munnichs & B. Miesen. Utrecht: Van Loghum,
p.11. See also Cassidy, J. (2000) The complexity of the caregiving system: a perspective from attachment theory.
Psychological Inquiry, 11(2), 86–91, p.88.
302 Bowlby, J. (1973) Letter to Scott Henderson, 30 July 1973. PP/Bow/J.9/98: the caregiving provided by the care-
giving system ‘includes rather more’ than the bodily contact of the retrieval response and the affects of ‘interest,
esteem and affection’. In particular, Bowlby urged Henderson to note that caregiving also included ‘encourage-
ment, support, help and protection’. On the ambiguity between the ‘retrieval’ and ‘nurturance’ models of caregiving
in Bowlby see Bell, D.C. & Richard, A.J. (2000) Caregiving: the forgotten element in attachment. Psychological
Inquiry, 11(2), 69–83.
303 Bowlby, J. (1978) Caregiving. In ‘Emotion and feeling’. PP/Bow/H.5.
304 Bowlby, J. (1990) Interviewed in The Nuts and Bolts of Ben Bowlby (Channel 4). https://licensing.screen-
ocean.com/r/216106.
system are scarce. And the conflation of specific and general set-goals of the caregiving
system is a major limitation of his remarks, given their significant differences. Not least, re-
trieval, following, and attachment are all homeostatic systems at root, whereas the provision
of care and nurturance in a wider sense aims to support growth, not a return to equilib-
rium. Retrieval and encouragement/support also seemed to be modulated and their suc-
cessful achievement met by potentially quite different affects in the caregiver, including quite
different senses of commitment and pleasure.305 Indeed, subsequent to Bowlby, ethologists
have come to distinguish a consoling system from a caregiving system, given their distinctive
behavioural repertoires and conditions of activation and termination.306
Despite conceptual ambiguity in Bowlby’s writings about caregiving, neither the narrow
nor the broad notions of the caregiving system implied, as Vicedo has mistakenly claimed,
that Bowlby discussed mothers as ‘unthinking and natural’, acting ‘just out of instinct’, in

behaviour devoid ‘of rationality, of choice and of moral value’.307 Vicedo’s description is a
reductive characterisation of even Bowlby’s populist writings, and as a characterisation of
his theoretical texts, it is essentially polemical. Bowlby did not regard parenting behaviours
as characteristic only of mothers—though Ainsworth was more critical than Bowlby of the
cultural values that assigned responsibility for caregiving to mothers.308 As mentioned in
the section ‘The attachment behavioural system’, Bowlby regarded all behavioural systems
as neurologically primed, but dependent on experience and contextual support for their
elicitation, elaboration, and expression. He stated explicitly in relation to caregiving behav-
iour that his ‘view of behavioural development contrasts sharply with both of the older para-
digms, one of which, invoking instinct, over-emphasised the preprogrammed component
and the other of which, reacting against instinct, overemphasised the learned component.
Parenting behaviour in humans is certainly not the product of some unvarying parenting in-
stinct.’309 In a presentation of his ideas for radio in 1969, Bowlby drew a comparison between
parenting behaviour and language. In both cases, humans are predisposed to pay attention
to certain cues, and elaborate systems of meaning and practice. However, the nature of these
meanings and practices depends on our opportunities for learning, and our induction into
305 On the important role of the affect ‘delight’ in the functioning of the caregiving system see e.g. Bernard, K.
& Dozier, M. (2011) This is my baby: foster parents’ feelings of commitment and displays of delight. Infant Mental
Health Journal, 32(2), 251–62. Delight and other positive communicated affect may also be a significant relay be-
tween the caregiving and attachment behavioural systems. Dozier, M. & Bernard, K. (2019) Coaching Parents of
Vulnerable Infants: The Attachment and Biobehavioral Catch-up Approach. New York: Guilford: ‘In a personal com-
munication (August 2001) Mary Main suggested that parental delight communicates to children how important
they are to their parents, a sentiment she indicated Mary Ainsworth shared’ (57).
306 E.g. Burkett, J.P., Andari, E., Johnson, Z.V., Curry, D.C., de Waal, F.B., & Young, L.J. (2016) Oxytocin-
dependent consolation behavior in rodents. Science, 351(6271), 375–8.

307 Vicedo, M. (2011) The social nature of the mother’s tie to her child: John Bowlby’s theory of attachment in
post-war America. British Journal for the History of Science, 44(3), 401–26, p.423. For an interesting illustration
of how the concept of ‘maternal instinct’ is now treated as a cultural form rather than part of human nature by
attachment researchers see Murphy, A., Steele, M., & Steele, H. (2013) From out of sight, out of mind to in sight
and in mind: enhancing reflective capacities in a group attachment-based intervention. In J.E. Bettmann & D.D.
Friedman (eds) Attachment-Based Clinical Work with Children and Adolescents (pp.237–57). New York: Springer.
For a plea for a more measured and less polemical discussion of attachment research, and its heterogeneity see
Duschinsky, R., van IJzendoorn, M., Foster, S., Reijman, S., & Lionetti, F. (2019) Attachment histories and fu-
tures: reply to Vicedo’s ‘Putting attachment in its place’. European Journal of Developmental Psychology, 17(1).
308 Ainsworth, M.D.S. (1991) Attachments and other affectional bonds across the life cycle. In: C.M. Parkes, J.
Stevenson-Hinde, & P. Marris (eds) Attachment Across the Life Cycle (pp. 33–51). London: Routledge: ‘Nowadays
in Western societies these traditional roles are being challenged, and many couples are experimenting with alter-
native ways of providing adequate care to their infants and young children. The more successful solutions seem to
involve the male taking more responsibility for direct caregiving to the children’ (42).
309 Bowlby, J. (1980, 1988) Caring for children. In A Secure Base (pp.1–21). London: Routledge, p.5.
56 John Bowlby
cultural systems. For instance, the development of retrieval response/caregiver behaviour is

shaped by what opportunities an adult has had to touch children in the past, and the child’s
own reciprocal signals.310
The role of both the motivational and behavioural aspects of the caregiving system, and
of the role of learning and culture, can be illustrated by two phenomena that especially con-
cerned Bowlby. A first was the use of threats as part of caregiving, a behaviour that had inter-
ested him from his earliest work as a clinician working with families.311 In Separation, threats
by caregivers are a central theme. Bowlby acknowledged that threats might well have the
ultimate end of retrieval or protection, the two suggested set-goals of the caregiving system
depending on whether it was narrowly or broadly construed. For instance, a threat of aban-
donment by a parent might be used effectively to keep a child nearby or ensure their obedi-
ence, where this is considered of special importance.

However, the same behaviour might express the anger behavioural system, or be a com-
pound fed by both the anger and caregiving systems. Bowlby emphasised that threatening
as a cultural practice is quite complex for a young child to understand: it is intensely felt
and expressed by the parent, but in fact localised and not fully serious. To the degree that
the serious-but-not-serious quality of threats can be difficult for children to understand or
trust, he worried about the long-term harm verbal threats could have on a child’s percep-
tion of caregiver availability in times of need.312 And he argued that verbal threats to hurt
or abandon the child by a parent can contribute to conflict between the attachment and fear
systems, since threats with a marked aggressive quality by a caregiver would elicit a desire for
both approach and avoidance by their child (Chapter 3).313 In a letter to Henry Hansburg in
1979 he observed that parents may ‘insist that no one loves a child who behaves so and so, and
that no one will ever love you unless etc. etc. I believe that when a child is exposed to these
pressures over many years, the notion of his being unlovable becomes deeply engrained.’314
A second phenomenon that especially interested Bowlby was what he termed ‘compulsive
caregiving’ in The Making and Breaking of Affectional Bonds, published in 1979.315 This was
the term Bowlby used when caregiving behaviour was shown by an individual as an alter-
native to seeking care. Children or adults may receive clear signals from their attachment
figures that asking for comfort is not permitted or would backfire.316 However, in providing
care themselves, they at least gain some closeness with and availability from their attachment
figures. The culturally mandated and practical closeness that caregiving permits allows in-
dividuals to achieve the set-goal of the attachment system, even if in a contorted and limited
way.317 The caregiving system is therefore used in the service of the attachment system, a
310 Bowlby, J. (1969) Ape and apex. BBC Radio, recorded 16 October 1969. PP/Bow/F.5/5: ‘Our capacity to learn
diversity of behaviour is itself genetically determined e.g. speech and language . . . Development of maternal be-
haviour dependent on appropriate experience, cf. Hinde on false dichotomy. Development of mothering greatly
helped by 1. Opportunity to touch, examine infant; 2. Smiling of infant.’
311 Bowlby, J. (1933–36) Anxiety, guilt, etc: old papers. PP/Bow/D.1/2/13
312 Bowlby, J. (1973) Separation. London: Penguin, pp.231–2.
313 Ibid. p.117. On threatening caregiver behaviour as a predictor of approach/avoidance conflict in infants see
Jacobvitz, D., Leon, K., & Hazen, N. (2006) Does expectant mothers’ unresolved trauma predict frightened/fright-
ening maternal behavior? Risk and protective factors. Development & Psychopathology, 18(2), 363–79.
314 Bowlby, J. (1979) Letter to Henry Hansburg, 21 February 1979. PP/Bow/J.9/90.
Affectional Bonds (pp.150–88). London: Routledge.

316 Bowlby, J. (1986) Letter to John Birtchnell, 14 October 1986. PP/Bow/J.9/22: ‘Compulsive caregiving is based
mainly on a fear of seeking love for fear of being rejected.’

317 Bowlby, J. (1980) Loss. London: Penguin, p.411.
phenomenon that highlights both that caregiving behaviour in humans is not the product of
some unvarying parenting instinct, and that the silent expression of attachment needs may
be present even lodged inside child-to-adult caregiving behaviours.
In his book A Secure Base, published in 1988, Bowlby gave particular attention to condi-
tions that may reliably elicit caregiving in the place of attachment behaviour, at least from
toddlerhood, and the increased cognitive resources for attending to the mind of the care-
giver.318 Child-to-parent caregiving in the place of attachment may be seen when a parent
has very severe depression such that the child’s attachment signals elicit little or no response
except when the child him-/h erself helps the parent to cope, for instance through retrieval
behaviours shown towards the parent. Child-to-parent caregiving may additionally be
elicited if a parent’s own attachment system is activated and becomes directed, through cir-
cumstances or misdirection as a consequence of conflict, towards the child.319 Bowlby ob-

served that child-to-parent caregiving responses may also be elicited, and combined with
components of the fear system, in placatory behaviour. In such cases, for instance, ‘children
learn early that it is possible to placate a disturbed and potentially violent mother by constant
attention to her wishes. Such apparently placatory behavior has been observed in young
abused children, some less than two years old.’320
Yet a problem with Bowlby’s characterisation of the caregiving system was the ambiguity
between its narrow reference, referring primarily to the retrieval response, and its flung
spray of wider meanings.321 Bowlby’s inattention to this system and ambiguities in concep-
tualizing its set-goal contributed to a neglect of the caregiving behavioural system by attach-
ment researchers until the mid-1990s. A re-examination of Bowlby’s position by George and
Solomon, in the context of a rising cultural concern with ‘motherhood as an experience and
institution’, led to renewed scientific discussion and debate of Bowlby’s position.322 An im-
portant point made by George and Solomon was that the mature balance of the caregiving
systems with other behavioural systems is critical for the provision of ‘good enough’ care to
a child. They contrasted their account to the image put forward by Bowlby, especially in his
318 Bowlby, J. (1988) A Secure Base. New York: Basic Books.

319 Personal communication by John Bowlby to Scott Henderson, cited in Henderson, S. (1974) Care-eliciting
behavior in man. Journal of Nervous and Mental Disease, 159(3), 172–81.
Journal of Psychoanalysis, 44, 9–27, p.18. Crittenden and DiLalla would later distinguish placatory behaviour from
caregiving, as different but related behavioural repertoires: Crittenden, P.M. & DiLalla, D.L. (1988) Compulsive
compliance: the development of an inhibitory coping strategy in infancy. Journal of Abnormal Child Psychology,
16(5), 585–99.
321 In the 2000s, social psychologists would highlight the significant limitations of Bowlby’s image in Attachment,
Volume 1 of retrieval as the central form and symbol of the caregiving system. Mikulincer, M., Shaver, P.R., Gillath,
O., & Nitzberg, R.A. (2005) Attachment, caregiving, and altruism: boosting attachment security increases compas-
sion and helping. Journal of Personality and Social Psychology, 89(5), 817–39, p.818; Collins, N.L., Guichard, A.C.,
Ford, M.B., & Feeney, B.C. (2006) Responding to need in intimate relationships: normative processes and indi-
vidual differences. In M. Mikulincer & G. S. Goodman (eds) Dynamics of Romantic Love: Attachment, Caregiving,
and Sex (pp.149–89). New York: Guilford.
322 Solomon, J. & George, C. (1996) Defining the caregiving system: toward a theory of caregiving. Infant Mental
Health, 17(3), 183–97. See also Stern, D. (1995) The Motherhood Constellation. New York: Basic Books; Lieberman,
A. (1996) Aggression and sexuality in relation to toddler attachment: implications for the caregiving system.
Infant Mental Health, 17(3), 276–92; Heard, D. & Lake, B. (1997) The Challenge of Attachment for Caregiving.
London: Routledge; Feeney, B.C. & Collins, N.L. (2001) Predictors of caregiving in adult intimate relationships: an
attachment theoretical perspective. Journal of Personality and Social Psychology, 80, 972–94. On changes in
public attention to the mother as ‘subject’ see e.g. Rich, A. (1995) Of Woman Born: Motherhood as Experience and
Institution. New York: Norton.
58 John Bowlby
early writing, of the caregiving system as the perfect match for the needs of the attachment
system and hence the child.323
George and Solomon also argued that economic and social supports are important for the
effective elaboration and functioning of the caregiving system. This is, in fact, exactly in line
with an argument made by Bowlby himself in his book from 1953, The Roots of Parenthood,
which was not published in America and so was unavailable to George and Solomon. There
Bowlby emphasised that caregiving is dependent on the material and social resources avail-
able to a parent, which support a caregiver’s energy, patience, and courage in the face of the
demands of caring for a child. Without support, a caregiver may well ‘give up trying’, no
matter that ‘they would like to give their children all that good parents do’.324 Bowlby con-
demned government inattention to ‘the poverty of mothers with young children’ and called
on his readers to ‘campaign unremittingly until it is remedied’.325 This was not a campaign

that he, however, would pursue over the coming decades. Instead, in his later work this con-
cern with the conditions of the effective functioning of the caregiving system focused on
psychological factors. He recognised the role of despair in hindering caregiving. However,
after the early 1950s, he did not spell out poverty and isolation as potential contributors to
despair. The traumatic loss of Bowlby’s close friend, the Labour MP Evan Durbin, in 1948
undoubtedly played a role in this transition in Bowlby’s thought.326
Besides George and Solomon, another important appraisal of Bowlby’s concept of the
caregiving system was presented by the anthropologist and primatologist Sarah Hrdy. In
her 1999 book Mother Nature, Hrdy observed that ‘Bowlby’s ideas will stand among the
greatest contributions made by evolutionary-minded psychologists to human wellbeing.’327
However, she questioned his assumption in Attachment, Volume 1 that in the environment
within which humans evolved, care would have been provided by a parent alone; the re-
sources required by a human infant are far more than any caregiver could be expected to
provide on their own. Hrdy suggested that Bowlby’s own biases and popularizing polemics
combined with the available ethological and anthropological evidence of the time to paint
a false image of mother–infant care, neglecting the necessity of social and economic sup-
ports for caregiving. Additionally, Bowlby drew on baboons and rhesus monkeys as his pri-
mate models, when in fact contrary evidence was shown from infant-sharing primates like
langurs.328 Whilst infants usually have a relatively small number of primary caregivers, the
network of people involved in offering some care to a child may be very wide. Certainly
323 George, C. & Solomon, J. (1999) The development of caregiving: a comparison of attachment theory and
psychoanalytic approaches to mothering. Psychoanalytic Inquiry, 19(4), 618–46. Seemingly unaware of earlier
critiques of Bowlby by Mary Main (Chapter 3) and Solomon and George, anthropologists have developed this
point in detail especially over the past decade, e.g. Carlson, V.J. & Harwood, R.L (2014) The precursors of at-
tachment security: behavioral systems and culture. In H. Otto & H. Keller (eds) Different Faces of Attachment.
Cambridge: Cambridge University Press, 278–303.
324 Bowlby, J. (1953) The Roots of Parenthood. London: National Children’s Home, p.16.
325 Ibid. p.14.
326 Mayhew, B. (2006) Between love and aggression: the politics of John Bowlby. History of the Human Sciences,
19(4), 19–35.
327 Hrdy, S.B. (1999) Mother Nature: A History of Mothers, Infants and Natural Selection. New York: Pantheon,
p.xiii.
328 Hrdy, S.B. (2005) Evolutionary context of human development: the cooperative breeding model. In L.A.C.S.
Carter, K.E. Grossmann, S.B. Hrdy, M.E. Lamb, S.W. Porges, & N. Sachser (eds) Attachment and Bonding: A New
Synthesis (pp. 9–32). Cambridge, MA: MIT Press. More recently, see Myowa, M. & Butler, D.L. (2017) The evolu-
tion of primate attachment: beyond Bowlby’s rhesus macaques. In H. Keller & K.A. Bard (eds) The Cultural Nature
of Attachment. Contextualizing Relationships and Development (pp.53–68). Cambridge, MA: MIT Press.
in human evolutionary history, care has not usually been provided by one person alone as
Bowlby sometimes implied, especially in his earlier writings.329
Hrdy offered no objection to Bowlby’s claim that the attachment behavioural system dis-
criminates caregivers, and perhaps even has a hierarchy of attachment figures. However, she
proposed that humans have evolved to engage in cooperative care, with mothers, fathers,
grandmothers, aunts, siblings, and adult friends all involved, depending on circumstances.
And when insufficient economic and social resources and supports are available, caregivers
can be expected to divest from their children, reducing their availability as sources of pro-
tection.330 The attachment system evolved in the context of this threat, and it is part of what
makes exploration a dangerous activity for an infant who is not sure of the caregiver’s avail-
ability. Likewise, the potential for an attachment figure to divest from their infant if other
demands are pressing is part of what makes a network of attachment figures part of the

evolutionary expectable conditions of infant survival.331 Dragged towards reductionism at
times by the narrow version of caregiving as retrieval, Bowlby did not adequately recognise
the extent to which the ordinary functioning of the human caregiving behavioural system
entails not only direct care, but also bringing additional supports on line for one we care for.
And he did not adequately recognise that an integral part of the functioning of the caregiving
system entails protecting the one we care about from ourselves, whether this is failures of
patience or the wish to abandon the other when we feel overwhelmed. These are important
‘anti-goals’ of the caregiving behavioural system (Chapter 5).
Hrdy’s critique has been well taken at a theoretical level by attachment researchers. In gen-
eral, attachment researchers emphasise that young children will likely not be significantly
harmed, and may even benefit, from high-quality care from several caregivers who develop
personal relationships with the child. The primary qualifications that have been offered are
that: (i) the child should not be passed around so many people that it proves difficult to de-
velop at least one stable relationship, as is often the case with institutional care; and (ii) in-
secure attachment is made more likely if an infant is in very extensive daycare (e.g. over 70
hours a week), or (iii) if nights are spent with relative strangers.332 A primary caregiver by
no means needs to be present to the degree that Bowlby claimed in his popular writings of
329 Hrdy, S.B. (1999) Mother Nature: A History of Mothers, Infants and Natural Selection. New York: Pantheon,
p.495, discussing Tronick, E.Z., Winn, S., & Morelli, G.A. (1985) Multiple caretaking in the context of human
evolution: why don’t the Efé know the western prescription for child care? In M. Reite & T. Field (eds) The
Psychobiology of Attachment and Separation (pp.293–322). Orlando, FL: Academic Press. For a more recent discus-
sion of multiple caregiving see Meehan, C.L. & Hawks, S. (2013) Cooperative breeding and attachment among the
Aka foragers. In N. Quinn & J.M. Mageo (eds) Attachment Reconsidered: Cultural Perspectives on a Western Theory.
London: Palgrave.
330 Hrdy appears to have been, understandably, unaware of Bowlby who made exactly this point in his 1953
book The Roots of Parenthood. The book was not distributed in the USA. Bowlby, J. (1953) The Roots of Parenthood.
London: National Children’s Home.
331 Hrdy, S.B. (1999) Mother Nature: A History of Mothers, Infants and Natural Selection. New York:
Pantheon, p.536.
332 Hazen, N.L., Allen, S.D., Christopher, C.H., Umemura, T., & Jacobvitz, D.B. (2015) Very extensive nonmater-
nal care predicts mother–infant attachment disorganization: convergent evidence from two samples. Development
& Psychopathology, 27(3), 649–61; van IJzendoorn, M.H., Bakermans-Kranenburg, M.J., Duschinsky, R., &
Skinner, G.C.M. (2019) Legislation in search of ‘good-enough’ care arrangements for the child: a quest for con-
tinuity of care. In J. Dwyer (eds) Handbook of Children and the Law. Oxford: Oxford University Press. Belsky and
colleagues have, however, found that the extent of non-familial care is associated with a small increase in child
conduct problems and impulsivity, with maternal sensitivity as a moderator: e.g. Burchinal, M.R., Lowe Vandell,
D., & Belsky, J. (2014) Is the prediction of adolescent outcomes from early child care moderated by later maternal
sensitivity? Results from the NICHD study of early child care and youth development. Developmental Psychology,
50(2), 542–53.
60 John Bowlby
the 1950s. And Bowlby’s use of the term ‘mother’ to mean a primary caregiver—following
common practice in psychoanalytic discourse of his day—should not be taken to imply that
a primary caregiver needs to be a mother.
It is false to say of the contemporary field of empirical attachment research, as do Keller
and Chaudhary, that it ‘aims at demonstrating the uniqueness of the mother–child bond’.333
If this were the aim, then, paradoxically, greater attention would have been paid to a var-
iety of caregivers in order to comparatively demonstrate the particular qualities of maternal
care. Other care providers have sometimes been studied by attachment researchers, for in-
stance daycarers and foster parents.334 However, these have hardly been comparisons to
demonstrate the uniqueness of maternal care, as Keller and Chaudhary allege. Yet Keller
and Chaudhary are undoubtedly correct in their criticism that, in practice, attachment re-
search after Bowlby in the developmental tradition has predominantly focused on practice

on mother–infant relationships. Comparatively few researchers have been able to access or
prioritise the resources for studies of the contributions of wider caregiving networks using
the labour-intensive measures of developmental attachment research,335 though certainly
some research has been done on caregiving networks by developmental researchers.336
Nonetheless, it is striking, for instance, that not a single study to date has examined
infant–grandparent attachment using the Strange Situation, given the widely acknowledged
importance of grandparents as alloparents in both human evolutionary history and contem-
porary societies.337 The foremost reason why attachment researchers have focused research
attention on mothers is that mothers have had primary care responsibilities for infants in
America and in Europe, where the majority of attachment researchers are based. However,
there are certainly enough infants living with their grandparents for this to have been a viable
topic for study at some point since the 1970s. A focus on maternal care may have been partly
self-perpetuating for attachment researchers after Bowlby in the developmental tradition,
as it made this population the standard for comparability and study of the effects of spe-
cific variables on care.338 A positive recent development, however, are plans over the coming
years by Or Dagan and Avi Sagi-Schwartz to pool the raw data from all studies of attachment
333 Keller, H. & Chaudhary, N. (2017) Is the mother essential for attachment? Models of care in different cul-
tures. In H. Keller & K.A. Bard (eds) The Cultural Nature of Attachment (pp.109–37). Cambridge, MA: MIT Press,
p.112. This claim rests on a characteristic rhetorical strategy among many critics, in which early statements by
Bowlby are used to characterize all subsequent empirical attachment research (Chapter 2).
334 E.g. Sagi, A., van IJzendoorn, M.H., Aviezer, O., Donnell, F., Koren-Karie, N., Joels, T., & Harl, Y. (1995)
Attachments in a multiple- caregiver and multiple- infant environment: the case of the Israeli kibbutzim.
Monographs of the Society for Research in Child Development, 60(2–3), 71–91.
335 Keller, H. & Chaudhary, N. (2017) Is the mother essential for attachment? Models of care in different cul-
tures. In H. Keller & K.A. Bard (eds) The Cultural Nature of Attachment (pp.109–37). Cambridge, MA: MIT Press.
336 E.g. van IJzendoorn, M.H. (2005) Attachment in social networks: toward an evolutionary social network
model. Human Development, 48(1–2), 85–8; De Schipper, J.C., Stolk, J., & Schuengel, C. (2006) Professional care-
takers as attachment figures in day care centers for children with intellectual disability and behavior problems.
Research in Developmental Disabilities, 27, 203–16. There has been a somewhat greater focus on wider networks
among attachment researchers in the social psychological tradition, but not with respect to caregiving. Gillath, O.,
Karantzas, G.C., & Lee, J. (2019) Attachment and social networks. Current Opinion in Psychology, 25, 21–5.
337 Gibson, M.A. & Mace, R. (2005) Helpful grandmothers in rural Ethiopia: a study of the effect of kin on child
survival and growth. Evolution and Human Behavior, 26(6), 469–82.

338 However, there has been a growing literature on fathers and on co-parenting (Chapter 4). An important
development has been the inclusion of fathers within research on attachment-based interventions: e.g. Iles, J. E.,
Rosan, C., Wilkinson, E., & Ramchandani, P.G. (2017) Adapting and developing a video-feedback intervention
for co-parents of infants at risk of externalising behaviour problems (VIPP-Co): a feasibility study. Clinical Child
Psychology and Psychiatry, 22(3), 483–99.
to multiple caregivers using standardised measures, to permit statistical analyses with good
depth and breadth, and to reconcile questions of comparability.
The exploratory behavioural system
The exploratory behavioural system was first detailed in Attachment, Volume 1, though the
discussion remained relatively cursory.339 Following Hinde’s earlier discussions, Bowlby ar-
gued that the activating conditions of the exploratory behavioural system are ‘stimuli that
are novel and/or complex’.340 An additional criterion implicit in Bowlby’s account of the ex-
ploratory system, but left unstated, is that the stimuli need to have some potential relevance
to the individual. In infancy, Bowlby argued that the exploratory system has three elements

that often form a sequence: orienting towards the stimulus to gain information and prime
responsiveness; approach to the stimulus to examine it; and manipulation or experimenta-
tion with the stimulus in order to understand what it can do.341 The characteristic affect that
supports and calibrates the expression of the exploratory system is ‘curiosity’.342 Like attach-
ment, exploration was conceptualised as a homeostatic system, since once relevant novelty
and complexity had been examined and understood, the system was expected to terminate.
Bowlby situated the exploratory system as equivalent to Winnicott’s concept of ‘play’, high-
lighting its importance as a system through which a child learns that meanings are partly
self-created and partly supported by what the world makes available.343 The experience of
joy requires a world that can still surprise us with things we cherish, especially when we have
attachment figures with whom to share discoveries. So Bowlby regarded the exploratory
system as implicated in the pleasures that can come from connecting and reconnecting the
339 The limited acknowledgement of the exploratory system in Bowlby’s work might be linked to the scholarly
context of the 1960s. Social learning was a pivotal discourse on child development, and one with which Bowlby
had some sympathies. Both Bowlby and social learning theorists were concerned with the consistency and respon-
siveness of caregiving. However, he was keen to distinguish attachment behaviour from the effect of behaviourist
conditioning, which was the dominant frame at the time for conceptualizing social learning. A more detailed
elaboration of the exploratory system would have confronted Bowlby with the need to tackle social learning theory
head-on, which would have been a demanding task, as demonstrated when it eventually fell to Ainsworth to clarify
the differences between the attachment and social learning paradigms. See Bowlby, J. (1961) Comment on paper
by Dr Gewirtz. In B.M. Foss (ed.) Determinants of Infant Behaviour (pp.301–304). London: Methuen; Ainsworth,
M. (1972) Attachment and dependency: a comparison. In J. Gewirtz (ed.) Attachment and Dependency (pp. 97–
137). Washington, DC: Winston. An additional contemporary dynamic was that creativity and play research were
just getting off the ground. Attention to the exploratory system may well have been fed by this source, as shown
by Bowlby’s appeal to Winnicott’s concept of creative play. However, theories of creativity and play were still weak
when Bowlby was working on Attachment, Volume 1.
340 Bowlby, J. (1969, 1982) Attachment, Volume 1. London: Penguin, p.238, discussing Hinde, R.A. (1954) Factors
governing the changes in strength of a partially inborn response. 1. The nature of the response, and an examination
of its course. Proceedings of the Royal Society B, 142, 306–31. Other important sources for Bowlby were Berlyne,
D.E. (1950) Novelty and curiosity as determinants of exploratory behaviour. British Journal of Psychology, 41, 68–
80; Rheingold, H.L. (1963) Controlling the infant’s exploratory behavior. In B.M. Foss (ed.) Determinants of Infant
Behaviour, Vol. 2 (pp.171–217). London: Methuen.
341 Bowlby, J. (1969, 1982) Attachment, Volume 1. London: Penguin, pp.237–38.
342 Ibid. p.197. See also Marvin, R.S., Britner, A.A., & Russell, B.A. (2016) Normative development: the on-
togeny of attachment in childhood. In J. Cassidy & P.R. Shaver (eds) Handbook of Attachment: Theory, Research,
and Clinical Applications 3rd edn (pp. 273–90). New York: Guilford.
Affectional Bonds (pp.150–88). London: Routledge, p.183–4; cf. Winnicott, D.W. (1967) The location of cultural
experience. International Journal of Psychoanalysis, 48, 368–72.
62 John Bowlby
self and world, in sustaining their dialogue.344 As the game ‘peek-a-boo’ demonstrates, even
the absence of the caregiver may offer the pleasure of manipulation and experimentation if
it occurs under the aegis of the exploratory behavioural system, and is brief enough not to
threaten the dominance of this system.
Ainsworth qualified Bowlby’s account through her attention to half-hearted forms of
exploration. She identified this kind of play in infants showing avoidant behaviour in the
Strange Situation Procedure (Chapter 2). She suggested that these infants were using play
to distract themselves from attachment-relevant internal and external cues, thereby sup-
pressing the expression of the attachment system.345 Likewise, Main later observed that
frame-by-frame video analysis of the reunion episodes of avoidant dyads ‘reveals that the in-
animate object generally has far from the infant’s full attention. The infant may, for example,
rather frantically turn toward a tale leg and finger it (but with eyes fixed blankly on the wall

ahead). Or, in a clumsily decisive move, it may suddenly drop a toy into a box but then close
the box on its hand. The general impression is, again, that the infant could not succeed in
maintaining its avoidance of the parent without the aid of the object seized.’346 Here explor-
ation is used as somewhere else to go, a projection of attention away from felt concerns and
into the environment.
The contrast between the lacklustre and distracted infant exploration seen in avoidant
dyads and true play burnished by lively curiosity led Ainsworth to criticise Bowlby for his
account of the exploratory system, which only infrequently drew distinctions between these
two forms.347 Ainsworth suggested that Bowlby had given a place to exploration, but did not
take it seriously enough as a behavioural system. His characterisation had focused too much
on its behavioural components, at the expense of affective and cognitive components that
lead a child to engage in ‘actively seeking’ information within a social and cultural context.348
She offered the example of an infant encountering a novel and potentially alarming situation.
Before the sequence described by Bowlby of orienting, approaching, and experimenting,
Ainsworth asserted that a first stage of the exploratory system in infancy will be social ref-
erencing of attachment figures. The caregiver helps offer shape and definition to the infant’s
world and will often be actively sought to serve as ‘referee’ for the meaning of the stimulus,
especially when there is a question whether its novelty should stimulate exploration and
344 Some of Bowlby’s examples are quite culturally specific, e.g. on novelty and pleasure in family holidays,
Bowlby, J. (1969, 1982) Attachment, Volume 1. London: Penguin, p.41. Whether exploration of complex/novel
stimuli can be interpreted as equivalent to Winnicott’s concept of ‘play’ depends on what is understood by the
latter term, an understanding shaped in important ways by culture. Lancy, D.F. (2007) Accounting for variability in
mother–child play. American Anthropologist, 109(2), 273–84.
345 Ainsworth, M. (1984) Attachment. In N.S. Endler & J. McVicker Hunt (eds) Personality and the Behavioral
Disorders. New York: Wiley, 559– 602, p.565. See also Grossmann, K., Grossmann, K.E., Kindler, H., &
Zimmermann, P. (2008) A wider view of attachment and exploration: the influence of mothers and fathers on
the development of psychological security from infancy to young adulthood. In J. Cassidy & P.R. Shaver (eds)
Handbook of Attachment: Theory, Research, and Clinical Applications, 2nd edn (pp.857–79). New York: Guilford.
346 Main, M. (1981) Avoidance in the service of proximity: a working paper. In K. Immelmann, G.W. Barlow,
L. Petrinovich, M. Biggar Main (eds) Behavioral Development: The Bielefeld Interdisciplinary Project (pp.651–93).
Cambridge: Cambridge University Press, p.554–5.
347 E.g. Bowlby, J., Robertson, J., & Rosenbluth, D. (1952) A two-year-old goes to hospital. Psychoanalytic Study
of the Child 7, 82–94, p.83. He also discussed the use of distraction activities in adults: Bowlby, J. (1980) Loss.
London: Pimlico: ‘One or more behavioural systems within a person may be deactivated, partially or completely.
When that occurs one or more other activities may come to monopolize the person’s time and attention, acting ap-
parently as diversions’ (64).
348 Ainsworth, M. (1992) A consideration of social referencing in the context of attachment theory and research.
In S. Feinman (ed.) Social Referencing and the Construction of Reality in Infancy (pp.349–67). New York: Plenum
Press, p.361.
affiliation or wariness. For Ainsworth, exploration took part of its significance from the
long infancy of human beings, and the dense, multifaceted urgency of the human society
that infants must learn to navigate. A behavioural system that makes matters that are novel
and/or complex attractive for exploration is therefore a great asset, since, alongside staying
safe moment-to-moment, the other great task of children is to encounter and learn from the
world around them (Chapter 2).349
The fear behavioural system
In Bowlby’s account, when a stimulus is judged as alarming, two systems may be acti-
vated: the attachment system and the fear system. The first third of his book Separation was

essentially a theory of the fear behavioural system.350 Bowlby urged recognition that attach-
ment and fear are ‘distinct behavioural systems that (a) have the same function, (b) may
be elicited by many of the same conditions, (c) are frequently compatible with each other,
but (d) can easily be in conflict’.351 The compatibility of the attachment system and the fear
system is that both are oriented towards increasing proximity to safety in contexts of per-
ceived threat. The difference, for Bowlby, is that the attachment system disposes attempts to
achieve the availability of attachment figures, whereas the fear system prompts attempts both
to escape the threat and to find people or places that have proven safe in the past. Bowlby
suggested that the term ‘alarm’ should be used for concern about a threatening stimulus, and
‘anxiety’ should be used for concern about the availability of safety, whether due to the pres-
ence of danger or questions about the availability of attachment figures.352 In distinguishing
the attachment and fear systems, however, Bowlby left unaddressed whether or how feelings
of safety and of security might inflect one another, and how the set-goal of the attachment
behavioural system might be conditioned by experiences of the fear behavioural system and
its resolution.
349 Ainsworth, M. & Bell, S. (1970) Attachment, exploration, and separation: illustrated by the behavior of one-
year-olds in a strange situation. Child Development, 41(1), 49–67, p.51. The underdevelopment of the idea of the
exploratory system in Bowlby’s work is widely acknowledged, e.g. Gullestad, S.E. (2001) Attachment theory and
psychoanalysis: controversial issues. Scandinavian Psychoanalytic Review, 24, 3–16. Subsequent to Ainsworth,
the major sustained interest in the relationship between exploration and attachment has been among clinically-
focused commentators, who have been concerned with what attachment research can suggest about moderators
of our capacity to explore the minds of others and learn from experience e.g. Fonagy, P., Luyten, P., Allison, E., &
Campbell, C. (2017) What we have changed our minds about: part 2. Borderline personality disorder, epistemic
trust and the developmental significance of social communication. Borderline Personality Disorder and Emotion
Dysregulation, 4, 9; Golding, K. & Hughes, D. (2012) Creating Loving Attachments: Parenting with PACE to Nurture
Confidence and Security in the Troubled Child. London: Jessica Kingsley; Powell, B., Cooper, G., Hoffman, K., &
Marvin, B. (2016) The Circle of Security Intervention. New York: Guilford.
350 Bowlby had also been very interested in human fear behaviour since his work as an army psychiatrist.
However, the proximal cause of Bowlby’s attention to the topic in the early 1970s appears to have been the pro-
posal of fear as a distinct behavioural system by the American developmentalist Gordon Bronson. Bronson, G.W.
(1968) The development of fear in man and other animals. Child Development, 39(2), 409–31. Bronson observed
that the threshold for activation of the fear behavioural system seemed to be lower for institutionally reared human
children and monkeys, and that fear of strangers seemed to develop at around the same time as discrimination
of attachment figures. However, Bowlby felt that Bronson’s distinctions were not adequately sharp between fear,
wariness, and anxiety. In a wider context, cultural discourses appealing to and distinguishing ‘fear’ and ‘anxiety’
were gaining ground from the late 1960s, and may have contributed to the salience of the topic for Bronson and
Bowlby. See Jenkins, P. (2006) Decade of Nightmares: The End of the Sixties and the Making of Eighties America.
Oxford: Oxford University Press; Bourke, J. (2006) Fear: A Cultural History. Emeryville, CA: Shoemaker & Hoard.
351 Bowlby, J. (1973) Separation. London: Penguin, p.117.
352 Ibid. Appendix III.
64 John Bowlby
Component behaviours of the fear system identified by Bowlby in his notes included the
startle response, withdrawal, fleeing, and hiding.353 Later, on the suggestion of Hinde, he
added freezing as another behavioural component of the fear system.354 This led him to dis-
cuss ‘three distinct kinds of predictable outcome’ of the fear behavioural system ‘(a) immo-
bility, (b) increased distance from one type of object, and (c) increased proximity to another
type of object’.355 These responses, he proposed, are highly responsive to context: ‘When a
chimpanzee is startled by a sudden noise or movement nearby, its immediate response is
to duck its head and to fling one or both arms across its face; alternatively, it may throw
both hands in the air. Occasionally these startle reactions are followed by a hitting-away
movement with the back of the hand towards the object, at other times by flight. When the
alarming object is another and more dominant chimpanzee, flight is accompanied by loud
screaming; when it is anything else, flight is quite silent.’356 In an unpublished paper from

1973 entitled ‘Wariness’, Bowlby discussed a continuum of fear responses. At one end was
flight and expressions of terror. At the other end was various forms of wary behaviour.357
In Separation, Bowlby proposed that the activating conditions for the fear system are
partly learnt on the basis of experience, but are also partly neurologically primed by human
evolutionary history. Children are certainly more likely to fear the dark if they have been told
that darkness is dangerous, they have seen others respond with caution to darkness, or if they
associate darkness with unhappy experiences. However, Bowlby argued that humans are
predisposed to have a lower bar for treating certain cues, such as darkness, as threatening.
Other such cues potentiated to have lower bars for activation include being cold and loud
sudden noises. Perhaps the most important such cue for danger, Bowlby observed, is the
feeling of being alone, or of blocked access to attachment figures.358 Furthermore, he argued
that the bar for activation of the fear behavioural system would fall dramatically where two
or more such cues were present at once, and especially when one of these was feeling alone.
Where multiple cues to danger occur at once, most adult humans respond with an activation
of the fear system.359
However, especially in childhood, we may not know how to interpret the cause of such an
activation: ‘Often, in fact, when we feel impelled to act in a certain way that is readily explic-
able in terms of biological function, we concoct “reasons” for doing so that bear little or no
relation to the causes of our behaviour. For example, a child or adult, who in order to reduce
risk is biologically disposed to respond to strange sounds in the dark by seeking his attach-
ment figure, gives as his reason that he is afraid of ghosts.’360 Bowlby contended that children’s
actual experiences in situations with such primed cues for danger are likely to be especially
important for the later development of mental health problems. For instance, the origins of
353 Bowlby, J. (1968) Types of fear response. PP/Bow/H.209.

354 Bowlby, J. (1973) Separation: Anxiety and Anger. New York: Basic Books: ‘Hinde (1970) reports a finding
by Hogan that, in chicks, withdrawal occurs from stimuli at high intensity (and some others) whereas freezing is
elicited by stimuli that are strange, novel, or surprising’ (154).
355 Ibid. p.114.
356 Ibid. p.157.
357 Bowlby, J. (1973) Wariness. Unpublished manuscript, June 1973. PP/BOW/J.9/39. The same distinction was
used by Bretherton, I. & Ainsworth, M.D.S. (1974) Responses of one-year-olds to a stranger in a strange situation.
In M. Lewis & L.A. Rosenblum (eds) The Origin of Fear (pp.131–64). New York: Wiley.
358 Bowlby, J. (1973) Separation: Anxiety and Anger. New York: Basic Books, p.119.
359 Bowlby, J. (1970, 1979) Self-reliance and some conditions that promote it. In The Making and Breaking of
Affectional Bonds (pp.124–49). London: Routledge, p.147.

360 Bowlby, J. (1981) Psychoanalysis as a natural science. International Review of Psycho- Analysis, 8,
243–56, p.248.
many phobias and paranoid symptoms can be understood as ‘intelligible, albeit distorted’ re-
sponses to historical experiences in contexts with multiple such cues, where a caregiver was
not available to help soothe the child and interpret the meaning of their response.361
On the basis of analysis of their activating and terminating conditions, Bowlby was clear
that the attachment and fear systems were distinct. With Robertson, he had considered the pre-
dicament of hospitalised children, who experienced fear but without being able to turn to their
caregivers. Part of Bowlby’s focus on long-term separations stemmed from his sense that the
clinical community and public were not able to acknowledge child abuse by parents. Yet in the
early 1970s, the tide was turning on the acknowledgement of child physical abuse within Britain,
and within the medical community in particular.362 The manuscript history of Separation sees
Bowlby edging slowly towards the claim that the attachment and fear behavioural systems could
conflict as a result of directly frightening or abusive actions by the attachment figure. In an early

draft of material for Separation Bowlby reported having seen a lamb attempting to cross directly
into the path of a car, having been startled by the sound of the car and seeing its mother on the
other side of the road.363 In correspondence with Ainsworth, and then in the typescript, Bowlby
added consideration of the case of an infant who is confronted with a threatening stimulus, such
as a ‘barking dog’, that lies in the path towards the attachment figure.364 Reflecting further, in
a pencil annotation on the typescript and then in the published version, Bowlby added that ‘a
special but not unusual situation in which there is conflict between attachment behaviour and
withdrawal is when the attachment figure is also the one who elicits fear, perhaps by threats or
violence. In such conditions young creatures, whether human or non-human, are likely to cling
to the threatening or hostile figure rather than run away from him or her.’365 A threatening or
violent caregiver, Bowlby observed, could be expected to activate the fear and attachment sys-
tems, producing conflict between the two behavioural systems (Chapter 3).
The aggression behavioural system
With one exception in Attachment, Volume 1, Bowlby did not generally describe aggression
as a behavioural system.366 Yet consideration of the case of the aggression system in closing
this section on behavioural systems may help shed light on the boundaries of this concept for
Bowlby. He gave aggression only passing and faltering attention.367 Despite the fact that the
subtitle of his book Separation is ‘Anxiety and Anger’, in fact Part I focuses on anxiety, Part
362 Parton, N. (1985) The Politics of Child Abuse. Basingstoke: Macmillan; Hacking, I. (1991) The making and
molding of child abuse. Critical Inquiry, 17(2), 253–88; Ferguson, H. (2004) Protecting Children in Time: Child
Abuse, Child Protection and the Consequences of Modernity. London: Palgrave.
363 Bowlby, J. (1973) Draft material towards Separation. PP/Bow/K.5./17.
364 Bowlby, J. (1973) Letter to Ainsworth, 29 October, 1973. PP/Bow/J.1/33; Bowlby, J. (1973) Separation: Anxiety
and Anger. New York: Basic Books: ‘Conflict can easily occur, for example, whenever a stimulus situation that
elicits both escape and attachment behaviour in an individual happens to be situated between that individual and
his attachment figure; a familiar instance is when a barking dog comes between a child and his mother’ (116). The
typescript is available at PP/Bow/K.5./17.
365 Bowlby, J. (1973) Separation: Anxiety and Anger. New York: Basic Books, p.117. The typescript is available at
PP/Bow/K.5./17.
366 Bowlby, J. (1969, 1982) Attachment, Volume 1. London: Penguin: ‘In Tom, it can be said, there is a tendency to
appraise certain situations in such a way that a behavioural system is activated that results in his attacking his little
sister and biting her’ (118).
367 Bowlby’s most sustained attention to aggression appears in Durbin, E.F.M. & Bowlby, J. (1939) Personal
Aggressiveness and War. London: Kegan Paul. Little from this book fed through into his later reflections on
66 John Bowlby
II focuses on fear, and Part III returns essentially to a consideration of anxiety, with the ex-
ception of a dozen pages in Chapter 17 (entitled ‘Anger, Anxiety, and Attachment’).368 Mary
Main later observed that Bowlby’s exclusion of aggression as a behavioural system ‘has puz-
zled many clinicians’, and emphasises the disparity this has left between theory and empirical
research on aggression.369 Shaver and Mikulincer expressed disappointment with Bowlby
for his poor attention to aggression, since it has left attachment theory without a specified
place for anger (Chapter 5).370 And van IJzendoorn and Bakermans-Kranenburg and col-
leagues felt forced to bypass Bowlby in their development of an intervention with caregivers
struggling with aggressive child behaviour.371
It may be speculated that aggression was refused the status of a behavioural system be-
cause Bowlby wanted to avoid even the slightest implication of acceptance of Melanie Klein’s
position. Whereas Klein regarded aggression as an innate human drive, Bowlby claimed that

most aggression is a response to the frustration of other behavioural system.372 Certainly
this is a point that he reiterates time and again when aggression is under discussion. Yet, as
Ainsworth observed, frustration could still then be a behavioural system: it is neurologically
primed as a result of human evolutionary history, but requires frustrating experiences for its
elaboration, and works to achieve its goals through various behavioural means.373
An additional factor may have contributed to Bowlby’s wariness in describing aggression
as a behavioural system. Bowlby’s younger colleagues at the Tavistock recall that he used
to think aloud in seminars and conversations about whether the idea of ‘aggression’ was
too much of a catch-all. He was far from certain that it was a single thing.374 In Separation,
he described two forms of aggression, only one of which resembled a behavioural system.
aggression in Separation. Those who knew Bowlby personally have often remarked that he did not seem at all com-
fortable with aggression, which may have contributed to a theoretical antipathy.
368 This emphasis has been echoed by later attachment researchers: for instance, Cassidy’s influential intro-
duction to the Handbook of Attachment discusses the attachment, caregiving, exploratory, sociable, and fear sys-
tems, but does not discuss anger as a behavioural system. Cassidy, J. (1999) The nature of the child’s ties. In J.
Cassidy & P.R. Shaver (eds) Handbook of Attachment: Theory, Research, and Clinical Applications (pp.3–21).
New York: Guilford. This selection of behavioural systems was maintained by Cassidy in the second (2008) and
third (2016) editions.
369 Main, M. (1993) Les bébés et leurs colères. In M.C. Busnel (eds) Le Langage des Bébés, Savons-nous Entendre?
(pp.17–91). Paris: Grancher; Main, M. (1995) Recent studies in attachment: overview, with selected implications
for clinical work. In S. Goldberg, R. Muir, & J. Kerr (eds) Attachment Theory: Social, Developmental and Clinical
Perspectives (pp.407–70). Hillsdale, NJ: Analytic Press, p.460. In work with Jude Cassidy, Main describes the role
of aggressive behaviour within displays of dominance from children towards their attachment figures (Chapter 3),
but without ever referring this back to Bowlby’s discussion of aggression. An important disjuncture lay in the fact
that Bowlby’s accounts of aggression characterized primarily frustration and protest in the service of attachment,
whereas what Cassidy and Main were seeing was more like use of aggression in the service of dominance in the
service of attachment.
370 Shaver, P.R., Segev, M., & Mikulincer, M. (2011) A behavioral systems perspective on power and aggression.
In P.R. Shaver & M. Mikulincer (eds) Human Aggression and Violence: Causes, Manifestations, and Consequences
(pp.71–87). Washington, DC: American Psychological Association, p.71.
371 Van Zeijl, J., Mesman, J., van IJzendoorn, M.H., et al. (2006) Attachment-based intervention for enhancing
sensitive discipline in mothers of 1-to 3-year-old children at risk for externalizing behavior problems: a random-
ized controlled trial. Journal of Consulting and Clinical Psychology, 74(6), 994–1005.
Mary Salter Ainsworth. Psychoanalytic Study of the Child, 52, 386–405, p.403.
374 E.g. Issroff, J. (2005) Donald Winnicott and John Bowlby: Personal and Professional Perspectives.
London: Karnac: ‘Bowlby disliked what he called sometimes “portmanteau” and sometimes “umbrella” or
“omnibus” words like “aggression”, which covered too many different broad possibilities and created confusion in
the way they were used’ (56).
The first form was what he termed ‘the anger of despair’.375 In this form of aggression the
behaviour occurs primarily as the expression of emotion, spinning within its own loops of
intensity, without the environmental responsiveness that would help behaviour to achieve
a specific set-goal. The anger of despair, Bowlby argued, ‘occurs whenever a person, child
or adult, becomes so intensely and/or persistently angry with his partner that the bond be-
tween them is weakened, instead of strengthened, and the partner is alienated. Anger with a
partner becomes dysfunctional also whenever aggressive thoughts or acts cross the narrow
boundary between being deterrent and being revengeful.’376 The anger of despair is a mood
that disbelieves hope; it is premised on the assumption that aggression cannot change the
environment.
He contrasted the anger of despair with ‘the anger of hope’.377 This form of anger is elicited
by frustration of another behavioural system. It may seek removal of the obstacle. More often,

though, it acts upon people or the environment in such a way as might coerce them to yield,
permitting the satisfaction of the other system. For instance, Robertson documented a great
increase in aggressive behaviour in children following their hospital separations. Bowlby felt
that this might be conceptualised as a kind of punishment of the caregivers for the separation
by the child, signalling that the experience should not be repeated: ‘In its functional form anger
is expressed as reproachful and punishing behaviour that has as its set-goals assisting a reunion
and discouraging further separation. Therefore, although expressed towards the partner, such
anger acts to promote, and not to disrupt, the bond.’378 The anger of hope is a mood that disbe-
lieves despair; it is premised on the wish that aggression may change the environment.
The anger of hope is a dangerous strategy, and may readily backfire by provoking retali-
atory aggression or rejection from the partner or caregiver. As Waters and Sroufe later ob-
served, aggression risks (further) disrupting the trust and patience on which any secure
attachment relationship must rest.379 However, this risk may be better, or be felt to be better,
than alternatives, such as being unnoticed or despairing of achieving co-regulation within
a relationship.380 The anger of hope has a good claim to be a behavioural system, especially
in light of its clear set-goal. However, perhaps a mark against even the anger of hope as a
behavioural system is that it remains the servant of other behavioural systems, and when
it achieves independence from them it is as the dysregulated anger of despair. The anger of
hope may have seemed to Bowlby as a component behaviour of other behavioural systems,
rather than having the status of a behavioural system in itself. However, as Waters and col-
leagues have argued, such a stance ascribes too little independence to aggressive behaviours,
not least since they may be fed by more than one source of frustration.381 Moreover, sev-
eral second-generation attachment researchers including Shaver and Fonagy later criticised
376 Ibid. p.288.
377 Ibid. p.285.
378 Ibid. p.287.
379 Waters, E. & Sroufe, L.A. (1983) A road careened into the woods: comments on Dr Morrison’s commentary.
Developmental Review, 3(1), 108–14, p.223.

380 Knox, J. (2005) Sex, shame and the transcendent function: the function of fantasy in self development.
Journal of Analytical Psychology, 50, 617–39: ‘I was fortunate enough to hear John Bowlby lecture on one occasion
and one comment he made, almost in passing, remains imprinted on my mind—it was that children can survive
the experience that their hate may drive a parent away but to have one’s love rejected is intolerable’ (625).
381 Waters, E., Posada, G., Crowell, J., & Lay, K.L. (1993) Is attachment theory ready to contribute to our under-
standing of disruptive behavior problems? Development & Psychopathology, 5(1–2), 215–24, p.220.
68 John Bowlby
Bowlby on the grounds that aggression can be evoked in the absence of the frustration of a
behavioural system.382
The meaning and structures of symptoms
Defence
A central motivation—perhaps the central motivation—for Bowlby’s development of a

theory of behavioural systems was to account for the development of mental health symp-
toms in humans. His new model allowed him to consider symptoms that might arise from

the interplay or conflict of different systems, from poor alignment between behavioural
systems and their social context, or from aspects of the functioning or misfiring of indi-
vidual systems themselves. Reflecting on his clinical experiences with adults, Bowlby was
impressed by the psychoanalytic insight that even an inhibited response not only continues
to exist but can develop, ‘putting out derivatives and establishing connections’ even if these
may be contorted by the inhibition.383 For instance, inhibited behavioural systems, even if
they do not influence our behaviour, may nonetheless exert a powerful influence on our
judgement and mood, as well as the activation of memories and forms of imagination or
daydream.384 Such processes may influence individuals’ experiences even of their own mo-
tivations. In Attachment, Volume 1, Bowlby wrote that ‘reports are usually trustworthy, but
psychoanalysts know well that that is not always so. A subject may in fact mis-identify the
set-goal of a behavioural system currently active—and such mis-identification may itself be
the result of interference by an active system that has a set-goal which is incompatible with
the first. This leads to the concept of unconscious wish.’385
However, in discussions of such phenomena, Bowlby felt that the concept in common use
in the psychoanalytic community, ‘defence’, was often confusing. The term absorbed a variety
of psychological and behavioural activities that aimed to reduce or eliminate experiences
that might threaten the integrity or stability of the individual. Bowlby felt that this usage was
lazy, and left nameless and soundless many critical psychological processes.386 He addressed
382 Shaver, P., Schwartz, J., Kirson, D., & O’Connor, C. (1987) Emotion knowledge: further exploration of a proto-
type approach. Journal of Personality and Social Psychology, 52(6), 1061–86, p.1078; Fonagy, P. (2003) The violence
in our schools: what can a psychoanalytically informed approach contribute? Journal of Applied Psychoanalytic
Studies, 5(2), 223–38, p.230.
383 Annotations by Bowlby dated 1958 on Freud’s essay ‘Repression’ in The Standard Edition of the Complete
Psychological Works of Sigmund Freud, Vol. 14. Copy held in the library of Human Development Scotland.
Highlighted and underlined: ‘Repression does not hinder the instinctual representative from continuing to exist
in the unconscious, from organising itself further, putting out derivatives and establishing connections’ (p.149).
Bowlby marginalia: ‘Very much my position.’
384 Bowlby, J. (1980) Loss. London: Pimlico: Information about experience ‘can be retained long enough out-
side consciousness in a temporary buffer store for it to influence judgement, autonomic responses and, I believe,
mood’ (49).
386 Bowlby’s concerns resembled closely those of Laplanche and Pontalis: ‘When operations as diverse as, say,
rationalisation, which brings complex intellectual mechanisms into play, and turning against the self, which is a
“vicissitude” of the instinctual aim, are attributed to a single function, and when the same term “defence” connotes
such a truly compulsive operation as “undoing what has been done” as well as the search for a form of “working-
off ” after the fashion of certain kinds of sublimation, then it may well be asked whether the concept in question is
a really operational one.’ Laplanche, J. & Pontalis, J.B. (1973) The Language of Psycho-Analysis, trans. D. Nicholson-
Smith. London: Karnac, p.109.
The meaning and structures of symptoms 69
these concerns in a short unpublished book from 1962 entitled Defences that Follow
Loss: Causation and Function. (‘Defences that Follow Loss’ recently appeared in Trauma and
Loss: Key Texts from the John Bowlby Archive, edited by Duschinsky and White.)387 In this
work, Bowlby disparaged the concept of ‘defence’ as a confused mix of description and ex-
planation. He felt that it also served to mask the important distinction between the inhibition
of a motivation and the inhibition of behaviour, and the distinction between the immediate
cause and the ultimate function of inhibition. All too often, Bowlby felt, psychoanalysts dis-
cussed ‘defences’ as if they were initiated in order to avoid some foreseeable consequence.388
For instance, psychoanalytic discussions of obsessive-compulsive symptoms at the time
tended to situate them as a defence against Oedipal conflicts. Klein argued that ‘obsessional
neurosis is an attempt to cure the psychotic conditions which underlie it’, which arise from
a predicament of experiencing both love and hate for the mother.389 Bowlby felt that such

claims failed to clarify the initiating conditions of obsessional symptoms, the contexts that
led them to become enduring, and the evolutionary function obsessive-compulsive behav-
iours may have served.
In failing to draw such distinctions Bowlby argued that Klein and other contemporary
psychoanalysts were unable to distinguish effectively between defensive strategies and strat-
egies for coping.390 Some forms of defence are evidently under our conscious control; others
indicate that conscious control is breaking down. Some forms of defence are helpful for in-
dividuals in responding to their environment; others are highly destructive. Yet psychoana-
lysts were making these distinctions remarkably rarely (with a few, inconsistent exceptions
like Donald Fairbairn and Joseph Sandler).391 In general, Bowlby felt that in psychoanalytic
theory ‘the relation of defence to healthy control, or to coping processes, has never been
clarified. Like Melanie Klein, most analysts hold the view that there are no great differences
between them.’392 The primary current of psychoanalytic thought directed attention away
from the question of which defences were able to contribute to individual coping, for in-
stance through offering short-term adaptation to an adverse environment for an individual.
To address this issue Bowlby appealed to the four questions of ethology: (i) the contribu-
tion a behaviour may have for species survival or reproduction; (ii) how the behaviour came
about in the course of natural selection; (iii) the behaviour’s mechanism or how it works; and
(iv) how it develops in the individual. To Bowlby, the processes and behaviours subsumed
under the term ‘defence’ likely have very different answers to these questions. However, what
they likely have in common, he argued, is that whilst at a species or population level they may
387 Duschinsky, R. & White, K. (eds) (2019) Trauma and Loss: Key Texts from the John Bowlby Archive.
London: Routledge.
389 Klein, M. (1932) The Psychoanalysis of Children. London: Hogarth, p.226.
390 The concept of coping strategies was just entering the academic literature in the early 1960s. E.g. Murphy,
L.B. (1960) Coping devices and defense mechanisms in relation to autonomous ego functions. Bulletin of the
Menninger Clinic, 24(3), 144–53. The concept would gain ground in academic psychology, until it detonated
into widespread use with Lazarus. R.S. & Folkman, S. (1984) Coping and adaptation. In W.D. Gentry (ed.) The
Handbook of Behavioral Medicine (pp.282–325). New York: Guilford.
391 Fairbairn, W.R.D. (1929, 1994) Dissociation and repression. In E. Fairbain & D.E. Scharff (eds) From Instinct
to Self: Selected Papers of W.R.D. Fairbairn (Vol. II). New York: Aronson Publishing; Sandler, J., Joffe, W.G., Baker,
S., & Burgner, M. (1965) Notes on obsessional manifestations in children. The Psychoanalytic Study of the Child,
20(1), 425–38. Bowlby’s annotations on and correspondence with Sandler are of special interested in this regard.
Bowlby was intrigued by the fact that Sandler was not always able to hold on to the distinction between cause and
function even despite his best efforts. See Bowlby, J. (1965) Comments on Joffe and Sandler 1965 ‘Notes on Pain,
Depression and Individuation’. PP/Bow/J.9/168-9.
70 John Bowlby
help at times to contribute to survival or reproduction, for a given individual they may be
baffling and sometimes counterproductive: ‘defensive processes come into action in certain
conditions and that they do so without the individual having any more idea of their bio-
logical function than the ordinary man has of the function served by his temperature rising
when he contracts an infection’.393 Later, in Loss, Bowlby urged that ‘the effects of defensive
activity must be judged on a number of distinct scales. For example we can ask: what are its
effects, beneficial or otherwise, on the personality concerned? what are its effects, beneficial
or otherwise, on the members of the person’s family? what are its effects, beneficial or other-
wise, on the community at large?’394 However, Bowlby felt that the concepts available for
even asking these questions were overloaded and confused in psychoanalytic discourse. This
prompted his decision to introduce the concept of segregated systems.

Segregation
Another problem that Bowlby had with the concept of defence as used by the psychoanalytic
community of his day was that it seemed to him oriented by the basic image of withdrawal
from an aversive source of excitation. However, whereas in some ways the concept of defence
was too absorptive, in this regard Bowlby found it too restrictive. Picking apart the concept,
Bowlby narrowed in on the idea of the inhibition of behavioural systems.395 In Defences that
Follow Loss: Causation and Function in 1962, he introduced for the first time the concept of
‘segregated systems’. This concept would be pulling the strings as the governing concept of
Bowlby’s later thought, but in fact was only introduced in print 18 years later in Loss, in a
discussion so brief that subsequent researchers have generally not found it clear or usable.396
Defences that Follow Loss, by contrast, engaged in an extended and elaborate discussion of
the concept. In this book Bowlby argued that behavioural systems are ‘segregated’ if there
is reduced, distorted, or blocked communication with perception, memory, and/or other
behavioural systems. Though there was, intentionally, some overlap between Bowlby’s term
‘segregation’ and Freud’s concept of ‘repression’, Bowlby preferred the former term since he
felt that ‘repression’ had accrued too much baggage, such as implying processes that keep
something unconscious. By contrast, segregation could occur between two conscious sys-
tems, or between a conscious system and memory or perception.397 And whereas for Freud,
393 Ibid.
395 Bowlby, J. (1961) Letter to Dr Robert Hinde, 24 January 1961. PP/Bow/H224: ‘The whole psychoanalytic no-
tion of defence is confused and I want to explore the notion that at the time of onset of what is later called repres-
sion one motivational system is blocked of expression and another, incompatible with it, evoked: both continue
active but out of communication with each other. In the human, one of them is likely to be more accessible to con-
sciousness than the other . . . Supposing this is a correct picture of events, the problem now becomes to define the
conditions that give rise to this state of affairs and their mode of operation.’
396 Among the few attachment researchers to have applied the concept are Bretherton, and Solomon and
George. It is revealing about the underdeveloped status of the concept in Bowlby’s published writings that their
treatments differ vastly from one another, and that Bretherton expressed hesitancy as to whether she fully grasped
Bowlby’s meaning. See Bretherton, I. (2005) In pursuit of the internal working model construct and its relevance
to attachment relationships. In K.E. Grossmann, K. Grossmann, & E. Waters (eds) Attachment from Infancy to
Adulthood: The Major Longitudinal Studies (pp.13–47). New York: Guilford; Solomon, J. & George, C. (2011)
Disorganization of maternal caregiving across two generations. In J. Solomon & C. George (eds) Disorganized
Attachment & Caregiving (pp. 25–51). New York: Guilford.
397 Bowlby, J. (1962) Defences that Follow Loss: Causation and Function. PP/Bow/D.3/78; cf. Bowlby, J. (1956)
Annotations on Charles Brenner ‘The nature and development of the concept of repression in Freud’s writings’. PP/
BOW/J.9/33.
repression worked against the dynamic unconscious as a ‘cauldron full of seething excita-
tions’,398 for Bowlby segregation was an abstraction used to describe any process that re-
sulted in a blockage or inhibition of communication within or between behavioural systems.
It might be thought that segregation would be regarded by Bowlby as a bad thing. And
this has been the impression of Bowlby’s readers who have access to only his published dis-
cussions of segregation: an apparent denigration of segregation has been part of Bowlby’s
legacy to later researchers and clinicians, among the very few who have engaged with the
concept.399 Bailey and colleagues, for instance, seem to presume that all segregation must
be an effect of trauma.400 However, such denigration ignores the potential for helpful segre-
gation, through processes that buffer or segment and creatively recompose what is taken in
from outside. In fact, in Defences that Follow Loss, Bowlby argued in some detail that long-
term mental health would be supported by effective communication between mental sys-

tems on the basis of relative and flexible forms of segregation, rather than those that were
strictly held. He was attentive to the role of segregation within typical development as well
as atypical development, and its contribution to mental health, as well as mental ill health.401
He was specifically interested in ‘the functions that in health the segregation processes serve’
and gave careful consideration to the idea that segregation should be regarded ‘in the same
light that Claude Bernard taught us to view physiological illness, namely as the outcome of
processes that are beneficial in kind but faulty in amount’.402
Bowlby used the term ‘selective exclusion’ to refer to strategies of behavioural and at-
tentional aversion or withdrawal that enact a flexible and relatively minimal segregation.
‘Selective exclusion’ may contribute to the mutual enrichment of behavioural systems.
Bowlby gives the example of engrossment: selective exclusion of other thoughts, distrac-
tions, and daydreams may insulate and protect the effective operation of the exploratory
system. Other cases are easy to imagine. For instance, selective exclusion can help preserve
sources of private and sustaining joy, warped and wonderful against the flattening bustle of
the world. Or to take another example, selective exclusion may be helpfully deployed to keep
worries away during relaxation or sleep. The direction and quality of attention would need to
be flexible enough to change once work began again. Where this can be achieved, communi-
cation between systems ensures that benefits of physical and attentional rest are transferred
in the form of feeling genuinely refreshed.
Segregation essentially means that a system is not accepting information from another
source, or potentially from any source. Sometimes, as in the case of selective exclusion, this
398 Freud, S. (1933, 2001) New introductory lectures on psychoanalysis. In The Standard Edition of the Complete
Psychological Works of Sigmund Freud, Vol. 22 (pp.1–182). London: Vintage, p.73. In Loss, Bowlby acknowledged
some parallels between his concept of segregation and Freud’s concept of ‘splitting’, though he felt that the latter
term had accrued too much baggage to offer much clarity. ‘Splitting’ would later feature within Crittenden’s re-
interpretation of Bowlby’s information processing model: Crittenden, P.M. (2016) Raising Parents: Attachment,
Representation, and Treatment (2nd edn). London: Routledge.
399 See e.g. Lemma, A., Target, M., & Fonagy, P. (2011) Brief Dynamic Interpersonal Therapy: A Clinician’s Guide.
Oxford: Oxford University Press: ‘Although there is never a direct correspondence between external and internal,
as what is internal reflects the operation of defensive processes that distort what is taken in from the outside’ (95).
400 Bailey, H.N., Redden, E., Pederson, D.R., & Moran, G. (2016) Parental disavowal of relationship difficul-
ties fosters the development of insecure attachment. Revue Canadienne des Sciences du Comportement, 48(1),
49–59, p.50.
401 In this regard, Defences that Follow Loss is a forerunner of the priorities and concerns of developmental psy-
chopathology (Chapter 4).

402 Bowlby, J. (1962) Defences that Follow Loss: Causation and Function. PP/Bow/D.3/78. On Bernard see Gross,
C.G. (1998) Claude Bernard and the constancy of the internal environment. The Neuroscientist, 4(5), 380–85.
72 John Bowlby
is a generally beneficial filtering process. However, segregation can also occur because other-
wise useful information is experienced as difficult to accept or incompatible with currently
held values.403 This is likely the meaning of Bowlby’s claim, cited earlier in the ‘Introduction’,
that the conditions of lack of integration between aspects of the self rest on conditions of
forgivability or unforgivability: the critical question for the segregation or desegregation of
systems is whether they are willing to accept information from one another, from memory
and from the world, recognizing it as content available for inclusion within the system. If the
information is regarded as unacceptable or dangerous, it may not be reconciled, causing seg-
regation of the behavioural system.
Defensive exclusion

Bowlby was not concerned about segregation and the incompatibility of behavioural systems
so long as ‘when they conflict, as habitually they do, regulation is tolerably smooth and effi-
cient’.404 Some segregation is an inevitable part of being human, and localised and controlled
incompatibility can provide a foundation of imagination, creativity, and work–life balance.
By contrast, Bowlby felt that rigid and intense segregation can lead to very significant prob-
lems.405 One fundamental consequence of rigid and/or intense segregation is that the system
then does not have access to new perceptions, or to reciprocal development with other
behavioural systems. Not only does the cognitive component of a segregated behavioural
system not receive the benefits of the learning provided by the exploratory system, or con-
scious reflection, but also the opportunities for feedback and revision are blocked that would
usually occur when a behavioural system is expressed.406 Without such feedback, behav-
ioural systems cannot fully benefit from the opportunities and lessons of the environment.
When the behavioural system does get expressed, the result may be too strong or extreme, or
clumsily formulated or ineffective.407
Furthermore, in Bowlby’s account, behavioural systems become active only when
their activating conditions are met. As a consequence, to the degree that a behavioural
system is segregated from information about these conditions, the system will only be
partially activated or not activated at all. When strategies of behavioural and attentional
aversion or withdrawal are used to inhibit memories that would otherwise activate a
403 Main would later criticize Bowlby on this point, suggesting that he placed too much emphasis on informa-
tion that was difficult to accept because integration would be painful, and not enough emphasis on information
that was difficult to accept simply as a result of human cognitive biases and developmental processes, such as
the difficulties of a three year old in holding in mind contradictory qualities of a single person. Main, M. (1991)
Metacognitive knowledge, metacognitive monitoring, and singular (coherent) vs. multiple (incoherent) models of
attachment: some findings and some directions for future research. In P. Marris, J. Stevenson-Hinde, & C. Parkes
(eds) Attachment Across the Life Cycle (pp.127–59). New York: Routledge, p.138–9.
405 Indeed, the strength of his emphasis on segregation as a problem in his book on Loss would begin a diver-
gence between two rather different trends in subsequent attachment theory: one that focused on whether internal
working models were positive, and one that focused on whether behavioural systems were flexibly used and open
to revision. The first to call attention to this ambiguity was van IJzendoorn, M.H., Tavecchio, L.W.C., Goossens,
F.A., & Vergeer, M.M. (1982) Opvoeden in geborgenheid: Een kritische analyse van Bowlby’s attachmenttheorie.
Amsterdam: Van Loghum Slaterus, pp.61–2.
406 Bowlby, J. (1964) Segregation of psychic systems. PP/Bow/H10. See also Reisz, S., Duschinsky, R., & Siegel,
D.J. (2018) Disorganized attachment and defense: exploring John Bowlby’s unpublished reflections. Attachment &
Human Development, 20(2), 107–34.
Table 1.2 Selective exclusion of motivation and levels of motivation
Motivation Exclusion
Level 1: Low Slight, e.g. daydreaming, free-association

Level 2: Moderate Moderate. Some degree of concentration but still
possible to attend to other Input
Level 3: High Considerable. Strong concentration and resulting
in exclusion of all other input
Level 4: Very high (or persistence of high) Erratic. Whilst exclusion continues on a
considerable scale, there is also a tendency for
responses to a wider class of objects than in other
conditions of motivation (cf Hinde)

Source: Theory of Defence, JB notes, 1960–63, PP/Bow/H10.
behavioural system, Bowlby referred to ‘cognitive disconnection’.408 The main effect, he

proposed, was that strong emotions would seem to appear out of nowhere or be associ-
ated with inappropriate objects, since they have been disconnected from their original,
provoking source. If cognitive disconnection is the segregation of memory, Bowlby
offered the term ‘defensive exclusion’ to refer to the segregation of external perception.
Cognitive disconnection keeps memory, reflection, and other cognitive processes from
causing the activation of a behavioural system. Defensive exclusion keeps new experi-
ences from achieving this same consequence. Both are dynamic properties that block
relevant information from having implications for a behavioural system or systems. In
turn, ‘segregation’ is the structure that such processes create or hold in place, a state in
which information or its meaning is blocked, thinned, or distorted in such a way as to
shape the activating or terminating conditions of a behavioural system. As Crittenden
and Bretherton both observed, the exact nature of what is being blocked, thinned,
or distorted, and by what agency, is not clear in Bowlby’s account of segregated sys-
tems.409 It would seem that segregation can be maintained by various processes. Among
these, however, Bowlby gave defensive exclusion special importance. External percep-
tion offers the potential to gain from the opportunities and lessons of the environment.
Without this, segregation becomes rigid, a situation with particular implications for
mental health.
In unpublished notes from the time of Defences that Follow Loss: Causation and Function,
Bowlby created a table to distinguish between four levels of exclusion (Table 1.2).410
Level 1 is exclusion with only the barest drop of defence. It may, for instance, be little
more than engrossment in something novel and/or complex, where the alternative would
be rumination. Or it might be thinking about food or work when we might otherwise find
408 Ibid. p.66.

409 Crittenden, P.M. (1997) Truth, error, omission, distortion, and deception: the application of attachment
theory to the assessment and treatment of psychological disorder. In S.M. Clancy Dollinger & L.F. DiLalla (eds)
Assessment and Intervention Issues Across the Lifespan (pp.35–76). London: Lawrence Erlbaum; Bretherton, I.
(2005) In pursuit of the internal working model construct and its relevance to attachment relationships. In K.E.
Grossmann, K. Grossmann, & E. Waters (eds) Attachment from Infancy to Adulthood: The Major Longitudinal
Studies (pp.13–47). New York: Guilford.
410 Bowlby, J. (1960) Theory of Defence. JB notes, 1960-63. PP/Bow/H10.
74 John Bowlby
ourselves getting upset, bringing one behavioural system partly online in order to exclude
another a little. Or exclusion with a drop of defence might be identified in the absorption fa-
cilitative of endurance during exercise or sport. What Bowlby termed ‘the ordinary everyday
things of life’,411 its familiar, conventionalised rhythms and language, can also be character-
ised as Level 1 defensive exclusion. They keep us connected to others but without demanding
vulnerability or deeper availability to others—or to ourselves. ‘How are you?’ ‘Fine thanks,
how are you?’ Such are the ‘outer rings’ of social stabilisation, Bowlby argued, that keep indi-
vidual behavioural systems, such as the attachment system, from needing to be activated.412
Level 2 is described by Bowlby as ‘moderate’ exclusion. It is ‘still possible to attend to other
input’. There is still some allowance, at least, for untidy, troubling experiences. Laugh-or-you’ll-
cry humour might be offered as an illustration of this level of defensive exclusion, since the
distress is walled off yet, if incipiently, personally and socially acknowledged. In Level 3 there

is ‘exclusion of all other input’. The information permitted through the filter is rigorously and
rigidly policed. Finally, at Level 4, ‘whilst exclusion continues on a considerable scale, there is
also a tendency for responses [tobe made] to a wider class of objects than in other conditions of
motivation’. The exclusion becomes extended beyond its original objects, for instance excluding
not just perception but whole related classes of emotions as well. An example of this exclusion
of a ‘wider class of objects’ is provided by a patient described by Bowlby in the chapter entitled
‘Psychoanalysis as Art and Science’, who until she was ten years old had ‘been terrified of another
separation; but then she had “switched off ” her anxiety “like a tap”, as she put it, and with the
anxiety had disappeared most of her emotional life as well.’413
In Defences that Follow Loss: Causation and Function, Bowlby theorised that defensive ex-
clusion and cognitive disconnection enact a weakening of the integration between or within
behavioural systems. Defensive exclusion and cognitive disconnection may, then, contribute
to coping when the alternative might otherwise be a greater or more enduring disruption to
the system, or to the psychological apparatus that underpins and orchestrates behavioural
responses more generally. This is why children are especially vulnerable to using segrega-
tive processes, because their psychological apparatus is still immature and less resilient to
disruption. Defensive exclusion and cognitive disconnection, then, permit a certain kind of
resilience in the face of disintegrative threats precisely by accepting some determinate and
limited degree of segregation.
On this basis, Bowlby distinguished avoidance and dissociation as different intensities and
kinds of defensive exclusion. Some kinds of avoidant response may involve little segregation,
just perhaps a turning of attention away from a person, situation, or thing that might other-
wise activate a behavioural system.414 There is some loss of information from reality, but not
411 Bowlby, J. (1940) The influence of early environment in the development of neurosis and neurotic character.
International Journal of Psycho-Analysis, 21, 154–78, p.173.

412 Bowlby, J. (1973) Separation: Anxiety and Anger. New York: Basic Books: ‘the regulatory systems that
maintain a steady relationship between an individual and his familiar environment can be regarded as an “outer
ring” of life-maintaining systems complementary to the “inner ring” of systems that maintain physiological
homeostasis’ (180).
413 Bowlby, J. (1978, 1988) Psychoanalysis as art and science. In A Secure Base (pp.43–64).
London: Routledge, p.58.

414 Bowlby, J. (1960) Theory of Defence. JB notes, 1960-63. PP/Bow/H10: ‘Hypothetical function of narrowing
of attention. Narrowing of attention is achieved by reducing input, but sensory and cognitive. The advantages
are: a) to cut-out irrelevant and confusing input; b) to cut-out input that in fact evokes other motivational sys-
tems because such system would distract the organism from the task in hand by reducing relevant input; c) to
cut-out input that might require abandonment of current organisation of a motivational system, and consequently
reorganisation with its attendant layer of inability to act—disorganisation.’ Some commentators have assumed
a loss of contact with reality. For instance, we may tune out small occurrences that might
otherwise generate impatience or irritation, to keep the aggression system from coming on-
line and disrupting our overall plans. This would be Level 1 in Bowlby’s taxonomy. It is also
possible to have more thorough-going segregation with avoidance, as for instance when ac-
cess to anger is muted—the threshold for activating the aggression system is raised—for an
individual who was made to feel endangered when they displayed anger as a child. Muted
access to anger would be Level 2; if anger is essentially unreachable, this would be Level 3.415
In the 1950s Robertson observed that withdrawn and avoidant behaviours shown by hos-
pitalised children were often accompanied by disorientated behaviours. However, Bowlby
treated dissociation as, in contrast to avoidance, a more extreme form of segregation—more
of an emergency break.416 At a population level, forms of defensive exclusion such as dissoci-
ation may have evolved as a way to shut off and thereby protect behavioural systems or other

psychological processes from extreme perceptions or memories that might otherwise risk a
greater disruption. Instead of segregating certain information from a behavioural system,
perception in general is segregated from other mental processes. He speculated that it would
specifically be episodic information about occurrences within time and place that would be
lost to dissociative processes; general attitudes might well remain unaffected.417 Bowlby pro-
posed that such a process will produce responses that, unlike avoidance, are not responsive
to the environment. This produces the phenomenon of ‘coming to’ following dissociation.418
The generalisation of segregation to all sense perceptions for a time suggests that dissoci-
ation would be considered by Bowlby as one kind of Level 4 defensive exclusion.
Displays of conflict
Yet defensive exclusion is not always successful. Bowlby found parallels between psychoana-
lytic theory and the work of ethologists, who had observed odd behaviour by animals ex-
periencing a conflict of motivations, for instance when cued both to approach and to flee.419
that all ‘detached’ behaviours observed by Robertson represented dissociative phenomena, e.g. Barach, P. (1991)
Multiple personality disorder as an attachment disorder. Dissociation, 4, 117–23. Whilst avoidance may perhaps
have relevant prospective links with dissociation (Chapter 4), Bowlby’s thinking about different kinds of defensive
exclusion acknowledges their differences.
415 A remark by Ursula Bowlby offers an illustration. She described her husband as, for biographical reasons,
incapable of feeling the emotion of fear. She considered this an aspect of Bowlby’s general inhibition of nega-
tive feelings, and a contributing factor to his indomitable courage as a public intellectual and theoretician. Again,
this would be Level 3 defensive exclusion. Ursula Bowlby interview with Robert Keren, cited in Karen, R. (1998)
Becoming Attached: First Relationships and how they Shape our Capacity to Love. Oxford: Oxford University
Press, p.469.
416 Bowlby, J. (1962) Defences that Follow Loss: Causation and Function. PP/Bow/D.3/78: ‘I am introducing the
generic term “to segregate” and “segregated process”; they denote any process that creates barriers to communica-
tion and interaction between one psychic system and another . . . Other additional terms are required for the many
other particular sorts of segregating process.’ Handwritten marginal note: ‘dissociation’. See also Bowlby, J. (1961)
Processes of mourning. International Journal of Psycho-Analysis, 42, 317–40, p.336.
417 Bowlby, J. (1982) Outline of attachment theory. PP/Bow/H.260.
418 E.g. Bowlby J., Robertson, J., & Rosenbluth, D. (1952) A two-year-old goes to hospital. Psychoanalytic Study
of the Child 7, 82–94: ‘During the hospital experience these splits were relatively brief; after a few minutes of blank-
ness she “came to” and responded to her real mother . . . This, perhaps, helps us to understand how during longer
experiences of separation this split can develop to a point where integration on reunion with the mother is no
longer automatic and the child is unable to link his need for a good mother and his hatred of a frustrating one to an
individual woman’ (86).
Animal Models in Human Psychobiology (pp.27–36). New York: Plenum Press: Ethology and psychoanalysis fit
76 John Bowlby
Bowlby was fascinated that Hinde had discerned in his study of chaffinches that ‘paradox-
ically, strangeness evokes both escape and curiosity, and that there is a complex balance be-
tween the two competing response systems’.420 Likewise, Hinde observed that the following
response show by young birds towards their parents may be disrupted or discoordinated
where this competes with another response system, for instance if the parent is also in some
way a cue for danger, not uncommon given ‘the broad range of stimuli eliciting both fol-
lowing and fleeing at this age’.421 In such situations, animals show ‘conflict behaviours’ such
as the rapid transition between one tendency and the other, poorly coordinated combin-
ations, freezing in place, misdirected movements, or signs of confusion or tension.422 Bowlby
also considered that anger might be evoked by conflict between two behavioural tendencies,
presumably since conflict can obstruct the achievement of one or both tendencies, contrib-
uting to frustration.423

Bowlby’s interest in conflict behaviour was likely primed by having observed several kinds
of conflict response in his clinical work. For instance, as an army psychiatrist in World War
II, working with veterans on their return from the front, he observed amnesias, loss of bodily
control, misdirected and undirected behaviours, out of context anger or anxiety, signs of
confusion, and tic-like behaviours or other signs of tension. Bowlby interpreted these behav-
iours as reflecting veterans’ experiences of psychological conflict between feelings of duty
and lingering fear from their time at the front.424 Robertson also reported many of the same
behaviours in his description of children during hospitalisation and on their return home.
There were also discussions of conflict behaviour in the psychanalytic community. For in-
stance, in 1956, Anna Freud described the case of a 13-month-old infant whose behaviour
towards her mother was disrupted by screaming and states of withdrawal. Since these behav-
iours were only shown in the relationship with the mother, Freud anticipated that they sug-
gested ‘some traumatic event’ in the history of this relationship, a hypothesis that was later
confirmed.425
Bowlby came to believe that his concept of segregated systems could account for some
forms of conflict behaviour, including when displayed towards attachment figures. On the
one hand, conflict could occur between a behavioural system and its defensive exclusion,
especially when information from perception or memory for the activation of a behavioural
so well together because ‘both were interested in the effect of early experience on later development’ and ‘in
conflict arising from social situations’ (28). As well as Robert Hinde, another important influence on Bowlby’s
reflections on conflict behaviour was von Holst, E. & von Saint Paul, V. (1963) On the functional organisation
of drives. Animal Behaviour, 11, 1–20. They detail seven kinds of response to the conflict of behavioural re-
sponses: display of one toned by the other; averaging of the responses; alternation between the two; each cancel-
ling the other out; the production of a third rather different kind of response (e.g. pecking and feeing becomes
threat screeching); the dominance of one which masks weak expressions of the other; display of only one or
the other.
421 Hinde, R.A. (1961) The establishment of the parent– offspring relation in birds, with some mamma-
lian analogies. In W.H. Thorpe & O.L. Zangwill (eds) Current Problems in Animal Behaviour (pp.175–93).
Cambridge: Cambridge University Press, p.185. Passage underlined as of particular note in Bowlby’s personal copy
of the text. PP/Bow/H.226.
422 Hinde, R. (1970) Animal Behavior, 2nd edn. New York: McGraw-Hill.
423 Bowlby, J. (1965) Motivation. PP/Bow/H.128.
424 Bowlby, J. & Soddy, K. (1940) War neurosis memorandum. PP/Bow/C.5/1; Bowlby, J. (1942) Selection and
diagnosis in Army: notes for a talk. PP/BOW/C.5/2/3.

425 Freud, A. (1956, 1969) The assessment of borderline cases. In The Writings of Anna Freud, Vol. 5 (pp.301–14).
New York: International Universities Press, p.310.

system is particularly salient. For instance, a child who is hurt may seek to exclude distress-
ing information that might otherwise trigger the attachment system, if they have learnt that
their caregiver will likely punish them for being upset or displaying a desire for comfort.
They may ignore their injury, pretend it does not cause them distress, or keep their attention
away from information about their caregiver’s whereabouts to avoid the desire to seek the
caregiver. However, this exclusion will become increasingly precarious to the degree that the
child’s injury is especially painful, is experienced as upsetting, or the caregiver is on hand.
The result will be conflict between the attachment system and its inhibition, an approach–
avoidance conflict, which may be visible in the child’s behaviour.
On the other hand, Bowlby was also interested in the forms of conflict that could arise
when two behavioural systems were active at the same time. Often, of course, different be-
havioural systems are easily compatible. This is why, as Ainsworth repeatedly argued, it is

a mistake to study behavioural systems in isolation: ‘In any situation, the extent to which
attachment behaviour occurs depends not only on the strength to which it is activated, but
also on the strength of activation of other competing or compatible behavioural systems.’426
A favourite example for both Bowlby and Ainsworth was that the fear system and the attach-
ment system are often easily reconciled and integrated, in directing an infant away from a
perceived threat and towards a haven of safety.427 Sometimes behavioural systems, whilst
somewhat different in their demands, can be combined: for instance where courtship sys-
tems and wariness combine to produce flirting behaviour, or the anger and caregiving sys-
tems combine to produce harsh forms of discipline.
In other cases, however, Bowlby argued that behavioural systems may come into signifi-
cant conflict, especially if there has been segregation of one or both, hindering communica-
tion and effective compromise.428 For example, conflicts may arise between the attachment
and caregiving systems. The injured child discussed above may have caring responsibilities
for a sick parent. This may be one of the reasons why the child feels that their attachment
system should be inhibited where possible, since they have learnt that their parent is most
responsive when they seem happy and available to meet the parent’s wishes.429 In such a case,
however, pain or distress associated with the injury would come into conflict with the child’s
caregiving system. The child may attempt to achieve the ends of both systems at once. The
two systems do have some compatible behaviour, like physical approach to the caregiver. If
the child attempts to maintain his caregiving role, this may be undermined or interrupted
by distress from the injury. If the child lets the parent know about his injury and conveys his
426 Ainsworth, M.D.S. (1977) Attachment theory and its utility in cross-cultural research. In: P.H. Leiderman,
S.R. Tulkin, & A. Rosenfeld (eds) Culture and Infancy: Variations in the Human Experience (pp. 49–67).
New York: Academic Press, p.59.
427 Bretherton, I. & Ainsworth, M.D.S. (1974) Responses of one-year-olds to a stranger in a strange situation. In
M. Lewis & L.A. Rosenblum (eds) The Origin of Fear (pp. 131–64). New York: Wiley.
428 This is discussed in the section ‘Incompatible behavioural systems: results of simultaneous activation’ of
Bowlby, J. (1969, 1982) Attachment, Volume 1. London: Penguin, pp.97–101.

429 Bowlby, J. (1986) Transcript of interview by Lange. Wellcome Trust Library Archive. PP/ BOW/A.5/16:
‘Anxiously avoidant children who have a parent who tends towards rejection . . . They are really caught in a classic
approach–avoidance conflict. On the one hand they really would love to have affection and support and comfort.
They would love to have that sort of relationship. But their attempts to develop such a relationship were met with
so many painful rejections in the past that they dare not attempt it again. And they pretend, sometimes convince
themselves and convince everyone else, that they get along very well, thank you, without any of that sentimental
nonsense . . . They are incredibly out of touch with their feelings and out of touch with the situations which evoke
their feelings.’
78 John Bowlby
feelings of distress, this may nonetheless be inflected, even distorted, by the child’s habitual
caregiving response and relative exclusion of the attachment system.
As a general principle, Bowlby suggested that ‘whenever a system that has been deacti-
vated becomes in some degree active, such behaviour as is then shown is likely to be ill-
organised and dysfunctional’.430 The smooth resolution of conflicts depends upon systems
being able to communicate and compromise. When one behavioural system is active at the
expense of a conflicting one that has been segregated, trapdoors are to be found inside our
intentions and actions:
Not only are information and motor response relevant to any one goal narrowly restricted
but information and motor responses relevant to some other and perhaps incompatible
goal may be allowed through. It is as though an enquiry clerk, when asked about trains

to Cornwall, gave information endlessly about the night express to Plymouth, with occa-
sional intrusions about a plane to Rome.431
Psychoanalytic theory proposed that conflict between incompatible psychological demands

is part of being human. However, the strength of these demands, the extent of their incom-
patibility, and other challenges may turn such conflict into a source of symptoms, in which
one or both of these demands gets expressed in a distorted form. For instance, in a pas-
sage highlighted by Bowlby in his personal copy, Karl Abraham suggested that a conflict
between the sexual response and its suppression could find outlet in the intensification of
hunger: ‘The great frequency of hunger attacks in frigid women is very striking . . . Strong li-
bidinal impulses, against the undisguised appearance of which consciousness protects itself,
can be unusually well masked by a feeling of hunger. For hunger is a sensation that can be
admitted to oneself and to others.’432
Bowlby proposed that conflict behaviours would become stabilised as symptoms of
mental ill health to the extent that the situation that evoked the conflict was stable, and
especially when it related to a close and enduring relationship. Signs of ‘tension, anxiety
and depression’ may be especially expected to the degree that contradictory responses (or
conflict between a response and its defensive exclusion) are felt towards an attachment
figure.433 When access to attachment figures is uncertain, as may occur in conflict situ-
ations, this may evoke tension and anxiety. Conflict situations, when they are sustained,
also block the satisfaction of at least one behavioural system (and perhaps more). This
predicament can contribute to the onset and/or maintenance of depression by provoking
a sense that the achievement of the set-goal is impossible, no matter what strategy is
deployed.434

432 Annotations by Bowlby on Abraham, K. (1927) Selected Papers. London: Hogarth Press, p.263; not dated
but handwriting and reference on p.74 suggest annotations are from 1933. Copy held in the library of Human
Development Scotland.
WTO Study Group of the Psychobiological Development of the Child, Vol. 1. London: Tavistock, p.186.
434 Bowlby, J. (1974) Psychological processes evoked by a major psychosocial transition. Presented to Tavistock
Research Workshop, March 1974. PP/Bow/F.3/90; Bowlby, J. (1989) Foreword to Emmy Gut’s Productive and
Unproductive Depression (pp.xiii–xv), London: Routledge.
Bowlby anticipated that encounters with attachment figures that evoke conflict are
powerful sources of anxiety and depression. He identified several reasons for this. Part of the
importance of the attachment system in this regard is that it has implications for the individ-
ual’s basic sense of being intrinsically worthy, acceptable, and capable of being cherished. It
therefore has a developmental role in informing the components of other behavioural sys-
tems. Another reason for the important role of conflict in relation to attachment figures in
prompting anxiety and depression is that our attachment figures are often socially difficult
to avoid, not only but especially in childhood. In the chapter ‘The making and breaking of
affectional bonds’, Bowlby speculated that the intensity of emotion in close relationships
may serve especially to activate early behavioural, cognitive, and affective components of
behavioural systems, so that forgotten wishes and disappointments from childhood be-
come incorporated into present-day behaviour and expectations.435 This was one aspect of

what Bowlby termed the ‘risks of intimacy’.436 It is in good part in response to these risks
that closeness gets so thoroughly modulated and ritualised in everyday family life. This re-
duces occasions that might risk disturbing the others’ composure by activating early and
raw components of a behavioural system rather than the more mature and well-conditioned
components.
Bowlby’s interest in the contribution of psychological conflict to depression and anx-
iety made him enthusiastic about the work of Aaron Beck, and the emergent paradigm of
cognitive behavioural therapy. In Bowlby’s terms, to the degree that thoughts such as hope-
lessness in relationships, personal responsibility, and the unacceptability of the self become
segregated from feedback, they would in turn reinforce depression and anxiety.437 In print,
Bowlby stated that his model of depression and anxiety was ‘cast in the same mould’ as that
of Beck, and that ‘his theory is compatible with mine’.438 He was especially impressed by the
fact that Beck appeared to distinguish between two forms of depression—one anxious and
needy, the other avoidant and with reduced affect—that corresponded to the two classes of
response to separation that he and Robertson had identified.439 However, Bowlby felt that,
more than Beck, attachment theory highlighted the importance of early experiences in con-
tributing to psychological conflicts. In particular, the theory highlighted the relationship ex-
pectations and forms of defensive exclusion that could perpetuate conflicts even after the
original situation evoking conflict had ceased.440
Affectional Bonds (pp.150–88). London: Routledge: ‘The stronger the emotions aroused in a relationship the more
likely are the earlier and less conscious models to become dominant’ (168).
436 Bowlby, J. (1985) Letter to John Byng-Hall, 12 April 1985. PP/Bow/J.9/45: ‘You need to clarify what the risks of
intimacy are. One risk, which may motivate either child or mother, is fear of rejection; another risk, which may also
motivate either partner, is fear of being held captive by the intense attachment behaviour of the other.’
437 This point is elaborated with case studies in Bowlby, J. (1980) The place of defensive exclusion in depressive
disorders. PP/Bow/K.7/94; see also Bowlby, J. (1987) Notes on depression, towards correspondence with Emmy
Gut. PP/Bow/B.3/15.
438 Bowlby, J. (1980) Loss. London: Pimlico, p.249–50. For a recent discussion see Bosmans, G. (2016) Cognitive
behaviour therapy for children and adolescents: can attachment theory contribute to its efficacy? Clinical Child
and Family Psychology Review, 19(4), 310–28.
439 Bowlby, J. (1981) Letter to Aaron Beck, 8 October 1981. PP/Bow/J.9/16. See Beck, A. (1983) Cognitive therapy
of depression: new perspectives. In P.J. Clayton & J.E. Barrett (eds) Treatment of Depression: Old Controversies and
New Approaches (pp.265–90). New York: Raven.
159–71, p.167.
80 John Bowlby
In the clinic
Mrs Q
In his last article, written with Mary Ainsworth, Bowlby highlighted especially his work as a clin-
ician with children and families, and additionally the decades in which he ran a mother’s group
in a wellbaby clinic, ‘learning much from his informal observations of mother–child interaction
there’.441 Bowlby was mindful that his theory addressed general tendencies at the level of popula-
tions, and had only a probabilistic relationship with any individual, whereas his clinical practice
addressed the concrete dynamics of individual lives, each with their particular vitality and equi-

librium.442 In a sense, Bowlby observed in private notes, the clinician experiences the complex
human being in front of his as like a work of art: ‘The criteria of a work of art are a) sensory and
formal vitality (tension, life); and 2) order (balance, equilibrium, and symmetry) . . . What is true
of a work of art is also true of a human personality.’443 Bowlby emphasised the need to distin-
guish between a clinical understanding of tension and order in the lives of individuals and the
generalisations of diagnostic or research frameworks, and to establish a productive relationship
between the two:
In the past there has been a deplorable tendency for the experimentalist to despise the clini-
cian’s lack of precision and the clinician to reciprocate with contempt for the experimentalist’s
lack of insight into human nature. Each has stoutly maintained that his own method was the
one true way to knowledge. These claims are absurd: each method is indispensable. It is the
clinician who usually has the earliest insights, defines the problem, and formulates the first
hypotheses. But [through] the detailed minute study of the feelings and motivations of his pa-
tients, and the complicated intellectual and emotional repercussions to which they give rise,
the clinical worker provides information regarding the relations of psychic and environmental
forces which can be obtained in no other way.444
During Bowlby’s lifetime, there was already a good deal of discussion of the clinical impli-
cations of attachment theory.445 Since then, commentators have described an ‘explosion’ of
texts offering exposition of Bowlby’s ideas for practitioner audiences.446 These works for
441 Ainsworth, M. & Bowlby, J. (1991) 1989 APA award recipient addresses: an ethological approach to person-
ality development. American Psychologist, 46(4), 333–41, p.336. The Bowlby Archive contains a wonderful de-
scription of the clinic, entitled ‘Mothers Discussion Group’ from 1967 (PP/Bow/C.6/3). In justifying the benefits
of the group Bowlby wrote: ‘It is from other mothers that a beginner can learn most. Professional people can add
information about different sorts of behaviour and the range of ages at which different developments are likely to
occur—but how to cope best is learned from others who are actually confronted with the job.’
243–56, p.253.
443 Bowlby, J. (1957–59) Untitled notes responding to the work of John Alford. PP/Bow/B.3/19.
444 Bowlby, J. (1951) Maternal Care and Mental Health. Geneva: World Health Organisation, p.61.
445 E.g. Belsky, J. & Nezworski, T. (eds) (1988) Clinical Implications of Attachment. Hillsdale, NJ: Lawrence
Erlbaum; West, M., Sheldon, A., & Reiffer, L. (1989) Attachment theory and brief psychotherapy: applying current
research to clinical interventions. Canadian Journal of Psychiatry, 34, 369–75; Byng-Hall, J. (1990) Attachment
theory and family therapy: a clinical view. Infant Mental Health Journal, 11, 228–37.
446 An important recent contribution to and review of this literature is Slade, A. & Holmes, J. (2017) Attachment in
Therapeutic Practice. London: SAGE. For a thematic review, calling for a shift in emphasis from theory-to research-
based guidance for attachment-inspired therapies see Berry, K. & Danquah, A. (2016) Attachment-informed
In the clinic 81
practitioner audiences offer both exposition and revision of Bowlby’s ideas.447 However,
close readings of Bowlby’s clinical remarks have been rare. One contributing factor may have
been that the reader of Bowlby’s works is presented with his general reflections on his clin-
ical experience, interspersed only occasionally with a paragraph describing a clinical case.
These cases generally describe the childhood factors that predisposed a patient to mental
health problems, and the precipitating adult context that led them to enter therapy. Bowlby’s
own efforts, approaches, successes, and failures with patients are almost never reported. The
reason appears to be, alongside Bowlby’s characteristic reserve, that unlike many psychoana-
lysts of his day he did not want to give details about his patients in publications without their
permission.448 In the Bowlby Archive in the Wellcome Collection, Bowlby’s case files are
embargoed until a century after the clinical work took place; as a result, for the most part, his
earliest clinical cases will become available in around 2035.449

A marked exception, however, is the case of ‘Mrs Q’, which is discussed repeatedly by
Bowlby in six different published treatments, and several unpublished manuscripts, over the
span of decades, beginning from Defences that Follow Loss. Given its clear exceptionality,
Bowlby presumably sought approval from the patient to discuss the case. In both Separation
and Loss the case receives attention on multiple occasions. In lectures on the clinical impli-
cations of attachment research, across different countries, he positioned this case as para-
digmatic.450 Putting together the jigsaw puzzle of Bowlby’s remarks on the case across his
different writings offers an unusual opportunity to see his clinical work with a patient over
time.451
Bowlby described his introduction to Mrs Q as follows:
Some years ago the doctor at a maternity and child welfare clinic asked me to see a little boy
of 18 months who was not eating and was losing weight. His mother was intensely anxious
and depressed and had been so since the boy’s birth. On enquiry I found that she was ter-
rified lest her son die and was therefore pestering him to eat. She also told me that she had
therapy for adults: towards a unifying perspective on practice. Psychology and Psychotherapy: Theory, Research
and Practice, 89(1), 15–32. For descriptions of the boom in therapeutic approaches claiming genesis in attachment
theory as an ‘explosion’ see Magnavita, J.J. & Anchin, J.C. (2013) Unifying Psychotherapy: Principles, Methods, and
Evidence from Clinical Science. New York: Springer, p.67; Johnson, S.M. (2019) Attachment Theory in Practice.
New York: Guilford, p.22.
447 E.g. McCluskey, U. (2005) To Be Met as a Person: The Dynamics of Attachment in Professional Encounters.
London: Karnac Books; Lyons-Ruth, K. (2007) The interface between attachment and intersubjectivity: perspective
from the longitudinal study of disorganized attachment. Psychoanalytic Inquiry, 26(4), 595–616; Crittenden, P.M.,
Dallos, R., Landini, A., & Kozlowska, K. (2014) Attachment and Systemic Family Therapy. London: McGraw-Hill.
448 For instance, Bowlby’s request of Emmy Gut to use anonymized material from her case, discussed and ap-
praised in Ross, L.R. (2006) Talking theory, talking therapy: Emmy Gut and John Bowlby. Issues in Mental Health
Nursing, 27(5), 475–97.
449 There is one early case which is, however, available: Bowlby’s clinical notes of a case from 1938 to 1939 (PP/
Bow/C.4/23). The reason why these are available appears to be the lack of identifying details in these notes. There
are also a series of undated short case histories under the title ‘Maternal Behaviour: Humans’ (PP/Bow/H.136) in
the Bowlby Archive. However, these contain few details about Bowlby’s therapeutic practice, and remained un-
published. The case histories were likely written in the early 1960s, and describe cases seen by Bowlby at the Child
Guidance Clinic.
450 Bowlby, J. (1981) Clinical applications: material for lectures. PP/Bow/F.3/103.
451 The first discussion of ‘Mrs Q’ is in a 1962 manuscript. The first published mention is 1963, discussing an ana-
lysis that has lasted three years. This suggests that the patient is the ‘Mrs K’ with whom Bowlby conducted detailed
clinical interviews between 1960 and 1964, keeping these interviews for reference. These are the only clinical notes
Bowlby kept in this way. The ‘reports of interviews with Mrs K (mother) (Mother and Child Welfare Clinic)’ (PP/
BOW/C.6/6-8) are embargoed until the 2060s.
82 John Bowlby
sometimes had impulses to throw the baby out of the window and to commit suicide. Only
some months later did she tell me that on occasion she became hysterical, smashed the
dishes and battered the baby’s pram.452
Bowlby characterised Mrs Q as ‘one of the most anxious patients I have ever treated’.453
Mrs Q described her commitment to giving her son a happy childhood, in contrast to
her own. In many respects, Bowlby felt, she succeeded very well. However, Mrs Q was
dismayed and confused by her own outbursts of violence, which seemed to coincide espe-
cially with occasions when her mother came over to visit.454 Over the first three years of
analysis, Bowlby worked with Mrs Q to trace how her experiences as a child had been an
important contributory factor to her adult anxiety and aggression. Her parents had been
physically violent and threatened to kill one another. Her mother would also, at times, seek

to coerce her husband and children by threatening to leave the family unless they did as
she said.
Mrs Q experienced feelings of irrational distrust of Bowlby, which puzzled her. In the
clinical work, Bowlby used these feelings as a basis for exploring Mrs Q’s experiences of at-
tachment relationships. He gave particular attention to Mrs Q’s experiences of her mother’s
threats of suicide: ‘Mrs Q. described how on two occasions she had returned from school to
find her mother with her head in the gas oven and how at other times her mother would pre-
tend to have deserted by disappearing for half a day. Naturally, Mrs Q. grew up terrified that
if she did anything wrong her mother would go.’455 Additionally, Bowlby explored with Mrs
Q how she had become scared of her own capacity for anger, which she had learnt to redirect
‘either towards herself or towards something which, or someone who, could not retaliate.
When a child, Mrs Q. recalled, she retreated to her room and bit herself severely or attacked
her dolls.’456
The clinical work was slow. Mrs Q ‘claimed for a long while not only that her feelings
for her mother were of love, which was true since her mother had many good qualities,
but that that must exclude hatred’.457 However, an important point in the therapy came
when she acknowledged that, when she became angry with Stephen, ‘she said the most
dreadful things, the very same things, in fact, that her mother had said to her when she
was a girl’:
Once the facts were known it was possible to arrange some joint sessions with mother and
son during which mother, with real regret, acknowledged making the threats and Stephen
explained how terrified they made him. Mother assured Stephen that she would never do it
really. All was not well thereafter, but recognition that Stephen’s fears were well based and
an opening of communication between mother and son eased the situation.458
Lecture.) Animal Behaviour, 28(3), 649–56, p.653.

453 Bowlby, J. (1969) Psychopathology of anxiety: the role of affectional bonds. British Journal of Psychiatry, 3,
80–86, p.85.

Lecture.) Animal Behaviour, 28(3), 649–56, p.653–4.

456 Ibid.
458 Ibid. p.270.
In the clinic 83
Over three years of therapeutic work, Bowlby helped Mrs Q come to acknowledge the anger
she felt towards her mother. As this became less segregated, the violent outbursts towards
Stephen reduced. Mrs Q came to understand that, as a child, there had been good reason for
segregating her anger. Expressions of anger could otherwise provoke her mother to threaten
suicide or to abandon her. Additionally, Mrs Q was strictly told not to let anyone else know
about her parents’ behaviour, which contributed to the segregation by blocking opportun-
ities for feedback and affirmation of her experiences.
Towards the end of the third year of therapeutic work Mrs Q’s father died unexpectedly,
following an elective operation for cataracts. In the first years of therapy, Mrs Q’s account of
her father had been ‘overtly negative’, hiding the positive feelings that she also felt towards
him.459 Following her loss, Mrs Q became intensely depressed, had thoughts of suicide,
and also described dissociative symptoms. Among these were anomalous beliefs about her

father’s death and its cause:
During the weeks following her father’s death, she now told me, she had lived in the half-
held conviction that the hospital had made a mistake in identity and that any day they
would phone to say he was alive and ready to return home. Furthermore, she had felt spe-
cially angry with me because of a belief that, had I been available, I would have been able to
exert an influence on the hospital and so enabled her to recover him.460
She experienced the sense that her father’s home had to be kept exactly as it was because he
was very much alive and would be displeased to find anything changed when he returned.461
Bowlby was struck that Mrs Q appeared to have two distinct sets of thoughts and feelings
in operation: one that led her to act as if her father was dead, and another that led her to act
as if her father was alive.462 Yet neither organisation was strictly unconscious: Mrs Q could
discuss her experience of holding both views. In Defences that Follow Loss, Bowlby observed
that Mrs Q’s symptoms could not be explained using the conventional psychoanalytic con-
cept of repression, which suggests a division between conscious and unconscious material.
‘Mrs Q. is, however, little different,’ he mused. Like with repression, ‘there is evidence of a
psychic system with its accompanying affects and fantasies that is alien to the one with which
we as analysts are in communication’. However, in contrast to repression, this system was not
unconscious, but nonetheless blocked from communicating with other conscious systems. It
is from this set of reflections on the case of Mrs Q that Bowlby, in fact, went on in the manu-
script to use the concept of ‘segregated systems’ for the first time.463
Bowlby as therapist
Three key principles can be gleaned from the treatment of Mrs Q and more generally from
Bowlby’s late writings about his approach to clinical practice. A first was the importance
for Bowlby of the clinical transference: the behaviours, affects, and cognitions displayed by
459 Bowlby, J. (1963) Pathological mourning and childhood mourning. Journal of the American Psychoanalytic
Association, 11(3), 500–41, p.517.

460 Ibid. p.517.
462 Ibid. pp.348–9.
84 John Bowlby
the patient towards the therapist, which may find their origin in the plans developed on the
basis of earlier experiences of family and intimate relationships. In the 1960s, Bowlby had
assumed that relationships with attachment components would be formed by children solely
on the basis of familiarity with the caregiver. By contrast, in his late writings on therapeutic
technique, he situated the successful longer-term therapeutic relationship for adult patients
as one with attachment components, in the provision of a secure base.464 In this, he seems to
acknowledge that attachment components could and should form in the relationship with
the therapist, not merely due to familiarity, but due to other factors as well. Bowlby appeared
to see something about the intimacy of communication within a therapeutic relationship
contributing to its status as a relationship with attachment components. This activated the
patient’s expectations about close relationships, and at the same time provided an oppor-
tunity to reflect on them.465

In the case of Bowlby’s clinical work with Mrs Q, the transference relationship was used
to help the patient notice her difficulties around trust. It was on the basis of this recogni-
tion about her relationship with Bowlby as therapist that Mrs Q could then be supported to
see that trusting a parental figure was a justifiable fear, given her early experiences. Bowlby
stated that ‘my therapeutic approach is far from original,’466 and indeed a special attention
to transference phenomena was common to the psychoanalysis of his day, especially for
Klein and her followers. However, a significant point of contrast was a second principle of
his therapeutic technique. Bowlby believed that the therapist’s task, where possible, is to ad-
dress stable patterns of interaction within close relationships, rather than solely problems
‘inside’ particular individuals.467 Even though Bowlby tended to treat behavioural systems
as properties of individuals, recovery from symptoms of mental ill health is rarely achieved
without considering the resourcing of relevant systems, and the contexts and interpersonal
meanings that calibrate their circumstances of activation and termination. The case of Mrs
Q began with her son, who was not eating. Bowlby sought the origins and eliciting condi-
tions of these symptoms in the meaning of the boy for his mother, Mrs Q, and in her behav-
iour towards him—and, in turn, the meaning of Mrs Q’s behaviours in her biographical and
464 Whether provision of a relationship with attachment components should be the goal of shorter-term work by
helping professionals has been debated, e.g. Charles, G. & Alexander, C. (2014) Beyond attachment: mattering and
the development of meaningful moments. Relational Child and Youth Care Practice, 27(3), 26–30.
465 Bowlby’s copy of The Standard Edition of the Writings of Sigmund Freud still retains his bookmark, in Volume
11, marking the following underlined passage: the patient’s ‘symptoms, to take an analogy from chemistry, are
precipitates of earlier experiences in the sphere of love (in the widest sense of the word), and it is only in the
raised temperature of his experience of the transference that they can be resolved and reduced to other psychical
products.’ Freud, S. (1910) Five lectures on psycho-analysis. In The Standard Edition of the Complete Psychological
Works of Sigmund Freud, Vol. 11 (pp.1–56). London: Hogarth, p.51; Bowlby’s copy is held in the library of Human
Development Scotland.
466 Bowlby, J. (1979, 1988) On knowing what you are not supposed to know and feeling what you are not sup-
posed to feel. In A Secure Base. London: Routledge, pp.111–33, p.132.

467 Bowlby, J. (1949) The study and reduction of group tensions within the family. Human Relations 2, 123–8,
p.123; Bowlby, J. (1969, 1982) Attachment, Volume 1. London: Penguin: ‘Nothing in child psychiatry has been
more significant in recent years than the increasing recognition that the problems its practitioners are called upon
to treat are not often problems confined to individuals but are usually problems arising from stable interactional
patterns that have developed between two, and more often several, members of a family’ (349). Potential links
between attachment theory and systemic family therapy have been widely discussed. See e.g. Cowan, P.A. (1997)
Beyond meta-analysis: a plea for a family systems view of attachment. Child Development, 68(4), 601–603; Akister,
J. & Reibstein, J. (2004) Links between attachment theory and systemic practice: some proposals. Journal of Family
Therapy, 26(1), 2–16; Crittenden, P.M., Dallos, R., Landini, A., & Kozlowska, K. (2014) Attachment and Systemic
Family Therapy. London: McGraw-Hill; Vetere, A. (2016) Systemic theory and narratives of attachment: integra-
tion, formulation and development over time. In M. Borcsa & P. Stratton (eds) Origins and Originality in Family
Therapy and Systemic Practice (pp.129–39). New York: Springer.
In the clinic 85
relational context. From the 1950s until the end of his life, Bowlby referred to himself as a
‘family psychiatrist’, since he put emphasis on seeing and helping the whole family, rather
than only the member of the family showing the symptoms of a mental health problem.468
In his attention to the familial context, Bowlby felt that his approach was truly one that
avoided blaming parents, but instead considered their actions, thoughts, and feelings in
turn in a wider context of predisposing, triggering, and sustaining factors. In ‘The making
and breaking of affectional bonds’, Bowlby argued that when connected to the history or
situation that has predisposed or elicited it, much parental behaviour that might otherwise
seem simply inexplicable comes in fact to make sense—either as a response to a truly dis-
tressing situation or as an attempt to avoid reacting to this situation.469 In a letter to Joan
Stevenson-Hinde, he expressed his strong disagreement with the idea common in family
therapy circles that ‘patterns of interaction has some particular purpose e.g. that of keeping

the family together’. Of course, actions to keep a family together may be a conscious inten-
tion of individuals. But Bowlby was hostile to interpretations by therapists that symptoms
shown by patients could be understood in terms only of their present function. He felt that
such accounts were often generated when the therapist did not yet understand enough about
the history of the family, its members, and interactions.470
A third principle in Bowlby’s late writings on technique was that no advice or guidance
should be offered until it is clear that the patient or family is able to understand this as an
attempt to be supportive rather than critical. Early in his career, Bowlby regarded himself
as more non-directive as a therapist than his peers. However, in his late writings on clinical
technique, he made more of a space for challenging and guiding patients and families. Such
directive interventions were only recommended, though, on the condition that the patient
understands that the therapist’s ‘concern is to help the patient review his own life, to look at
his problems in his own way’.471 Bowlby was worried that unless therapy leads with support,
the encounter may in fact harm the patient or family by reinforcing feelings of guilt and des-
pair. Additionally, where the therapist is perceived as showing disregard for the world as the
patient sees it, the potential for trust is eroded. The therapist should not be quick to cast their
role as the ‘representative of reality’, contradicting patients’ own perspectives.472 As well as
potentially harmful, Bowlby felt such approaches were ineffective at helping change to occur.
468 Bowlby, J. (1986) ‘Attachment Theory: New Directions’. ACP-Psychiatric UPDATE 7(2), panel discussion,
Washington. PP/BOW/A.5/1. On Bowlby’s position in relation to the history of family therapy and systems ap-
proaches to family relations see Byng-Hall, J. (1991) An appreciation of John Bowlby: his significance for family
therapy. Journal of Family Therapy, 13(1), 5–16. For a discussion of societal factors contributing to ‘the family’ as
the unit of concern and intervention see Weinstein, D. (2013) The Pathological Family: Postwar America and the
Rise of Family Therapy. Ithaca, NY: Cornell University Press.

470 Bowlby, J. (1990) Letter to Joan Stevenson-Hinde, 26 March 1990. PP/Bow/J.9/186.
471 Bowlby, J. (1990, 2011) John Bowlby: interview by Leonardo Tondo. Clinical Neuropsychiatry, 8(2), 159–71,
p.169. Bowlby was reacting to the emphasis in the Kleinian tradition on transference interpretations. He gave
particular emphasis to the importance of support, and helping the patient or family get to a point where they can
make sense and use of interpretations—much like the concept of ‘developmental help’ within the Anna Freudian
tradition. See Edgcumbe, R. (1995) The history of Anna Freud’s thinking on developmental disturbances. Bulletin
of the Anna Freud Centre, 18(1), 21–34. Bowlby’s emphasis, however, led some second-generation attachment re-
searchers to assume that he was arguing only for support, without challenge, in therapy. See e.g. Bretherton, I.
(1991) Pouring new wine into old bottles: the social self as internal working model. In M.R. Gunnar and L.A.
Sroufe (eds) Self-Processes and Development: The Minnesota Symposia on Child Development, Vol. 23 (pp.1–41).
Hillsdale, NJ: Lawrence Erlbaum, p.32.
472 E.g. Bowlby, J. & Parkes, C.M. (1970, 1979) Separation and loss within the family. In The Making and Breaking
of Affectional Bonds (pp.99–123). London: Routledge, p.114.

86 John Bowlby
In ‘Constructions in Analysis’, Freud wrote that ‘no damage is done if, for once in a way, we
make a mistake and offer the patient a wrong construction as the probably historical truth’.
Bowlby, in his marginalia, described Freud’s stance as ‘complacent’.473
Given the importance of laying the ground through support, Bowlby divided therapy into
two ‘phases’.474 In the first, the therapist should primarily seek to offer companionable support,
combined with open-ended exploratory questions. The focus in this first phase should be on
present-day experiences and the wider social context, seeking to identify the kinds of situation
that repeatedly tend to be difficult or cause problems for the patient or family. These are likely
to be those that led to the initiation of therapy, though the patient may or may not be aware of
the pattern. For instance, in cases where a behavioural system has been chronically suppressed
(Level 3 defensive exclusion), the first task is for the patient to recover a sense of what it might
mean for this system to be activated. Where the attachment system of a patient has been subject

to full defensive exclusion, the therapist might explore current experiences outside of therapy
and within the transference where this system would otherwise be activated, and the thoughts,
feelings, and behaviours that occur instead.475 This entails slow, careful work to support pa-
tients to articulate the ecology of their ordinary lives, and the circumstances that prompt the
activation or deactivation of behavioural systems and their associated affects.
Both the attachment system and the exploratory system may need to be coaxed online
by this combination of support and attention to present-day experiences: the attachment
system bringing fruitful material for the transference, and the exploratory system engaging
and integrating this material. Past experiences will have already been at least implicit in the
first phase, since both the present and our perceptions of it always hold the past within them.
Yet, only once the patient or family feel supported to engage the exploratory system to ad-
dress present-day experiences should the therapist, in the second phase, seek to explore the
nature of these feelings in depth, and then consider past experiences that may have contrib-
uted to these difficulties. This should include helping the patient or family sort through their
feelings about the past and present, to reduce the extent to which these contaminate one an-
other and become confused when decisions are made in the present.476 In the case of Mrs Q,
Bowlby described patiently supporting her towards recognition of the anger she felt towards
her mother. The aspect of the therapeutic task which ultimately ‘can best wait is consider-
ation of the past since its only relevance lies in the light it throws on the present’:
The sequence may often be for the therapist and patient, working together, first to recog-
nise that the patient tends habitually to respond to a particular type of interpersonal situ-
ation in a certain self-defeating way, next to examine what kinds of feeling and expectation
such situations commonly arouse in him, and only after that to consider whether he may
473 Annotations by Bowlby dated 1964 on The Standard Edition of the Complete Psychological Works of Sigmund
Freud, Vol. 23, p.261. Copy held in the library of Human Development Scotland.
474 Bowlby, J. (1973) Separation: Anger and Anxiety. New York: Basic Books, p.354.
243–56, p.251.
476 See the three unpublished case histories in ‘Maternal behaviour: humans’ (PP/Bow/H.136), where this is the
common central feature of his therapeutic approach. See also Bowlby, J. (1971) Letter to Graham Davies, 15 April
1971. PP/Bow/H.5: ‘I see psychoanalytic therapy as an attempt to help a patient explore his own motives, his own
model of the world and himself in it, and also to reconsider the validity of that model. Often I believe it is more
valuable to raise questions for a patient than to attempt to inform him by means of interpretations . . . In helping a
patient make these reappraisals, I believe it useful for us to have some knowledge of the conditions that commonly
lead an individual to grow up in ignorance of his motives and with more or less distorted models of his world and
himself.’
In the clinic 87
have had experiences, recent or long past, which have contributed to his responding with
those feelings and expectations.477
Bowlby offered a metaphor: ‘If a ball has gone down a dark passage, a child may be frightened
to go there and get the ball, but if I say “Look, I will come with you”, he may be quite happy.
In psychotherapy we act as a companion to a patient who is too frightened to look at what
has happened to him in the past. So we accompany him in the exploration.’478 Bowlby be-
lieved that this companion and social point of reference is important because it gives patients
the confidence to unlatch their personal hopes and fears from matters of fact, through close
scrutiny of both. Bowlby clearly saw the discovery of historical truth in therapy as making
some contribution to the reduction in symptoms, since it permitted more effective informa-
tion-processing in the present.479 However, just as important was his emphasis on therapy as

supporting curiosity and courage within and beyond the therapeutic setting.
Attachment disorders
Also relevant to his work as a clinician, Bowlby made a number of important remarks on the
meaning and function of diagnosis. His overall position was that diagnosis is a relevant and
valuable clinical tool for medical professionals, but that it should not be mistaken for explan-
ation. From early in his career, he emphasised that development was multiply determined,
and would therefore often exceed static characterisations that depicted particular symptoms
as manifestations of specific disturbances.480 In an undated text, archived with material from
around 1939 to 1940, Bowlby reflected carefully on the purposes of psychiatric classification
in an attempt to develop diagnostic groupings for children under five:
Classification and diagnosis in child psychiatry is at present in a state of anarchy. Very few
of the children seen correspond to any of the classical psychoneuroses or psychoses—the
fact unjustly being simply “character-problems”. For these character cases there is as yet no
good classification for adults, let alone for children . . . Even in adults it is sometimes dif-
ficult to distinguish clearly between the habitual personality and the particular syndrome
of symptoms from which the patient is suffering. This difficulty is increased in childhood.
Consequently the classification used here is only provisional.481

159–71, p.163.
479 Fonagy has regularly criticized strands of psychoanalytic theory for treating the discovery of historical truth
as the mechanism and target of therapeutic action, e.g. Fonagy, P. & Tallandini-Shallice, M. (1993) On some prob-
lems of psychoanalytic research in practice. Bulletin of the Anna Freud Centre, 16, 5–22. Bowlby’s emphasis on the
contribution of defensive exclusion to mental illness, and the curative powers of the reintegration of reality within
information processing, do at times veer in this direction. However, suggesting that this may be a structural issue
faced by psychotherapy as an activity, in his more recent work on therapeutic technique Fonagy has himself at times
become tangled in the same problem, e.g. Fonagy, P. (2016) The role of attachment, epistemic trust and resilience
in personality disorder: a trans-theoretical reformulation. DMM News, 22 September 2016. http://www.iasa-dmm.
org/images/uploads/DMM%20%2322%20Sept%2016%20English.pdf. ‘Feeling recognized opens the epistemic
path necessary to update the neural nets which in turn enable accurate (resilient) interpretation of reality’ (6).
480 Bowlby, J. (1954) The diagnosis and treatment of psychological disorders in childhood. Health Education
Journal, 12(2), 59–68, p.62.

481 Bowlby, J. (undated, c.1939) Diagnostic Groupings for Children Over 5 Years. PP/Bow/C3/9.
88 John Bowlby
In his early book Personality and Mental Illness from 1940, Bowlby acknowledged that ‘few
people remain the same throughout their lives’, and he explored this lack of continuity in the
case of mental health symptoms. He pointed out that only rarely do individuals exactly fit
the criteria of psychiatric classification, and generally only those with relatively less severe
symptoms. Among patients with more severe problems, ‘many show at successive periods
an unstable personality, symptoms of psycho-neurosis and of psychosis’. Bowlby therefore
called it ‘absurd’ to regard such a patient ‘in terms of any one condition. He must be thought
of as an individual of certain potentialities, a unity of which the particular traits and symp-
toms shown at any one moment are but fleeting expression.’482 Symptoms are not, in such an
account, the manifest effect of an underlying and discrete disorder. For instance, low mood,
sleeplessness, and a lack of energy are not the expressions of a disorder of ‘depression’, as dis-
tinct from other mental health problems. Rather, Bowlby’s account was transdiagnostic. He

suggested that the individual represents a unity of potentialities, in this case towards mental
health or ill health. Particular symptoms, or even clusters of symptoms that can be grouped
as mental disorders, are best considered as consequences of this set of potentialities. He ar-
gued that, where early childhood experiences have been positive, this set of potentialities
will predispose mental health. By contrast, early adverse experiences will support the in-
tensification and expression of potentialities towards mental ill health across the lifespan.
In his mature thought, Bowlby developed these reflections further. In ‘Developmental
psychiatry comes of age’, published 1988, he insisted that ‘a sharp distinction must be drawn
between current functioning, measured in terms of presence or absence of psychiatric dis-
order and personality structure, measured in terms of greater or less vulnerability to adverse
life events and situations’.483 Furthermore, ‘the features of personality to which we draw at-
tention are different to those that most clinical instruments are designed to measure and not
necessarily correlated with them’.484 Bowlby considered factors that might contribute to such
transdiagnostic, developmental vulnerability. He was especially concerned with the way that
the defensive exclusion of attachment-related motivations may deplete possibilities for experi-
encing life as ‘emotionally rich and varied’, and contribute variously to distress, aggression, and
depression.485 Bowlby’s position implied that attachment processes can readily impact other
areas of functioning, but this did not mean that the resulting behaviours are therefore to be con-
sidered as attachment.486 Though he was particularly concerned with attachment, Bowlby was
clear that defensive exclusion of other behavioural systems could contribute to a vulnerability
to mental illness. Defensive exclusion of the exploratory system could make it difficult for an
individual to learn from experience, or from therapy.487 And loss of access to the prompts of
the fear system could lead an individual to enter cruel and painful relationships. Without itself
constituting a form of mental illness, Bowlby anticipated that such processes would contribute
to the initiation or the maintenance of various mental health symptoms.
From Separation onwards, Bowlby emphasised that clinicians and researchers should note
the importance of ‘developmental pathways’, self-reinforcing patterns in children’s trajectories
482 Bowlby, J. (1940) Personality and Mental Illness. London: Kegan Paul, p.187.
483 Bowlby, J. (1988) Developmental psychiatry comes of age. American Journal of Psychiatry, 145, 1–10, p.6.
484 Bowlby, J. (1980) Loss. London: Pimlico, p.202–203.
486 For a later statement well aligned with this position see Waters, E., Posada, G., Crowell, J., & Lay, K.L. (1993)
Is attachment theory ready to contribute to our understanding of disruptive behavior problems? Development &
Psychopathology, 5, 215–24.
487 See also Fonagy, P. & Allison, E. (2014) The role of mentalizing and epistemic trust in the therapeutic rela-
tionship. Psychotherapy, 51(3), 372–80.

In the clinic 89
towards or away from mental health. Bowlby was inspired by Waddington’s description of
how cells can initially develop in a variety of different ways, but that once they do begin to de-
velop, they become canalised such that a change from the established pathway requires greater
and greater intervention.488 Bowlby reflected that, with human development conceptualised
in terms of pathways, we can expect ‘adverse childhood experiences489 [to] have effects of at
least two kinds. First they make the individual more vulnerable to later adverse experiences.
Secondly they make it more likely that he or she will meet with further such experiences.’490
This pathways metaphor suggested a nuanced model of continuities over time:
All pathways are thought to start close together so that, initially, an individual has access to a
large range of pathways along any one of which he might travel. The one chosen, it is held, turns
at each and every stage of the journey on an interaction between the organism as it has devel-

oped up to that moment and the environment in which it then finds itself. Thus at conception
development turns on interaction between the newly formed genome and the intrauterine
environment; at birth it turns on interaction between the physiological constitution, including
germinal mental structure, of the neonate and the family, or non-family, into which he is born;
and at each age successively it turns on the personality structure then present and the family
and, later, the wider social environments then current . . . As development proceeds and struc-
tures progressively differentiate, the number of pathways that remain open diminishes.491
This model implied for Bowlby that neither knowledge of general patterns nor knowledge
of the specific case should be abandoned. Today, this might be discussed in terms of the
distinction between ‘diagnosis’ and ‘formulation’.492 On an individual level, diagnostic
488 Waddington, C.H. (1957) The Strategy of Genes. London: Allen & Unwin.
489 This term would later become the label for an influential self-report measure of adversity and trauma. Felitti,
V.J., Anda, R.F., Nordenberg, D., et al. (1998) Relationship of childhood abuse and household dysfunction to many
of the leading causes of death in adults: the Adverse Childhood Experiences (ACE) Study. American Journal of
Preventive Medicine, 14(4), 245–58. If there is a debt to Bowlby’s earlier use, it is not acknowledged by Felitti and
colleagues. This is discussed by Partridge, S. (2019) Review of ‘The Deepest Well: Healing the Long-term Effects of
Childhood Adversity’ by Nadine Burke Harris. Attachment, 13(1).
490 Bowlby, J. (1981, 1988) The origins of attachment theory. In A Secure Base (pp.22–42). London: Routledge.
491 Bowlby, J. (1973) Separation. London: Pimlico, p.412. The importance of the concept of developmental path-
ways for Bowlby’s late thought cannot be overestimated. See for instance the prominent place it receives in his plan
for the unpublished book, representing his final position (on the model of Freud’s ‘Outline of Psychoanalysis’),
Bowlby, J. (1982) Outline of attachment theory. PP/Bow/H.260:
Chapter 1. Historical. A way of conceptualising family influence.
Chapter 2. Main features: developmental pathways; description of attachment, caregiving, exploration; abuse
Chapter 3. Cognitive models
Part II: Pathways of development
Chapter 4. Pathways to health, including healthy parenting
Chapter 5. Pathways to anxious attachment & phobia
Chapter 6. Pathways to depressive disorders. Suicide?
Chapter 7. Pathways to false self, masked depression
Chapter 8. Pathways to delinquency & psychosis
Chapter 9.
Part III: Applications
Chapter 10. Prevention and crisis intervention
Chapter 11. Treatment
Part IV: Problems of theory
Chapter 12. Evolutionary theory, control systems, activation, termination function
Chapter 13. More re cognitive models and defence. Consciousness and unconsciousness
Chapter 14. Attachment and science’
492 Johnstone, L. & Dallos, R. (2013) Formulation in Psychology and Psychotherapy: Making Sense of People’s
Problems. London: Routledge.

90 John Bowlby
classifications tell clinicians about what one child may have in common with other children
with similar symptoms. They can help clinicians to develop an integrative account of the
child’s difficulties, encourage more consistent care-planning, and encourage professionals
to consider the treatments that may help. Formulation, on the other hand, is an individual
approach that considers the ways in which the child is unique from others. Within formula-
tion, assessment of the attachment system and its functioning serves as one part of nuanced
thinking about a child’s behavioural profile, potential risk factors, and the role of the child’s
external environment.
In 1980, the ‘Infancy, Childhood and Adolescent Disorders’ committee of the American
Psychiatric Association Diagnostic and Statistical Manual of Mental Disorders (III) (DSM-
III) introduced the category of ‘reactive attachment disorder in infancy’ as a recognised
diagnosis. This diagnosis was an attempt to bring within medical assessment practice ob-

servations that had been made of specific forms of behaviour shown by institutionalised and
former institutionalised children, influenced by Bowlby’s 1951 report to the World Health
Organisation, as well as other reports.493 The disorder was, as it was initially proposed, to be
diagnosed on the basis of weak infant physical growth, poor social responsiveness, and emo-
tional apathy, as a consequence of grossly inadequate experiences of caregiving. In 1987, the
diagnosis was revised to remove the physical growth criterion and adjust the age criterion,
and has since seen other changes.494 It is notable that in the ten years between 1980 and his
death in 1990, Bowlby made no public statement discussing this new official diagnosis os-
tensibly drawing inspiration from his work and using his headline concept of ‘attachment’.
Some have treated this as implying that the diagnosis was simply a natural expression of
Bowlby’s position and that he generally approved.495 However, Bowlby’s silence on the ad-
vent of ‘attachment disorders’ more likely reflected discomfort.
The introduction of the attachment disorder diagnosis as a problem of ‘attachment’ ap-
pears to have been based solely on Bowlby’s earliest writings on institutionalisation and his
diffuse claims about its socioemotional implications, essentially prior to the development
of attachment theory from the late 1950s. Justin Call, the primary member of the ‘Infancy,
Childhood and Adolescent Disorders’ committee to discuss Bowlby’s work in print, justified
the new diagnostic category in 1982 by arguing that ‘attachment disorders of infancy are
characterised by the absence, disruption, or distortion of normally occurring developmental
sequences of attachment behaviors’.496 Yet he gave no reference to Bowlby’s work after 1958.
493 Bakwin, H. (1949) Emotional deprivation in infants. Journal of Pediatrics 35, 512‒21; Provence, S. & Lipton,
R. (1962) Infants in Institutions. New York: International Universities Press; Tizard, R. & Rees, J. (1975) The effect
of early institutional rearing on the behaviour problems and affectional relationships of four year old children.
Journal of Child Psychology and Psychiatry, 16, 61–73.
494 There has been a growing body of research on attachment disorders since 1998, and especially since the in-
creased prominence and detail of the diagnosis in DSM-IV in 1994 and DSM-5 in 2013. These developments in
diagnostic practice have been influenced by the pivotal series of studies of children adopted from Eastern European
orphanages. See Zeanah, C.H., Smyke, A.T., & Settles, L.D. (2006) Orphanages as a developmental context for early
childhood. In K. McCartney & D. Phillips (eds) Blackwell Handbook of Early Childhood Development (pp.424–54).
Hoboken, NJ: Wiley-Blackwell; Zeanah, C.H., Chesher, T., Boris, N.W., & American Academy of Child & and
Adolescent Psychiatry (2016) Practice parameter for the assessment and treatment of children and adolescents
with reactive attachment disorder and disinhibited social engagement disorder. Journal of the American Academy
of Child & Adolescent Psychiatry, 55, 990–1003.
495 E.g. Kanieski, M.A. (2010) Securing attachment: the shifting medicalisation of attachment and attachment
disorders. Health, Risk & Society, 12(4), 335–44.

496 Call, J. (1982) Attachment disorders of infancy. In H.I. Kaplan, A.M. Freedman, & Z.B.J. Sadock (eds)
Comprehensive Textbook of Psychiatry, Vol. 3 (pp.230–68). Baltimore: Williams and Wilkins, p.2587. Repeated in
Call, J.D. (1984) Child abuse and neglect in infancy: sources of hostility within the parent–infant dyad and dis-
orders of attachment in infancy. Child Abuse & Neglect, 8(2), 185–202, p.190.
In the clinic 91
This would account for the fact that Call and the DSM interpret the ‘disruption’ or ‘distortion’
of attachment to mean primarily extreme deprivation. Call’s lack of familiarity with Bowlby’s
work would also explain why the DSM-III initially limited attachment disorders to infants
under eight months—when, in Attachment, Volume 1, Bowlby had argued that the attach-
ment system is not likely formed until at least nine months. The DSM criteria displayed no
concern to assess disruption to the attachment system, for instance in distinguishing be-
tween behaviour in contexts where the attachment system would be activated and behaviour
in other contexts. The criteria also situated ‘attachment disorder’ as a property of the indi-
vidual child, in contrast to Bowlby’s emphasis on the dyadic status of the attachment system
in infancy.497
The DSM-III was a major event and sent shock-waves through the psychiatric establish-
ment. Even if particulars were regarded as arguable, it appeared to offer the basis for a valid

and reliable category-based diagnosis system, in which all symptoms find their logic and
cause in an underlying disorder.498 However, the ‘attachment disorder’ diagnosis was not
a part of this general upsurge of interest and support for diagnostic categorisation. No em-
pirical studies used the category within research until after Bowlby’s death.499 If the idea
of an ‘attachment disorder’ diagnosis without consideration of Bowlby’s writings on attach-
ment theory had been put forward in another document, it seems probable that it would
have been ignored. Certainly, Call’s chapter ‘Attachment Disorders in Infancy’ in the 1982
Comprehensive Textbook of Psychiatry, written to accompany the new DSM diagnosis, has
rarely been cited.500 And in the next edition Call’s chapter was replaced by one that largely as-
similated the new attachment disorder category into the frame of infant failure to thrive and
psychosocial dwarfism.501
However, the diagnosis of ‘attachment disorder’ was hooked inside one of the single most
influential medical documents of the twentieth century. The DSM-III contributed to a trans-
formation in mental health assessment to give prominence to diagnosis and diagnostic path-
ways.502 Diagnostic categories have been made central to the delivery and administration
of clinical services, to clinical resource allocation, and to clinical training. Throughout his
career Bowlby was concerned with clean distinctions, where possible, between diagnostic
activity and reflection on developmental processes. Furthermore, it is evident that his overall
impulse was towards attention to developmental pathways in understanding children’s
symptoms, and the use of diagnosis only for specific, limited purposes. Bowlby addressed
these matters further in a paper presented to the British Psychological Society three years
after the publication of the DSM-III (included as Chapter 4 in Trauma and Loss: Key Texts
497 Additionally, the Robertson and Bowlby book discussing observations in the 1950s of anomalous and ap-
parently dissociated behaviours of institutionalized children remained unpublished, so Call and colleagues on
the ‘Infancy, Childhood and Adolescent Disorders’ committee did not have access to this subterranean current
of attachment theory, which would find full recognition only in the 1980s in the work of Mary Main and Judith
Solomon on the ‘disorganized/disoriented’ attachment classification (Chapter 3).
498 Mayes, R. & Horwitz, A.V. (2005) DSM-III and the revolution in the classification of mental illness. Journal of
the History of the Behavioral Sciences, 41(3), 249–67.

499 Richters, M. & Volkmar, F. (1994) Reactive attachment disorder of infancy or early childhood. Journal of the
American Academy of Child and Adolescent Psychiatry, 33, 328–32.

500 Call, J. (1982) Attachment disorders of infancy. In H.I. Kaplan, A.M. Freedman, & Z.B.J. Sadock (eds)
Comprehensive Textbook of Psychiatry, Vol. 3 (pp.230–68). Baltimore: Williams and Wilkins.

501 Green, W. (1985) Attachment disorders of infancy and early childhood. In H.I. Kaplan & Z.B.J. Sadock (eds)
Comprehensive Textbook of Psychiatry, Vol. 4 (pp.1722–31). Baltimore: Williams and Wilkins.

502 Greco, M. (2016) What is the DSM? Diagnostic manual, cultural icon, political battleground: an overview
with suggestions for a critical research agenda. Psychology & Sexuality, 7(1), 6–22.
92 John Bowlby
from the John Bowlby Archive). Bowlby denounced the focus of the psychological establish-
ment on category-centric practice:
The categorists are still searching for diagnostic criteria that distinguish the mentally ill
from the normal, though today their search is more likely to be for genetically determined
biochemical anomalies than for any behavioural criterion. [On the other hand, there are]
those others who, like myself, believe continuity to be a more fruitful perspective.503
Bowlby’s remarks to the British Psychological Society suggest that he would have endorsed
Sroufe’s view that ‘the circumscribing of attachment problems to specific disorders reveals a
failure to grasp the developmental significance of attachment history and the potential power
of a developmental approach to psychopathology in general’ (Chapter 4).504 As we saw in

the section ‘Disorientation’, Robertson documented behaviours that resembled the ‘attach-
ment disorders’ in the children he observed during and following long-term hospitalisation.
However, Bowlby decided not to publish his book with Robertson containing these observa-
tions, and did not integrate them into his subsequent theorizing. He also kept silent when the
attachment disorder diagnosis was introduced into the DSM. These choices, in retrospect,
appear to have helped initiate a split between widely circulating clinical discourses of attach-
ment disorder and the bulk of empirical and theoretical work on attachment.
In the context of diagnosis-focused clinical practice, and continued observations of be-
haviours such as disinhibited social engagement by clinicians—particularly among those in
adoption and fostering contexts—later psychiatrists interested in infant mental health, such
as Charles Zeanah, acted as if that they had to work with and elaborate the diagnosis that was
‘on the books’.505 Likewise, the World Health Organisation’s International Classification of
Diseases largely took on the disorder as characterised in the DSM. The Zero to Three manual
introduced in 1994 for the assessment of the mental health and development of young chil-
dren is an exception, with uses of attachment ideas in the manual informed by Bowlby’s later
work and subsequent attachment research.506 The Zero to Three manual, however, is grossly
subordinate to the DSM in status. It is rarely possible for American clinicians to seek recom-
pense from insurance companies for their work using Zero to Three. The standard remains
the DSM diagnoses.
The primary use of the DSM classification in the 1990s was to describe children reared
in institutional care, where two profiles could sometimes be seen: withdrawal from all
caregivers, and indiscriminate attachment behaviours shown towards caregivers and
503 Bowlby, J. (1983) Darwin: Psychiatry and Developmental Psychology. Contribution to a Symposium on
Darwin and Psychology held at the conference of the British Psychological Society, London, December 1983. PP/
BOW/F.3/132.
251–68, p.263. Bowlby’s position would perhaps align with developments in the UK, where clinical guidelines now
advise professionals to consider a broader range of problems described as ‘attachment difficulties’ for children
who are adopted or are at risk of going into care, outside of the constraints of the ‘attachment disorder’ diagnoses.
National Institute for Health & Care Excellence (2015) Children’s attachment: attachment in children and young
people who are adopted from care, in care or at high risk of going into care. NICE Guideline (NG26). https://www.
nice.org.uk/guidance/ng26.
505 E.g. Zeanah, C.H., Mammen, O., & Lieberman, A. (1993) Disorders of attachment. In C.H. Zeanah (ed.)
Handbook of Infant Mental Health (pp.332–49). New York, Guilford.

506 Zero to Three/ National Center for Clinical Infant Programs (1994) Diagnostic Classification of Mental
Health and Developmental Disorders of Infancy and Early Childhood (Diagnostic Classification: 0-3). Washington,
DC: Zero to Three; Lyons-Ruth, K., Zeanah, C.H., & Benoit, D. (1996) Disorder and risk for disorder during in-
fancy and toddlerhood. In E.J. Mash & R.A. Barkley (eds) Child Psychopathology (pp.457–91). New York: Guilford.
In the clinic 93
non-caregivers. Over subsequent years, the latter has been redescribed as disinhibited social
engagement. In general, the relationship between the DSM disorder and the broader field
of attachment theory and research has remained unclear, except the repeated observation
that attachment disorder may be more likely when children have not had an opportunity
to form a selective attachment or attachments.507 Some clinicians have argued for the nar-
rowing of the attachment disorder to exclude disinhibited social engagement, since its rela-
tionship with the attachment system is unclear.508 Other clinicians have argued for a broader
applicability for the concept of attachment disorder.509 However, mostly the established gen-
eration of attachment researchers, with a few exceptions such as Zeanah, Lyons-Ruth, and
Spangler,510 have largely ignored the clinical diagnosis. Several attachment researchers, for
instance Sroufe, van IJzendoorn, and Bakermans-Kranenburg, have expressed exasperation
with the attachment disorder diagnosis, given its ill-fit with Bowlby’s attachment theory or

the predominant categories used in mainstream attachment research.511 There is a sense in
these writings that attachment researchers feel powerless to alter the diagnosis, now that it
exists. Yet it is also not clear that they would want a diagnostic entity that better reflected
attachment theory. In commentaries on the attachment disorder diagnosis by attachment
researchers, attachment is frequently depicted as a dyadic phenomenon, incommensurate
with individual-focused diagnostic systems like the DSM.512
The poor integration between clinical discourses of attachment disorder and the pre-
dominant focus of researchers on patterns of attachment has had several consequences. One
has been the still relatively weak research base on disinhibited social engagement, and es-
pecially the correlates and implications of these behaviours when shown by home-reared
507 Atkinson, L. (2019) Reactive attachment disorder and attachment theory from infancy to adolescence: re-
view, integration, and expansion. Attachment & Human Development, 21, 205–17.
508 Lyons-Ruth, K., Zeanah, C.H., & Gleason, M.M. (2015) Commentary: should we move away from an attach-
ment framework for understanding disinhibited social engagement disorder (DSED)? A commentary on Zeanah
and Gleason (2015) Journal of Child Psychology and Psychiatry, 56(3), 223–7.
509 Minnis, H., Marwick, H., Arthur, J., & McLaughlin, A. (2006) Reactive attachment disorder—a theoretical
model beyond attachment. European Child & Adolescent Psychiatry, 15(6), 336–42.
510 E.g. Zeanah, C.H., Smyke, A.T., Koga, S.F., Carlson, E., & Bucharest Early Intervention Project Core Group
(2005) Attachment in institutionalized and community children in Romania. Child Development, 76(5), 1015–28;
Lyons-Ruth, K., Zeanah, C.H., & Gleason, M.M. (2015) Commentary: should we move away from an attachment
framework for understanding disinhibited social engagement disorder (DSED)? A commentary on Zeanah and
Gleason (2015) Journal of Child Psychology and Psychiatry, 56(3), 223–7; Lyons-Ruth, K., Riley, C., Patrick, M.P., &
Hobson, R.P. (2019) Disinhibited attachment behavior among infants of mothers with borderline personality dis-
order, depression, and no diagnosis. Personality Disorders, 10(2), 163–72; Spangler, G., Bovenschen, I., Jorjadze,
N., et al. (2019) Inhibited symptoms of attachment disorder in children from institutional and foster care samples.
Attachment & Human Development, 21(2), 132–51. Spangler, like van IJzendoorn and Cassidy, is perhaps best con-
sidered an intermediate case for the heuristic contrast between second-and third-generation attachment researchers.
511 E.g. Sroufe, L.A., Duggal, S., Weinfield, N., & Carlson, E. (2000) Relationships, development, and psycho-
pathology. In A.J. Sameroff, M. Lewis, & S.M. Miller (eds) Handbook of Developmental Psychopathology (pp.
75–91). New York: Kluwer Academic/Plenum Press, p.83; van IJzendoorn, M.H. & Bakermans-Kranenburg,
M.J. (2003) Attachment disorders and disorganized attachment: similar and different. Attachment & Human
Development, 5(3), 313–20. See also Allen, B. (2016) A RADical idea: a call to eliminate attachment disorder and
attachment therapy from the clinical lexicon. Evidence-based Practice in Child and Adolescent Mental Health, 1,
60–71; Lyons-Ruth, K., Bureau, J.F., Riley, C.D., & Atlas-Corbett, A.F. (2009) Socially indiscriminate attachment
behavior in the Strange Situation: convergent and discriminant validity in relation to caregiving risk, later behavior
problems, and attachment insecurity. Development & Psychopathology, 21(2), 355–72; Granqvist, P., Sroufe, L.A.,
Dozier, M., et al. (2017) Disorganized attachment in infancy: a review of the phenomenon and its implications for
clinicians and policy-makers. Attachment & Human Development, 19(6), 534–58.
512 In this, researchers commenting on the attachment disorder diagnosis rather neglect the question of whether,
by late childhood and adolescence, expectations about the availability of attachment figures become a partial prop-
erty of individuals in much the same way as mental health symptoms.
94 John Bowlby
children.513 A second consequence has been that, in the gap left between the clinical and re-
search communities, the development of treatments for ‘attachment disorder’ have taken on a
life of their own, very largely unmoored to research.514 Clinicians wishing to identify insecurity
and attachment-related needs within a diagnosis-focused clinical culture have also been drawn
to use of the ‘attachment disorder’ label.515 Such clinical uses align with, and may have been sup-
ported by, occasional use of the term ‘disordered attachments’ by researchers to mean simply
problems in attachment relationships.516 Third, the disjointed relationship between clinical sys-
tems of diagnosis and the non-diagnostic categories of attachment has hindered the informed
integration of attachment research with diagnosis-focused clinical services and clinical training.
It can be especially hard for generalist clinicians to know what meaning attachment should have
for their work, given that it figures as both a rare clinical diagnosis and a transdiagnostic de-
velopmental perspective.517 Commentators have warned, however, that slippage between the

broad and circumscribed uses of the term ‘attachment disorder’ has contributed in some quar-
ters to an overdiagnosis of attachment disorders, misuse of appeal to attachment disorder in
psychological assessments for family courts, and neglect of children’s potential other psycho-
logical needs.518 The gap between clinical discourses and the research paradigm has also been
filled at times by inappropriate uses of the disorganised attachment classification, forced to play
the role of a quasi-diagnostic category in child welfare practice (Chapter 3).519
Some remaining questions
Continuity
Three limitations of Bowlby’s theorizing can be identified as of special significance for sub-
sequent attachment theory. A first limitation is Bowlby’s assumption that early adverse
513 Though see Scheper, F.Y., Groot, C.R., de Vries, A.L., Doreleijers, T.A., Jansen, L.M., & Schuengel, C. (2019)
Course of disinhibited social engagement behavior in clinically referred home-reared preschool children. Journal
of Child Psychology and Psychiatry, 60(5), 555–65.
514 Chaffin, M., Hanson, R., Saunders, B.E., et al. (2006) Report of the APSAC task force on attachment therapy,
reactive attachment disorder, and attachment problems. Child Maltreatment, 11(1), 76–89. Allen and Mercer
have criticized attachment disorder and interventions to resolve it as, at present, more pseudoscience than sci-
ence: Allen, B. (2018) Misperceptions of reactive attachment disorders persist: poor methods and unsupported
conclusions. Research in Developmental Disabilities, 77, 24–9; Mercer, J. (2019) Conventional and unconventional
perspectives on attachment and attachment problems: comparisons and implications, 2006–2016. Child and
Adolescent Social Work Journal, 36(2), 81–95. Nonetheless, perhaps it can be said that ‘attachment disorder’ has
been integrated into attachment theory as a characterization of children who have had little chance to form attach-
ments due to neglect or institutionalization.
515 Barnes, G.L., Woolgar, M., Beckwith, H., & Duschinsky, R. (2018) John Bowlby and contemporary issues of
clinical diagnosis. Attachment, 12(1), 35–47.

516 E.g. Fonagy, P., Target, M., Steele, M., et al. (1997) Morality, disruptive behavior, borderline personality dis-
order, crime, and their relationships to security of attachment. In L. Atkinson & K.J. Zucker (eds) Attachment and
Psychopathology (pp. 223–74). New York: Guilford, p.224.
517 Turner, M., Beckwith, H., Duschinsky, R., et al. (2019) Attachment difficulties and disorders. InnovAiT,
12(4), 173–9.
518 Woolgar, M. & Baldock, E. (2015) Attachment disorders versus more common problems in looked after
and adopted children: comparing community and expert assessments. Child and Adolescent Mental Health, 20(1),
34–40; White, S., Gibson, M., Wastell, D., & Walsh, P. (2019) Reassessing Attachment Theory in Child Welfare.
Bristol: Psychology Press.
519 Reijman, S., Foster, S., & Duschinsky, R. (2018) The infant disorganised attachment classification: ‘Patterning
within the disturbance of coherence’. Social Science & Medicine, 200, 52–8.
Some remaining questions 95
experiences would have strong continuities with later mental health, not only in the extent of
symptoms but also in the kinds of symptoms shown. This reflected Bowlby’s residual but strong
commitment to a psychoanalytic concept of ‘identification’ in which the parent is set up as a
model for the child, together with influence from Bandura’s Social Learning Theory and clin-
ical experiences: ‘I strongly suspect that the particular form of atypical care-eliciting behaviour
selected by a patient is greatly influenced by modelling, the term introduced by Bandura, and
roughly equivalent to identification, to describe adopting the same behaviour that one has ob-
served engaged in by others . . . More and more in work with parents I have been struck by the
extent to which they have adopted the same disciplinary procedures towards their children as
they themselves were subjected to—often despite their wish to behave quite otherwise.’520
Bowlby felt that there are many factors that keep developmental pathways stable, or even
intensify them, once they have begun to develop. Some of these are individual: Bowlby

was especially interested in the way that children’s expectations, once shaped, guide their
subsequent behaviour. He was confident that, shaped by past experiences, ‘present cogni-
tive and behavioural structures determine what is perceived and what [is] ignored, how a
new situation is construed, and what plan of action is likely to be constructed to deal with
it’.521 Bowlby highlighted the role of social factors in strengthening and steadying develop-
mental pathways: he had theorised that the effects of early long-term separations on later
mental health occur at least in part as a result of the fact that the clingy and difficult be-
haviour children show following such separations may elicit rejecting or hostile responses
from caregivers. Yet, beyond their independent influence, Bowlby felt that the reciprocal
reinforcement of psychological and social processes was especially critical to the stability
of developmental pathways: ‘Whatever family pressures led the development of a child to
take the pathway he is now on are likely to persist and so to maintain development on that
same pathway’,522 and ‘whatever expectations are developed during those years tend to per-
sist relatively unchanged throughout the rest of life’.523
Throughout his career, Bowlby was attentive to factors that might mediate, buffer, or break
continuity between early experiences and later mental health. Sroufe and colleagues are right
that ‘Bowlby is not deterministic but, rather probabilistic’ in his claims.524 Nonetheless,
Bowlby’s characterisation of how the attachment system changes in the context of matur-
ation or changes in social environment remained underdeveloped, with the amorphous con-
cept of ‘internal working models’ picking up the slack, but often not doing so effectively.525
He profoundly underspecified the relationship between attachment and personality traits: it
520 Bowlby, J. (1973) Letter to Scott Henderson, 30 July 1973. PP/Bow/J.9/98. See also Bowlby’s undated notes on
the concept of identification, PP/BOW/H.117-121. For instance, the note ‘Identification—Klein’, in which Bowlby
writes ‘Mrs K. assumes that the copying follows and is a manifestation of introjection. It seems more likely that the
identification is in fact the resultant of copying’ PP/Bow/H.118 (parentheses suppressed). For Bandura and the
concept of modelling see Bandura A. & Menlove F.L. (1968) Factors determining vicarious extinction of avoidance
behavior through symbolic modeling. Journal of Personality & Social Psychology, 8, 99–108.
521 Bowlby, J. (1973) Separation: Anger and Anxiety. New York: Basic Books, p.417.
522 Ibid.
523 Ibid. p.235; see also Bowlby, J. (1984) No such thing as a baby, March 1984. PP/Bow/F.3/136: ‘How to explain
persistence: a) Parents commonly continue in the way they started; b) virtuous or vicious circles pattern perpetu-
ates itself.’
524 Sroufe, L.A., Duggal, S., Weinfield, N.S., & Carlson, E. (2000) Relationships, development, and psycho-
pathology. In M. Lewis & A. Sameroff (eds) Handbook of Developmental Psychopathology, 2nd edn (pp.75–92).
New York: Kluwer Academic/Plenum Press, p.87.
525 Gilmore, K. (1990) A secure base. Parent– child attachment and healthy human development: by John
Bowlby. Psychoanalytic Quarterly, 59, 494–8; Crittenden, P.M. (1997) Truth, error, omission, distortion, and
96 John Bowlby
is not clear whether attachment sets early personality, whether it forms the first point in a
loose and interruptible chain of social experiences, whether it moderates the relationship
between innate qualities and behaviour, or all three (Chapter 5).526 Implicitly contrasting her
position to that of Bowlby, in 1988 Ainsworth reported that ‘increasingly, we are concerning
ourselves with increasing our understanding of change, and with defining the conditions
under which it takes place’.527 Later attachment researchers would find only partial support
for the strength of Bowlby’s emphasis on continuity, and would, as Ainsworth hoped, have
much to say about change (Chapter 4).
Furthermore, Bowlby’s overstrict opposition between ‘actual experiences’ and ‘fantasy’
led him to assume that experiences would be mirrored in symptoms, with adults ‘continuing
to respond in social situations with the very same patterns of behavior that they had devel-
oped during early childhood’.528 So, for instance, in his 1944 ‘Forty four juvenile thieves’

paper he claimed that children who do not receive affection develop an incapacity for af-
fection, which in turn contributes to their delinquency.529 And in ‘Violence in the family’,
from 1984, he proposed that children who experience one kind of abuse will then show this
same kind of abuse to their spouse and children.530 It is true that later research would sup-
port the idea that early experiences of neglect, violence, and highly unstable caregiving ar-
rangements will shape developmental pathways and contribute towards the probability that
an individual will develop mental health symptoms. However, Bowlby’s claim for a specific
mirror between early experiences and later symptoms has not been supported. His claims in
the 1940s linking early separations and stealing have not been supported. And later research
has found that early experiences of abuse make later abusive behaviour more likely, but not
necessarily the same form of abuse (Chapter 4).
Emotion
A second consequential limitation of Bowlby’s work also stemmed from his overstrict dis-
tinction between ‘actual experience’ and ‘fantasy’: his underelaboration of the emotional
deception: the application of attachment theory to the assessment and treatment of psychological disorder. In
S.M. Clancy Dollinger & L.F. DiLalla (eds) Assessment and Intervention Issues Across the Lifespan (pp.35–76).
London: Erlbaum; Bretherton, I. (2005) In pursuit of the internal working model construct and its relevance
to attachment relationships. In K.E. Grossmann, K. Grossmann, & E. Waters (eds) Attachment from Infancy to
Adulthood: The Major Longitudinal Studies (pp.13–47). New York: Guilford.
526 This concern has also been discussed by Weinfield, N.S., Whaley, G. & Egeland, B. (2004) Continuity, discon-
tinuity, and coherence in attachment from infancy to late adolescence: sequelae of organization and disorganiza-
tion. Attachment & Human Development, 6(1), 73–97; Fraley, R.C. & Shaver, P.R. (2008) Attachment theory and
its place in contemporary personality research. In O. John & R.W. Robins (eds) Handbook of Personality: Theory
and Research, 3rd edn (pp. 518–41). New York: Guilford. The problem stems in part from the fact that much of
Bowlby’s early thinking in the 1940s had taken ‘personality’ as its object; and this thinking was then transferred to
‘attachment’, without full elaboration of the relationship between the two constructs. On the concept of personality
in Bowlby’s early work see Bowlby, J. (1940) Personality and Mental Illness. London: Kegan Paul.
527 Ainsworth, M.D.S. (1988) Security. Unpublished discussion paper prepared for the Foundations of
Attachment Theory Workshop, convened for the New York Attachment Consortium by G. Cox-Steiner &
E. Waters, Port Jefferson, NY. http://www.psychology.sunysb.edu/attachment/online/mda_security.pdf.
Journal of Psychoanalysis, 44, 9–27, p.21. An early critic of Bowlby on this point was Rutter, M. (1972) Maternal
Deprivation Reassessed. London: Penguin.
Psycho-Analysis, 25, 107–28.


components of behavioural systems. In his writings, Bowlby was clear that affects are im-
portant components of behavioural systems, and attention to feelings do emerge from time
to time in his scholarly writings. He would often repeat the claim that ‘A person’s whole emo-
tional life—the underlying tone of how he or she feels—is determined by the state of these
long-term, committed relationships. As long as they are running smoothly the person is con-
tent; when they are threatened, he or she is anxious and perhaps angry; when the person has
endangered them by his or her own actions the person feels guilty; when they are broken,
the person feels sad; and when they are resumed he or she is joyful.’531 Such passages have
sometimes been cited, especially by those seeking a model for therapeutic practice in at-
tachment theory, to situate Bowlby as the quintessential theorist of human emotional life.
However, such accounts depend upon anachronism. Johnson, for example, describes attach-
ment for Bowlby as oriented by a striving for ‘felt security’ and ‘emotion regulation’, when in

fact these represent the ideas of the later attachment researchers such as Sroufe, Waters, and
Cassidy (Chapter 2, 3, and 4).532 In fact, Bowlby gives the causal role of emotions little space
compared to the behavioural and cognitive aspects of his theory, and as a consequence his
account is often oversimple. It is never quite clear what relationship Bowlby perceived be-
tween emotion and motivation, perhaps because he was often extrapolating across species
from ethological models.533 Compared to Sroufe (Chapter 4), Bowlby generally offered little
acknowledgement that moods and other felt states, such as confidence and fatigue, may arise
or be held in place by multiple factors.534 The caregiving system is a clear example: it is some-
times hard to recognise the actual dynamics of caregiving in Bowlby’s characterisation of the
caregiving behavioural system, for instance in how he neglects to give attention to both the
worries and relief of being needed.
Bowlby highlighted the role of emotion only: (i) in modulating the activation and ter-
mination of behavioural systems; and (ii) as a consequence of the satisfaction or non-
satisfaction of set-goals. Implicit in this is recognition that moods can offer us partially
incompatible visions of the world, including of the viability and meaning of other emotions.
Yet Bowlby rarely recognised, at least explicitly, the extent to which the fundamental shape
and deployment of behavioural systems, including the calibration of their set-goal, can be
conditioned by emotions. In his later writings especially, he came to consider the effects of
depression and hopelessness. However, the emotional components of behavioural systems
make them more varied and fragile than Bowlby generally acknowledged. This fragility can
be seen with horrible clarity if we consider how a behavioural system looks when it becomes
invested by shame or boredom, two emotions left aside by Bowlby (and, following him, by
531 Ibid. p.11.

532 Johnson, S.M. (2019) Attachment Theory in Practice. New York: Guilford, Chapter 2. See for instance the
anachronistic attribution of a focus on ‘felt sense’ of emotions such as security to Bowlby, rather than Sroufe and
Waters, on p.34.
533 Bell, D.C. & Richard, A.J. (2000) Caregiving: the forgotten element in attachment. Psychological Inquiry,
11(2), 69–83, p.73. It is interesting to watch Bowlby ponder the relationship between motivation and emotion in
his notes on ‘Emotion and feeling’, PP/Bow/H.5. These include reflections from the 1970s on emotion theorists
such as Tomkins and Izard. His most determinate statement on the relationship between motivation and emotion
in these notes seems to be the claim that the two are distinct, and that feelings contribute to the rise of motiv-
ations: ‘The feeling of being oriented & drawn toward some end gives rise to the sensation of desiring or wanting
to achieve that end.’ However, he criticized Izard for insufficient attention to ‘causal factors’, seemingly besides feel-
ings, ‘that activate an actual sequence of behaviour’.
534 One of Bowlby’s most sustained considerations of mood is actually in his notes on history-taking from the
mid-1930s: see ‘History taking; methods of examining’, PP/BOW/D.2/13. However, there he was interested in the
role of moods as indices of personality, rather than in their cause or effects.
98 John Bowlby
later empirical attachment research).535 Indeed, except in relation to outright conflict be-
tween behavioural systems, in general Bowlby was consistently poor in acknowledging and
characterizing states in which motivation becomes half-hearted, bendy, or sleepy, states of
velleity rather than concrete intention or outright inhibited action. He also did not con-
sider the way that some affects simultaneously undermine or sustain different behavioural
systems. For instance, excitement can facilitate the exploratory system, but qualifies fear.
Even during his lifetime, albeit writing in German, Mary Main criticised Bowlby’s account
of emotions for treating them as ‘precipitates’: an invisible part of a liquid that becomes a
visible solid only after a chemical reaction. The metaphor suggests the role of emotion as a
necessary part of behavioural systems, and actually integral to shaping their form—but as
only visible, at least to Bowlby, at the end of the reaction as a ‘consequence’ of behavioural
systems.536

In undated notes, likely from around 1955, Bowlby asked himself why he continually
underplayed fantasy and emotion in his writing. He concluded that the ‘reason is partly be-
cause we think psychobiological reactions have been neglected by psychoanalysis & partly
because research method has led this way. This is not to say that phantasy is unimportant—
obviously it comes in & tends to complicate things still further. Roots of phantasy are in the
psychobiological responses.’537 Bowlby evidently had a strong aversion to appeals to the idea
of ‘fantasy’. Speaking in an interview in 1986, he observed that ‘nowadays, the word fantasy
has become used in the analytic world to mean any cognitive process. It means conscious
wish, it means a daydream, it means expectation.’538 He regarded this loose use of the con-
cept as obscuring phenomena that he regarded as both important and distinct. He also did
not like the way that the term ‘fantasy’ collapsed distinctions between different ways of re-
lating to the future, such as expectation, wishing, and planning. Such distinctions, he felt,
were important for acknowledging the accounts of patients, and maintaining a ‘neutral and
empathetic position’ in relation to what they said about their experiences.539 In conversa-
tions with junior clinicians at the Tavistock, Bowlby used to ask that conversations of fantasy
were ‘parked’, to see how or whether the matter in question could be addressed with greater
precision in other ways.540
Bowlby’s opposition between ‘actual experience’ and ‘fantasy’ was criticised by Ainsworth
in their correspondence.541 Ainsworth felt that Bowlby neglecting the emotional aspects of
535 By contrast, shame has been a central theme in clinical elaborations of Bowlby’s ideas by Allan and Judy
Schore and by Dan Hughes and colleagues: Schore, A.N. (1991) Early superego development: the emergence of
shame and narcissistic affect regulation in the practicing period. Psychoanalysis and Contemporary Thought, 14(2),
187–250; Hughes, D., Golding, K.S., & Hudson, J. (2015) Dyadic developmental psychotherapy (DDP): the devel-
opment of the theory, practice and research base. Adoption & Fostering, 39(4), 356–65.
536 Main, M. (1977) Sicherheit und wissen. In K.E. Grossmann (ed.) Entwicklung der Lernfahigheit in der
sozialen Umwelt (pp.47–95). Munich: Kindler.

537 Bowlby, J. (undated) Unpublished notes from his filing cabinets. c. 1955. PP/Bow/H115.
538 Bowlby, J. (1986) An interview with John Bowlby on the origins and reception of his work. Free Associations,
1, 36–64, p.42.
539 Bowlby, J. & Parkes, C.M. (1970, 1979) Separation and loss within the family. In The Making and Breaking of

540 This is described amusingly in Issroff, J. (2005) Donald Winnicott and John Bowlby: Personal and Professional
Perspectives. London: Karnac: ‘Bowlby . . . “parked” the term “fantasy” in what I always imagine to be a quite cap-
acious parking lot in his mind’ (56).
541 Ainsworth, M. (1971) Comments on the manuscript of Attachment Volume 2. Letter to Bowlby, 16 August
1971. PP/BOW/K.5/62: ‘You yourself, naturally, push your own view. And yet I really am convinced (and I am sure
that you basically agree) that in many cases both kinds of “dynamics” are at work . . . It would take only a few alter-
ations of turns of phrase to leave the classical Freudians and Kleinians a little more opportunity to feel that you are
asking them to extend their view rather than to abandon it.’
behavioural systems, ‘quite deliberately’, in order to avoid ceding ground to those trends in
psychoanalytic thought that privileged fantasy over past experiences.542 In the 1980s, this
motivation likely aligned with wider trends in cognitive science, in which thought and feeling
were opposed, downplaying their interrelation.543 To give an example, in ‘Developmental
psychiatry comes of age’ in 1988 Bowlby stated that ‘to Sigmund Freud is due the credit for
having emphasised the influence on how people think, feel, and behave that is exerted by
their internal world—namely, by the way they perceive, construe, and structure the events
and situations they encounter’.544 However, Bowlby was well aware that how people ‘think,
feel and behave’ is not equivalent to how people ‘perceive, construe and structure’. He equated
them deliberately and strategically, to ensure that the cognitive aspect of behavioural sys-
tems, and the role of these in shaping emotions, would not be missed. The price, which ap-
peared to be knowingly accepted by Bowlby, was that the role of emotion within behavioural

systems, and the relationship between emotion and motivation, would remain hazy. This
would leave problems for later attachment theory, for instance in how exactly to conceptu-
alise affects at odds with expressed attachment behaviour (Chapter 3).
Predation and evolution
Bowlby’s familiarity with ethology and developments in evolutionary biology, by his own
admission, dropped away sharply after the publication of Attachment, Volume 1. A conse-
quence was that aspects of his theory remained partially trapped in amber, and especially his
knowledge of developments in evolutionary biology.545 His account of attachment behaviour
as a repertoire developed in human evolutionary history to sustain survival was a powerful
cross-disciplinary integration of forms of knowledge, and an important plank in the emer-
gence of developmental psychology as a subdiscipline informed by evolutionary theory. Yet
Bowlby tended to rigidly emphasise that the evolutionary purpose of the attachment system
was to save young children from predation by ensuring proximity to at least one adult.
Downplayed in such an account, however, were the other evolutionary advantages of being
physically close to a caregiver—advantages that Bowlby apparently knew. One such advan-
tage is that closeness with the caregiver helps support psychobiological regulation. Bowlby
stated this himself in ‘The nature of the child’s tie to his mother’ in 1958, citing with approval
a discussion by Winnicott.546 The notion is also implied in Bowlby’s use of a thermostat as the
central model or metaphor for the attachment system. Another evolutionary advantage of
the attachment system is that proximity with a caregiver offers opportunities for nurturance
543 Waters, E., Corcoran, D., & Anafarta, M. (2005) Attachment, other relationships, and the theory that all good
things go together. Human Development, 48(1–2), 80–84: ‘Bowlby’s attachment theory remains a work in pro-
gress . . . his notion of attachment working models was limited to what could be done with the cognitive psychology
of the day’ (82).
545 Bowlby, J. (1990, 2011) John Bowlby: interview by Leonardo Tondo. Clinical Neuropsychiatry, 8(2), 159–
71: ‘The main thing about the monkey work has been that, with fairly rigorous experimental designs and methods,
they have demonstrated the ill effects of separation and its obvious consequences (Hinde 1966). It is a huge literary
reserve which I was fairly familiar with in the 1960s because it was very dramatic but I haven’t kept up’ (166).
546 Bowlby, J. (1958) Nature of the child’s tie to his mother. International Journal of Psycho-Analysis, 39, 350–73,
p.356, citing Winnicott, D.W. (1948) Pediatrics and psychiatry. British Journal of Medical Psychology 21, 229–40.
100 John Bowlby
and learning. In 1965, Ainsworth wrote to Bowlby arguing against the exclusive focus on
predation as the evolutionary function of attachment, and giving the examples of nurturance
and learning.547 These points were not taken up by Bowlby, except tacitly in his characterisa-
tion of attachment behaviour as directed towards someone ‘stronger and/or wiser’. However,
it would appear from correspondence that he regarded ‘wiser’ to be less a description of the
actual properties of an attachment figure, and more a cognitive bias in the perception of the
attachment figure when the attachment system is activated.548
In 1970, whilst still a doctoral student with Ainsworth, Mary Main nonetheless took
Bowlby’s side and argued that a desire for learning could not account for the clinging re-
sponse as part of the attachment system. Opportunities for learning might be ‘correlates
that “ride on” ’ the evolutionary function of proximity for protection.549 However, Main’s
support for Bowlby’s position still offers no argument against physiological regulation

and nurturance as possible evolutionary advantages of the attachment system. From the
1980s, physiological regulation and nurturance would be discussed by Myron Hofer as two
‘hidden regulators’ missed by Bowlby’s exclusive focus on the attachment system as protec-
tion against predation.550 Hofer claimed that Bowlby had ignored the ‘chronic’ advantages
that physiological regulation and nurturance offer. The repetitive sequences of physiological
regulation and nurturance may even play a role in constituting attachment as a coherently
organised system, and infant separation anxiety may reflect a withdrawal response to these
sources of regulation as much as a desire for protection. Furthermore, Hofer’s perspective
implied that attachments are not formed solely on the basis of familiarity, as Bowlby gener-
ally suggested, but with familiar sources of biosocial homeostasis. Even an abusive caregiver
can be an attachment figure, if they are familiar and provide sufficient physiological regula-
tion and nurturance to be felt as a source of regulation.
Bowlby entirely accepted Hofer’s claims. In a warm letter to Hofer in 1983 he agreed on all
points. Furthermore, he offered the notable claim that in humans as well as animals, ‘internal
representations are usually a poor second best’ within the operation of behavioural systems,
547 Ainsworth, M. (1965) Letter to John Bowlby, 7 October 1965. Mary Ainsworth papers, Box M3168, Folder 2: ‘I
was very pleased with your chapter on Instinct Theory . . . There was one point at which I was jerked out of my “yes”
attitude. On page 60 you state that the main function of attachment behaviour is to ensure safety of the young—
safety from predators. Of course, you will develop this point fully in your revision of the “Child’s Tie” paper. I did
not say “no” to your statement, but I did think about functions of maternal behaviour other than protection from
predators—specifically nurturance of the young, but also training.’ Several of Ainsworth’s students have followed
her lead on this matter rather than Bowlby’s position. See e.g. Waters, E. (2008) Live long and prosper: a note on at-
tachment and evolution. http://www.psychology.sunysb.edu/attachment/gallery/live_long/live_long.html.
548 Bowlby, J. (1988) A Secure Base. London: Routledge p.121. Bowlby, J. (1985) Letter to Malcom West, 13
November 1985. PP/Bow/J.9/201: ‘I suspect that, in the specific situations in which A’s attachment behaviour is ac-
tivated, A always regards the partner B as definitely the more competent to deal with them. And also that, in other
situations, the roles of the partners are reversed. Thus, whilst overall each may regard the other as no more compe-
tence, etc. than the self, whenever attachment behaviour is activated the roles become complementary rather than
reciprocal.’
549 Main, M. (1970) Infant play and maternal sensitivity in primate evolution. PP/Bow/J.4/1: ‘Alternate bio-
logical functions have been proposed, chief among them that mother and infant will come together in order that
the infant may learn from the mother the behaviours which will promote its survival. The conditions which acti-
vate and terminate attachment behaviour and caretaking behaviour, such as alarming events in the environment
and the actual attainment of contact or proximity, indicate that Bowlby’s explanation is preferable.’
550 Hofer, M.A. (1983) On the relation between attachment and separation processes in infancy. In R. Plutnik
(ed.) Emotion: Theory, Research and Experience, Vol. 2 (pp.199–219). New York: Academic Press; Hofer, M.A.
(1984) Relationships as regulators: a psychobiologic perspective on bereavement. Psychosomatic Medicine, 46(3),
183–97. For a review of the literature on hidden regulators and attachment see Polan, H.J. & Hofer, M. (2016)
Psychobiological origins of infant attachment and its role in development. In J. Cassidy & P.R. Shaver (eds)
Handbook of Attachment, 3rd edn (pp.117–32). New York: Guilford.
Conclusion 101
which likely much more ‘rely on sensorimotor pathways’.551 However, Bowlby did not inte-
grate such insights into his subsequent published writings on the evolutionary function of
the attachment behavioural system. A hint of a somewhat qualified position on the function
of the attachment system can be seen only in Bowlby’s last publication, and even this is quite
limited: ‘It contributes to the individual’s survival by keeping him or her in touch with one or
more caregivers, thereby reducing the risk of harm, for example from cold, hunger or drowning
and, in the human’s environment of evolutionary adaptedness, especially from predators.’552
Bowlby’s almost exclusive emphasis on proximity and predation would be criticised in
turn by nearly all of the second generation of attachment researchers as part of marking
their own emerging voices in the field. Despite this, in the decades following Bowlby’s death
in 1990, it took a long while for the theory to receive even partial update in light of devel-
opments in evolutionary biology.553 It has not helped that ethology, the original alloparent

of the attachment paradigm, has gone into decline since the 1990s, as interest shifted from
behavioural sequences to other levels of analysis such as sociobiology and, more recently,
gene–environment coevolution.554 However, the failure of further engagement with chan-
ging paradigms in biology has been fed by an internal dynamic within the field of attachment
research: since Bowlby, authority about the biological function of behaviour has been part
of leadership of the developmental tradition of attachment research, since in part it defines
the nature of attachment as a research object and therefore the questions to be asked. With
biological theory mixed up with authority and power, this dynamic has helped insulate at-
tachment theory against a continued update of its models, in Bowlby’s era and subsequently.
Bowlby’s ossified account of the relationship between attachment and evolutionary theory
contained in the predation model would eventually lead to disagreements between Bowlby
and Main, with lasting consequences for attachment as a research programme (Chapter 3).
It would later also be significant ammunition for Peter Fonagy and Chloe Campbell’s call for
a new paradigm to replace attachment research whilst absorbing its strengths (Chapter 6).
Conclusion
See Table 1.3 in the Appendix to this c hapter for consideration of key concepts discussed
here.
551 Bowlby, J. (1983) Letter to Myron Hofer, 29 July 1983. PP/Bow/J.9/102. This correspondence supports the
claim of Cassidy and colleagues that Bowlby downplayed non-representational affective and physiological pro-
cesses as part of his focus on cognitive processes: Cassidy, J., Ehrlich, K.B., & Sherman, L.J. (2013) Child–parent
attachment and response to threat: a move from the level of representation. In M. Mikulincer & P.R. Shaver (eds)
Nature and Development of Social Connections: From Brain to Group (pp.125–44). Washington, DC: American
Psychological Association.
552 Bowlby, J. (1991) Ethological light on psychoanalytic problems. In P. Bateson (ed.) The Development and
Integration of Behaviour: Essays in Honour of Robert Hinde (pp.301–14). Cambridge: Cambridge University
Press, p.306.
553 An update in light of evolutionary biology would partially take place from the late 1990s especially through
the work of Jay Belsky. Belsky, J. (1999) Modern evolutionary theory and patterns of attachment. In J. Cassidy &
P.R. Shaver (eds) Handbook of Attachment (pp.141–61). New York: Guilford.
554 Griffiths, P.E. (2008) Ethology, sociobiology, and evolutionary psychology. In S. Sarkar & A. Plutynski (eds)
A Companion to the Philosophy of Biology (pp.393–414). Oxford: Blackwell. There remain some ethologists who
have continued to conduct empirical studies with a focus on sequences of behaviour in the manner of Lorenz and
Tinbergen. See e.g. Suomi, S. (2016) Attachment in rhesus monkeys. In J. Cassidy & P.R. Shaver (eds) Handbook of
Attachment (pp.133–54), 3rd edn. New York: Guilford; Polanco, A., Díez-León, M., & Mason, G. (2018) Stereotypic
behaviours are heterogeneous in their triggers and treatments in the American mink, Neovison vison, a model car-
nivore. Animal Behaviour, 141, 105–14.
102 John Bowlby
Sroufe argued that ‘as is always characteristic of development, whether in an individual

or in a scientific field, Bowlby’s work both integrates and transforms what went before
creating an alternative way of viewing the world without leaving behind critical insights
contained in previous viewpoints’.555 Bowlby described himself as speaking a ‘hybrid,
bastard language’, bringing together concepts from ethology and psychoanalysis, as well
as terms from ordinary language.556 In creating this hybrid combination of discourses,
Bowlby’s theory could benefit from the precision of Hinde’s conceptualisation of behav-
ioural responses, their conditions and their conflicts; the emotional punch and drama
of Kleinian theory of intense emotions and defences invested with his own personal and
clinical experiences; and the explosive obviousness of the ordinary language connota-
tions of ‘attachment’, ‘separation’, ‘loss’, ‘mother’, ‘love’, and others.557 Attachment theory
was able to make an iconic assemblage of the precision, emotional power, and intuitive-

ness of these elements in ways that quickened the potential of each.558 Yet, in making an
identity, it also retained qualities of these constituent differences. Various faces of this
assemblage offered the theoretical basis for an empirical research paradigm, an approach
to clinical work with children and adults, and a popular discourse regarding child devel-
opment and family life.
If Bowlby had remained a theorist only attentive to the precisely defined ‘following re-
sponse’, and had not plugged these ethological observations of behavioural sequences into
the evocative world of psychoanalytic concerns, the theory would have been much less rich.
Hinde himself would later concede this in reflections many years after Bowlby’s death.559
And if Bowlby had achieved his synthesis of ethology and psychoanalysis, but had com-
municated only in the language of these two disciplines, his ideas would have had a very
limited audience. Bowlby’s appeals to ordinary language and cultural stereotypes in writing
for popular forums helped set his theory alight; it glowed to widespread visibility, even as its
qualifications and technical subtlety burned away as fuel. The manner of Bowlby’s popular
writings helped create what Bourdieu termed ‘allodoxia’, a ‘light’, commodified version of a
more complex cultural form, appealing to a wider base of constituents without the tools or
means to access the original.560 This is in contrast to forms of popularisation that attempt
to convey the intricacy and technical quality of the form, generally by unpacking terms
and explaining in longhand or prioritizing the most essential elements, using an accessible
style and stories to keep the audience engaged.561 Certainly not all works of popular science
are allodoxia. What characterises allodoxia in psychology is the circulation of a simplified
account of the human mind as if it had the same meaning as the technical account of empir-
ical researchers.
555 Sroufe, L.A. (1986) Bowlby’s contribution to psychoanalytic theory and developmental psychopathology.
Journal of Child Psychology and Psychiatry, 27, 841–9, p.841.

556 Bowlby, J. cited in J.M. Tanner & B. Inhelder (eds) (1956) Discussions on Child Development: Proceedings of
the WTO Study Group of the Psychobiological Development of the Child, Vol. 1. London: Tavistock, p.184.
557 Bowlby’s passionate focus on what an assemblage can do, and not of the price of its formation or use, can be
seen clearly in a late interview: Bowlby, J. (1986) An interview with John Bowlby on the origins and reception of
his work. Free Associations, 1, 36–64: ‘What’s important about a person’s work is what they have contributed. I don’t
care two pins about their mistakes or their shortcomings or their omissions—what have they contributed?’ (63).
558 Cf. Bartmanski, D. (2012) How to become an iconic social thinker: the intellectual pursuits of Malinowski
and Foucault. European Journal of Social Theory, 15(4), 427–53.

559 Hinde, R. (2005) Ethology and attachment theory. In K. Grossman, E. Waters, & K. Grossman (eds)
Attachment from Infancy to Adulthood: The Major Longitudinal Studies (pp.1–12). New York: Guilford, p.9. See
also Petters, D.D. (2019) The attachment control system and computational modeling: origins and prospects.
Developmental Psychology, 55(2), 227–39.
560 Bourdieu, P. (1979, 1984) Distinction, trans. R. Nice. Cambridge, MA: Harvard University Press.
561 Bucchi, M. (2013) Style in science communication. Public Understanding of Science, 22(8), 904–15.
Conclusion 103
All science has a ‘price for full entry’ in terms of theoretical and technical competence and
commitment.562 This price may be paid by specialists through their training as part of formal
qualifications or accreditation. Some or all access may also be gained by non-specialists, for
instance through reading, depending on the structural barriers to entry and how much so-
cialisation in tacit skills is required.563 In both cases there remains some recognition of the
distinction between technical and non-technical use of concepts. By contrast, the acces-
sibility of elements of Bowlby’s popularizing discourses from the 1950s and 1960s offered
cut-price tickets to attachment theory, even though these only granted access to a fairly anti-
quated portion of the fairground, and—the source of much later trouble—looked much the
same as full entry tickets. The cut-price popular discourse of attachment was evocative and
underdetermined, as well as having the appearance of scientific credibility. This gave it flexi-
bility, urgency, and reach for these diverse constituents concerned with speaking about the

nature of family relationships and child development.564 The language of attachment theory
can be experienced as running towards us, calling our name. This was deliberate. Both the
circulation of half-truths about attachment and their rhetorical insistence can be regarded,
in part, as a by-product of Bowlby’s marketing strategy for his scholarly thinking.565
In one of his final papers, Bowlby argued that science has two stages, drawing ideas
from the philosopher of science Karl Popper. In the first stage, the task is to frame hypoth-
eses. Where possible, this process benefits from ‘detailed and first-hand knowledge of the
problem . . . together with a dose of intuition and imagination’.566 Raw materials such as or-
dinary language and clinical anecdote can be relevant elements for this stage of work, along-
side all the other resources of the humanities. Bowlby, in fact, chided behaviourism for failing
to see such resources as relevant, making the paradigm rather arid and, at times, ‘barren’ des-
pite its rigour.567 Nonetheless, even the thin theory of behaviourism, he felt, was better than
atheoretical research in psychology, since theory is necessary for the effective planning and
interpretation of empirical research as a cumulative endeavour, as well as the design of inter-
ventions and public health policies.568 Looking back over his long career, Bowlby offered the
self-criticism that his work had displayed an ‘absence of a follow-through’, remaining in the
first stage of hypothesis-generation.569 It would only be with the work of Bowlby’s colleague
Mary Ainsworth that attachment theory would become the basis for attachment research
as an empirical research paradigm. Ainsworth’s contribution remained exploratory. It was
also work in the context of hypothesis-generation, though it was sometimes misrecognised
as hypothesis-testing research. However, Ainsworth’s development of procedures for meas-
uring individual differences in attachment and related constructs would prove the bedrock
of a hypothesis-testing tradition of attachment research within developmental science.
562 Bourdieu, P. (2004) The Science of Science and Reflexivity, trans. R. Nice. Chicago: University of Chicago
Press, pp.53–4.
563 Collins, H. & Evans, R. (2018) A sociological/philosophical perspective on expertise: the acquisition of
expertise through socialization. In K. Anders Ericsson, R.R. Hoffman, A. Kozbelt, & A.M. Williams (eds) The
Cambridge Handbook of Expertise and Expert Performance (pp.21–32). Cambridge: Cambridge University Press.
564 On the practical logic of underdetermined discourses: de Certeau, M. (1984) The Practice of Everyday
Life, trans. S.F. Rendall. Berkeley: University of California Press; Bourdieu, P. (2000) Pascalian Meditations.
Stanford: Stanford University Press; Mercer, J. (2015) Revisiting an article about dyadic developmental psycho-
therapy: the life cycle of a Woozle. Child and Adolescent Social Work Journal, 32(5), 397–404; Alexander, J.C. (2016)
Dramatic intellectuals. International Journal of Politics, Culture, and Society, 29(4), 341–58; Pettit, M. & Young, J.L.
(2017) Psychology and its publics. History of the Human Sciences, 30(4), 3–10.
565 See also Duniec, E. & Raz, M. (2011) Vitamins for the soul: John Bowlby’s thesis of maternal deprivation, bio-
medical metaphors and the deficiency model of disease. History of Psychiatry, 22(1), 93–107.
566 Bowlby, J. (1988) Where science and humanism meet. Group Analysis, 21, 81–8, p.81.
567 Ibid.
568 Bowlby, J. (1982) Attachment and loss: retrospect and prospect. American Journal of Orthopsychiatry, 52(4),
664–78, p.676.
569 Bowlby, J. (1988) Where science and humanism meet. Group Analysis, 21, 81–8, p.81.
104 John Bowlby
Appendix
Table 1.3 Some key concepts in Bowlby’s writings
Concept Mistaken for Technical meaning
Attachment The instinctive The ‘attachment system’ is a way of describing a form of

system relationship motivation. The motivation is activated when a person is
with a familiar alarmed. When the person feels that a particular, familiar
caregiver person—or familiar people—is available and responsive to
their concerns, the motivation is reduced. Where the system
is strongly activated, some form of contact is generally sought

(though this contact may still be verbal rather than touch).
This motivation has some basis in evolution, and for this
reason is especially easy for humans to develop. However,
a great deal about this motivation, including exactly the
conditions that prompt and terminate it, are shaped deeply
by experiences in relationships. This is why it is misleading to
think of attachment as an ‘instinct’.
The attachment system has some characteristic behaviours,
but in principle any behaviour can be recruited that helps
achieve the goal of attachment figure availability.
Attachment researchers have debated the conditions that lead
to the satisfaction of the attachment system and a reduction in
display of attachment behaviour. The physical and attentional
availability of a familiar caregiver has been emphasised
as terminating conditions in infancy; other attachment
researchers, since Sroufe and Waters, have emphasised the
infant’s ‘felt security’ as the terminating conditions for the
system.
In general it is agreed that experience, circumstance,
and culture can all shape the conditions under which the
attachment system is activated, the forms of behaviour
recruited by the attachment system, and how these are
expressed.
Attachment Pre-set Anything in principle can be an attachment behaviour. All
behaviour behaviours that is required is that the behaviour should be clearly directed
that express towards gaining the availability and responsiveness of a
attachment as familiar person or familiar people. It is not an ‘instinctual’
an instinct pre-set pattern of behaviour.
Attachment behaviours are observable, whereas the
motivation they are presumed to express is inferred.
The term ‘attachment behaviour’ was used in two different
ways by Bowlby. His most common use of the term was
to refer to proximity-seeking and contact-maintaining
behaviours, such as smiling, crawling towards the caregiver,
clinging, and directed cries to attract the caregiver’s attention.
These behaviours were understood as direct expressions of the
attachment behavioural system.
Appendix 105
Table 1.3 Continued

Sometimes, however, Bowlby also used the term ‘attachment
behaviour’ to refer to any behaviour that occurs in the context
of the activation of the attachment system. This could even
include withdrawal from the caregiver and attempts at self-
reliance by a child who had found that seeking proximity
with their caregiver when alarmed or distressed would be
counterproductive.
Often these two meanings are aligned. However, some
behaviours, such as when a child shows caregiving behaviour
towards a parent, are attachment behaviours in the second
sense, but not the first sense. At times this has caused a lack of

clarity in discussing such behaviour, its eliciting conditions,
and its relationship with the attachment behavioural system.
Attachment Parent–child Bowlby characterised relationships and their qualities as
bond bonding diverse. Attachment dynamics characterised only some
relationships and not others. He therefore distinguished the
broad class of affectional bonds, in which members are special
to one another and seek to remain in contact. Within this
broad class are relationships with attachment dynamics. These
are characterised by the fact that the other person is taken to
be the object of the attachment behavioural system: there is a
disposition to seek this person under conditions of alarm, and
a sense of security when this person is reliably available and
responsive to concerns.
The attachment bond is distinct from ‘parent–child bonding’,
the process by which parents develop an affectionate bond
with their child and take the child as the familiar target of the
caregiving behavioural system.
Attachment An absolute Being an attachment figure is not a yes/no situation. Bowlby
relationship state proposed that an attachment relationship is present to the
characteristic extent that an individual is disposed to seek the availability of
of a child’s a familiar other when alarmed. This disposition may exist even
relationship if the other is rejecting or abusive.
with their Bowlby felt that an individual could have a variety of
mother attachment relationships—including wider kin (e.g.
grandparents), divine beings, and also a person’s relationship
with their physical home. However, he believed that evolution
had primed humans to develop these dynamics especially with
our familiar caregivers from childhood. Other relationships
would be more contingent in the degree to which these
dynamics would be expected.
(continued )
106 John Bowlby
Table 1.3 Continued

Major Occasions when Attachment researchers sometimes discuss children
separation the child and experiencing ‘major separations’. This is a technical term,
parent are not which can easily be confusing. What makes a separation
together ‘major’ is that the child is alarmed by the absence of their
attachment figure, and this alarm continues for long enough
that the attachment behavioural system then becomes
chronically unresponsive for a long period. In effect, the child
appears to give up searching for, calling, or expecting the
parent to return. The result is that even when the caregiver
is available, the child is not able to use them—at least for
a time—to regulate distress. Or, in Bowlby’s terms, the

behavioural system becomes chronically unresponsive for a
period to cues for its activation and/or termination.
The classic case of a major separation was the long-term
hospitalisations observed by James Robertson in the 1950s,
in which there were no or few visits to young children over
several months.
Attachment researchers do not regard some use of daycare as a
‘major separation’ in the technical sense.
Monotropy The exclusivity Bowlby introduced the term ‘monotropy’ in 1958 with the
and priority of intention that it would refer to particular, special relationships,
child–mother shaped by time and habit. Unfortunately, the literal meaning
attachment of the term is ‘mono’ (one) + tropy (turning to). This gave the
mistaken impression he meant the exclusive importance of
one caregiver for children.
Bowlby later mostly abandoned the term, given the extent of
misapprehension of his meaning.
Caregiving The natural The ‘caregiving system’ is a way of describing a kind of
system capacity of motivation. A motivation to help is activated when a child or
parents, other person in our care is alarmed, and terminated when we
especially have identified and responded to what we understand to be
mothers, to their concerns.
care for their In his initial description of caregiving as a behavioural system,
children Bowlby focused on the caregiver’s motivation to retrieve
infants who are alarmed or in trouble. However, later in his
career he described caregiving as more broadly concerned
with encouragement, support, help, and protection.
Effects The notion that In his early writings, Bowlby sometimes made claims that
of early early social implied that every child who receives poor care or who
experience experience can experiences major separations will develop social and
be expected at emotional problems. From the 1970s onwards, he was more
an individual careful, claiming that—on average—poor care or major
level to strongly separations are likely to increase the chances of later social and
determine emotional problems.
later emotional Later attachment researchers have synthesised findings from
and social many studies through meta-analysis, indeed finding that early
experience care does have effects on later socioemotional development,
but that early experience does not determine later outcome,
and that there are important mediators and moderators.
Appendix 107
Table 1.3 Continued

Internal Representations This is perhaps the single most confusing concept used
working of caregivers, by attachment researchers. Bowlby used the term in two
model which become different ways.
generalised to Firstly, he intended it only to mean that the way the
all relationships attachment system works depends on expectations based
with on previous experiences of interaction with caregivers in
development childhood—and with partners and friends in adulthood. So
a synonym for the internal working model, in this sense, in
ordinary language is simply ‘expectations’. Bowlby’s point was
that expectations about early relationships can play a role in
shaping our assumptions about later social relationships and

interactions.
Both humans and non-humans will have expectations about
our caregivers or partners and their availability. However,
humans also develop elaborated cognitive and cultural
representations about ourselves and our attachment figures.
These include narratives and images about the availability
of attachment figures, and how we think they feel about us.
A second use of the term ‘internal working model’ by Bowlby
was therefore to refer to the specific symbolic and affective
representations made by humans about attachment figures
and their availability, and the efficacy of attempts to seek them
when alarmed.
When Mary Main and colleagues introduced the Adult
Attachment Interview (AAI) in the 1980s, they documented
individual differences in the coherence of autobiographical
accounts by participants of their childhood. Initially she
referred to these differences in speakers’ narratives as
reflecting differences in ‘internal working models’ about
attachment. By the 1990s, she had abandoned and criticised
the use of the term ‘internal working model’ to refer to these
differences. Given the two different meanings above, she felt
that the term was confusing and misleading for describing
what the AAI was measuring. Main preferred to characterise
individual differences in the AAI as reflecting ‘states of mind
regarding attachment’. However, many attachment researchers
still refer to the AAI as measuring internal working models.
Segregation The inhibition In Bowlby’s later writings the term ‘segregation’ is used to
of information; refer to a coping strategy in which some information is filtered
essentially out of experience. This can be minor and remain flexible: the
the same as filter can be raised or dropped as needed. Or the flow of
dissociation information to or from whole behavioural systems can be
blocked over a long period, regardless of the circumstances.
Bowlby distinguished two forms of segregation. A first was
‘defensive exclusion’. Here the filter is placed on perception. So,
certain things in the world may not be noticed. Or if noticed,
they may not prompt a response. The paradigmatic form of
defensive exclusion is the infant in an avoidant attachment
relationship, who directs attention away from their caregiver
on reunion. In doing so, they filter out information about their
situation that might otherwise prompt the activation of the
attachment system.
(continued )
108 John Bowlby
Table 1.3 Continued

A second form of segregation was ‘cognitive disconnection’.
Here the filter is placed on memory. So, certain memories may
not be available. Or if available, they may not be tagged with
well-defined and accurate meanings.
The segregation of information about attachment figures was
anticipated by Bowlby to contribute to an individual holding
multiple incompatible perceptions and expectations of these
figures.
Bowlby’s primary book on segregation, defensive exclusion,
and cognitive disconnection remained unpublished. The
terms appear only briefly in his published works. As a

result, these terms are only used rarely now by attachment
researchers.
Multiple Bowlby felt In his early writings Bowlby sometimes made claims that
attachments that a child suggested a child should always be cared for by their mother.
should always However, he subsequently regretted these claims. In his
be cared for by mature writing Bowlby saw value in a child having access to
their mother multiple secure bases and safe havens, and did not think that
one attachment would be at the expense of another.
Bowlby’s final statement was that the attachment system
‘contributes to the individual’s survival by keeping him or her
in touch with one or more caregivers’. The idea of attachment
relationships as a network was developed by subsequent
researchers such as Avi Sagi-Schwartz and Marinus van
IJzendoorn.
Illustrative statement: ‘Bowlby argued that early experiences of care within monotropic attachment relationships,
such as potential separations, contribute to the later integration or segregation of mental processes, and to the
child’s own caregiving behaviours when they reach adulthood.’
Mistaken for: Bowlby argued that the quality of the mothers’ care, including separations such as maternal employ-
ment, will determine children’s later mental health and how they parent their own children.
Technical meaning: Bowlby argued that a child’s experiences in specific relationships with familiar caregivers will
shape and calibrate the operation of the attachment behavioural system. This process may be disrupted, however,
by major separations such as being hospitalised without visitation. Disruptions to the attachment system may in-
fluence the formation and operation of later mental processes, including coping strategies. Where disruptions lead
to fixed or extreme distortions or blockages of attachment-relevant information, a predisposition to mental health
problems may be anticipated—though this will be seen at a population level rather than in any given individual.
Behaviours, affects, and cognitions that form components of the attachment behavioural system may influence the
caregiving system when it develops, since elements may be inherited by or inform the latter system (e.g. expect-
ations about what intimacy entails).
2
Mary Ainsworth and the Strange
Situation Procedure
Biographical sketch
Mary Salter took her undergraduate and graduate degrees in psychology in the 1930s at the

University of Toronto. Her mentor was the director of the Institute of Child Study, William
Emet Blatz. Salter completed her doctorate in 1940, based on Blatz’s ideas. Following time
in the Canadian Women’s Army Corps during World War II, she rejoined the University
of Toronto as assistant professor in psychology, and worked with Blatz in developing self-
report measures of security and insecurity. During World War II, she worked in personnel
for the Canadian Women’s Army Corps, attaining the rank of Major. She married Leonard
Ainsworth in 1950, and then followed him to London, which is where she met John Bowlby
(Chapter 1). In early 1954, Leonard Ainsworth accepted a position in Uganda. Whilst in
Uganda with her husband, Mary Ainsworth conducted an observational study of 26 moth-
ers and their infants living in six villages near Kampala. Mary and Leonard Ainsworth left
Uganda for Baltimore in late 1955. Mary Ainsworth gained a permanent academic position
at Johns Hopkins University in 1958. With funds from the William T. Grant Foundation,
she began a study in 1963 of Baltimore infants and their mothers, who were visited regu-
larly until the children were a year old. As a supplement to home observations, Ainsworth
invited the mothers and infants for a laboratory-based observational procedure, which
she called the Strange Situation. Ainsworth’s findings from this study were reported in nu-
merous articles. Drawing on additional results from her students, she co-authored Patterns
of Attachment in 1978, which presented a thorough report on the Strange Situation as a re-
search methodology.1 She was unable to secure funds to replicate or extend her results, des-
pite growing recognition of her work and election to Presidency of the Society for Research
in Child Development from 1977 to 1979. Yet her pioneering and profound work established
attachment as a paradigm within developmental science, offering a way beyond the oppos-
ition between frequency counts of behaviours and subjective judgement about relationships.
She also mentored an astonishing cohort of developmental psychologists and clinical re-
searchers, first at Johns Hopkins University and then at the University of Virginia.
Introduction
In a letter to Everett Waters in 1985, Bowlby wrote of his intense pride at having had the op-
portunity to work with Mary Ainsworth. He described Ainsworth and himself as horses in
1 Ainsworth, M., Blehar, M., Waters, E., & Wall, S. (1978, 2015) Patterns of Attachment: A Psychological Study of
the Strange Situation. Bristol: Psychology Press.

110 Mary Ainsworth
‘double harness’, pulling the cart along.2 This beautiful image of a sturdy, effortful partner-
ship glosses over the fact that, at times, Bowlby and Ainsworth pulled in different directions.
As Chapter 1 described, Ainsworth identified limitations in Bowlby’s ideas on several fronts.
She felt that Bowlby oversimplified matters when he claimed proximity as the set-goal of the
attachment behavioural system, and protection from predation as its evolutionary function.
She disliked his imprecision in discussions of separation, and particularly the way that the
term ‘maternal deprivation’ could absorb anything from occasional use of professional child-
care through to abuse and neglect. She was frustrated that her contributions to research on
hospitalised children in London in the 1950s resulted in few publications because Bowlby’s
lack of empirical expertise had led to poor choice of measures. Ainsworth also had concerns
about aspects of Bowlby’s account of behavioural systems, feeling that he had underplayed
the sexual, exploratory, and aggression behavioural systems, and neglected adequate atten-

tion to the emotional components of behavioural systems in humans. And yet the image of
two horses pulling the cart along is exactly appropriate. Ainsworth’s criticisms came from
her overall sense of common purpose with Bowlby, bringing her independent intellectual
perspective to shared problems, as well as her own interests.3 In a co-authored article com-
posed largely in the months before Bowlby died and then finished by Ainsworth, the two re-
searchers wrote that ‘their contributions to attachment theory and research interdigitated in
a partnership that endured for 40 years across time and distance’.4 The word ‘interdigitated’
is a characteristically stiff one; Bowlby, especially, was a person with significant capacities for
reserve. But the phrase ‘enduring across time and distance’ was actually a technical one for
these two researchers: it defines one of the qualities of an attachment relationship.5 It comes
from the kind of deep happiness in another person that does not require a smile or other
marks of informality.
As discussed in Chapter 1, Ainsworth had worked for Bowlby as a clinical postdoctoral re-
searcher within the Separation Research Unit from 1950 to 1953. During this time, they had
been colleagues, but Bowlby’s group was run according to a strict hierarchy, with Ainsworth
conducting empirical work but with little say on research design or theory development.
In an interview with Robert Karen, Ainsworth recalled her experiences of Bowlby during
this period: ‘He made no bones about the fact that he was single-handedly fighting the ana-
lytic establishment, that it pained him some, but that he was convinced he was on the right
track. It was a long time before I felt any sense of getting close to him or being a friend. But
I had no difficulty whatsoever making him into a surrogate father figure—even though he’s
not much older than I.’6 They remained in touch after this, including through Ainsworth’s
period in Uganda, but the correspondence had little impetus. This changed from 1960. In
this year, Ainsworth and her husband divorced.7 She became quite depressed, feeling that
2 Bowlby, J. (1985) Letter to Everett Waters, 30 October 1985. PP/Bow/B.3/40.

3 A definition of genuine partnership was provided by Bowlby as ‘when two more or less autonomous beings
share a common plan’. A sure sense of common plan is evident between Bowlby and Ainsworth from the 1960s
onwards. The definition appears in Bowlby, J. (1969) Letter to Robert Marvin, 10 December 1969. PP/Bow/J.9/132.
4 Ainsworth, M. & Bowlby, J. (1991) 1989 APA award recipient addresses: an ethological approach to personality
development. American Psychologist, 46(4), 333–41, p.333.

5 Bowlby J. (1951) Maternal Care and Mental Health. Geneva: WHO, p.53; Ainsworth, M.D. (1985) Attachments
across the life span. Bulletin of the New York Academy of Medicine, 61(9), 792–812.
6 Interview with Mary Ainsworth by Robert Karen cited in Karen, R. (1998) Becoming Attached: First
Relationships and How They Shape Our Capacity to Love. Oxford: Oxford University Press, p.133.
7 All seven men in the cohort of doctoral students who graduated with Ainsworth were married, stayed married,
and raised children; in contrast, only one of the six female academics had an enduring marriage, and it was only
Introduction 111
her life had become empty.8 Gradually, she began to work out new plans, and these brought
her into a much closer engagement with Bowlby. Ainsworth had read Bowlby’s ‘The nature
of the child’s tie to his mother’ paper in 1958. The account of attachment behaviours helped
her make sense of her observations of Ganda infant–caregiver dyads, in terms of both the
maturational processes associated with the following response and the role of caregiving in
shaping individual differences.9
In 1960 Bowlby came to visit Ainsworth in Baltimore, following his year at the Stanford
Institute for Advanced Study. Bowlby discovered Ainsworth’s enthusiasm for his recent
theoretical work, and the relationship was rekindled on changed terms. Bowlby remained
the senior colleague. However, compared to their years working together in London, the
relationship gained greater equality and affection, both of which continued to grow over
subsequent years. Where Bowlby had found in ethology the heuristic frame that integrated

his otherwise diverse observations, Ainsworth found this in Bowlby’s work, supporting her
thinking about infant behaviour and infant–caregiver interaction.10 Yet Ainsworth also
found in Bowlby’s ideas from 1958 onwards a deep and persuasive account of the human
condition, offering a unifying perspective on relatedness, development, and how we respond
when our needs are not met. She saw in attachment theory qualities that resembled existen-
tialist philosophy in its careful reflection on relationships, the uncomfortable feelings that
stem from them, and what these suggest about the nature of a human life.11
Following her divorce, Ainsworth also entered into what would be eight years of therapy,
which she later described as perhaps ‘the most important positive influence on my career’.12
It is hardly possible to understand Ainsworth’s intellectual orientation, and therefore her
contribution to developmental science from the 1960s onwards, without attention to this
‘most important positive influence’. For this reason, Ainsworth was herself candid about her
therapy in autobiographical writings as well as in interview. At the start, therapy initially pro-
vided ‘some core of stability in what would otherwise be a confused and confusing period’.13
Over the years, however, Ainsworth felt that she gained a greatly deepened understanding
of psychological processes, especially emotional life, its conflicts, and forms of defence or
inhibition.14 In a late interview, Ainsworth recalled the exploration and learning of her time
in therapy. She came to acknowledge and understand ‘the feelings of warmth, love and se-
curity’ she received from her relationship with her father. Her mother was jealous of this
closeness between father and daughter, and banned her from seeking physical proximity
after the breakdown of the marriages that their careers took off. Isaacson, K.L. (2006) Mary Ainsworth and John
Bowlby: the development of attachment theory. Unpublished doctoral dissertation, University of California, Davis.
8 Ainsworth, M. (1983) Mary D. Salter Ainsworth. In N. Felipe Russo & A.N. O’Connell (eds) Models of
Achievement: Reflections of Eminent Women in Psychology (pp.200–18). New York: Columbia University Press.
10 Ainsworth, M. (1995) On the shaping of attachment theory and research: an interview with Mary D.S.
Ainsworth. Monographs of the Society for Research in Child Development, 60(2–3), 2–21: ‘Mainly, however, findings
obtained with other species have made me feel that I have been on the right track rather than helping me under-
stand the specifics of human babies’ behaviour.’ (9)
11 Ainsworth, M. (1965) Letter to John Bowlby, 2 February 1965. Mary Ainsworth papers, Box M3168, Folder 2.
12 Ainsworth, M. (1983, 2013) An autobiographical sketch. Attachment & Human Development, 15(5–6),
448–59, p.456.
13 Ainsworth, M. (1960) Letter to John Bowlby, 18 October 1960. Mary Ainsworth papers, Box M3168, Folder 1.
14 Ainsworth, M. (1983) Mary D. Salter Ainsworth. In N. Felipe Russo & A.N. O’Connell (eds) Models
of Achievement: Reflections of Eminent Women in Psychology (pp.200–18). New York: Columbia University
Press, p.212.
112 Mary Ainsworth
with her father.15 Though her mother made Ainsworth feel rejected, anger in response to this
rejection was unacceptable, to the point that Ainsworth lost access to that emotion: ‘I got to
the point of not ever being able to feel angry. I would just feel hurt.’16
Therapy also helped Ainsworth think through the rubble and emotional fallout that fol-
lowed her divorce, and especially her grief that she had been unable to have a child.17 Her one
pregnancy had ended in a miscarriage. She would later reflect to Bowlby that she felt that her
grief and preoccupied longing for a child ultimately became transfigured into perceptive-
ness.18 This entailed an unusual ability to see things from the baby’s point of view, through
both an awareness of infants’ signals and communications and acuity in interpreting them.
In a sense, all subsequent attachment researchers after Ainsworth would, one by one, un-
knowingly light their own work with the spill from this transfigured loss.
As her therapy was coming to an end, Ainsworth composed an important article, ‘Object

relations, dependency and attachment’, published in 1969, comparing Bowlby’s ideas with
the mainstream psychoanalytic ideas of the day, and highlighting the strengths of both. In
particular, she argued that it was in considering the qualities of individual differences that
‘psychoanalysts have made a valuable contribution’:
They have not been concerned so much with the quantitative dimension of object rela-
tions—stronger or weaker love or attachment—as with the qualitative variations among
different object relations. How ambivalent is the relationship, what admixture of love and
hate, and how well is the ambivalence resolved? How anxious is the relationship? How is it
affected by the person’s defenses against anxiety?19
Ainsworth’s stable academic position at Johns Hopkins, her experiences of ethnographic

observation in Uganda, and her subsequent thinking about attachment all fed into plans
to develop a longitudinal study of infant development, based on detailed observations of
Baltimore infants and mothers at home. This plan was difficult to implement. Ainsworth
had great trouble finding funds for such a study. Reviewers did not appreciate her desire for
in-depth work with a small sample, or her intention to examine many aspects of interaction.
And they regarded as unscientific her wish to develop scales after having conducted qualita-
tive analysis to explore emergent findings.20
15 Main, M. (1999) Mary D. Salter Ainsworth: tribute and portrait. Psychoanalytic Inquiry, 19(5), 682–736, p.704.
17 This sadness remains powerful, however, in Ainsworth’s reference to her ‘vain longing’ for children
in Ainsworth, M. (1983, 2013) An autobiographical sketch. Attachment & Human Development, 15(5–6),
448–59, p.459.
18 Ainsworth, M. (1984) Letter to John Bowlby, 11 April 1984. PP/BOW/B.3/8: ‘Longing gave me some kind of
perceptive in terms of which I could understand mother–infant interaction.’ See also Maurer, D. (1998) Interview
with Mary Ainsworth: never miss an opportunity to hold a baby, 12 May 1998. Daily Progress. http://www.psych-
ology.sunysb.edu/attachment/gallery/never_miss/nevermiss.htm:‘Quite a lot of my wanting my own child played
into my life work.’ Among the most distinctive characteristics of Ainsworth’s written voice is the dignity she gives
to young children’s gestures and concerns, grounded in meticulous attentiveness to their behaviour, affect, and
interaction. Countless illustrations could be offered from works such as Ainsworth, M. (1967) Infancy in Uganda.
Baltimore: The Johns Hopkins Press. However, even in discussing adult autobiographical discourse, an imagina-
tive concern with the child this adult might once have been is palpable: Ainsworth, M.D.S. & Eichberg, C.G. (1991)
Effects on infant–mother attachment of mother’s experience related to loss of an attachment figure. In: C.M.
Parkes, J. Stevenson-Hinde, & P. Marris (eds) Attachment Across the Life Cycle (pp.160–83). New York: Routledge.
19 Ainsworth, M. (1969) Object relations, dependency and attachment. Child Development, 40, 969– 1025,
p.1002.
20 Ainsworth, M. (1962) Letter to John Bowlby, 3 January 1962. Mary Ainsworth papers, Box M3168, Folder 1.
Security and independence 113
She eventually received a fraction of the money that she requested from the William
T. Grant Foundation, and was able to begin a study in 1963. The multifaceted nature of this
longitudinal study allowed Ainsworth to use many of the skills and insights she had previ-
ously developed, including attention to feelings of security and insecurity (from Blatz); close
observational study (from work with Robertson in London, and then from her Uganda eth-
nography); and a personal interest in affection, anger, anxiety, the wish for physical contact,
and the inhibition of these feelings (from her therapy). Together, these skills and insights
combined to give Ainsworth the desire and ability to take on the challenge of empirically
examining Bowlby’s hypothesis that early relationships with attachment figures would
shape the expression of the attachment behavioural system. This was a radical project. Until
Ainsworth, hypotheses about defence mechanisms in young children had been mostly re-
garded as untestable; the skill of even young children in regulating and redirecting affect,

familiar to every clinician working with children, had been regarded as outside the domain
of science.
Security and independence
Ainsworth’s concern with feelings of security originated in the lectures she attended with
Blatz. Blatz hypothesised a number of distinct needs including food, sex, rest, and novelty.21
According to Blatz, feelings of security are generated if an individual feels their actions will
not harm access to the meeting of these needs, whether by the individual themselves or by
someone else. Security means that it is possible for an individual to try things out, even to
fail or retract a commitment, without this having relevance to whether their needs will be
met. By contrast, Blatz proposed that feelings of insecurity are caused by concern that needs
might be left unmet. Such concerns prompt anxiety and/or the use of defences.22 Blatz ar-
gued that when an individual feels secure this allows them to turn their attention to other
matters. As such, in early life, parents who are able to give children confidence in their avail-
ability to meet their needs in general will offer what Blatz referred to as a ‘secure base’ from
which to explore the world, headlong and fully, without the need for excessive caution or
control.
Ainsworth took from Blatz the idea that against this ‘generally secure background, the
infant or young child becomes able to tolerate some degree of insecurity’.23 Security allows
the child to accept the uncertainties inherent in human relationships without defensiveness,
and to seek support within relationships as needed. Ainsworth followed Blatz in the sugges-
tion that security also forms a feature of a broader experience of life: ‘the person feels that
he belongs, not only in his more or less intimate relationships, but in the world at large, and
that the contribution he has to make is somehow significant in the larger scheme of things.
This is in contrast to the feeling of insignificance, helplessness, and isolation that charac-
terises insecurity.’24 On Ainsworth’s interpretation, Blatz’s work highlighted that, in a deep
21 Blatz, W.E. (1934) Human Needs and How They Are Satisfied. Des Moines: State University of Iowa.
22 Blatz, W.E. (1940) Hostages to Peace: Parents and the Children of Democracy. New York: William Morrow,
p.182; see Ainsworth, M.D.S. (1988) Security. Unpublished discussion paper prepared for the Foundations
of Attachment Theory Workshop, convened for the New York Attachment Consortium by G. Cox-Steiner &
23 Ainsworth, M. & Ainsworth, L. (1958) Measuring Security in Personal Adjustment. Toronto: University of
Toronto Press, p.4.

24 Ibid. p.46.
114 Mary Ainsworth
sense, the people we need are ultimately independent of us. Others’ independence can be
regarded as a source of worry or as a source of reassurance, depending on what this freedom
has implied in the past. If the independence of others can be the basis of security and reduce
anxiety, it can nonetheless also be an irreducible threat, and may even expand the scope of
potential anxiety: ‘no matter how secure a person may be in his everyday life, his security will
be shaken when he first encounters catastrophe, serious illness, injury, or the possibility of
death, whether the threat is directed towards himself or towards other people on whom his
security depends’.25
Blatz’s security theory had a powerful appeal for Ainsworth, especially as compared with
other theories available at the time. It offered a model of thinking about development in
terms of both individual needs and environmental experiences, in which each had a role
in shaping the other. Furthermore, the model acknowledged the potential role of different

sources of security for one another, so that even if security in family relationships had a par-
ticular primacy since it related to a wide variety of needs, this sense of security in family
relationships was in turn influenced by various other sources of security or insecurity for a
person in childhood and over the lifespan26—not least economic insecurity, and the polit-
ical security of civil rights.27 For Ainsworth, following Blatz, a feeling that emergencies can
or cannot be confidently handled in social relationships, in athletic or cultural activities for
instance, can in turn influence the security of family relationships. In this, Blatz’s account
differed in important ways from the emphasis of the day in psychoanalytic theory on innate
drives and the singular primacy of the family, and the emphasis in behaviourist approaches
on learned responses.
Indeed, in one of her final publications, Ainsworth would take a Blatzian approach to
offer a criticism of attachment research: ‘By focusing so closely on intimacies some attach-
ment researchers have come to conceive of them as the only source of security—which is
a pity.’28 The ethological concept of security in the use of the caregiver as a secure base and
safe haven was, for Ainsworth, a particular form of a broader concept of security. Other
sources of security are not detailed in theoretical terms by Ainsworth, but might include
reliable experiences of successful exploration, and reliable experiences of safety when the
fear system is activated. Such a broader cross-domain concept of ‘security’ seems to have
been inherited only by Ainsworth’s direct students and immediate collaborators, presum-
ably as a result of oral transmission.29 Shaver and colleagues also later adopted a broader
conceptualisation of security, though seemingly without awareness of Ainsworth’s stance
(Chapter 5).
Toronto Press, p.44.

26 Ainsworth, M. (1990, 2010) Security and attachment. In R. Volpe (eds) The Secure Child: Timeless Lessons in
Parenting and Childhood Education (pp.43–53). Charlotte, NC: Information Age Publishing.
27 Ainsworth, M.D.S. (1980) Attachment and child abuse. In: G. Gerbner, C.J. Ross, & E. Zigler (eds) Child
Abuse: An Agenda for Action (pp.35–47). Oxford: Oxford University Press; Ainsworth, M. (1965) Letter to John
Bowlby, 2 February 1965. Mary Ainsworth papers, Box M3168, Folder 2.
28 Ainsworth, M. (1990, 2010) Security and attachment. In R. Volpe (ed.) The Secure Child: Timeless Lessons in
Parenting and Childhood Education (pp.43–53). Charlotte, NC: Information Age Publishing, p.49.
29 E.g. Main, M. (1977) Sicherheit und wissen. In K.E. Grossmann (ed.) Entwicklung der Lernfahigheit in der
sozialen Umwel (pp.47–95). Munich: Kindler; Sroufe, L.A. (1996) Emotional Development. Cambridge: Cambridge
University Press; Waters, E. & Cummings, E.M. (2000) A secure base from which to explore close relationships.
Child Development, 71(1), 164–72; Davies, P.T., Harold, G.T., Goeke-Morey, M.C., & Cummings, E.M. (2002)
Child emotional security and interparental conflict. Monographs of the Society for Research in Child Development,
67, 1–115.
Security and independence 115
Both psychoanalytic and behaviourist theories of the 1940s and 1950s presumed that in-
fants would be more clingy and dependent the more their needs were satisfied. They as-
sumed continuities in the form of behaviour with development. Blatz’s model led to the exact
opposite conclusion. Blatz’s perspective suggested that confidence and an appropriate level
of self-reliance would grow out of experiences of being able to rely and rest our weight upon
others, and of their availability to help us as needed. Though this was not a point made clear
by Blatz himself, Ainsworth drew the implication that mutual reliance within family rela-
tionships and an independent and confident attitude in other areas of life could be compat-
ible. In fact, Ainsworth concluded, forming close relationships is itself a human need. As a
result, insecurity will result if these are not available, and security will provide a springboard
for confident and flexible action in other areas such as in school and work.30
Harry Harlow and Robert Zimmermann had used the phrase ‘haven of safety’ to refer

to the way that an infant’s alarm and motivation to seek their caregiver would be termin-
ated once they achieved proximity with the caregiver.31 Ainsworth cultivated the concept of
‘secure base’ to refer to the way that an infant—or, indeed, humans in general—can feel free
to explore the world with confidence, as he or she knows that protection and care is available
if needed.32 A secure base permits negative experiences in the world, even pain, to feel more
bearable and less overwhelming.33 Harlow and Zimmerman’s ‘haven of safety’ was about
termination of the attachment behavioural system and its associated distress. By contrast,
the concept of ‘secure base’ was not, for Ainsworth, primarily about the achievement of inde-
pendent self-reliance, as has sometimes been presumed by anthropologist critics.34 Instead,
seen in the context of Ainsworth’s debt to Blatz, the secure base concept was more about the
role that a person can play in helping another to live a larger life than the latter would be able
to on their own, with the freedom to chase and tumble after the world without worry.35 This

31 Harlow, H.F. & Zimmermann, R.R. (1958). The development of affectional responses in infant monkeys.
Proceedings of the American Philosophical Society, 102(5), 501–509.

32 Ainsworth, M. (1964) Patterns of attachment behavior shown by the infant in interaction with his mother.
Merrill-Palmer, 10(1), 51–8, p.54. The distinction would be further established within attachment theory through
becoming the central theme of the Circle of Security intervention: Marvin, R., Cooper, G., Hoffman, K., & Powell,
B. (2002) The Circle of Security project: attachment-based intervention with caregiver–pre-school child dyads.
Attachment & Human Development, 4(1), 107–24.
33 Ainsworth, M. (1976) Attachment and Separation in Paediatric Settings. Unpublished manuscript. PP/Bow/
J.1/40: ‘Painful procedures may still be painful, but the pain is easier to endure with mother present to give com-
fort, and more easily recovered from when it is over. On the other hand, all the potential sources of fear and distress
may become overwhelming if the child must face them without the security given by the presence of an attachment
figure.’
34 E.g. Harwood, R.L., Miller, J.G., & Irizarry, N.L. (1995) Culture and Attachment: Perceptions of the Child in
Context. New York: Guilford; Rothbaum, F., Weisz, J., Pott, M., Miyake, K., & Morelli, G. (2000) Attachment and
culture: security in the United States and Japan. American Psychologist, 55(10), 1093–104; LeVine, R.A. (2004)
Challenging expert knowledge: findings from an African study of infant care and development. In U.P. Gielen
& J.L. Roopnarine (eds) Childhood and Adolescence: Cross-Cultural Perspectives and Applications (pp.149–65).
Westport, CT: Praeger.
35 Waters, E. & McIntosh, J. (2011) Are we asking the right questions about attachment? Family Court Review,
49(3), 474–82, p.475. Bowlby would later observe that ‘the values of western culture’ lead the benefits of a secure
base to be ‘overlooked, or even denigrated’. Both he and Ainsworth were angered that in the society around them,
autonomy and independence were mistakenly supposed to just be default states, unless there was a specific
problem. This had led to a neglect of the quiet enormity of the secure base role and an overvaluation of liberal self-
reliance. Bowlby, J. (1970, 1979) Self-reliance and some conditions that promote it. In The Making and Breaking
of Affectional Bonds (pp.124–49). London: Routledge, p.125. Nonetheless, a focus on the haven of safety at the
expense of the secure base has occurred within attachment theory and especially in its public representation. See
Wall, G. (2018) ‘Love builds brains’: representations of attachment and children’s brain development in parenting
education material. Sociology of Health & Illness, 40(3), 395–409.
116 Mary Ainsworth
may have been hidden somewhat by the overridingly spatial and territorial image of a secure
base,36 resulting from Ainsworth’s insertion of Blatz’s concept into the Hinde–Bowlby etho-
logical account of proximity-maintenance.
Ainsworth was the first to attempt to develop empirical measures based on Blatz’s idea of
security.37 In her 1958 book Measuring Security in Personal Adjustment, Ainsworth reported
findings from her use of self-report measures of security, drawing on the skills in measure
design and administration from her time as an Army Examiner (personnel selection) dur-
ing World War II.38 However, the self-report scales did not generate results that particularly
interested her. This work also led her to conclude that individuals with a chronic experience
of insecurity, especially from childhood, may develop anxiety and/or defences to such a de-
gree that self-report measures lose validity.39 Such a person may be ‘so handicapped in his
communication with others and in insight into his own needs and feelings that pencil-and-

paper tests cannot reflect the nature and extent of his maladjustment’.40 Ainsworth would
later conduct a study, which remained unpublished, utilizing the scales to assess patients in a
psychiatric hospital. She found that her scales ‘did indeed highlight depression. Those emer-
ging with highly insecure scores felt insecure and unhappy and readily said so.’ However,
her items failed to differentiate patients with anxiety disorders, paranoid and psychotic
symptoms, and those with personality disorders.41 Ainsworth came increasingly to question
whether security was, as Blatz had assumed, solely a conscious attitude, measurable in a valid
way using self-report methodology.
Caregiving
Uganda
In 1953, Leonard Ainsworth took up a job at the East Africa Institute of Social Research
in Kampala, Uganda. Ainsworth joined him for their two-year stay.42 This was a period of
36 The spatial and territorial underpinnings of the ‘secure base’ metaphor were materialized by the squares phys-
ically marked by Ainsworth on the floor in the original Strange Situation procedure, to help coders—in the ab-
sence of videotape—identify infant movement away from and back towards the caregiver.
37 Salter, M. (1939) The concept of security as a basis for the evaluation of adjustment. Unpublished doctoral
thesis, University of Toronto. Cf. Prichard, E. & Ojemann, R.H. (1941) An approach to the measurement of inse-
curity. Journal of Experimental Education, 10(2), 114–18.
Toronto Press.
E. Waters, Port Jefferson, NY. http://www.psychology.sunysb.edu/attachment/online/mda_security.pdf: ‘I felt
dissatisfied with the validity of my scales because of their inadequate coping with the whole matter of defensive
maneuvers.’

E. Waters, Port Jefferson, NY. http://www.psychology.sunysb.edu/attachment/online/mda_security.pdf
42 On the travel required by women psychologists such as Ainsworth to support their husband’s careers, see
Johnston, E. & Johnson, A. (2008) Searching for the second generation of American women psychologists. History
of Psychology, 11(1), 40–72. On the East African Institute of Social Research see Mills, D. (2006) How not to be a
‘Government House Pet:’ Audrey Richards and the East African Institute for Social Research. In M. Ntarangwi, D.
Mills, and M. Babiker (eds) African Anthropologies (pp.76–98). London: Zed Books.
Caregiving 117
growing demands for political independence. The colonial British government seemed to
be seeking to encourage Ugandan independence whilst fearing what it might bring.43 Social
scientists were being encouraged by the colonial government to pursue ethnographic and
social scientific studies in Uganda, in an attempt to understand and respond to these ten-
sions.44 Together with her husband, Ainsworth pursued some research explicitly studying
political attitudes in Uganda and sociological factors that might contribute to insurrection
against the colonial government.45 This was, in a sense, the ‘day job’.
However, Ainsworth’s primary concern was to continue the study of early infant–
caregiver relationships, which had been the focus of Bowlby’s research group. She received
funding for this from the anthropologist Audrey Richards, the director of the East Africa
Institute of Social Research.46 The condition of the funding was that Ainsworth pursue re-
search with a significant qualitative, ethnographic component. With Robertson’s detailed

notes on hospitalised children as a model, Ainsworth embarked on an ethnographic study
of 26 mother–infant dyads from villages near to Kampala, visiting families for two hours,
twice a month, over a nine-month period. Seeking to offer recompense that would disturb
as little as possible the opportunity for naturalistic observation, Ainsworth paid for the
healthcare of her participants. The political context of Uganda is notable in its absence from
Ainsworth’s Infancy in Uganda, published in 1967. She seems to have separated her research
with the mother–infant dyads from her attitudinal research with her by-then-former hus-
band. The attitudinal research may have had associations with her painful divorce, though it
also clearly interested her less. Throughout Infancy in Uganda, however, she displayed great
attentiveness to the effects of poverty on the care that families could offer their children, as
well as other observable aspects of the families’ social context.
One early discovery from Ainsworth’s observations was the variety of forms of attachment
behaviour. Whereas Bowlby had built from Hinde’s work in modelling the attachment be-
havioural system on the following response and approach through locomotion, Ainsworth
documented that the attachment behavioural system could be terminated by various other
behavioural sequences that predictably led to the caregiver’s availability. These included
crying, smiling, or vocalisation directed towards the caregiver; scrambling on the mother’s
body or nestling into her lap; raising arms or clapping in greeting; and crying when she left
the house.47 She found that the infants used these different behavioural sequences flexibly,
depending on present context, but that they seemed to have preferred forms of attachment
behaviour built up through routine interaction and experience.
Ainsworth suspected that human evolution had led many of these behaviours to be
especially easy for children to learn.48 However, she also emphasised the role of childcare
43 See Cohen, A. (1957) Uganda’s progress and problems. African Affairs, 56(223), 111–22.
44 On the relationship between ethnographic method and colonialism in the period see Asad, T. (1979)
Anthropology and the colonial encounter. In G. Huizer &B. Manheim (eds) Politics of Anthropology: From
Colonialism and Sexism Toward a View from Below (pp.85–96). The Hague: Mouton.
45 Ainsworth, L.H. & Ainsworth, M.D. (1962) Acculturation in East Africa. I. Political awareness and atti-
tudes toward authority. Journal of Social Psychology, 57(2), 391–9; Ainsworth, M.D. & Ainsworth, L.H. (1962)
Acculturation in East Africa. II. Frustration and aggression. Journal of Social Psychology, 57(2), 401–407.
46 Ainsworth, M.D. (1983, 2013) An autobiographical sketch. Attachment & Human Development, 15(5–6),
448–59: ‘I welcomed Dr Richards’ directive that there be an anthropological component to the study, for this en-
sured that I would view current mother–infant interaction and maternal care practices in their cultural context,
and I valued the opportunities presented by the institute again to interact with a multidisciplinary team’ (455).
Merrill-Palmer, 10(1), 51–8.

48 Ainsworth, M. (1967) Infancy in Uganda. Baltimore: The Johns Hopkins Press, p.438.
118 Mary Ainsworth
culture in shaping their possibility, frequency, and intensity of expression. The clearest ex-
ample was clapping hands in greeting on reunion: Ainsworth saw this frequently among
the Ugandan infants, who were enculturated to treat this as a way to express greeting.
By contrast, Ainsworth never saw this form of greeting behaviour towards attachment
figures among American infants.49 Ainsworth was also attentive to relationship-level dif-
ferences that could prompt differences in the display of attachment behaviours. Some
children, for example, seemed more or less inclined to physically follow their caregivers.
A large part of such preferences seemed to Ainsworth to be shaped by how the care-
givers responded when the infant followed them. Another influence seemed to be the
position of the relationship within the broader life of the infant: the same child might
show different configurations of attachment behaviour towards different caregivers,
and at different times. One infant, for instance, tended to preferentially follow her older

sister when she was home, even above her mother, but did not necessarily seek to be held.
However, when the infant was ill, she showed a strong preference for her mother, and
wanted to be held all the time.50
One important line of difference among the Ganda infants was that ‘there were some
babies, who seemed clearly attached to their mothers, who did not dependably cry, follow
or cling when their mothers showed signs of leaving’.51 Some of these infants appeared rela-
tively unruffled by signs of impending separation, seeming confident in the availability of
another caregiver, or in the expectation that the separation would be brief and unthreat-
ening. Another set seemed to have had insufficient interaction with their caregiver, for in-
stance if they were often away for long periods; Ainsworth would wonder whether they
had, in fact, developed an attachment relationship yet.52 A further group of infants, how-
ever, were clearly attached and seemed concerned about their caregiver’s availability, yet
did not show attachment behaviours on separation. These infants tended to be those whose
caregivers were less responsive to their signals when the child was distressed. However,
Ainsworth also noticed that some infants who seemed less confident in their caregiv-
er’s availability displaced insistent and frequent attachment behaviours. It appeared to
Ainsworth that a child’s lack of confidence in their caregiver’s responsiveness could be ex-
pressed in a variety of ways.
Ainsworth anticipated that an ordinary attachment relationship makes use of proximity-
seeking only occasionally, or under circumstances of alarm. Most of the time, interactions
to affirm caregiver availability are achieved through expression, movement, gesture, and
vocalisation at a distance. It was, on Ainsworth’s observation, only ‘the anxious infant who
requires close physical contact with his mother, and who is not content to maintain inter-
action through a middle distance at least part of the time’.53 For instance, one mother had
to work very long hours in a desperate effort to establish a new garden in order to ensure
49 Ibid. p.340.
50 Ibid. p.307.
Merrill-Palmer, 10(1), 51–8, p.52.
ality development. American Psychologist, 46(4), 333–41: In Uganda, Ainsworth ‘divided the babies into three
groups: securely attached, insecurely attached, and non-attached . . . Nonattached babies were left alone for long
periods by unresponsive mothers but, because they were the youngest in the sample, Ainsworth now believes that
they may merely have been delayed in developing attachment’ (337).

Caregiving 119
food for the family, and left the baby with a neighbour during this time. Even when mother
and baby were together, the mother was too tired to respond with patience to her child.
Ainsworth described ‘a vicious spiral’ in some of the dyads in her sample in which ‘the
baby’s fussy demands exasperated the mother, who then overtly or covertly rejected the
baby, who in turn responded to the rejection by anxiety and by increasing his demands’.54
Reflecting on these behavioural observations, Ainsworth identified that diverse forms of
attachment behaviour could generally have the predictable outcome of increasing prox-
imity with the caregiver, but that some repetoires seemed to risk alienating the caregiver.
However, at the time of writing Infancy in Uganda, Bowlby’s control system model of the
attachment behavioural system was not yet available to Ainsworth for interpreting her
observations.
Ainsworth’s study utilised qualitative observation and also the construction of quanti-

tative scales. As she analysed her quantitative data, an unexpected finding emerged. Two
variables stood out as predictors of infants who appeared to be able to effectively use their
caregiver as a secure base and safe haven.55 One was the quantitative amount of care pro-
vided by the mother. This supported Bowlby’s emphasis on the importance of hours of care
by a primary caregiver (Chapter 1). Ainsworth, however, distrusted the finding. Though
she felt she lacked firm data, her general impression was that it was not the amount of care,
but—at least above a minimum threshold—how the care was provided that most contrib-
uted to security. She interpreted the finding as indicating that whilst quality of care is most
important for security, the caregiver’s understanding of her child’s signals can be hindered
if there is insufficient opportunity to learn about them. The qualities and the quantitative
extent of care were both, she suspected, also shaped by the attitude of the caregivers towards
their infants.56
A second variable that stood out as a good predictor of use of the caregiver as a secure
base and safe haven in Ainsworth’s Uganda data was the extent to which the mothers could
give a lively account when interviewed about their infant. This was not so much details of
the child’s schedule as ‘idiosyncracies and sensitive little things that testify to a mother’s
interested perception’.57 This was unexpected: it was not obvious why a mother’s capacity
to talk freely and fully with Ainsworth should be associated with her infant’s behaviour
towards her when worried or on separation. Ainsworth concluded again that both the
mother’s capacity to offer an effective interview about her caregiving and her infant’s at-
tachment behaviour reflected the depth and fluency of her engagement in the caregiving
role.58 The relationship between maternal coherence in interview and infant attachment
behaviours would recur as a central theme in the work of Ainsworth’s student Mary Main
(Chapter 3).59
54 Ainsworth, M. (1967) Infancy in Uganda. Baltimore: The Johns Hopkins Press, p.392.
55 A third predictor was the mother’s enjoyment of breastfeeding.
56 Ainsworth, M. (1959) Letter to John Bowlby, 18 September 1959. Mary Ainsworth papers, Box M3168,
Folder 1.
57 Ainsworth’s ‘Excellence as Informant’ scale, unpublished, cited by Bretherton, I. (2013) Revisiting Mary
Ainsworth’s conceptualization and assessments of maternal sensitivity–insensitivity. Attachment & Human

Development, 15(5–6), 460–84, p.467.
58 Ainsworth, M. (1959) Letter to John Bowlby, 18 September 1959. Mary Ainsworth papers, Box M3168, Folder
1; Ainsworth, M. (1967) Infancy in Uganda. Baltimore: The Johns Hopkins Press, p.398.
59 See also Bailey, H.N., Redden, E., Pederson, D.R., & Moran, G. (2016) Parental disavowal of relationship
difficulties fosters the development of insecure attachment. Revue Canadienne des Sciences du Comportement,
48(1), 49–59.
120 Mary Ainsworth
The Baltimore study
Her observations in Uganda confirmed for Ainsworth the value of exploratory observational
research with mother–infant dyads in the home. As discussed in the ‘Introduction’, after
joining Johns Hopkins University in 1958, she was successful in obtaining a grant to begin a
short-term longitudinal study from 1963. The study was undertaken to examine the role of
caregiving factors in shaping the development of infant attachment relationships and attach-
ment behaviour. Fifteen families were observed from 1963 to 1964 by Ainsworth and her
assistant Barbara Wittig; eleven families were observed from 1966 to 1967 by Ainsworth’s
assistants Bob Marvin and George Allyn, thanks to additional funding from the National
Institute of Child Health and Human Development (NICHD). Baltimore during this period

was in the process of losing its industrial base, and the net population of the city was in
decline. However, this demographic shift comprised two trends: a rapid decline in the
white middle-class population of the city, and a less rapid but still substantial increase in
the African-American population, living in poverty and facing substantial discrimination.60
Ainsworth wanted to understand infant–caregiver interaction under favourable socio-
economic conditions in order to reduce confounds and complexity for her exploration of the
role of caregiver–infant interaction on attachment. As a result, she sought a sample of white
and middle-class families, recruiting through paediatricians. Visits were made approxi-
mately every three weeks from three weeks after the child’s birth to 54 weeks . They generally
took place during office hours, with the consequence that the mother was home with the
baby and the father was out at work. Ainsworth reported that visits, at least in the first wave
of work, lasted around four hours,resulting in 72 hours of home observation for each dyad.61
‘It is a very onerous and time-consuming research’, Ainsworth wrote, ‘but it has captured me
heart and soul’.62
Looking back a few years later, Ainsworth reported that she felt that her study had aligned
with a growing interest from the scientific community and the public in the naturalistic de-
velopment of children and in family relationships.63 Ainsworth’s observers were responsive
to overtures by the mother, baby, or others in the house, since to do otherwise would have
been rude and disruptive. However, they otherwise kept from interfering with ongoing ac-
tivities, and above all avoided any implied judgement of the mother’s behaviour. The obser-
vers wrote what they saw, and after the visit narrated an account onto tape. These records
used ‘common English usage in describing behaviour, with all of its advantages and disad-
vantages. The advantages were that the descriptions were vivid; the disadvantages were that
the same verbal labels might be used to describe different behaviours, with only the quali-
fying words and the context serving to differentiate them, whereas different verbal labels
might be used with reference to behaviours that were essentially similar.’64 Observers made
60 Levine, M.V. (2000) A third-world city in the first world: social exclusion, racial inequality, and sustainable
development in Baltimore, Maryland. In R. Stren & M. Polese (eds) The Social Sustainability of Cities: Diversity and
the Management of Change (pp.123–56). Toronto: University of Toronto Press.
ality development. American Psychologist, 46(4), 333–41, p.337.

62 Ainsworth, M. (1966) Letter to Bruno Klopfer, 22 September 1966. Mary Ainsworth papers, Box M3168,
Folder 8.
63 Ainsworth, M. (1967) Letter to Martin James, 23 February 1967. Mary Ainsworth papers, M3170, Folder 1.
64 Blehar, M. & Ainsworth, M. (1978) Close Bodily Contact. Unpublished manuscript. PP/Bow/J.1/49.
Caregiving 121
home visits individually and took written notes; observer reliability was not assessed, except
on a few occasions when Ainsworth accompanied one of her students on a visit.
Whilst the Baltimore sample was generally reported in print by Ainsworth as a randomly
selected community sample, albeit all middle-class, in fact in a late interview she acknow-
ledged that ‘the pediatricians who recruited potential participants for us tended to select
women who interested them—“this one is a charmer, that one puzzles me, I wonder how
motherhood will work out for this one”—and that this led to our getting a particularly di-
verse group’.65 This approach to recruitment may have somewhat increased potential lines
of difference between participants, making contrasts sharper. At a statistical level, it made it
more likely that Ainsworth’s data would appear to be categorically rather than dimensionally
distributed, compared to the population from which the sample is drawn. This effect would
likely have been intensified by the fact that Barbara Wittig only wrote up half of her observa-

tions many months after her home visits, contributing to sharper contrasts through potential
recollection biases.66 As Ainsworth and her group began to analyse the data, they began to
notice such a distribution. In a letter to Bowlby from 1967, Ainsworth wrote:
One of my impressions of my current sample is so bizarre that I hesitate to mention it.

I can dichotomise my mother–infant pairs. On one side are the good mothers with the
normally-attached infants and on the other side are the non-good mothers with infants
who are not normally attached. The not-good mothers are diverse and so are their infants.
But, and this is the bizarre part, there seems to be no truly middle or average group. Our
ratings reflect this. Usually rating scales tend to lump most cases in the middle . . . I find our
ratings tend towards the extremes with the middle of the scales scarcely represented in the
sample. At first I scolded my team for halo effect, and urged them to judge the variables
separately and to feel free to use the scale-points intermediate between the extremes—but
when we went into it more deeply, and really tried to get rid of the halo effect it looks as
though there is another genuine effect which truly does dichotomise the sample.67
The sharpening of certain contrasts in the data may have helped Ainsworth and colleagues
find order within the astonishing detail of the information they had collected. Yet it may also
have overstrengthened certain signals, such as perhaps the centrality of caregiver sensitivity
for secure attachment, discussed below.68
Ainsworth. Monographs of the Society for Research in Child Development, 60(2–3), 2–21, p.11–12.
66 Ainsworth, M. (1965) Letter to John Bowlby, 22 June 1965. Mary Ainsworth papers, Box M3168, Folder 2: ‘For
two years, I set myself to believing that I had found a good assistant, and to ignoring the obvious deficiencies in her
performance. Suddenly, there was a moment of truth. I found that she had written up fewer than half of the visits
that she had made to the babies in our sample. She has been catching up ever since, and will probably not finish
catching up until the end of September.’
67 Ainsworth, M. (1967) Letter to John Bowlby, 9 April 1967. Mary Ainsworth papers, Box M3168, Folder 2.
68 Another example, consequential for the next chapter, was that two-thirds of the infants classified as part of
avoidant dyads in the Ainsworth sample showed extensive conflict behaviours according to the home observation
data. These avoidant dyads were more troubled than, in retrospect, might be expected from a representative com-
munity sample. This may have contributed to Main’s initial assumption that avoidant attachment and conflict be-
haviour would go together, until she and colleagues came to the conclusion that conflict behaviour could cut across
the Ainsworth classifications (Chapter 3). Main, M. (1981) Avoidance in the service of proximity: a working paper.
In K. Immelmann, G.W. Barlow, L. Petrinovich, & M. Biggar Main (eds) Behavioral Development: The Bielefeld
Interdisciplinary Project (pp.651–93). Cambridge: Cambridge University Press, p.664.
122 Mary Ainsworth
Sensitivity
Perhaps the most important contribution made by the Ainsworth’s home observation study
was her development of a construct, ‘sensitivity’, which sought to capture this quality of care-
giving.69 This construct would have as much influence on the direction of subsequent attach-
ment research as Bowlby’s headline term ‘attachment’, though not on reception of the theory
by the public, practitioners, and researchers in other fields. The term ‘sensitivity’ was used
by Ainsworth from the late 1960s in a wholly technical sense, described in her unpublished
manuscript ‘Sensitivity vs. Insensitivity to the Baby’s Signals Scale’. This manuscript would
eventually appear in print only in the 2015 reissue of Patterns of Attachment by Waters.70 In
Ainsworth’s usage, the term ‘sensitivity’ referred to the ability of the caregiver to ‘perceive

and to interpret accurately the signals and communications implicit in her infant’s behavior,
and given this understanding, to respond to them appropriately and promptly’.71 Conversely,
insensitive caregiving had the same four components in reverse: less awareness of the child’s
signals; inaccuracies in interpreting them; inappropriate responses to the signals; and a lack
of timeliness in these responses. Until that point, researchers in developmental psychology
had tended to regard relationships as incalculable aspects of human life: too ‘squidgy’ and
amorphous to measure in themselves.
The first component of sensitivity was ‘awareness’. Ainsworth conceptualised awareness
in terms of the ‘threshold’ at which a caregiver would become responsive to infant cues: at
a higher threshold, only the baby’s most blatant and obvious communications prompt a re-
sponse, whereas caregivers with ‘the highest thresholds seem often oblivious, and are, in ef-
fect, highly inaccessible’.72 A high threshold for awareness of infant cues also contributes,
Ainsworth suggested, to poor accuracy in interpreting them, since when these signals do
break through to the caregiver’s awareness, they do so without contextual information such
as what prompted them. This means that interactions are often less satisfying, and can be
poorly coordinated, incomplete, or even somewhat fragmentary. However, even caregivers
with a low threshold for awareness may have poor accuracy for interpreting infant signals if
‘perception is distorted by projection, denial, or other marked defensive operations’.73 When
caregivers are able to perceive and to interpret signals accurately, the child’s whole behav-
ioural repertoire takes on signal value, as having relevance for understanding the infant’s
experience and the implications that stem from this.74
69 Ainsworth’s first use of the term ‘sensitivity’ was in Infancy in Uganda, but it was used there essentially as a de-
scriptor, without a technical meaning yet. Ainsworth, M. (1967) Infancy in Uganda. Baltimore: The Johns Hopkins
Press, p.397.
70 Ainsworth, M., Blehar, M., Waters, E., & Wall, S. (1978, 2015) Patterns of Attachment: A Psychological Study of
the Strange Situation. Bristol: Psychology Press.

71 Ainsworth, M. (1969) Sensitivity vs. insensitivity to the baby’s signals scale. http://www.psychology.sunysb.
edu/attachment/measures/content/ainsworth_scales.html. Some but not all aspects of the sensitivity scale and
construct were described in Stayton, D.J., Hogan, R., & Ainsworth, M. (1971) Infant obedience and maternal be-
havior: the origins of socialization reconsidered. Child Development 42(4), 1057–69, p.1060–61.
edu/attachment/measures/content/ainsworth_scales.html.
73 Ibid.
74 Ainsworth, M.D.S. (1977) Social development in the first year of life: maternal influences on infant–
mother attachment. In: J.M. Tanner (ed.) Developments in Psychiatric Research (pp.1–20). London: Hodder &
Stoughton, p.6.
Caregiving 123
As Waters and colleagues observed, a problem with the term ‘sensitivity’ is that it comes with
familiar, ordinary language connotations: ‘Sensitivity suggests warmth, tenderness, and atten-
tion to detail.’75 If this were what Ainsworth’s scale measured, these would be qualities that, in
theory, could be assessed with a checklist. They would also be overtly ethnocentric as a cultural
ideal of parenting.76 It is unsurprising that these ordinary language connotations were what
Ainsworth’s critics presumed that she meant, since the scale itself remained unpublished! In
fact, however, what Ainsworth operationalised with her scale was—mostly—something quite
different to what her critics presumed. It is true that at times she slid towards the everyday
language connotations of sensitivity in using ‘warm’ as a characterisation of the sensitive care-
giver; this led many later attachment researchers to include warmth in their assessments of
sensitivity or extrapolation of assessments of sensitivity of caregiving provided to older chil-
dren.77 However—contrary to Bowlby’s expectations78—Ainsworth herself found that ma-

ternal warmth was not associated with infant attachment security in her Uganda data.79
In fact, Main recalled that Ainsworth’s work on the sensitivity scale was actually prompted
by the finding that there was no association between maternal warmth and infant behaviour
in the Strange Situation.80 Later researchers, including Egeland and colleagues (Chapter 4),
confirmed that many caregivers can be warm and display tenderness without being sensitive
in Ainsworth’s technical sense.81 Ainsworth’s construct of ‘sensitivity’ primarily captured the
extent to which caregivers detect and successfully interpret behaviours that may convey their
child’s experience, and offer a relevant response in a timely manner. This was an assessment
more of the form than the content of interactions, allowing for a huge variety of ways in
which these formal features may be met within caregiving practices, whilst also anchoring
these features in concrete examples.82
75 Waters, E., Petters, D., & Facompre, C. (2013) Epilogue: reflections on a Special Issue of Attachment & Human
Development in Mary Ainsworth’s 100th year. Attachment & Human Development, 15(5–6), 673–81, p.676.
76 LeVine, R.A. & Norman, K. (2001) The infant’s acquisition of culture: early attachment reexamined in an-
thropological perspective. In C.C. Moore & H.F. Mathews (eds) The Psychology of Cultural Experience (pp.83–104).
Cambridge: Cambridge University Press; Mageo, J.M. (2013) Toward a cultural psychodynamics of attach-
ment: Samoa and US comparison. In N. Quinn & J. Mageo (eds) Attachment Reconsidered: Cultural Perspectives on
a Western Theory (pp.191–214). New York: Palgrave.
edu/attachment/measures/content/ainsworth_scales.html; Mesman, J. & Emmen, R.A. (2013) Mary Ainsworth’s

legacy: a systematic review of observational instruments measuring parental sensitivity. Attachment & Human
Development, 15(5–6), 485–506; Bohr, Y., Putnick, D.L., Lee, Y., & Bornstein, M.H. (2018) Evaluating caregiver
sensitivity to infants: measures matter. Infancy, 23(5), 730–47. Among the studies to have included warmth in
indexing sensitivity, the most important is the NICHD study: NICHD Early Child Care Research Network (1999)
Child care and mother–child interaction in the first three years of life. Developmental Psychology, 35, 1399–413.
78 For Bowlby’s emphasis on maternal ‘warmth’ see e.g. Bowlby, J. (1953) Child Care and the Growth of Love.
Harmondsworth: Pelican, p.77.

80 Main, M. (1999) Mary D. Salter Ainsworth: Tribute and portrait, Psychoanalytic Inquiry, 19(5) 682–736.
81 See e.g. Egeland, B., & Farber, E. A. (1984). Infant-mother attachment: Factors related to its development and
changes over time. Child Development, 753–71; Bailey, H.N., Bernier, A., Bouvette-Turcot, A.A., Tarabulsy, G.M.,
Pederson, D.R., & Becker-Stoll, F. (2017). Deconstructing maternal sensitivity: Predictive relations to mother-
child attachment in home and laboratory settings. Social Development, 26(4), 679–93.
82 Grossmann, K. & Grossmann, K. (2012) Bindungen— das Gefüge psychischer Sicherheit Gebundenes.
Stuttgart: Klett-Cotta; Mesman, J., van IJzendoorn, M., Behrens, K., et al. (2016) Is the ideal mother a sensitive
mother? Beliefs about early childhood parenting in mothers across the globe. International Journal of Behavioral
Development, 40(5), 385–97. Dozier and Bernard conclude that sensitivity in Ainsworth’s technical sense is a uni-
versal good, which can be achieved in various culturally specific ways; they therefore take a principled stance
in adapting the delivery but not the basic tenants of their Attachment and Biobehavioral Catch-up intervention
124 Mary Ainsworth
Ainsworth’s concept of attending to the infant’s signals has been interpreted by some critics
as ethnocentric. They construe Ainsworth as ascribing a kind of liberal autonomy to the infant,
and valuing autonomy and individual will over connectedness and joint needs.83 Certainly there
is evidence that Ainsworth personally valued acknowledgement of infant autonomy: another of
her scales—‘Interference with baby’s ongoing behavior’—characterised the ‘highly interfering
mother’ as one with ‘no respect for her baby as a separate, active, and autonomous person’.84
However, it is not clear that criticism of Ainsworth’s liberal values applies to the sensitivity scale,
and may have been influenced by the connotation of the word ‘sensitivity’ as non-conflictual
interaction; the criticism does not appear to be grounded in observation of how coders actually
use the scale in practice.
Various critics have interpreted Ainsworth’s notion of sensitivity as mandating that ‘the nor-
mative imperative is to take the infant’s cue’,85 an oppressive demand on mothers. Again, this

interpretation seems shaped by the lack of availability of Ainsworth’s sensitivity scale, showing
actually what it measured. For a caregiver to respond to signals does not necessarily mean obedi-
ence to a child’s dictates. The sensitive caregiver, Ainsworth proposed, ‘acknowledges the baby’s
wishes even though she does not unconditionally accede to them’, since much of what is ‘for
the baby’s own good is done contrary to his wishes’.86 The caregiver is supporting the infant
to achieve a sense that there is contingency between the infant’s activity and the activity of the
world, even when this contingency and acknowledgement of agency comes together with a re-
sponse that is contrary to the baby’s wishes. There is some ambiguity in Ainsworth’s account on
this matter, however. Despite having acknowledged that much of what is done for the baby’s own
good is contrary to his wishes, the highest point on the sensitivity scale identifies a caregiver
who ‘nearly always gives B what he indicates that he wants, although perhaps not invariably so’.
The difference between acknowledgement of infancy signals and following of infant signals is a
somewhat unstable distinction for Ainsworth, in part because she underspecified what actually
is being signalled.
The signals Ainsworth anticipated that a sensitive caregiver will attend to include the
baby’s ‘tempo, state and communications’.87 However, tempo, state, and communications are
quite different phenomena, and responding to them has quite different challenges and con-
sequences. Not least, an important aspect of sensitivity is the active support of helping a child
interpret their state—for instance, whether they feel hurt after falling over—rather than the
when applying it in different cultures. Dozier, M. & Bernard, K. (2019) Coaching Parents of Vulnerable Infants: The
Attachment and Biobehavioral Catch-up Approach. New York: Guilford, p.231.
83 LeVine, R.A. (2004) Challenging expert knowledge: findings from an African study of infant care and de-
velopment. In U.P. Gielen & J.L. Roopnarine (eds) Childhood and Adolescence: Cross-Cultural Perspectives and
Applications (pp.149–65). Westport, CT: Greenwood Publishing. Keller, H. (2018) Parenting and socioemotional
development in infancy and early childhood. Developmental Review, 50, 31–41.
84 Ainsworth, M. (1969) Cooperation vs. interference with baby’s ongoing behavior. http://www.psychology.
sunysb.edu/attachment/measures/content/ainsworth_scales.html. Rothbaum and Morelli mistakenly attribute

this quote to Ainsworth’s sensitivity scale in their accusation that the latter is ethnocentric. Rothbaum, F. & Morelli,
G. (2005) Attachment and culture: bridging relativism and universalism. In W. Friedlmeier, P. Chakkarath, & B.
Schwarz (eds) Culture and Human Development: The Importance of Cross-Cultural Research to the Social Sciences.
Lisse: Swets & Zeitlinger, p.103.
85 White, S., Gibson, M., Wastell, D., & Walsh, P. (2019) Reassessing Attachment Theory in Child Welfare.
Bristol: Psychology Press. See also Vicedo, M. (2013) The Nature and Nurture of Love: From Imprinting to
Attachment in Cold War America. Chicago: University of Chicago Press, Chapter 7.
edu/attachment/measures/content/ainsworth_scales.html.
87 Ibid.
Caregiving 125
passive receipt of pre-formed signals from an ‘autonomous’ infant.88 Admittedly, even ad-
vocates of Ainsworth’s coding system acknowledged that it is relatively poor at explicitly
indexing caregiver behaviour that pre-empts infant signals, so that these are not shown.89
Ainsworth’s overarching point, however, was that detection and response to tempo, state,
and communications are all part of sensitivity, as are responses that attend to the baby’s ex-
perience as a whole, rather than necessarily either following their wishes or waiting for their
signals. Some forms of insensitivity can come from lack of awareness or inaccurate inter-
pretation of tempo, state, and communications. But Ainsworth believed that significant in-
sensitivity is most likely when caregivers are geared largely by their own experience, rather
than taking that of the infant into account. This did not imply ascription of full autonomy or
personhood to the baby or the assumption that a baby, like a liberal citizen, can be assumed
to know his or her own wishes.90 It did, however, imply some attribution to the baby of a cap-

acity for experience relevant to the caregiver’s actions, revealed to some extent in the baby’s
mood and behaviour.91
Ainsworth identified that her measure of sensitivity had a high degree of stability over
time in her Baltimore sample, a finding replicated by later researchers working with sam-
ples of caregiver–infant dyads drawn from the community.92 Later researchers also found
that, remarkably, even if Ainsworth’s sensitivity scale was built for observations of infant–
caregiver interactions, the principles could readily be extrapolated to different ages without
needing to be recast. The construct of sensitivity as the perception and accurate interpret-
ation of signals, together with a prompt and appropriate response, has as much relevance
for setting boundaries with a toddler as it does in identifying the needs of an infant.93 Yet,
already in infancy there are varied important aspects of caregiving that are not reducible to
88 Ainsworth, M. (1992) A consideration of social referencing in the context of attachment theory and
research. In S.Feinman (ed.) Social Referencing and the Construction of Reality in Infancy (pp.349–67).
89 Kondo-Ikemura, K. (2001) Insufficient evidence. American Psychologist, 56(10), 825. On the issue of forms
of pre-emptive sensitivity see also Keller, H. & Otto, H. (2009) The cultural socialization of emotion regulation
during infancy. Journal of Cross-Cultural Psychology, 40(6), 996–1011; Shai, D. & Belsky, J. (2017) Parental em-
bodied mentalizing: how the nonverbal dance between parents and infants predicts children’s socio-emotional
functioning. Attachment & Human Development, 19(2), 191–219.
90 For a comparison of attachment theory and liberal political theory see Duschinsky, R., Greco, M., & Solomon,
J. (2015) Wait up! Attachment and sovereign power. International Journal of Politics, Culture, and Society, 28(3),
223–42. On the history of concepts of agency see Smith, R. (2015) Agency: a historical perspective. Annals of
Theoretical Psychology, 12, 3–29.
91 Keller has argued that Ainsworth’s sensitivity scale is ethnocentric, since there are societies in which ‘care-
givers do not take the infant’s point of view because infants have not (yet) attained personhood status, and it makes
no sense to take the perspective of someone who is not yet a person’. Keller, H. (2018) Parenting and socioemo-
tional development in infancy and early childhood. Developmental Review, 50, 31–41, p.38. However, the ascrip-
tion of personhood is by no means necessarily the same as taking the experience of another into account, which
was Ainsworth’s concern.
92 E.g. Lindhiem, O., Bernard, K., & Dozier, M. (2011) Maternal sensitivity: within- person variability and
the utility of multiple assessments. Child Maltreatment, 16(1), 41–50; Joosen, K.J., Mesman, J., Bakermans-
Kranenburg, M.J., & van IJzendoorn, M.H. (2012) Maternal sensitivity to infants in various settings predicts harsh
discipline in toddlerhood. Attachment & Human Development, 14(2), 101–17.
93 Britner, P.A., Marvin, R.S., & Pianta, R.C. (2005) Development and preliminary validation of the caregiving
behavior system: association with child attachment classification in the preschool Strange Situation. Attachment
& Human Development, 7(1), 83–102; Mesman, J. & Emmen, R.A. (2013) Mary Ainsworth’s legacy: a systematic
review of observational instruments measuring parental sensitivity. Attachment & Human Development, 15(5–6),
485–506; Hallers-Haalboom, E.T., Groeneveld, M.G., Endendijk, J.J., Linting, M., Bakermans-Kranenburg, M.J.,
& Mesman, J. (2017) Mothers’ and fathers’ sensitivity with their two children: a longitudinal study from infancy to
early childhood. Developmental Psychology, 53(5), 860–72.
126 Mary Ainsworth
awareness and accurate interpretation of the child’s signals.94 Some of these may be valuable
in particular ecological and cultural contexts.95 Others may have some claim to more general
relevance.96 For instance, an important later addition to Ainsworth’s concern with sensitivity
has come with growing attention amongst attachment researchers to the caregiving provided
by traumatised or abusive parents (Chapter 3).
Ainsworth found that caregiver sensitivity predicted children’s cooperativeness, distress,
and aggression on brief everyday separations within the home, and other positive aspects of
their home behaviour.97 Later researchers confirmed these findings, and contributed other
associations of sensitivity with psychological, linguistic, neurological, and even immuno-
logical correlates.98 For example, Manning, Davies, and Cicchetti documented that caregiver
sensitivity fully buffered the association between toddlers’ exposure to partner violence and
their later behavioural problems and prosocial behaviour. As Manning and colleagues ob-

served, this finding was theoretically expectable since sensitivity signals the capacity of the
parent to provide a safe base and secure haven, a capacity that can be anticipated to help
regulate the difficult feelings evoked for children in violent family contexts.99 Such associ-
ations have provided evidence of predictive validity by confirming Ainsworth’s interpret-
ation of sensitivity as relevant to a child’s later development.
Yet, as well as supporting Ainsworth’s concern with sensitivity, later researchers altered
some aspects of Ainsworth’s approach to measuring it. Caregiver sensitivity has often sub-
sequently been assessed in the context of play; this was a significant part of the activity ob-
served by the naturalistic observation of Ainsworth and her students.100 But it is not clear that
play is a good environment for understanding attachment processes specifically, especially
given that subsequent attachment researchers observed dyads for much briefer periods than
Ainsworth’s study.101 It is possible that some infant signals are more attachment-relevant
94 Van IJzendoorn, M.H. & Bakermans-Kranenburg, M.J. (2019) Bridges across the intergenerational trans-
mission of attachment gap. Current Opinion in Psychology, 25, 31–6.

95 Yovsi, R.D., Kärtner, J., Keller, H., & Lohaus, A. (2009) Maternal interactional quality in two cultural en-
vironments: German middle class and Cameroonian rural mothers. Journal of Cross-Cultural Psychology, 40(4),
701–707.
96 E.g. Whipple, N., Bernier, A., & Mageau, G.A. (2011) Broadening the study of infant security of attach-
ment: maternal autonomy-support in the context of infant exploration. Social Development, 20(1), 17–32. The
question of whether support for child exploration is a universal or a culturally specific contributor to infant attach-
ment security is one that has been debated in the cross-cultural research literature, though on the basis of ethno-
graphic observation rather than research findings using the standardized attachment measures. See e.g. LeVine,
R.A. (2004) Challenging expert knowledge: findings from an African study of infant care and development. In U.P.
Gielen & J.L. Roopnarine (eds) Childhood and Adolescence: Cross-Cultural Perspectives and Applications, 149–65.
Westport, CT: Praeger.
97 For a summary of all correlates of sensitivity in the Ainsworth Baltimore study see Bretherton, I. (2013)
Revisiting Mary Ainsworth’s conceptualization and assessments of maternal sensitivity-insensitivity. Attachment

& Human Development, 15(5–6), 460–84, Table 4.
98 E.g. Tamis-LeMonda, C.S., Briggs, R.D., McClowry, S.G., & Snow, D.L. (2009) Maternal control and sen-
sitivity, child gender, and maternal education in relation to children’s behavioral outcomes in African American
families. Journal of Applied Developmental Psychology, 30(3), 321–31; Vermeer, H.J., van IJzendoorn, M.H.,
Groeneveld, M.G., & Granger, D.A. (2012) Downregulation of the immune system in low-quality child care: the
case of secretory immunoglobulin A (SIgA) in toddlers. Physiology & Behavior, 105(2), 161–7; Bernier, A., Dégeilh,
F., Leblanc, É., Daneault, V., Bailey, H.N., & Beauchamp, M.H. (2019) Mother–infant interaction and child brain
morphology: a multidimensional approach to maternal sensitivity. Infancy, 24(2), 120–38.
99 Manning, L.G., Davies, P.T., & Cicchetti, D. (2014) Interparental violence and childhood adjustment: how
and why maternal sensitivity is a protective factor. Child Development, 85(6), 2263–78.
100 Mesman, J. & Emmen, R.A.G. (2013) Mary Ainsworth’s legacy: a systematic review of observational instru-
ments measuring parental sensitivity. Attachment & Human Development, 15(5–6), 485–506.
101 It may also not be the best context for making cross-cultural comparisons: Lancy, D.F. (2007) Accounting
for variability in mother–child play. American Anthropologist, 109(2), 273–84. It should be noted that samples
with longer observations, for instance the Minnesota Longitudinal Study of Risk and Adaptation (Chapter 4),
Caregiving 127
than others, though this will be a matter of degree. Leerkes and colleagues drew a distinc-
tion between sensitivity to signals suggesting infant distress and sensitivity to non-distress
signals. They found that only the former predicted later child attachment, conduct problems,
and social competence, as Ainsworth expected. The effect was particularly strong for chil-
dren who appeared temperamentally inclined to be easily distressed. By contrast, caregiver
sensitivity to non-distress signals did not have this effect, at least in the shorter observations
used in studies after Ainsworth.102 Though a qualification of Ainsworth’s operationalisation
of sensitivity, this finding is exactly in line with her theory, since signals suggesting infant
distress would have particular relevance to the attachment behavioural system, and to the
provision of a safe haven in particular.
Bowlby vs Ainsworth on feeding

Ainsworth’s inquiries in Uganda and in Baltimore sought to examine the caregiving behav-
iours associated with the development of attachment behaviour and an infant’s confidence in
his or her caregiver. One form of caregiving behaviour of particular salience at the time was
infant feeding, given the emphasis psychoanalysis had placed on the pleasure of feeding in
shaping the infant–caregiver relationship. Chapter 1 described how the rejection of Bowlby’s
‘Child’s tie’ paper by Klein and her followers in 1957 was a defining moment in his career
and contributed in a profound way to his expectation that his ideas would be rejected. In
particular, the Kleinians made repeated attacks on Bowlby for underplaying the importance
of feeding in the origins of ambivalence within family relationships. As a consequence, only
in the most passing references did Bowlby acknowledge that feeding deserved the status of
a behavioural system.103 For the rest of his career, Bowlby heaped scorn on anyone who as-
signed importance to feeding interactions for later psychological development: ‘It seemed to
me the feeding variable was totally irrelevant, or almost totally irrelevant.’104
However, this put Bowlby at odds with Ainsworth. In Uganda, Ainsworth observed that
reductions in breastfeeding, such as weaning, would lead to an increase in attachment behav-
iour by infants. Her Baltimore home observations led her to the conviction that social inter-
actions around feeding were relevant to the development of the child–parent attachment
relationship. In January 1967, she wrote to Bowlby to say that she was ‘astonished and—yes—
horrified at the tension and anxiety which attended feeding in the majority of infant–mother
pairs’. The feeding interaction was clearly one that both caregiver and infant regarded as im-
portant, complex, and troubling.105 Ainsworth did not agree with the Kleinian perspective,
or regard feeding as offering insight into an infant’s motivational underpinnings. However,
she saw the way that infant feeding served as a close interactive behaviour that required skill,
have tended to find an association between sensitivity and attachment even in contexts such as bathing, where
attachment-specific signals are less frequent.
102 Leerkes, E.M. & Zhou, N. (2018) Maternal sensitivity to distress and attachment outcomes: interactions with
sensitivity to nondistress and infant temperament. Journal of Family Psychology, 32(6), 753–61.
103 E.g. Bowlby, J. (1980, 1988) Caring for children. In A Secure Base (pp.1–21). London: Routledge: ‘I regard it
as useful to look upon parenting behaviour as one example of a limited class of biologically rooted types of behav-
iour of which attachment behaviour is another example, sexual behaviour another, and exploratory behaviour and
eating behaviour yet others’ (6).
104 Bowlby, J. (1977–79) Interview with Alice Smuts and Milton J.E. Senn. Wellcome Trust Library Archive. PP/
BOW/A.5/2.
105 Ainsworth, M. (1967) Letter to John Bowlby, 16 January 1967. Mary Ainsworth papers, Box M3168, Folder 2.
128 Mary Ainsworth
prompted complex emotions in caregivers, and afforded opportunities to observe a higher

rate of interaction than during many other activities.
In April, having received an early draft of Attachment, Volume 1, Ainsworth was full of en-
thusiasm for the book. However, she was critical of Bowlby’s perception that in order to show
the importance of attachment, he must downplay feeding interactions:
While agreeing with all that you say in the chapters you sent to me, I feel that there is still
something to be said about feeding and especially about mother–infant interaction in the
feeding situation. I hope that there may be room in your Chapter 10 to restore the balance.
I think you have pushed feeding behaviour very much out of the picture . . . Far too many
people confuse what happens in the so-called ‘oral phase’ with orality. There is obviously
much that goes on in the first year of life that is not linked in any way with hunger, feed-

ing behaviour, nurturance, dependence and the link. Nevertheless in my American sample
such a large proportion of the interaction between infant and mother during the first three
months of life took place in the feeding situation or relevant to it.106
In Ainsworth’s home observation data, infant signals related to feeding were a powerful pre-
dictor of later attachment.107 She agreed with Bowlby that this was not because the child’s
tie to his or her mother occurs because of a need for food. However, Ainsworth’s impres-
sion from her data was that when infants were hungry, attachment behaviour, not just food-
seeking behaviour, became activated.108 Furthermore, in her observations of infant care
practices in Uganda, breastfeeding served as both the major source of infant nutrition and
a primary means of soothing infant distress. She wrote to Bowlby that early feeding inter-
actions were emotionally charged, and the extent to which this was handled with sensitivity
had ramifications for other forms of interaction in the first year: ‘I do think that feeding
can become entangled with the development of attachment, and something more is needed
here.’109
In the final version of Attachment, Volume 1, Bowlby discussed the development of the
feeding response in infants, and conflict behaviour shown by animals when alarmed by a
threat whilst feeding. However, he ignored Ainsworth’s concerns. The power imbalance that
had characterised their early relationships remained at least partly in place here, as Ainsworth
publicly accepted Bowlby’s position even though her empirical data on this matter ran con-
trary. At least in part as a consequence, later attachment researchers generally followed
Bowlby’s lead, and did not discuss the specific qualities of feeding interactions even when
they were used instrumentally to measure sensitivity or infant secure base behaviour.110
Though specifically influenced by her Uganda ethnography and Baltimore study, Ainsworth’s emphasis on the
feeding interaction can be placed in the broader context of American parenting discourses in the 1960s, in which
the challenges of infant feeding and its value were being emphasized. See Foss, K.A. (2010) Perpetuating ‘scientific
motherhood’: infant feeding discourse in Parents magazine, 1930–2007. Women & Health, 50(3), 297–311.
107 Ainsworth, M.D.S. & Bell, S.M. (1969) Some contemporary patterns of mother–infant interaction in the
feeding situation. In A. Ambrose (ed.) Stimulation in Early Infancy (pp.133–70). London: Academic Press.
Disorders (pp.559–602). New York: Wiley, p.566.

110 An early exception is Egeland, B. & Brunnquell, D. (1979) An at- risk approach to the study of child
abuse: some preliminary findings. Journal of the American Academy of Child Psychiatry, 18(2), 219–35. However,
even in this article, the importance of feeding interactions in predicting child abuse appears clearly in the re-
sults but is downplayed in the discussion. Additional more recent exceptions include Britton, J.R., Britton, H.L.,
& Gronwaldt, V. (2006) Breastfeeding, sensitivity, and attachment. Pediatrics, 118(5), 1436–43; Woolley, H.,
The Strange Situation 129
Ainsworth would later write that Bowlby’s neglect of the topic had ultimately won out: ‘the
feeding situation has been neglected as a context for mother–infant interaction’.111 And the
direct role of food as a safe haven for many adults, or its role in family life as a symbol of care-
giving, has been ignored by researchers.112 Yet even if the particular issue of feeding inter-
actions was lost, Ainsworth’s deeper point was that certain kinds of interaction, like feeding,
offer an especially valuable window into the attachment relationship. Another such form of
interaction, as we shall see, was brief separations.
The Strange Situation
Origins of the procedure

The Strange Situation procedure was not planned when Ainsworth’s project was originally
proposed. However, Ainsworth decided to supplement her naturalistic observations with
a structured observation more intelligible to the academic psychology journals of the time.
She began privately to use the term ‘critical situations’ as the generic characterisation for any
predicament that activated the attachment behavioural system, thereby allowing ‘both oc-
currence and nonoccurrence of expected behaviors’ to be observed: ‘A baby does not spend
his day consistently manifesting a certain degree of attachment to this, that and the other
person. The quality and strength of his attachment is likely to be seen only in certain critical
situations.’113 Ainsworth and colleagues had seen feeding operate as just such a critical situ-
ation: the interplay of feeding and attachment was intense, complex, and often challenging,
and it was this interactional demand on the dyad that made it predictive of later attachment
behaviour by the child. Her home observations also led Ainsworth and colleagues to regard
the departure of a person from the room as another ‘critical situation’, and therefore a useful
vantage for relevant observation.114 Furthermore, Bowlby’s work strongly emphasised that
Hertzmann, L., & Stein, A. (2008) Video-feedback intervention with mothers with postnatal eating disorders
and their infants. In F. Juffer, M.J. Bakermans-Kranenburg, & M.H. van IJzendoorn (eds) Promoting Positive
Parenting: An Attachment-Based Intervention (pp.111–38). New York: Psychology Press; Tharner, A., Luijk, M.P.,
Raat, H., et al. (2012) Breastfeeding and its relation to maternal sensitivity and infant attachment. Journal of
Developmental & Behavioral Pediatrics, 33(5), 396–404; Messina, S., Reisz, S., Hazen, N., & Jacobvitz, D. (2019)
Not just about food: attachments representations and maternal feeding practices in infancy. Attachment & Human
Hevelopment, 23 April, 1–20. Illustrative of the underelaborated position of eating: the relevance of meal-times
is acknowledged by Poslawsky and colleagues in the choice of this potentially fraught setting for measuring sen-
sitivity among parents with children with autism. However, the researchers offer no reflection on the relation-
ship between attachment and meal-time practices. Poslawsky, I.E., Naber, F.B., Bakermans-Kranenburg, M.J., van
Daalen, E., van Engeland, H., & van IJzendoorn, M.H. (2015) Video-feedback intervention to promote positive
parenting adapted to autism (VIPP-AUTI): a randomized controlled trial. Autism, 19(5), 588–603.
111 Ainsworth, M. (1979) Infant–mother attachment. American Psychologist, 34(10), 932–7, p.934.
112 An exception is McCormack, M. (2012) Investigating the association between attachment and binge eating.
Unpublished doctoral dissertation, Deakin University, Victoria.

113 Ainsworth, M. (1963) Letter to John Bowlby, 18 July 1963. Mary Ainsworth papers, Box M3168, Folder 1;
Ainsworth, M., Bell, S., & Stayton, D. (1972) Individual differences in the development of some attachment behav-
iors. Merrill-Palmer, 18(2), 123–43: ‘In the case of attachment behaviors other than crying the coding did not begin
with the behavior itself, but rather with a “critical” situation that seemed likely to activate the behavior, so that both
occurrence and nonoccurrence of expected behaviors could be counted. Among such critical situations was the
departure of a person from the room in which an infant was situated’ (126).
114 Ainsworth, M., Bell, S., & Stayton, D. (1972) Individual differences in the development of some attachment
behaviors. Merrill-Palmer, 18(2), 123–43.

130 Mary Ainsworth
evolution made children disposed to experience unanticipated separations, even brief, as a

potential source of threat—what he would later term a ‘natural cue for danger’ (Chapter 1).
This stress was anticipated to increase the frequency and thus predictability and reliability
with which observers could directly examine attachment behaviour. Ainsworth decided to
bring her sample into the laboratory to participate in a study entailing brief separations of a
few minutes.
Ainsworth was impressed with the work of Bettye Caldwell’s Syracuse group, who had
combined interviews with mothers and observation of infant behaviour under standardised
laboratory settings.115 However, she felt that Caldwell and colleagues had conducted their
study in such a way that it was difficult to cleanly distinguish the role of different behavioural
systems in the interpretation of behaviour:

One of their projects—a study of visual following in the first year of life—is so neatly con-
trolled that its equivocal results give one little chance of sorting out the variables. (The
problem is that visual fixation, searching and following may indicate a) attachment; b)
keeping the secure base in view; c) curiosity and exploration of a new stimulus; and d)
keeping a wary eye on the stranger who is a threat.) In ordinary free observation with the
context clearly in mind, it is no real problem to differentiate between these four possibil-
ities. But it is in an experiment in which the four possibilities were not clearly envisaged to
begin with.116
In 1964, when the first wave of her sample of infants were 11 months old, Ainsworth at-
tempted a study to cleanly distinguish prompts for behavioural systems.117 Van Rosmalen
and colleagues documented that the term ‘strange situation’ was already in circulation before
Ainsworth, to describe a procedure in which the responses of young children to an unfamiliar
environment were observed, and compared with other information known about the child’s
life.118 ‘Strange’ here referred to the novelty of the environment for the infant. Ainsworth’s
Strange Situation was especially indebted to the ‘strange situation’ of Jean Arsenian, who
had examined infant behaviour in response to the novel environment of the laboratory, and
in the presence and absence of their mother.119 Arsenian’s sample was drawn from mothers
115 Caldwell, B.M., Hersher, L., Lipton, E.L., et al. (1963) Mother–infant interaction in monomatric and poly-
matric families. American Journal of Orthopsychiatry, 33(4), 653; Caldwell, B.M. & Hersher, L. (1964) Mother–
infant interaction during the first year of life. Merrill-Palmer, 10(2), 119–28.
116 Ainsworth, M. (1963) Letter to John Bowlby, 27 June 1963. Mary Ainsworth papers, Box M3168, Folder 1.
Ainsworth. Monographs of the Society for Research in Child Development, 60(2/3), 2–21: ‘I had seen a lot of separ-
ations and reunions in the homes, a lot of exploration, a lot of proximity seeking, and a lot of differences in how the
baby and the mother behaved in these situations. So constructing the episodes of the Strange Situation wasn’t hard
at all; as I recall, it took just about half an hour of talking with Barbara Wittig to decide on the episodes and their
sequence—it just came naturally’ (12).
118 Van Rosmalen, L., Van der Veer, R., & Van der Horst, F. (2015) Ainsworth’s strange situation procedure: the
origin of an instrument. Journal of the History of the Behavioral Sciences, 51(3), 261–84. See Shirley, M.M. (1942)
Children’s adjustments to a strange situation. Journal of Abnormal and Social Psychology, 37, 201–17. Shirley’s
work is not cited by Ainsworth in any of her publications but is cited by Arsenian, and by Bowlby, J. (1969, 1982)
Attachment, Volume 1. London: Penguin, p.247. Michael Lamb (personal communication) recalls that Ainsworth
recommended that he read Shirley’s paper in 1973. The procedure seems to have also been independently devel-
oped by Harlow, who applied the approach to study the exploratory and care-seeking behaviour of baby rhesus
monkeys in an unfamiliar environment: Harlow, H.F. (1958) The nature of love. American Psychologist, 13, 673–85.
119 Arsenian, J.M. (1943) Young children in an insecure situation. Journal of Abnormal and Social Psychology, 38,
225–49.
and children in a reformatory, and the mothers were permitted only limited access to their
children. Arsenian sought to confirm Blatz’s idea that, with the caregiver available, children
feel more secure and respond to the environment with more exploration and less distress.
The availability of security from the caregiver counteracted the ‘fear of the strange’ prompted
by the unfamiliar environment, as did more time in the setting. As well as demonstrating
the importance of the caregiver for providing the confidence for exploration, Arsenian also
showed that caregiver availability reduced the incidence of crying, and of stereotypic and
other anomalous behaviours without ‘goal-directedness’, and which primarily ‘appeared to
be determined by a condition of excess tension’.120
Following Arsenian, the presence of the caregiver as secure base was expected by
Ainsworth to serve as adequate reassurance for most infants. After the novelty of the en-
vironment itself, a second prompt was the availability of toys, a little distance away from

the caregiver’s chair. This was a cue for the exploratory system. Infants were given around
three minutes to acclimatise to the room and, should they wish, explore the toys, before
a stranger enters. In Uganda, Ainsworth saw that her own entrance, as a relative stranger
(and as a white Canadian), provided the most reliable prompt for the display of attachment
behaviour by infants.121 In the Ainworth Strange Situation the stranger begins by sitting
quietly and observing the dyad; the stranger then speaks with the caregiver, which conveys
to the infant that the caregiver does not regard the stranger as a threat; finally, the stranger
attempts interaction with the infant. This was expected to activate the attachment behav-
ioural system another increment. The caregiver then takes leave of and returns to the infant
twice. These separations, in turn, ratchet up the activation of the attachment behavioural
system further.
The Strange Situation used the combination of separations and the unfamiliar envir-
onment to achieve a functional equivalent of instances of availability or unavailability
embedded in the wider life of the child.122 With the behaviour of the caregiver partially
standardised, and the attachment behavioural system activated by careful increments,
Ainsworth aimed to mobilise the infant’s expectations based on what happened when he
or she has felt anxiety in the past about the availability of the attachment figure, and allow
a viewer to interpret these expectations from observed behaviour. As the episodes of the
procedure modulate the infant’s anxiety, Ainsworth anticipated that the infant’s movement
between behavioural systems would be displayed: the interplay of exploration of novelty and
attachment behaviour, in the presence and in the absence of a parent.123
120 Ibid. p.227. Arsenian’s findings were replicated by Rheingold, H.L. (1969) The effect of a strange envir-
onment on the behavior of infants. In B.M. Foss (ed.) Determinants of Infant Behavior, Vol. 4 (pp.137–66).
London: Methuen. Ainsworth, M. (1998) Harold Stevenson—SRCD oral history interview. http://srcd.org/sites/
default/files/documents/ainsworth_mary_interview.pdf: ‘One day, Harriet [Rheingold] and I met at a meeting and
I said, “Oh, Harriet, you’d be interested in something I’m currently doing, um, the strange situation . . .” (Laugh)
“You are too? I’m just starting mine”.’

of the Strange Situation. Bristol: Psychology Press: ‘Tight control of maternal behaviour is impossible and indeed
undesirable. The compromise represented in our procedures turned out to have effected a reasonable degree of
standardisation of the situation, while allowing most mothers to behave naturally and fairly comfortably’ (41).
Cf. Brown, S. (2012) Abstract experimentalism. In N. Wakeford & C. Lury (eds) Inventive Methods (pp.61–75).
London: Routledge.
year-olds in a strange situation. Child Development, 41(1), 49–67, pp.50–53.

132 Mary Ainsworth
The Ainsworth Strange Situation was, then, a means for coaxing to visibility infants’
expectations about the availability of their caregiver as a secure base and safe haven, and
arraying these expectations and their associated affects within physical space and over epi-
sodes to make them available for analysis. In this way, the Strange Situation was intended to
dramatise a predicament faced in an ordinary, low-level way by the infant–caregiver dyad in
everyday life: the question of the extent to which the infants’ experiences led them to believe
that the caregiver was available when needed. The highly contrived situation was intended to
intensify and display specific aspects of real life experience, to be interpreted in the context of
home observations of these dyads.124
Ainsworth anticipated that, with the attachment behavioural system activated through
‘cumulative stresses’, infants would be increasingly disposed to seek their caregiver as a safe
haven:125

The results are very much in accordance with expectation. Nearly all children explore vig-
orously when mother is there and not when she is absent; nearly all protest and attempt
to follow when she leaves; stranger anxiety is variable, but mother is used as a secure base
and/or a haven of safety when a stranger is there.126
In general terms, Bowlby’s description of the expectable behavioural expression of the at-
tachment system was confirmed. Ainsworth was therefore all the more intrigued, however,
by the fact that some of the Baltimore infants made no approach to their caregiver after the
first reunion. However, the second separation seemed to activate the attachment behavioural
system to an intensity that they abandoned this task, and instead sought their caregiver:
Two little girls faced the strange situation with remarkable poise, to the extent of interact-
ing with the stranger and offering her toys—only to disintegrate when mother returned for
the second time, to cry and cling and carry on, as though they had borne as much as they
could, and now could give delayed expression of their distress.127
Yet several infants did not display distress even after the second separation, and Ainsworth
also noted the display of tension behaviours during reunion, suggesting the strain of holding
back the expression of the attachment behavioural system. She wrote to Bowlby:
A couple of babies who are clearly attached to their mothers showed relatively little stranger
anxiety and separation-disturbance, although showing subtle differences in behaviour in
the various phases of the strange situation, but they manifested the strain that had been
placed upon them by disturbance when the mother returned.128
124 Writing under Ainsworth’s influence, see Waters, E. & Sroufe, L.A. (1983) Social competence as a develop-
mental construct. Developmental Review, 3, 79–97: ‘Any single sample of naturalistic behavior, especially if brief,
could be unrepresentative and, paradoxically, less revealing of the child’s competence in the “real world” than a
strategically designed laboratory assessment, in which a child must cope with a problem that regularly (though
rarely) occurs in the natural environment’ (85). In fact, the similarity of the Strange Situation to ordinary expect-
able brief infant–caregiver separations has regularly been used by attachment researchers in research ethics appli-
cations over the decades.

126 Ainsworth, M. (1965) Letter to John Bowlby, 2 February 1965. PP/Bow/D3/69.
127 Ibid.
128 Ibid.
The apparent lack of distress on separation was reminiscent of some of the infants Ainsworth
had observed in Uganda, who showed few attachment behaviours in response to separations
and reunions with their caregivers. These had often been infants with relatively less-sensitive
caregivers, by Ainsworth’s ethnographic assessment. The unruffled behaviour of these in-
fants also resembled the avoidant or ‘detached’ behaviour of some of the long-term hospital-
ised children seen by Robertson when observed in reunions with their caregivers. Ainsworth
quickly concluded that these individual differences in infant behaviour reflected differences
in the history of the caregiver–infant relationship.
Forming categories

Within Bowlby’s research group at the Tavistock in the early 1960s, Rudolph Schaffer studied
60 infants, who were observed at four-weekly intervals until they were one year of age.129
Ainsworth acknowledged the importance of this study in documenting the development of
attachment behaviour towards the infant’s primary caregivers through the first year of life.
However, Ainsworth’s grounding in Blatz’s ideas made her troubled by Schaffer’s approach.
Schaffer assumed that more attachment behaviour directed towards a figure would indicate
more attachment—except in situations like foster-care where an attachment bond may still
be in formation.130 Ainsworth wrote that such a quantitative approach missed important
qualitative differences resulting from security, anxiety, and defences.131 On the one hand,
her basis in Blatz’s work led her to anticipate that a secure attachment would be associated
with less clinging, crying, and following, except when the child needed comfort. When using
the caregiver as a secure base for exploration, Ainsworth anticipated that babies would show
little attachment behaviour towards the caregiver, except to periodically check in with them
and confirm their availability.132 On the other hand, Ainsworth worried that counting at-
tachment behaviours would be a treacherous research strategy, since a child unsure about
the availability of their attachment figure may intensify attachment behaviours, and a child
who has learnt that attachment behaviours will be ignored or punished by an attachment
figure may show fewer.133
129 Schaffer, H.R. & Emerson, P.E. (1964) The development of social attachments in infancy. Monographs of the
Society for Research in Child Development, 29(3), 1–77.

130 See e.g. Ainsworth, M., Blehar, M., Waters, E., & Wall, S. (1978, 2015) Patterns of Attachment: A
Psychological Study of the Strange Situation. Bristol: Psychology Press. Ainsworth and colleagues propose that
the Strange Situation can be used in custody and forensic decision-making: ‘A practical situation in which the
issue is whether or not to remove a child from his natural parents and place him in a foster or adoptive home, it
might be of moment to ascertain whether he has become strongly enough attached to his parent(s) that it would
be more traumatic to him to be separated from them or to remain with them’ (291). However, Ainsworth’s meas-
ures did not allow for assessment of strength of attachment. A scale for assessing the strength of attachment in
the Strange Situation would later be developed by Betty Carlson, though it has seen little use outside of research
contexts. See Zeanah, C.H., Smyke, A.T., Koga, S.F., Carlson, E., & Bucharest Early Intervention Project Core
Group (2005) Attachment in institutionalized and community children in Romania. Child Development, 76(5),
1015–28.
131 Ainsworth, M. & Wittig, B. (1969) Attachment and exploratory behaviour of one-year-olds in a Strange
Situation. In B.M. Foss (ed.) Determinants of Infant Behaviour, Vol. 4 (pp.111–36). London: Metheun, p.112–13.
132 Ainsworth, M. (1973) A Secure Base. Unpublished manuscript. PP/Bow/J.1/33. This claim was soon after
supported by Carr, S., Dabbs, J., & Carr, T. (1975) Mother–infant attachment: the importance of the mother’s visual
field. Child Development, 46, 331–8.
133 Ainsworth, M D.S. (1985) Attachments across the life span. Bulletin of the New York Academy of Medicine, 61,
792–812, p.805.
134 Mary Ainsworth
Ainsworth therefore stressed that researchers should not count behaviours in order to as-
sess the ‘strength’ of an attachment, but take note of qualitative differences in attachment rela-
tionships under conditions where the attachment system was anticipated to be activated. When
the second wave of her sample reached 11 months, Ainsworth and her team conducted the
Strange Situation again with these infant–caregiver dyads. In total, 23 of the 26 dyads in her
sample were seen in the Strange Situation. On the basis of these further observations, Ainsworth
distinguished three groups. Initially she termed them ‘prematurely independent’ (6 dyads),
‘secure’ (13 dyads), and ‘disturbed’ (4 dyads). However, Bowlby urged that these terms were ‘shot
through with value judgements & hidden predictions’.134 He suggested that the labels ‘A’, ‘B’, and
‘C’ should be used instead to avoid prejudging what the individual differences would mean.135
This was a strategy used by Ainsworth, Robertson, and Bowlby from their earliest work together,
analysing Robertson’s notes to distinguish different groups of children based on their response

to reunion after hospitalisation.136
In a letter to Bowlby from 1967, Ainsworth described Group B as ‘normally attached’. The
attachment behavioural system was activated and expressed in infant behaviour according to
the expected increments, and could be deactivated again by the presence of the caregiver. The
attachment behaviours shown were various. Infants could crawl or waddle to their caregiver
on reunion, lift their arms to be picked up, signal with a directed cry, or clamber up onto the
caregiver. Group A initially comprised those dyads with infants who did not show separation
anxiety as expected; Ainsworth observed that these infants had a variety of different histories of
care. But—at least by the standards of her low-risk sample—her home observations suggested
‘a deprivingly disturbed relationship with their mothers’.137 In 1969 Ainsworth changed the dis-
tinction, so that Group A was no longer defined by the lack of separation anxiety but rather clas-
sified those infants who did not show more attachment behaviour as the situation contributed
greater anxiety, and instead directed their attention and their movements away from their care-
giver.138 Ainsworth found no infants who avoided both the mother and the stranger, contrary
the idea that ‘avoidance’ might simply be regarded as a trait of the individual infant. The groups
were not ultimately defined by countable behaviours, but by whether the attachment behav-
ioural system was inferred to have found expression in behaviour (see Table 2.1 for Ainsworth’s
Strange Situations classifications).
In Patterns of Attachment, reporting results from Ainsworth’s original samples plus
doctoral projects by her first students, 66% of the total of 106 infants were classified as
134 Bowlby, J. (1967) Letter to Mary Ainsworth, 19 April 1967. Mary Ainsworth papers, Box M3168, Folder
2: ‘Terminology. All your terms—securely attached, prematurely independent & disturbed—are shot through with
value judgements & hidden predictions.’
135 Ainsworth, M. (1967) Letter to John Bowlby, 6 August 1967. Mary Ainsworth papers, Box M3168, Folder 2.
136 Ainsworth, M., Robertson, J., & Bowlby, J. (1953) ‘Reunion after prolonged separation’, chapters drafted for
an unpublished book. PP/BOW/D.3/21; see also Van Rosmalen, L., Van der Veer, R., & Van der Horst, F. (2015)
Ainsworth’s strange situation procedure: the origin of an instrument. Journal of the History of the Behavioral
Sciences, 51(3), 261–84.
137 Ainsworth, M. (1967) Letter to John Bowlby, 6 August 1967. Mary Ainsworth papers, Box M3168, Folder
2. For Ainsworth’s reflections having seen a greater diversity of samples, including higher risk samples and cases
of serious child neglect, see Ainsworth, M.D.S. (1980) Attachment and child abuse. In: G. Gerbner, C.J. Ross,
& E. Zigler (eds) Child Abuse: An Agenda for Action (pp.35–47). Oxford: Oxford University Press; Crittenden,
P.M. & Ainsworth, M.D.S. (1989) Child maltreatment and attachment theory. In D. Cicchetti & V. Carlson (eds)
Child Maltreatment: Theory and Research on the Causes and Consequences of Child Abuse and Neglect (pp.432–63).
Cambridge: Cambridge University Press.
Situation. In B.M. Foss (ed.) Determinants of Infant Behaviour, Vol. 4 (pp.111–36). London: Metheun, p.126.
Table 2.1 The Ainsworth Strange Situation Classifications, as outlined in Patterns

of Attachment (1978)
Attachment Strange Situation Behaviour

classification
A Lower proximity-seeking and contact-maintaining on reunion

than B or C, together with some proximity-avoiding behaviours.
The infant’s behaviour, attention and affect are integrated in a
coherent way to downplay the communication of distress and keep
focus away from the caregiver, e.g. by attention to the toys.
A1 Lowest proximity-seeking and contact-maintaining on reunion
than B or C; strongest proximity-avoiding behaviours.
A2 Low to moderate proximity-seeking on reunion. Marked

proximity-avoiding behaviours.
B Strong proximity-seeking and contact-maintaining on reunion
compared to A. Low contact-resisting compared to C. The infant’s
behaviour, attention and affect integrate in a coherent way which
allows distress to be communicated to the caregiver and assuaged,
allowing the child to then return calmly to play.
B1 Weak proximity-seeking and contact-maintaining. Weaker
proximity-avoiding behaviours than A1. Strong communication
and affective sharing with their caregiver from a distance.
Conceptualised as intermediate between the A and B infants.
B2 Low to moderate proximity-seeking and marked proximity-
avoiding on first reunion. But then strong proximity-seeking and
contact-maintaining on second reunion.
B3 Strong proximity-seeking and contact-maintaining on reunion.
No contact-resisting or proximity-avoiding.
B4 Some proximity-seeking and contact-maintaining prior to
separation from the caregiver. Strong proximity-seeking and
contact-maintaining prior to separation from the caregiver on
reunion. Some contact-resisting.
C Marked contact-resisting behaviour. The infant’s behaviour,
attention and affect integrate in a coherent way which strongly
communicates their distress and frustration to the caregiver.
C1 Strong proximity-seeking and contact-maintaining on reunion.
Strong contact-resisting behaviour punctuates the contact-
maintaining, as the child switches between communicating
distress and a desire for contact, anger, and a desire to be put down.
C2 Weak proximity-seeking but moderate to strong contact-
maintaining, particularly on second reunion. Moderate
contact-resisting.
Group B.139 The largest proportion of dyads showed a smooth balance between attachment
and exploration: with increasing prompts for the attachment behavioural system, attach-
ment behaviour increased; when the caregiver was present, the child was comforted and
could return to play. Ainsworth labelled dyads where this pattern of behaviour was shown

136 Mary Ainsworth
as B3. Out of 106 infants, 42% were classified B3, compared to 23% other kinds of B.140 One
kind of Group B response that differed from the prototypical B3 was evident in the differ-
ence between infant behaviour on first and second reunion. From the very first, Ainsworth
had been interested in the fact that some children showed avoidance on the first reunion,
and then attachment behaviours on the second reunion. Their behaviour conveyed a sense
that with the increasing activation of the attachment behavioural system, these infants felt
that they were no longer able to manage their distress on their own, and that their caregiver
would be receptive under such circumstances. Their avoidance thawed as their desire for
comfort increased. These dyads were labelled B2.141
Another group of infants did not display much separation anxiety or proximity-seeking
on reunion, but were unmistakably happy to see their caregiver again on reunion. And
the attachment system seemed to be able to be terminated through distance interaction.

Ainsworth termed dyads with such infants B1.142 This ran counter to Bowlby’s assumption
that proximity would be the set-goal of the attachment behavioural system in infancy, and
ultimately led to his qualification in his final writings that the set-goal is caregiver avail-
ability (Chapter 1). Nonetheless, Ainsworth and colleagues described B3 as the short and
most direct expression of the attachment behavioural system, whereas B1 and B2 were
regarded as ‘complicated’ expressions of the behavioural system.143 The Ainsworth labora-
tory considered B1 and B2 as, ultimately, intermediate between Group A and Group B in
their Strange Situation behaviour.144 In work using the Strange Situation by Ainsworth’s
student Sylvia Bell, an additional subgroup was added for infants who displayed more
distress and resistance than the infants of B3 dyads, but who ultimately were able to be
comforted by the presence of their caregiver and return to exploration within the Strange
Situation.145
Ainsworth and her group also distinguished subtypes of Group A. These subtype
groupings evolved over time, but had stabilised by the mid-1970s. A1 dyads were char-
acterised by the infant’s rigidly held avoidance; A2 dyads were characterised by a partial
approach by the infant, succeeded by avoidance. In both cases, the infants ultimately
140 Ibid.
141 Throughout the 1970s and early 1980s, Ainsworth repeatedly considered splitting the B2 group up, especially
after seeing more 18-month Strange Situations. One group of B2 dyads would contain children who were confident
in their capacity to self-regulate following the first reunion, but who knew they could approach their caregiver as
needed when their anxiety and distress became greater on the second reunion. Another group of B2 dyads con-
tained children who seemed anxious and unhappy, and so avoided on the first reunion, but who could not sustain
their avoidance into the second reunion. See e.g. Ainsworth, M. (1981) Letter to Michael Lamb. Mary Ainsworth
Archive, Box M3173, Folder 4. It is quite possible that where an avoidant strategy seems bent or snapped rather
than relaxed into proximity-seeking, this would now generally get coded as D/A, since it would come with other
markers of tension. Certainly Ainsworth’s concern about the insecure B2s was no longer mentioned after the intro-
duction of the D classification, though this may also have been because the coding system by that point was too
well established.

year-olds in a strange situation. Child Development, 41(1), 49–67, p.52.

144 Blehar, M., Lieberman, A., & Ainsworth, M. (1977) Early face-to-face interaction and its relation to later
infant–mother attachment. Child Development, 48(1), 182–94, p.186. Ainsworth would later conceptualize the
B1 classification as a kind of reserve, observable in other forms with later maturation—personal communication
cited in Cassidy, J., Marvin, R., with the Attachment Working Group of the John D. and Catherine T. MacArthur
Network on the Transition from Infancy to Early Childhood (1992) Attachment organisation in preschool chil-
dren: procedures and coding manual. Unpublished manual.
145 Ainsworth, M.D.S., Bell, S.M., & Stayton, D.J. (1971) Individual differences in strange-situation behavior of
one-year-olds. In H.R. Schaffer (ed.) The Origins of Human Social Relations (pp.17–58). New York: Academic Press.
did not engage in affective communication with their caregiver, even as the attachment
behavioural system was presumed to be incrementally activated by the episodes of the
Strange Situation. Instead, a characteristic of the group was that they would often en-
gage with the toys or point out toys to the caregiver precisely when another child showed
distress and attachment behaviour. Despite individual differences within the groups, at
base the predicaments faced by the Group C and Group A dyads differed from one an-
other. Ainsworth and colleagues offered the dictum that ‘the C baby fears that he will
not get enough of what he wants; the A baby fears what he wants’.146 In other words, C
babies are not confident in the availability of the caregiver in the Strange Situation to
offer the comfort and protection they desire; A babies are concerned that expression of
desire for the caregiver will not be effective or, indeed, will backfire by eliciting rebuff or
punishment. In Patterns of Attachment, 21% of the total of 106 infants were classified as

Group A.147
Ainsworth put dyads in Group C if the infants did not show the A or B responses.148 It is
curious to see that even in Patterns of Attachment as late as 1978, Ainsworth was still using
Group C in part as a residual category for generally ‘maladaptive’ behaviours.149 However,
this was in part a holdover, and was not how she discussed the category with her students
and collaborators. Though it began as a residual category, through the late 1960s Ainsworth’s
comparison of the Strange Situation and home observations led her to identify a common
theme in the behaviour of most of the Group C infants: ‘They are diverse, but they have in
common the trait of low frustration tolerance, and the experience that their own actions
have no consistent consequences, because so much that happens to them is the result of the
mother’s timing, not theirs’.150
In the 1960s, Ainsworth and her group came to distinguish two subtypes of Group C be-
haviour from within the ‘mixed bag’.151 The overall group became characterised as ‘babies
who were markedly distressed in both separation episodes, and whose behaviour throughout
the strange situation showed disturbance of a passive–aggressive nature’.152 The C1 clas-
sification was used for infants whose behaviour towards their caregiver clearly suggested
frustration or anger, most notably in resisting being held by the caregiver and less active in
maintaining contact. The C2 classification was used when infant behaviour was conspicu-
ously ‘passive’, though also had more of a tone of anger in their interactions with their care-
givers than the Group A or B infants.153 C1 infants would approach the caregiver on reunion
or reach for a pick-up, showing a mixture of attachment behaviours and signs of frustration.
By contrast, C2 infants would wail helplessly and gaze beseechingly at the caregiver, without

147 Ibid. p.230.
148 Ainsworth, M. (1967) Letter to John Bowlby, 6 August 1967. Mary Ainsworth papers, Box M3168, Folder 2.
of the Strange Situation. Bristol: Psychology Press, p.62. This role of Group C as a catch-all for anomalous behav-
iours would finally be officially eliminated by Ainsworth in the mid-1980s, following the introduction of the D
classification by Main (Chapter 3).
151 Ainsworth, M. (1967) Letter to John Bowlby, 16 January 1967. Mary Ainsworth papers, Box M3168, Folder
2: ‘I am sure that Group C will become at least two groups rather than the mixed bag it presently is.’
152 Ainsworth, M. (1967) Letter to John Bowlby, 17 October 1967. PP/Bow/K.4/12.
138 Mary Ainsworth
taking much determinate action to achieve their evident desire for closeness, and without
being fully comforted when that closeness was achieved. All of Ainsworth’s C2 infants also
displayed stereotypic behaviours, such as rocking to themselves;154 and Ainsworth later
wondered whether what she was seeing were infant ‘attempts to cope with a threat of psych-
otic fragmentation’ by quite mentally ill mothers.155
C1 and C2 infants had in common that the attachment behavioural system had a low
threshold for activation and termination: Group C infants were more wary of the stranger
than the other children seen in the Strange Situation, and might stop play and show a de-
gree of attachment behaviour even before the first separation. Additionally, following the re-
unions, they were not comforted or able to return to play. Whereas Group A infants seemed
unwilling to permit tension or drama, Group C infants seemed not to permit their reso-
lution. Ainsworth termed Group C ‘ambivalent/resistant’. Bowlby regarded this as an un-

helpful label since both Group A and Group C infants were ambivalent about contact and
comfort from their caregiver: Group A because they felt that they were not permitted to
seek their caregiver; Group C because they were not satisfied by the contact and comfort
they received.156 Ainsworth, on the other hand, did not use the term ‘ambivalent’ to refer
to the inner experience of conflict inferred to be common to Group A and Group C infants.
Instead, she used the term to refer to the child’s observable mingling of contact-seeking and
contact-resisting behaviour.157 In Patterns of Attachment, Ainsworth and colleagues re-
ported on the proportions of infants in different groups, drawing on the original Baltimore
study and further Strange Situations from studies by Silvia Bell and Mary Main (Chapter 3).
Thirteen percent of the 106 infants were classified as C.158
Strange Situation scales
Reflecting on the subtypes, Ainsworth came to the conclusion that infants could be distin-
guished by four kinds of behaviour. She developed scales that took account of ‘1) the degree
of activity and initiative of the behaviour; 2) promptness of the behaviour; 3) frequency of
the behaviour; and 4) duration of the behaviour’.159 These scales have only recently been
published as an appendix to the 2015 Psychology Press edition of Patterns of Attachment.
In the decades before that, they circulated as an unpublished manuscript, passed to indi-
viduals attending a training institute. The Ainsworth scales are, in practice, partly a written
and partly an oral tradition. Richters, Waters, and Vaughn found that without training in
154 Ainsworth, M.D.S., Bell, S.M., & Stayton, D.J. (1971) Individual differences in strange-situation behavior
of one-year-olds. In H.R. Schaffer (ed.) The Origins of Human Social Relations (pp.17–58). New York: Academic
Press, p.39.
155 Ainsworth, M. (1980) Infant attachment and maternal care: some implications for psychoanalytic concepts
of development. PP/Bow/J.1/53.
156 Bowlby, J. (1990) Letter to Sonia Monteiro de Barros, 6 August 1990. PP/Bow/B.3/40: ‘You are quite right to
link the two patterns of attachment you refer to, namely “anxious resistant” and “anxious ambivalent”. In fact, Mary
Ainsworth herself sometimes used “anxious ambivalent” as synonymous with “anxious resistant”. I thought that
was a mistake since all the insecure patterns of attachment are characterised by ambivalence, sometimes overt and
obvious, at others (e.g. avoidant & compulsive caregiving) only covert and potential.’
157 Ainsworth, M. (1969) Letter to John Bowlby, 23 February 1971. Mary Ainsworth Archive, Box M3168,
Folder3.

159 Ibid. p.51.
using these scales, inter-rater reliability is no better than chance.160 When the written text
is combined with training, three of the scales have incredible clarity and usability, a kind of
deftness of touch. These are the proximity-seeking, contact-maintenance, and avoidance
scales. Based on examples from a very small sample, Ainsworth managed to characterise
infant behaviour in terms of (i) initiative, (ii) promptness, (iii) frequency, and (iv) duration
within single dimensions. And this measure has captured the behaviour of the large ma-
jority of infants in all subsequent samples with a surprising degree of effectiveness. There
is certainly some shoehorning that takes place as coders work with samples with very dif-
ferent caregiving cultures; but, as Behrens observed, what is curious is that there is much
less than might be expected.161
In the case of proximity-seeking, higher scores have to do with the efforts the infant puts
into getting proximity. In the highest score on the scale, ‘the baby purposively approaches the

adult, creeping, crawling, or walking. He goes the whole way and actually achieves the con-
tact through his own efforts, by clambering up on or grasping hold of the adult.’162 A lower
score is assigned when the initiative, promptness, frequency, or duration is lower. So, for
instance, if the baby makes three full approaches to the caregiver, but without completing
contact, this scores 4 out of 7. In the case of contact-maintenance, higher scores reflect the
initiative of the baby and the duration of contact. As such, even a baby who is held for a long
time can receive the lowest possible score—1 out of 7—on contact-maintenance, if when
picked up ‘he neither clings nor holds on, and when he is put down he makes no protest; if he
is not put down he may still be coded (1) if he seems indifferent to being held’.163 By contrast,
a lower score is assigned when desire to maintain contact is less visible, or relatively less effort
is engaged to achieve it. For instance, a score of 4 out of 7 is assigned when ‘the baby has been
held, perhaps clinging a little, perhaps having diminished his crying when picked up; when
put down he decisively protests, giving more than a brief cry’ or when ‘The baby, having been
held, is released; he resists release briefly, by attempting to hold on or by clinging briefly, but
when this is ineffective he accepts the release without protest and without further effort to
maintain contact.’
The avoidance scale emphasises especially promptness, frequency, and duration of at-
tempts to direct attention or behaviour away from the caregiver. However, the initiative
taken by the child is also emphasised. The highest score on the scale can only be achieved by
an infant who ignores a caregiver attempting to directly attract his or her attention: ‘The baby
does not greet the mother upon her return in a reunion episode neither with a smile nor with
a protest. He pays little or no attention to her for an extended period despite the mother’s
efforts to attract his attention. He ignores her, and may turn his back on her. If his mother
nevertheless picks him up he remains unresponsive to her while she holds him, looking
around, interested in other things.’164 A lower score is assigned if avoidance is persistent but
low-keyed or only occasional. A score of 4 out of 7 is assigned, for instance, if ‘the baby fails
160 Richters, J.E., Waters, E., & Vaughn, B.E. (1988) Empirical classification of infant–mother relationships from
interactive behavior and crying during reunion. Child Development, 59(2), 512–22, p.520.
161 Behrens, K.Y. (2016) Reconsidering attachment in context of culture: review of attachment studies in Japan.
Online Readings in Psychology and Culture, 6(1), 7.

162 Ainsworth, M., Blehar, M., Waters, E., & Wall, S. (1978, 2015) Appendix: Coding of infants’ interactive
behaviour in the Strange Situation. In Patterns of Attachment: A Psychological Study of the Strange Situation.
163 Ibid.
164 Ibid.
140 Mary Ainsworth
to greet his mother and ignores her for a time and then takes the initiative in making contact
or undertaking interaction, even though the mother has not sought his attention’.165
Ainsworth also developed a resistance scale, which measures the intensity, frequency, and
duration of frustration or aggression directed towards the caregiver, including frustrated re-
sistance to being held. This scale is somewhat less polished and well characterised, likely
because incidence of aggression towards the caregiver was less frequent in the Ainsworth
sample than proximity-seeking, contact-maintenance, and avoidance. However, the resist-
ance scale has appeared to have equivalent inter-rater reliability to the others. The relevant
behaviours are: ‘pushing away, throwing away, dropping [toys passed to the infant by the
caregiver], batting away, hitting, kicking, squirming to be put down, jerking away, stepping
angrily, resistance to being picked up or moved or restrained. More diffuse manifestations
are angry screaming, throwing self about, throwing self down, kicking the floor, pouting,

cranky fussing, or petulance.’166 The highest score can be assigned, for instance, if the coder
sees the infant enter into ‘a full-blown temper tantrum, with angry screaming—the baby ei-
ther being rigid and stiff or throwing himself about, kicking the floor, batting his hands up
and down, and the like’. A lower score is assigned if displays of anger are persistent but low-
key or only occasional, for instance if the infant engages in ‘persistent low-intensity pouting
or cranky fussing’ or ‘one strong but isolated behavior, accompanied by a cry’.
Why categories?
Infants in dyads classified as Group B were characterised especially by strong proximity-

seeking and contact-maintenance following reunion, and relatively low levels of avoidant
and resistant behaviour. Infants in dyads classified as Group A were distinguished by low
levels of proximity-seeking and contact-maintenance, and high levels of avoidant behav-
iour. Infants in dyads classified as C1 were distinguished by high levels of infant resistant or
frustrated behaviour. C2 dyads were poorly characterised by the four scales, and no inde-
pendent ‘passivity’ scale was developed by Ainsworth. Over time, they have subsequently
become a rare subclassification—though they are more common in samples drawn from
some countries, such as Israel and Japan.167 Some samples with a significant proportion of
neglecting parents also had several C2 infants. One instance is the Minnesota Longitudinal
Study of Risk and Adaptation (Chapter 4). In Patterns of Attachment, Ainsworth and col-
leagues reported a discriminant function analysis, drawing on Everett Waters’ expertise with
this procedure. They found that a two-function model performed extremely well.168 The
first function essentially comprised scores on the avoidance scale for the first and second
165 Ibid.
166 Ibid.
167 Crittenden, P.M. (2001) Organization, alternative organizations, and disorganization: competing perspec-
tives on the development of endangered children. Contemporary Psychology, 46, 593–6: ‘I am reminded of a per-
sonal conversation that I had with Ainsworth around the time that samples for training on infant attachment
classification were being gathered. Ainsworth lamented a general lack of competence in discerning Type C, fearing
that the pattern was being lost, especially the passive C2 subpattern’ (595).
of the Strange Situation. Bristol: Psychology Press, Chapter 6. The identification of a two-function model was later
replicated by other researchers: Richters, J.E., Waters, E., & Vaughn, B.E. (1988) Empirical classification of infant–
mother relationships from interactive behavior and crying during reunion. Child Development, 59(2), 512–22;
Lamb, M., Thompson, R.A., Gardner, W., & Charnov, E.L. (1985) Infant–Mother Attachment: The Origins and
Developmental Significance of Individual Differences in the Strange Situation. Hillsdale, NJ: Lawrence Erlbaum.
reunions. This distinguished the A from the B and C dyads. Avoidance in the second reunion
made a large additional contribution to variance, over and above the first reunion. There was
also a negative relationship with the proximity-seeking and contact-maintaining, especially
in the second reunion.
The second function was mainly constituted by scores on the resistance scale on first and
second reunion, and with crying through the two reunion episodes. This function distin-
guished the C from the A and B dyads. Contrary to the coding protocols, which give par-
ticular weight to the second reunion, in fact the discriminant function analysis revealed that
both episodes made the same contribution to variance in classification. Though rare, the best
predictor of a C classification was displays of distress before any separation in the Strange
Situation, indicating little ability to use the caregiver as a safe base to deal with the novel set-
ting. However, overall, the second function was not as effective as the first. The two-factor

model could almost always predict whether a case would be A or B, but misclassified 30%
of C dyads. Whereas A/not-A was defined cleanly by the avoidance scale, the C/not-C dis-
tinction appeared to include distress and anger, and appeared to also include other elements.
For example, Ainsworth developed no scale for passivity, and was not surprised when the C2
infants were poorly characterised by the discriminant function analysis.
Despite limitations in characterizing the C dyads, Ainsworth and colleagues regarded
the two-function model as having performed very well. This relative translatability between
categories and scales might imply that either can be used by researchers in analysing their
data. In Patterns of Attachment, Ainsworth and colleagues stated clearly and explicitly that in
adopting a categorical approach to their data they did not assume a ‘rigid typological concept
of the way in which human behaviour is organised, with implications of discontinuity in the
various quantitative dimensions’.169 They held that ‘it is inconceivable that any system based
on a relatively small sample could comfortably accommodate all patterns’.170 Nonetheless,
they argued in favour of group categorisations in running analyses and reporting data, whilst
keeping an eye out for anomalies that suggest the need for revision.
As mentioned above, there was a marked bimodal distribution in Ainsworth’s ini-
tial data, with few children occupying middle positions on scales of proximity-seeking,
contact-maintenance, avoidance, and resistance. This would have made the use of cat-
egories rather than dimensions especially intuitive, since both ends of the spectrum were
sharpened. The data reported in Patterns of Attachment may also have retained a bimodal
distribution. For instance, it would seem unimaginable today to recruit a sample which,
like Ainsworth’s, had 42% B3s. By way of comparison, in the milti-site NICHD sample col-
lected in the 1990s as part of a study of the effects of daycare, there were 224 B3 dyads out
of the 1281 seen in the Strange Situation (17.5%). And even this is a substantially higher
proportion than most samples that have reported subtype classifications.171 Against claims
by colleagues and contemporaries that attachment phenomena were likely best measured

170 Ibid. p.229.
171 Umemura, T. & Jacobvitz, D.B. (2014) Nonmaternal care hours and temperament predict infants’ proximity-
seeking behavior and attachment subgroups. Infant Behavior and Development, 37(3), 352–65; on the lower rates
of B3 in other samples see for instance 12% B3 reported in Van IJzendoorn, M.H., Goossens, F.A., Tavecchio,
L.W.C., Vergeer, M.M., & Hubbard, F.O.A. (1983) Attachment to soft objects: its relationship with attachment to
the mother and with thumbsucking. Child Psychiatry & Human Development, 14(2), 97–105.
142 Mary Ainsworth
dimensionally using scales,172 in print Ainsworth defended her advocacy of categories

with four arguments:
1. A first argument was that categorical measures are appropriate when equivalence is
assumed between behaviours with a common goal.173 This is an effective argument
against approaches to the Strange Situation that merely counted the frequency of par-
ticular behaviours, a popular approach in the early 1970s.174 Ainsworth appeared not
to have noticed that this argument is not an effective one against the use of her own
scales for coding the Strange Situation. The scales already captured the fact that there
were a diversity of ways that infants could seek proximity, retain contact with their
caregiver, avoid expression of the attachment behavioural system, or display aggres-
sion. The fact that Ainsworth’s protocols mandated that coders should first score the

scales, and then use them in informing a categorical judgement, meant that over time
the field accumulated a vast amount of largely unpublished data on scale scores both
for Ainsworth’s Strange Situation coding system and for all coding measures based
on it. As a by-product, an archive of data on scale scores was produced that would, in
the 2000s, be a fundamental resource in the revolt of several younger attachment re-
searchers against the Ainsworth categories (Chapter 3).
2. Ainsworth proposed that categories are useful in helping to identify the relevant
dimensions.175 Indeed, Ainsworth’s identification of proximity- seeking, contact-
maintenance, avoidance, and resistance came out of her initial distinction between
Group A and Group B, and then her subsequent attempt to find order within Group
C. However, if this was the only function of categories, then it would seem that they
would be superseded by the scales. No further dimensional scales have been developed
for the Ainsworth Strange Situation (with the exception of the D scale; Chapter 3), so
this argument would appear to no longer hold.
3. A third advantage of a category-based system, according to Ainsworth, was that cat-
egories sharpen attention to potential causal factors.176 This was, she believed, in
contrast to scales, which flatten different causes of behaviour. So, for instance, both
B1 and Group A infants do not show separation anxiety on separation or proximity-
seeking on reunion. However, Ainsworth believed that the reason for this is different.
B1 infants are able to terminate their attachment behavioural system through distance
interaction, whereas Group A infants inhibit signals of their wish to approach their
caregiver. However, later attachment researchers would identify that many decisions
on the A/B boundary are arbitrary, reducing inter-rater reliability, as the distinc-
tion seems dimensional in certain regards. It also remains an open question whether
a category-based system has indeed contributed to a better identification of causes.
Some second-generation attachment researchers have remained firm defenders of
172 For a review of early arguments in favour of dimensionality see Lamb, M., Thompson, R.A., Gardner, W.,
& Charnov, E.L. (1985) Infant–Mother Attachment: The Origins and Developmental Significance of Individual
Differences in the Strange Situation. Hillsdale, NJ: Lawrence Erlbaum, p.281.
of the Strange Situation. Bristol: Psychology Press, p.xliv.

174 E.g. Feldman, S. & Ingham, M. (1975) Attachment behavior: a validation study in two age groups. Child
Development, 46, 319–30.


176 Ibid.
Ainsworth’s distinction between avoidant and ambivalent/resistant patterns. But

Fonagy has argued that Ainsworth’s advocacy of categories precisely directed atten-
tion away from the causal mechanisms underlying the behavioural clusters, and away
from important psychometric questions about the phenomena.177 And recently it was
remarkable to see Alan Sroufe, one of the primary defenders of a category-based ap-
proach to the Strange Situation, writing to acknowledge that ‘there also are very few
data regarding experiences that lead to resistant versus avoidant attachment. There is a
modicum of data suggesting that avoidance results from rejection precisely when the
infant signals a tender need (e.g., Ainsworth et al., 1978; Isabella, 1993),178 but the ori-
gins of these two patterns—if indeed they are coherent and distinctive—is not really
established.’179
4. A fourth advantage proposed by Ainsworth was that categories capture salient infor-

mation in a ‘picture’, some of which ends up lost in quantitative scales.180 A category-
based coding system appeared to offer a kind of restricting lighting to focus, highlight,
and burnish the scene of observation, keeping contrasts in view even if sometimes they
were oversharpened. For instance, a B2 infant shows avoidance and then proximity-
seeking. The average of the two proximity-seeking scores may be little higher than a
Group A infant, who engages in some proximity-seeking on first reunion, but inhibits
attachment behaviour more firmly on second reunion in response to stronger activa-
tion of the attachment behavioural system. The Strange Situation prompts activation
and deactivation of the attachment behavioural system carefully across episodes, and a
purely quantitative approach, at least an unweighted one, would miss this process and
its implications.181 How much difference this would make to prediction was unclear,
and the question soon fell away as the Ainsworth coding protocols became taken for
granted within developmental science.182
177 Fonagy, P. (1999) Points of contact and divergence between psychoanalytic and attachment theories: is psy-
choanalytic theory truly different? Psychoanalytic Inquiry, 19(4), 448–80, p.469.

of the Strange Situation. Bristol: Psychology Press; Isabella, R.A. (1993) Origins of attachment: maternal interactive
behavior across the first year. Child Development, 64(2), 605–621.
179 Sroufe, L.A. (2016) The place of attachment in development. In J. Cassidy & P.R. Shaver (eds) Handbook of
Attachment: Theory, Research, and Clinical Applications, 3rd edn (pp.997–1011). New York: Guilford, p.1008.
of the Strange Situation. Bristol: Psychology Press, p.57. See also Ainsworth, M. (1981) Letter to Michael Lamb, 8
November 1981. Mary Ainsworth Archive, Box M3173, Folder 4: ‘I do not think that all of the relevant basis for
classification judgment has as yet been captured by the variables we have so far identified and scored.’
181 In fact, curiously, the only study to investigate this matter empirically and in detail found that the second
reunion conveyed only 10% more information than the first reunion. The advantage of the two reunions is that
it strengthens the signal received by researchers about the functioning of the attachment behavioural system.
Kroonenberg, P.M., Dam, M.V., IJzendoorn, M.H., & Mooijaart, A. (1997) Dynamics of behaviour in the strange
situation: a structural equation approach. British Journal of Psychology, 88(2), 311–32: ‘The second sequence does
not add qualitatively new information to what is observed in the first sequence but merely intensifies the behav-
ioural pattern. The replicated nature of the Strange Situation procedure may be one of the reasons for its robustness
and its validity despite its relatively short duration’ (327–8).
182 An illustration of this transition can be seen in discussions of the B4 category. In the early 1980s, van
IJzendoorn was highly concerned about B4, and urged the need for larger samples to investigate Ainsworth’s sub-
types, e.g. van IJzendoorn, M.H., Goossens, F.A., Kroonenberg, P.M., & Tavecchio, L.W.C. (1985) Dependent at-
tachment: B-4 children in the strange situation. Psychological Reports, 57(2), 439–51. However, by the 2000s when
he had data available from many more and much larger samples, van IJzendoorn appears not to have even run
the analyses he himself called for two decades earlier. This shift in van IJzendoorn’s position offers an especially
clear illustration, perhaps even a microcosm, of the broader direction of travel of developmental science in the
period. See Roisman, G.I. & van IJzendoorn, M.H. (2018) Meta-analysis and individual participant data synthesis
in child development: introduction to the special section. Child Development, 89(6), 1939–42. A commitment
144 Mary Ainsworth
In the 1980s, as researchers in the developmental tradition of attachment research were

inheriting Ainsworth’s measure, they frequently commented that the category-based system
captured additional information about the operation of the attachment behavioural system
that was not available through the scales alone. Kroonenberg and van IJzendoorn, discussing
this argument, expressed concern, however, that no one seemed to know exactly what infor-
mation exactly was being added, making it a matter of faith.183 Equally, they worried that no
one seemed to know exactly what additional information might be being captured by the
scales, compared to categories, other than the fact that continuous measures tend to deal
better with individual variation. Both approaches might have pragmatic advantages, but
without explicit discussion and testing, it would not always be clear why.184
Furthermore, even if categories offer a ‘picture’, capturing more information than scales,
additional information is not always a blessing. In psychology, when overarching and

encompassing categories are reified, it is then difficult to separate out the relevant elements
from all the other information. The clarity with which relationships with other variables can
be picked out is therefore weakened. An important case would be the Group C classification,
which is coded on the basis of information about anger, inconsolability, and/or passivity.
Yet only inconsolability—understood as anxiety about the availability of the attachment
figure—was directly taken up by the social psychology tradition of attachment research. This
contribution to differences between the measures of adult attachment has been an under-
recognised source of confusion and miscommunication between the social psychology and
developmental traditions (Chapter 5).
Considered as a whole, it must be acknowledged that Ainsworth’s own writings were am-
bivalent as to whether categories were pragmatic tools to be taken up or put down as needed,
or whether they should be regarded as reflecting truths cut into the nature of individual dif-
ferences in attachment and a requisite of orthodox attachment research.185 On the one hand,
in Patterns of Attachment, Ainsworth and colleagues stated that advocacy of a category-
based system was not meant to imply a rigid typological concept of the way in which human
behaviour is organised, with implications of discontinuity in the various quantitative di-
mensions. Yet at other times Ainsworth seemed to imply that the attachment patterns repre-
sent distinct kinds of relationship. When questioned by Michael Lamb in correspondence,
Ainsworth described herself as ‘stubborn’ in her conviction that scales will never ‘be able to
capture everything that should be taken into account when assigning an individual infant to
a classification’.186 Many of Ainsworth’s students, including Main (Chapter 3), came to the
to subclassifications was retained in the work of Ainsworth’s student Patricia Crittenden, prompting recent de-
bate with van IJzendoorn and colleagues who argued that the fine-grained information captured by subclassifi-
cations put at risk scientific credibility, which must be based on aggregation. See Crittenden, P.M. & Spieker, S.J.
(2018) DMM vs. ABC+D assessments of attachment in child protection and treatment: reply to van IJzendoorn,
Bakermans, Steele, & Granqvist. Infant Mental Health Journal, 39(6), 647–51.
183 Kroonenberg, P.M. & van IJzendoorn, M.H. (1987) Exploring children’s behavior in the Strange Situation.
In L.W.C. Tavecchio & M.H. van IJzendoorn (eds) Attachment in Social Networks. Contributions to the Bowlby–
Ainsworth Attachment Theory (pp.379–426). New York: Elsevier, pp.380, 409.
184 Van IJzendoorn, M.H. & Bakermans-Kranenburg, M.J. (2014) Confined quest for continuity: the categor-
ical versus continuous nature of attachment. Monographs of the Society for Research in Child Development, 79(3),
157–67.
185 Fraley, R.C. & Waller, N.G. (1998) Adult attachment patterns: a test of the typological model. In J.A. Simpson
& W.S. Rholes (eds) Attachment Theory and Close Relationships. New York: Guilford: ‘Ainsworth et al. (1978) are
somewhat ambiguous regarding the ontological status of the classificatory groups’ (108).
186 Ainsworth, M. (1981) Letter to Michael Lamb, 8 November 1981. Mary Ainsworth Archive, Box M3173,
Folder 4.
conclusion that she had discovered ‘natural kinds’, representing qualitatively different forms
of relationships and patterns of child socioemotional development.187 They acknowledged
that scientific constructs are always approximations and simplifications of reality. However,
discourses that situated attachment as by nature divided into categories influenced and infil-
trated activities such as research design and coding.188
The Ainsworth categories were initially important in the 1970s and 1980s in countering
social learning theorists, who argued that secure attachment behaviour was caused simply by
the mother having reinforced approach when her infant cried. Yet Ainsworth could counter
by showing that neither distressed approach nor the absence of distressed approach de-
fined Group B, but rather the use of the caregiver as a secure base and safe haven. Yet, subse-
quently, the category-based system helped contribute to both the popularity and reifications
of attachment theory. A tale about ‘three kinds of infants’ is one that can carry a tune, and it

fitted well with the ascendence of diagnosis-based thinking about psychological processes in
the wake of DSM-III (Chapter 1). Even if it was not Ainsworth’s intention, then, a significant
part of what has rippled out from the coding system for the Strange Situation was an impres-
sion of ultimate certainty.189 The Ainsworth categories were taken as part of, or at least close
to, the inner core of the attachment paradigm as a cumulative research endeavour, perhaps
with some role as a symbol of the field of attachment research as a differentiated entity. And
when the categories were questioned, attachment researchers tended to circle the wagons.
An illustrative case was Chris Fraley and Sue Spieker’s 2003 paper ‘Are infant attachment
patterns continuously or categorically distributed?’190 From the late 1990s, there had been
growing concern across psychological science to replace categories with dimensions in the
interests of psychometric precision and statistical power.191 Influenced by these discussions,
Fraley and Spieker held that individual differences in infant attachment were likely influ-
enced by a variety of factors. It would therefore be expectable for these differences to occur
by degrees, depending on how much one factor or another was in play. This implied quan-
titative, not simply qualitative, variability. Fraley, especially, was concerned that a category-
based system would not only neglect this variability, but also misdirect discussions of the
meaning of attachment: ‘Even professional scholars have often misunderstood the theory
as implying a strong continuity between early experiences and adult romantic relationships
187 For other interpretations of Ainsworth that treated her categories as reflecting or like natural kinds, see e.g.
Bretherton, I. (1990) Communication patterns, internal working models and the intergenerational transmis-
sion of attachment relationships. Infant Mental Health Journal, 11, 237–51; Crittenden, P.M. (2000) A dynamic–
maturational model of the function, development, and organization of human relationships. In R.S.L. Mills & S.
Duck (eds) Developmental Psychology of Personal Relationships (pp.199–218). New York: Wiley.
188 See Beauchaine, T.P. & Waters, E. (2003) Pseudotaxonicity in MAMBAC and MAXCOV analyses of
rating-scale data: turning continua into classes by manipulating observer’s expectations. Psychological Methods,
8(1), 3–15.
189 In the hands of some critics the Strange Situation was treated as some kind of marine mammal, with jaws
wide as it moved through the water, ingesting individual differences like krill and exhaling pre-established cat-
egories of infants. See e.g. Knudson-Martin, C. (2012) Attachment in adult relationships. Journal of Family Theory
& Review, 4(4), 299–305; Gaskins, S. (2013) The puzzle of attachment. In N. Quinn & J.M. Mageo (eds) Attachment
Reconsidered (pp.33–66). London: Palgrave.
190 Fraley, R.C. & Spieker, S.J. (2003) Are infant attachment patterns continuously or categorically distributed?
A taxometric analysis of strange situation behavior. Developmental Psychology, 39(3), 387–404.

191 E.g. Waller, N.G. & Meehl, P.E. (1998) Multivariate Taxometric Procedures: Distinguishing Types from
Continua. Newbury Park, CA: Sage. For a review of this development see Kendler, K.S., Zachar, P., & Craver, C.
(2011) What kinds of things are psychiatric disorders? Psychological Medicine, 41(6), 1143–50. This perspective
has influenced the design of subsequent attachment measures, for instance Steele, H., Steele, M., & Kriss, A. (2009)
The Friends and Family Interview (FFI) Coding Guidelines. Unpublished manuscript.
146 Mary Ainsworth
styles. We think that the typological approach . . . may help promote the widespread belief
that there is a single etiology.’192
To support their claims, Fraley and Spieker used taxometric analysis to show that
Ainsworth’s three patterns of attachment were better modelled as two dimensions: a di-
mension between avoidance and security, crossed by a dimension between resistance and
security. Part of the attractiveness of this proposal was that dimensional scales might well
contribute to greater statistical power; differences between dyads in the middle-range would
be captured, rather than forcing cases artificially into categories. Another part of the attract-
iveness of the proposal was that these scales had been coded by researchers as part of making
judgements about the categories. So the field could use the already-existing data on the
scales from decades of work. Proximity-seeking and the absence of resistance and avoidance
could offer an approximation of a dimensional characterisation of security, even if this was

imperfect. A disadvantage was that scales are not coded independently, and likely influence
the scoring of one another. At least, however, there would be no need to start from scratch.193
The Fraley and Spieker paper was initially rejected flat out by Child Development as too
heretical. Eventually published in Developmental Psychology, the paper was met by hostile
peer-reviews, and accompanied in print by discouraging replies from other attachment re-
searchers.194 This may have been prompted by the fact that the initially submitted draft of
the Fraley and Spieker paper did, by the authors’ own admission, at points adopt quite a
strident evaluative tone.195 Yet, as Fraley set out in his letter to the editor responding to the
peer-review feedback, he regarded his position as a defence rather than attack on attachment
theory. He was primarily concerned that ‘if variation in attachment organisation is continu-
ously distributed—and if it plays a strong role in shaping developmental outcomes—then
attachment organisation will appear to have weak effects on other outcomes when studied
categorically . . . It would be an unsatisfactory state of affairs if the larger field of develop-
mental psychology eventually became disenchanted with the study of attachment for rea-
sons, such as measurement imprecision, that have little to do with the validity of the theory
per se.’196
& W.S. Rholes (eds) Attachment Theory and Close Relationships. New York: Guilford, p.101.
193 In fact, until the 2000s, many laboratories destroyed their data regarding the Ainsworth scales, since the focus
for publications was on categories alone. This was noted by Lamb, M., Thompson, R.A., Gardner, W., & Charnov,
E.L. (1985) Infant–Mother Attachment: The Origins and Developmental Significance of Individual Differences in the
Strange Situation. Hillsdale, NJ: Lawrence Erlbaum, p.138.
194 Cummings, E.M. (2003) Toward assessing attachment on an emotional security continuum: comment on
Fraley and Spieker (2003). Developmental Psychology, 39, 405–408; Cassidy, J. (2003) Continuity and change in
the measurement of infant attachment: comment on Fraley and Spieker (2003). Developmental Psychology, 39(3),
409–12.
195 Fraley, C. (2002) Response to Reviewers, Letter to Douglas M. Teti, Associate Editor of Developmental
Psychology, 30 January 2002. Unpublished text shared by the author: ‘The reviewer seems to be supposing hos-
tile and destructive ambitions for our paper that are simply not true’ . . . “Unfortunately, in the previous draft we
did phrase things in a way that might lead a reader to conclude that attachment research based on the categorical
model is faulty. As reviewer D noted, for example, we stated that “the typological model is invalid” on page 30 of the
previous draft. In that context, we were arguing that the categorical assumption (i.e., that attachment “categories”
exist) is not supported by the data. That does not imply, however, that the kinds of factors that are captured by the
categorical system are invalid. As reviewer D notes, it is probably the case that the categories have been successful
because they capture the relevant dimensions underlying the patterning of attachment behavior. We have revised
the manuscript in order to make it clear that we are not challenging or calling into question the significance of at-
tachment theory and research. It is precisely because we believe that this area of inquiry is important and valuable
that we have posed the questions that we have in this paper.’
196 Ibid.
Interpretation of the Strange Situation 147
There is often discussion today about opposition between a ‘category’ and a ‘dimension’
camp to Strange Situation data. However, in fact, this is something of an artefact. Both sides
seem to agree that there are underlying dimensions. The question is whether a dimensional
approach to the data would offer better prediction. Author of one of the critical replies to
Fraley and Spieker, Alan Sroufe stated in 2000 that ‘traditionally, the measurements of se-
curity of attachment have been categorical, although it would seem conceptually that there
are underlying dimensions’.197. In fact, from the late 1980s Sroufe’s close colleague Byron
Egeland even used group exercises to encourage parents in their STEEP intervention
(Chapter 4) to think about aspects of care and parenting as dimensions, as a dimensional
perspective was regarded as likely to improve attachment security in the caregiver–infant
dyads by contributing to a more ‘informed, realistic understanding’.198
It is curious that so many years after Fraley and Spieker’s paper, the field is yet to see a pub-

lished comparison of the relative merits of dimensions versus categories in predicting later
correlates of interest, such as externalizing behaviours. A first comparison of their relative
merits for understanding antecedents of attachment was published only recently, and re-
ported better prediction for the analysis in terms of two latent dimensions.199 The reasons
for the delay in pursuing comparative analysis are not clear, given that such inquiry could
have been pursued already on the data available to Fraley and Spieker. One reason may have
been that large datasets are needed to adequately address the question.200 Nonetheless, it
can be anticipated that the issue of individual differences in attachment as dimensionally
or categorically distributed is likely to be one of the major objects of attention for the third
generation of attachment researchers over the coming years, given their greater concern for
the psychometric standing of attachment methods and coding systems, and the recent avail-
ability of pooled datasets (Chapter 6).
Interpretation of the Strange Situation
The identification of distinct attachment classifications raised the question of their defining
characteristics and antecedents. In interpreting the behaviour shown by infants in the
197 Rutter, M. & Sroufe, L.A. (2000) Developmental psychopathology: concepts and challenges. Development
& Psychopathology, 12(03), 265–96. A few years earlier, in work with Gail Fury and Elizabeth Carlson, Sroufe had
criticised Main and Kaplan’s category-based system for coding family drawings at age 6 (Chapter 3) and replaced
it with a series of dimensional scales. Fury, G., Carlson, E.A., & Sroufe, A. (1997) Children’s representations of
attachment relationships in family drawings. Child Development, 68(6), 1154–64. Another illustration is offered
by Hesse and van IJzendoorn, generally regarded as key defenders of the category-based approach, who—with
category-based data available—nonetheless reported their findings in terms of a 9-point scale ‘LapseTr’, com-
prising the highest score obtained on either the loss or abuse scale of the Adult Attachment Interview. Hesse,
E. & van IJzendoorn, M. (1999) Propensities towards absorption are related to lapses in the monitoring of rea-
soning or discourse during the Adult Attachment Interview: a preliminary investigation. Attachment & Human
198 Egeland, B., Weinfield, N., Bosquet, M., & Cheng, V. (2000) Remembering, repeating, and working
through: lessons from attachment-based interventions. In J. Osofsky (ed.) WHIMH Handbook of Infant Mental
Health, Vol. 4 (pp.35–89). New York: Wiley, p.64–5.
199 Groh, A.M., Propper, C., Mills-Koonce, R., Moore, G.A., Calkins, S., & Cox, M. (2019) Mothers’ physio-
logical and affective responding to infant distress: unique antecedents of avoidant and resistant attachments. Child
Development, 90(2), 489–505.
200 Another contributing factor may have been that Main and Solomon decided to code disorganization using a
single encompassing scale rather than developing subscales for forms of disorganization, as Main and Hesse later
did for frightening/frightened behaviour (Chapter 3). This has made it more difficult to appraise the taxonicity of
disorganization.
148 Mary Ainsworth
Strange Situation, Ainsworth and colleagues drew on observations of the dyads at home and
particularly caregiving behaviour by the mother. The strong relationship between the two
sources of information gave Ainsworth’s team confidence that the Strange Situation tapped
patterns of attachment, since results were congruent with the history of infant–mother inter-
action.201 In publications, Ainsworth felt obliged by the genre of academic writing in devel-
opmental psychology in the 1970s and 1980s to present her research as setting out to test
hypotheses and, on this basis, discovering correlations. This gave many readers the impres-
sion that Ainsworth had more confidence in her findings than she did, and that she was
emphasizing the importance of infant–mother interaction over all other factors. In fact,
Ainsworth regarded her work as exploratory, seeking to identify previously unnoticed as-
sociations between infant attachment behaviour and the infant’s history of receiving care, by
wading around, up to the knees in her hundreds of hours of observational data.202 In a letter

from 1969 she wrote to Bowlby:
To discover the interrelationships implicit in these data is my chief talent and all-absorbing
aim. I realised that when my own intuitive feel for the data is blocked by our elaborate
reliability machinery and do not hesitate to go beyond it. Thus with our attachment–ex-
ploration balance classification. I was more concerned to find the “right” basis of classifica-
tion than to stop with the semi-satisfactory basis that was subject to our reliability-checks.
Research is always a compromise—and presumably the most important things are to know
what one’s own compromise has been and not to attempt to convince either oneself or
others that one has done the impossible.203
In a letter from 1970, she added: ‘Our horrible hypothesis-testing traditional leads us (me
and especially my co-authors) to lead the reader to the conclusion that we are claiming suc-
cessful hypothesis testing, whereas in fact we are presenting a new viewpoint together with
one small but “telling” body of evidence that seems congruent with it and hence to offer some
support.’204 Main later recalled that Ainsworth repeatedly applied over subsequent years for
funding to conduct a replication of her Baltimore study without success:
Worrying about the possibility of contaminations among variables which were only iden-
tified during the course of the study—she had intended it as a pilot investigation. In her
second, planned replication study, she would make no changes, develop no new infant or
maternal variables, re-conduct the strange situation procedure with no revisions in her
sub-groupings, and hence properly and completely re-test her initial results. However,
her applications to granting agencies to conduct this new Baltimore study were repeatedly
turned down.205
201 Ainsworth. M.D.S. (1983) Patterns of infant–mother attachment as related to maternal care: their early
history and their contribution to continuity. In D. Magnusson & V.L. Allen (eds) Human Development: An
Interactional Perspective (pp.35–55). New York: Academic Press, p.52.
202 The ethos of Ainsworth’s approach is revealed with remarkable clarity in her letters to Silvia Bell, especially
those sent during Ainsworth’s sabbatical at Stanford: Ainsworth, M. (1968) Letter to Silvia Bell, 2 January 1968.
Mary Ainsworth Archive, Box M3169, Folder 6: ‘Part of the luck has been capitalisang on differences within the
sample. I work back and forth from “cause” to “effect” in the same sample. Under these circumstances it is relatively
easy to get everything fitting into place.’
203 Ainsworth, M. (1970) Letter to John Bowlby, 7 August 1969. PP/Bow/K.4/12.
204 Ainsworth, M. (1970) Letter to John Bowlby, 1 September 1970. PP/Bow/K.4/12.
205 Main M. (1999) Mary D. Salter Ainsworth: tribute and portrait. Psychoanalytic Inquiry, 19, 682–776, p.722.
It is important to highlight that in pursuing an interpretation of the Strange Situation in light

of infant–mother interaction, Ainsworth was not ruling out the role of fathers on theoret-
ical grounds, though neither did she encourage it. In part this was a distal reflection of the
cultural values of the time, which emphasised the importance of maternal care. However,
proximally, inattention to fathers was a by-product of Ainsworth’s research design: she did
not have data on infant–father interaction, since she had conducted her study during office
hours in a sample where the fathers all worked away from home. She was also not ruling out
that infant temperament could play a role in their behaviour during the Strange Situation
procedure. ‘Everybody knows that . . . there is something innate that each child brings’,
Ainsworth observed.206 However, she anticipated that these qualities would interact with
the caregiving the child received in important ways, and that it would be caregiving that
would ultimately make the more important contribution to the Strange Situation classifi-

cation. Ainsworth was sympathetic, for example, to the idea that some children might be
predisposed to resistant behaviour in the Strange Situation as a result of a difficult or fussy
temperament.207 Some colleagues also held that it is probable that ‘ambivalent babies differ
from others from the beginning of life’.208 However, Ainsworth and her group felt that the
history of caregiving in the dyad was also an important contributing factor to ambivalent/
resistant attachment, and that it was of overwhelming importance for secure and avoidant
attachment.
Later empirical research supported Ainsworth’s general position. Attachment research
was an early adopter of a meta-analyses, a technique for the quantification of the combined
effect of a set of study outcomes. An especially important contribution to attachment re-
search were the conclusions regarding the role of infant temperament, in the context of ra-
ging debates about nature vs nurture in child development. Combining the effect sizes of
studies to date, in a paper from 1997 van IJzendoorn and De Wolff reported a correlation of
r = .17 between infant–mother and infant–father attachment across 14 independent sam-
ples, indicating less than 3% overlap in variance.209 This supported Ainsworth’s assumption
that only a small proportion of Strange Situation behaviour could be explained by child tem-
perament, unmodified by the particular attachment relationships infants had experienced.
Indeed, a later meta-analysis of 69 independent samples by Groh, van IJzendoorn, and col-
leagues found that infant temperament did have an association with resistant attachment
(r = .15). However, as expected, it had a very weak association with security (r = .04), and no
association was found for the avoidant attachment classification.210 The researchers flagged,
however, that assessments of infant temperament are often conducted with the parent pre-
sent. One of the defining characteristics of the resistant attachment classification is that
the attachment behavioural system has a low threshold for activation, and is accompanied

208 Main, M. (1980) Avoidance of attachment figures: index of disturbance. PP/Bow/J.4/1, discussing Waters,
E., Vaughn, B.E., & Egeland, B.R. (1980) Individual differences in infant–mother attachment relationships at age
one: antecedents in neonatal behavior in an urban, economically disadvantaged sample. Child Development, 51(1),
208–16.
209 Van IJzendoorn, M.H. & Wolff, M.S. (1997) In search of the absent father—meta-analyses of infant–father
attachment: a rejoinder to our discussants. Child Development, 68(4), 604–609.

210 Groh, A.M., Narayan, A.J., Bakermans-Kranenburg, M.J., et al. (2017) Attachment and temperament in the
early life course: a meta-analytic review. Child Development, 88(3), 770–95.

150 Mary Ainsworth
by fussing and distress already when the stranger seeks to engage the child in the Strange
Situation before the separations. As such, the temperament assessment itself may serve to
elicit the resistant attachment pattern, confounding the assessment of temperament with at-
tachment. In line with this supposition, resistance was found to be more strongly associated
in the meta-analysis with greater levels of fearful distress, rather than lower levels of the ex-
pression of positive emotions.
Secure attachment
Rather than infant temperament, Ainsworth argued that the antecedents of individual dif-
ferences in attachment lay primarily in experiences of caregiving. This argument was based

on close comparison of elements of infant behaviour in the Strange Situation with the infant
and the mother’s behaviour at home. This comparison revealed that Group B infants cried
less than non-B infants, especially in response to the ordinary brief separations of everyday
life. They responded more positively to being picked up, and less negatively to being put
down. And they were more cooperative to their mother’s requests.211 Ainsworth and col-
leagues also found an astonishingly strong association between the Group B classification
and caregiver sensitivity: r(21) = .78. The association was even stronger for the prototypical
B3 subtype. This was in contrast, in the data Ainsworth had available, to the non-significant
association between child behaviours in the first few months of life and the Group B classifi-
cation at 11 months.212 The strong correlation with caregiver sensitivity implied that Group
B represented infants who anticipated that their signals would be heeded by their caregiver.
Ainsworth therefore reinstated the label ‘secure’ as the name for Group B, which she had ini-
tially left aside on Bowlby’s urging as too value-laden.
In retrospect, this was an unfortunate decision in some ways, as the term ‘security’ has its
own connotations that differ from Ainsworth’s intended meaning of the term. Or at least,
Ainsworth failed to clarify that she intended the term in a technical sense, one that departed
from ordinary language. Admittedly there is no ready alternative single word in English that
conveys a sense of confidence in the other’s availability and responsiveness. Nonetheless,
other choices could have been made by Ainsworth, such as to retain the label ‘Group B’, or to
discuss ‘care-confident’ or ‘availability-trusting’ infants, though both are ungainly construc-
tions. Yet the term ‘secure’ was already value-laden in Ainsworth’s time, and furthermore has
subsequently been infiltrated by a whole range of connotations.213 Not least, the rapid rise of

212 Ibid. The exact tests conducted by Ainsworth are not described, so what child behaviours she examined
remains unclear. And other researchers, in fact, found associations between early child orienting behaviours
and later attachment, even with caregiving included in the model. The earliest such finding was Grossmann, K.,
Grossmann, K.E., Spangler, G., Suess, G., & Unzner, L. (1985) Maternal sensitivity and newborns’ orientation re-
sponses as related to quality of attachment in northern Germany. Monographs of the Society for Research in Child
Development, 50(1–2), 233–56.For a more recent discussion see Spangler, G. (2013) Individual dispositions as
precursors of differences in attachment quality: why maternal sensitivity is nevertheless important. Attachment &
Human Development, 15(5–6), 657–72.
213 Harwood, R.L., Miller, J.G., & Irizarry, N.L. (1995) Culture and Attachment: Perceptions of the Child in
Context. New York: Guilford: ‘The concept of “security”, although technically similar to a sense of psychological
safety, has become laden with an array of values and ideals peculiar to mainstream U.S. discourse: the “secure”
person is self-confident, independent, and able to utilise his or her talents and abilities to the fullest, but also
has the capacity to be empathetic and to relate to others. In short, the “secure” individual is one who embodied
U.S. ideals of optimal socioemotional development—ideals that may or may not translate well into the meaning
morally laden discourses about security in contemporary ‘risk society’ has helped failures
of security connote danger and destruction.214 A semantic mapping exercise conducted by
Waters with psychology students found that they used the connotations of the word to make
spurious assumptions. For instance, it was assumed that ‘security’ for Ainsworth meant con-
fident, and therefore someone socially dominant.215 Sociologists have observed that the con-
notations of Ainsworth’s terms have helped support both the popularisation and popular
misconceptions of attachment ideas, including moralizing narratives in which insecure
babies have been broken by their caregivers.216 Ainsworth’s students, especially those with
clinical training, have made much the same point. The eminent clinician Alicia Lieberman,
a graduate student of Ainsworth’s, offered a rare criticism of her teacher for failing to ad-
equately clarify that the meaning of ‘security’ differed from ordinary language. Lieberman
alleged that attachment researchers since Ainsworth have slid about unsteadily between

various connotations of the term ‘secure attachment’.217 Other attachment researchers also
trained as clinicians, for instance Pasco Fearon, have made the same point.218
Though warned about the unhelpful connotations of the term ‘secure’ by Bowlby,
Ainsworth felt that this Blatzian concept captured the infant confidence in the caregiver’s
availability, a confidence that seemed to be reflected in behaviour and that was at least some-
what stable over time. Use of a term from everyday language also perhaps appealed to help
signal that there would be multiple contributories to a sense of confidence in the availability
of others, not just early care. Furthermore, use of a term with strong and evocative mean-
ings in ordinary language, even if they were rather misleading, may have been attractive for
supporting interest in the nascent area of attachment research, though there is no evidence
to suggest that this was intentional on Ainsworth’s part. Ultimately, Ainsworth felt that she
could rest on the etymological meaning of ‘security’ as being without concern or worry.219
Not only were Group B infants able to use their caregiver effectively as a secure base and safe
systems of other cultural groups’ (143–4); Weisner, T.S. (2005) Attachment as a cultural and ecological problem
with pluralistic solutions. Human Development, 48(1–2), 89–94: ‘Using the word “secure” assumes that there is, in
cultures everywhere, a positive valence for development associated with that behavior profile’ (91). Bowlby did not
help matters in his characterization of the diverse perceived virtues of the Apollo 13 astronauts in terms of their
attachment security. Bowlby, J. (1970, 1979) Self-reliance and some conditions that promote it. In The Making and
Breaking of Affectional Bonds (pp.124–49). London: Routledge, p.129. See Laubender, C. (2019) From the bomb to
Apollo 13: Bowlby and the Cold War. The Psychologist, 32, 76–9.
214 E.g. Huysmans, J. (1998) Security! What do you mean? From concept to thick signifier. European Journal of
International Relations, 4(2), 226–55; Harrington, C. & Shearing, C. (2017) Security in the Anthropocene: Reflections
on Safety and Care. New York: Columbia University Press.
215 Everett Waters, personal communication, July 2019.
216 Thornton, D.J. (2011) Neuroscience, affect, and the entrepreneurialization of motherhood. Communication
and Critical/Cultural Studies, 8(4), 399–424; Wall, G. (2018) ‘Love builds brains’: representations of attachment
and children’s brain development in parenting education material. Sociology of Health & Illness, 40(3), 395–409.
217 Lieberman, A.F. & Van Horn, P. (2008) Psychotherapy with Infants and Young Children. New York: Guilford,
p.11. Liberman would make the same point about Main’s term ‘disorganised attachment’ (Chapter 3).
218 E.g. Fearon, R.M.P., Bakermans- Kranenburg, M.J., & van IJzendoorn, M.H. (2010) Jealousy and at-
tachment: the case of twins. In S.L. Hart & M. Legerstee (eds) Handbook of Jealousy. Theory, Research, and
Multidisciplinary Approaches (pp.362–86). New York: Wiley, p.372.
219 Ainsworth, M. (1969) CPA oral history of psychology in Canada interview. Unpublished. http://www.fem-
inistvoices.com/assets/Women-Past/Ainsworth/Mary-Ainsworth-CPA-Oral-History.pdf. Working on her thesis,

in discussion with Bott: ‘We came to the word “security” which I had defined in my own way, because it was a key
concept, and I had defined it in a sort of Blatzian way. He said, “That is not the meaning of the word ‘security’. Don’t
you know the original meaning of the word security?” And I shouted, “It doesn’t matter what the original meaning
was!” I was really angry at this point. He said, “It does matter.” He implied that a word was a word and it never really
lost that original meaning. And you know, he was right. He told me the word “security” had a Latin derivation and
meant sine cura (without care).’
152 Mary Ainsworth
haven in the Strange Situation, but also this security seemed intelligible in the context of the
sensitive caregiving the infants received at home, which would make them unconcerned or
not worried about the caregiver’s availability. As a consequence, these infants could imple-
ment the ‘short version’ of the expression of the attachment behavioural system, since the
system was not complicated by inhibition, anger, or other forms of conflict or guardedness.
Looking back, researchers such as Sroufe and van IJzendoorn recalled that early reports
of the strength of the association between Strange Situation behaviour and caregiver sen-
sitivity reported in Patterns of Attachment contributed to an intense interest in attachment
theory and to use of the Strange Situation among a generation of younger researchers.220 It
was, though, very rare for extensive naturalistic observations at home to take place; such
an expenditure of resources would have been reckless for a developmental psychologist in
a field increasingly focused from the 1970s onwards on quantification and rapid research.

The naturalistic observations conducted by Ainsworth and colleagues were, perhaps in part
as a result, treated as sufficient, especially since the association between secure attachment
and caregiver sensitivity would replicate time and again through the 1980s. In the 1990s,
Ainsworth’s conclusions were also backed up by findings that naturalistic or intervention-
based changes in caregiver sensitivity had significant effects on the frequency of secure at-
tachment, supporting the idea of a close causal relationship.221
The relationship between caregiver sensitivity and infant attachment appeared to be well
replicated. However, few studies found an association with anything like the strength of
Ainsworth’s original Baltimore study. A meta-analysis in 1997 by Wolff and van IJzendoorn
found a much lower association than Ainsworth, with r(835) = .24 for mothers and
r(544) = .13 for fathers.222 Van IJzendoorn commented that it has been disheartening for
younger researchers to find such modest effect sizes, when they had been led by Ainsworth’s
initial work to expect vast correlations.223 Nonetheless, he emphasised that these are still
very notable findings, reflecting processes that, over time, are threaded through thousands
upon thousands of child–caregiver interactions. Furthermore, many of the studies in-
cluded in the meta-analysis depended on very short periods of observation; the study with
the second longest period of observation after the Baltimore study found associations be-
tween sensitivity and security similar to those of Ainsworth.224 Recent work by Madigan
and colleagues found that the association between caregiver sensitivity and infant Strange
Situation classification was moderated by the measure used. The Maternal Behaviour Q-sort
220 Sroufe, L.A. (1985) Attachment classification from the perspective of infant–caregiver relationships and in-
fant temperament. Child Development, 56(1), 1–14, p.7; Wolff, M.S. & van IJzendoorn, M.H. (1997) Sensitivity
and attachment: a meta-analysis on parental antecedents of infant attachment. Child Development, 68(4), 571–91,
p.585. On the ‘winner’s curse’ of unrepresentatively strong findings in early studies leaving a legacy for later re-
search see Young, N.S., Ioannidis, J.P.A., & Al-Ubaydli, O. (2008) Why current publication practices may distort
science. PLoS Med, 5 (10), e201.
221 Van IJzendoorn, M., Juffer, F., & Duyvesteyn, M. (1995) Breaking the intergenerational cycle of insecure at-
tachments: a review of attachment-based interventions on maternal sensitivity and infant security. Journal of Child
Psychology and Psychiatry, 36, 225–48; Howes, C., Galinsky, E., & Kontos, S. (1998) Child care caregiver sensitivity
and attachment. Social Development, 7(1), 25–36.
222 Wolff, M. S. & van IJzendoorn, M.H. (1997) Sensitivity and attachment: a meta-analysis on parental ante-
cedents of infant attachment. Child Development, 68(4), 571–91; Van IJzendoorn, M.H. & Wolff, M.S. (1997)
In search of the absent father—meta-analyses of infant–father attachment: a rejoinder to our discussants. Child
Development, 68(4), 604–609.
223 Van IJzendoorn, M.H. (2019) Replication crisis lost in translation? Paper presented at International
Attachment Conference, Vancouver, 20 July 2019.

224 Pederson, D.R., Gleason, K.E., Moran, G., & Bento, S. (1998) Maternal attachment representations, maternal
sensitivity, and the infant–mother attachment relationship. Developmental Psychology, 34(5), 925–33.
developed by David Pederson, Greg Moran, and Sandi Bento,225 and the CARE-Index devel-
oped by Patricia Crittenden during graduate study with Ainsworth226 had the strongest as-
sociations, whereas other assessments of sensitivity, including Ainsworth’s original measure,
had weaker associations.227
Researchers have generally not returned to Ainsworth’s methodology of naturalistic ob-
servation to reconsider the sensitivity–attachment link inductively. Instead, the approach
adopted by attachment researchers has generally been dedictive identification of other fac-
tors. In subsequent years, other factors besides sensitivity have been identified deductively
and then found to be important. One is the emotional climate of the home in higher-risk
samples, which appears to exert direct influence on child security unmediated by the be-
haviour of the parent towards the child.228 Researchers also found important moderators of
the sensitivity–attachment link. Child genotype may play a role, with a gene × environment

interaction proposed by Barry, Kochanska, and Philibert, though evidence to date has not
consistently confirmed this proposal.229 Another moderator of caregiver sensitivity for at-
tachment security may be the extent and manner of parental involvement with the child,
which have substantial gender differences in many samples.230 Further study of moderators
of the relationship between sensitivity and attachment has flourished recently, with the avail-
ability of large datasets permitting comparison of the correlates of sensitivity in low-risk and
high-risk samples.231
Researchers such as Elizabeth Meins have proposed that at least some of the association
between caregiver sensitivity and infant attachment can be explained by the caregiver’s
225 Pederson, D.R. & Moran, G. (1995) Appendix B: maternal behavior Q-set. Monographs of the Society for
Research in Child Development, 60(2–3), 247–54.

226 Crittenden, P.M. (1979) CARE Index: coding manual. Unpublished manual, Miami, FL. The CARE Index
scales false positive affect and a construct of ‘compulsivity’. It also differentiates insensitive intrusiveness from in-
sensitive unresponsiveness. Madigan and colleagues reported a much larger range for the CARE Index than for
other measures in terms of associations with sensitivity across studies. This may reflect the wider lens of the CARE
Index, which can be anticipated to contribute to measurement variance between studies with populations charac-
terized by different caregiving profiles.
227 Madigan, S., Verhage, M.K., Schuengel, C., et al. (2019) Parental sensitivity and mentalization as predictors
of attachment quality: a meta-analysis. Paper presented at International Attachment Conference, Vancouver, 20

July 2019. Length of of time of observation was not a moderator. But this was likely clouded by differences between
the measures. Madigan and colleagues also reported that when coders of sensitivity and infant Strange Situations
were not independent, r = 1.08. An important appraisal of the different measures of sensitivity, and an attempt
to synthesize their strengths through differentiated subscales, has recently been published: Heinisch, C., Galeris,
M.G., Gabler, S., et al. (2019) Mothers with postpartum psychiatric disorders: proposal for an adapted method to
assess maternal sensitivity in interaction with the child. Frontiers in Psychiatry, 22 July, 10.
228 Raikes, H.A. & Thompson, R.A. (2005) Links between risk and attachment security: models of influence.
Journal of Applied Developmental Psychology, 26(4), 440–55. Raikes and Thompson contrasted the emotion cli-
mate of the home with the role of socioeconomic factors, which they found influenced the security of attachment
as mediated by caregiver behaviour.
229 Barry, R.A., Kochanska, G., & Philibert, R.A. (2008) G× E interaction in the organization of attach-
ment: mothers’ responsiveness as a moderator of children’s genotypes. Journal of Child Psychology and Psychiatry,
49(12), 1313–20; Luijk, M.P., Roisman, G.I., Haltigan, J.D., et al. (2011) Dopaminergic, serotonergic, and oxyton-
ergic candidate genes associated with infant attachment security and disorganization? In search of main and inter-
action effects. Journal of Child Psychology and Psychiatry, 52(12), 1295–307.
230 Lamb, M.E. (2002) Infant– father attachments and their impact on child development. In C.S. Tamis-
LeMonda & N. Cabrera (eds) Handbook of Father Involvement: Multidisciplinary Perspective (pp.93–117).
Mahwah, NJ: Lawrence Erlbaum; Lucassen, N., Tharner, A., van IJzendoorn, M.H., et al. (2011) The association
between paternal sensitivity and infant–father attachment security: a meta-analysis of three decades of research.
Journal of Family Psychology, 25(6), 986–92.
231 Verhage, M.L., Fearon, R.P., Schuengel, C., et al. (2019) Collaboration on attachment transmission synthesis.
Does risk background affect intergenerational transmission of attachment? Testing a moderated mediation model
with IPD. Paper presented at Biennial Meeting of Society for Research in Child Development, Baltimore, MD.
154 Mary Ainsworth
attention to and interest in the child’s emotional experience.232 Fonagy and colleagues even
argued that this association is an artefact, with both caregiver sensitivity and individual
differences in the Strange Situation reflecting the caregivers’ capacity to imagine, perceive,
and interpret their child’s behaviour in terms of intentional mental states (e.g., needs, de-
sires, feelings, beliefs, goals, and reasons), as well as their own behaviour towards the child
in such terms. They have termed this general capacity ‘mentalisation’, and as ‘reflective
function’ when it is applied within an attachment relationship.233 Recently, Fonagy and
colleagues proposed that individual differences in infant attachment reflect forms of trust
or distrust in information given by caregivers about the environment and their own avail-
ability. Whereas Ainsworth argued that Bowlby missed the value of learning in thinking
about the evolutionary function of attachment relationships, Fonagy and colleagues took
this argument further. They speculated that the most important evolutionary function of

attachment relationships is that young children learn from their caregiver whether trust or
distrust is the safer response to personally relevant information.234 Evaluating the criticisms
of Ainsworth by Fonagy and colleagues, Zeegers and colleagues conducted a meta-analysis
to investigate the relative contributions of sensitivity and mentalisation to infant attachment
classifications. They found that together the two predictors accounted for 12% of variance
in attachment classsifications. After controlling for sensitivity, the relationship between
parental mentalisation and infant–caregiver security was r = .24. And after controlling for
mentalisation, the relationship between parental sensitivity and infant–caregiver security
was r = .19. Sensitivity also partially mediated the association between mentalisation and
infant–caregiver security (r = .07).235 Such findings suggest that Meins, Fonagy, and others
were right to argue for the importance of mentalisation, but that sensitivity is not reducible
to mentalisation.236
Besides her emphasis on the causal role of caregiver sensitivity, another source of later
controversy lay in Ainsworth’s description of Group B, and the prototypically secure B3 sub-
group in particular, as ‘normative’:
Subgroup B3 is the largest in the sample, and accounts for 42% of the total sample. We con-
sider it to be the normative group, not merely because it is the largest, but also because, as
232 Meins, E., Fernyhough, C., Fradley, E., & Tuckey, M. (2001) Rethinking maternal sensitivity: mothers’ com-
ments on infants’ mental processes predict security of attachment at 12 months. Journal of Child Psychology and
Psychiatry and Allied Disciplines, 42(5), 637–48; Koren-Karie, N., Oppenheim, D., Dolev, S., Sher, E., & Etzion-
Carasso, A. (2002) Mothers’ insightfulness regarding their infants’ internal experience: relations with maternal
sensitivity and infant attachment. Developmental Psychology, 38(4), 534–42.
233 Fonagy, P. Steele, H., Steele, M., & Holder, J. (1997) Attachment and theory of mind: overlapping constructs?
Association for Child Psychology and Psychiatry Occasional Papers, 14, 31–40.
234 Fonagy, P. & Allison, E. (2014) The role of mentalizing and epistemic trust in the therapeutic relationship.
Psychotherapy, 51(3), 372–80.

235 Zeegers, M.A., Colonnesi, C., Stams, G.J.J., & Meins, E. (2017) Mind matters: a meta-analysis on parental
mentalization and sensitivity as predictors of infant–parent attachment. Psychological Bulletin, 143(12), 1245–72.
236 On the interpretation of effect sizes in psychology see Funder, D.C. & Ozer, D.J. (2019) Evaluating effect size
in psychological research: sense and nonsense. Advances in Methods and Practices in Psychological Science, 2(2)
156–68: ‘An effect-size r of .05 indicates an effect that is very small for the explanation of single events but poten-
tially consequential in the not-very-long run, an effect-size r of .10 indicates an effect that is still small at the level of
single events but potentially more ultimately consequential, an effect-size r of .20 indicates a medium effect that is
of some explanatory and practical use even in the short run and therefore even more important, and an effect-size
r of .30 indicates a large effect that is potentially powerful in both the short and the long run. A very large effect size
(r = .40 or greater) in the context of psychological research is likely to be a gross overestimate that will rarely be
found in a large sample or in a replication’ (156).
we subsequently show, it is the subgroup whose members have the most harmonious inter-
action with their mothers.237
‘Normative’ is a tricky word. As Cicchetti and Beeghly observed, the term confusingly hinges
judgements about what differs from a constant or average with assumptions about deviance
or defectiveness.238 The frequency of B3, the fact that it seemed a prototypical expression
of the attachment behavioural system, uncomplicated by avoidance or resistance, and the
strong relationship between B3 and caregiver sensitivity led Ainsworth and colleagues to
speculate that this is the natural state of mothers and infants. They argued that the human
attachment behavioural system is ‘adapted (in the evolutionary sense) to include a mother
whose reciprocal maternal behaviours are sensitively turned to infant signals’.239 Ainsworth
and her group accepted Bowlby’s dictum on this matter that ‘natural is better’ (Chapter 1),240

and hence that sensitive caregiving was both evolutionarily expectable and had the best im-
plications for a child’s mental health, in both the short and long term. In Ainsworth’s view,
‘the ordinary expectable social environment for a young child is both responsive and pro-
tective. These assumptions imply a fundamental compatibility between man and society’;241
in contrast, Groups A and C were believed to represent “developmental anomalies’.242
This is a conclusion that Robert Hinde contested on two counts.243 In criticizing
Ainsworth’s position, Hinde drew upon developments in ethology in the 1970s that were
subsequent to Bowlby’s development of his theory of the evolutionary function of attach-
ment. John Maynard Smith and Richard Dawkins introduced the idea that evolution may
have selected for a repertoire of behavioural patterns for achieving reproductive success in
diverse circumstances (Chapter 3).244 The ‘optimal strategy’ would be the one preferred when
circumstances were favourable, and would generally be the most direct way of achieving re-
productive success. However, ‘conditional strategies’ would be available as alternatives that
would have greater likelihood of success under less-favourable circumstances. On the basis
of these developments in the theory of evolutionary biology, Hinde argued that just because

238 Cicchetti, D. & Beeghly, M. (1990) An organizational approach to the study of Down syndrome: contribu-
tions to an integrative theory of development. In D. Cicchetti & M. Beeghly (eds) Children with Down syndrome: A
Developmental Perspective (pp.29–62). Cambridge: Cambridge University Press, p.32. See also Canguilhem, G.
(1966, 1989) The Normal and the Pathological, trans. C.R. Fawcett & R.S. Cohen. New York: Zone Books.
239 Ainsworth, M.D.S. (1976) Discussion of papers by Suomi and Bowlby. In: G. Serban (ed.) Animal Models in
Human Psychobiology (pp.37–47). New York: Plenum Press, p.43.

240 When Bob Marvin, a former student and close colleague of Ainsworth, wrote to raise with Bowlby the ques-
tion of whether infants could be adapted to various forms of caregiving environment, the reply (copied to Mary
Ainsworth) was uncompromising: ‘I need a lot of convincing that all these variations optimise infants’ chances
of survival.’ Bowlby, J. (1975) Letter to Robert Marvin, 5 November 1975. PP/Bow/J.9/132. Nonetheless, Bowlby
would ultimately be convinced (Chapter 3).
241 Stayton, D.J., Hogan, R., & Ainsworth, M. (1971) Infant obedience and maternal behavior: the origins of
socialization reconsidered. Child Development, 42(4), 1057–69, p.1059. The term ‘ordinary expectable social en-
vironment’ may have meant ordinary expectable social environment for forming an attachment relationship, ra-
ther than expectable within human evolutionary history, or expectable for twentieth century mothers. Ultimately,
Ainsworth’s intentions with the term are difficult to identify from her writings.
242 Ainsworth, M.D.S. (1979) Attachment as related to mother–infant interaction. In J.S. Rosenblatt, R.A. Hinde,
C. Beer, & M. Busnel (eds) Advances in the Study of Behavior, Vol. 9. (pp.1–51). New York: Academic Press, p.44.
243 Hinde, R.A. (1982) Attachment: some conceptual and biological issues. In J. Stevenson-Hinde & C. Murray
Parkes (eds) The Place of Attachment in Human Behavior (pp.60–78). New York: Basic Books.
244 Smith, J.M. (1979) Game theory and the evolution of behaviour. Proceedings of the Royal Society London, B,
205(1161), 475–88; Brockmann, H.J., Grafen, A., & Dawkins, R. (1979) Evolutionarily stable nesting strategy in a
digger wasp. Journal of Theoretical Biology, 77(4), 473–96.
156 Mary Ainsworth
it was common, and a direct expression of the attachment behavioural system, B3 could not
be treated as an ideal appropriate to all circumstances. Nor could sensitivity be assumed to
be expectable for human infants in the environment within which humans evolved; quite
the opposite. Even if the secure pattern of attachment might have the better long-term out-
comes, the other patterns of attachment may offer short-term advantages for achieving care-
giver availability. Ever cautious about presenting proposals about the evolutionary origins of
behaviour, Hinde did not press on to discuss what the advantages might be of the non-secure
patterns of attachment.245 His point was primarily to offer caution regarding the assumption
that security is always best, both in the short and long term. Ainsworth was displeased with
Hinde’s remarks.246 However, as Chapter 3 shows, Main came to much the same conclusions
as Hinde, and took these conclusions further.247 Though the debate has continued, Hinde’s
position seems certainly the more common one today among attachment researchers.

Avoidant attachment
Avoidant behaviour on reunion with the caregiver in the Strange Situation may last just a few
seconds. Nonetheless, Ainsworth thought that the behaviour seemed important. Comparing
Group A behaviour in the Strange Situation with their home observation data, Ainsworth
and colleagues found apparently paradoxical results. Though infants in Group A dyads were
precisely those that did not show distress in the Strange Situation, at home they were the
most frequently distressed and aggressive.248 The avoidance shown in the Strange Situation
was definitively not a stable trait between the laboratory and the home setting. Curiously this
point was missed by many of the second generation of attachment researchers, who assumed
that lack of distress would characterise Group A infants across contexts in their theory and
even in their design of measurement instruments.249
245 Griffiths, P.E. (2008) History of ethology comes of age. Biology and Philosophy, 23, 129–34: ‘At some stage in
the mid-50s Hinde and Tinbergen explicitly discussed a division of the “four questions” of ethology (Tinbergen
1963), with the Oxford program focusing on “survival value” and “evolution”, and the Cambridge department on
“development” and “causation” (Hinde, personal communication)’ (132).
246 Ainsworth offered a long and detailed critical discussion of Hinde’s remarks in Ainsworth, M. (1983) Letter
to Klaus Grossmann, 13 January 1983. Mary Ainsworth Archive, Box M3173, Folder 4. Her main counterargu-
ments were as follows: (i) the language of ‘adaptation’ used to refer to the insecure strategies by Hinde risked
implying that they are equally good, when in fact longer-term security surely tends to be better; (ii) ‘I see no reason
for arguing that it is part of that normal time-frame for human mothers to reject their infants in the interests of in-
fant autonomy and/or maximisang their own reproductive potential at 6 months or even at 1 year of age’; ‘my point
is that in the human species efforts to foster independence through withholding close contact from a baby when he
is upset and much wants contact do not foster a healthy kind of independence. Perhaps such efforts, if gradual ra-
ther than abrupt, may foster self-reliance without destroying the security of attachment when begun in the second
year of life. But I think that any time in the first year is too soon’; (iii) ‘to understand the relationship between a
given parent and a particular child obviously attachment cannot be the sole focus of attention’.
247 For a recent review of critiques of the ‘normative’ position assigned to security by Bowlby and Ainsworth
see Simpson, J.A. & Belsky, J. (2016) Attachment theory within a modern evolutionary framework. In J.
Cassidy & P.R. Shaver (eds) Handbook of Attachment: Theory, Research, and Clinical Applications (pp.91–116).
New York: Guilford.
248 Ainsworth, M.D.S. (1977) Social development in the first year of life: maternal influences on infant–mother
attachment. In J.M. Tanner (ed.) Developments in Psychiatric Research (pp.1–20). London: Hodder & Stoughton,
p.17. Ainsworth’s initial findings were elaborated and further confirmed by a secondary analysis of Ainsworth’s
data by Main: Main, M. & Stadtman, J. (1981) Infant response to rejection of physical contact by the mother.
Journal of the American Academy of Child Psychiatry, 20(2), 292–307.
249 E.g. Pederson, D.R. & Moran, G. (1995) A categorical description of infant–mother relationships in the
home and its relation to Q-sort measures of infant–mother interaction. Monographs of the Society for Research in
A secondary analysis conducted by Main revealed that Group A infants exceeded both the
other groups in terms of the number of times contact was initiated at home.250 However, this
proximity-seeking tended to have a ‘tentative’ quality, ‘making partial approaches followed
by moving off, or by going the whole way and then merely touching her’.251 Yet when they did
achieve close physical contact, they did not show active contact behaviour such as sinking in
or relaxing comfortably against the mother’s body. ‘When put down they were more likely
than infants of other groups to protest or to signal to be picked up again’, despite the fact that
protesting on being put down was the Group C signature move in the Strange Situation.252
Group A behaviour in the Strange Situation was also associated with relatively less-
sensitive behaviour by caregivers towards their infants at home. The strongest association
between Group A and caregiver behaviour in the home observation data was ‘picking the
baby up in an abrupt and interfering manner’. There was also a substantial negative rela-

tionship between Group A classification in the Strange Situation and a measure of ‘affec-
tionate behaviour while holding the baby’ by the mothers in the home environment over
the first year.253 For instance, Tracey and Ainsworth reported that the mothers in Group
A dyads kissed their babies more frequently than other dyads, but cuddled their babies
proportionately less.254 Ainsworth noted, however, that the baby also contributed to these
dynamics. Assessments of ‘tender, careful holding’ of the baby predicted positive infant re-
sponse in subsequent quarters, whereas positive infant response to holding did not predict
maternal careful holding going forward. That is, the mothers’ tender holding seemed to be
driving the infants’ response to holding and not vice versa.255 This profile of findings to-
gether suggested to Ainsworth and colleagues that Group A attachment, and perhaps all
attachment patterns, could be regarded as partly the result of a spiral between caregiver and
infant signals.
Whilst, in general, ‘tender, careful holding’ aligned with Ainsworth’s measure of sensi-
tivity, an exception was the B1 subgroup of dyads. These dyads were classified as secure since
infants greeted their caregiver with joy on reunion in the Strange Situation. However, like
Group A, the infants did not approach their caregiver. In the home observations, B1 care-
givers were found to be relatively sensitive: the average score out of 9 on the sensitivity scale
was 7.4 for B3 dyads and 4.5 for B1 and B2 dyads, compared to scores 2–3 for Group A and
C dyads.256 However, in contrast to the other Group B dyads, the caregivers in B1 dyads
received high ratings on apparent aversion to physical contact with their infant, though
Child Development, 60(2–3), 111–32; Pederson, D.R. & Moran, G. (1996) Expressions of the attachment relation-
ship outside of the strange situation. Child Development, 67(3), 915–27.
250 Reported from Main’s secondary analysis of Ainsworth’s home observation data: Main, M. (1980) Avoidance
of attachment figures: index of disturbance. PP/Bow/J.4/1.

251 Ainsworth, M.D.S. (1977) Social development in the first year of life: maternal influences on infant–
mother attachment. In: J.M. Tanner (ed.) Developments in Psychiatric Research (pp.1–20). London: Hodder &
Stoughton, p.18.

253 Ibid. p.145; Ainsworth, M. (1979) Infant–mother attachment. American Psychologist, 34(10), 932–7. Note 1,
citing unpublished data. The researchers did not explain how they operationalized ‘affectionate behaviour while
holding the baby’.
254 Tracey, R.L. & Ainsworth, M. (1981) Maternal affectionate behaviour and infant–mother attachment pat-
terns. Child Development, 52(4), 1341–3.

255 Ainsworth, M.D.S. (1979) Attachment as related to mother–infant interaction. In J.S. Rosenblatt, R.A. Hinde,
C. Beer, & M. Busnel (eds) Advances in the Study of Behavior, Vol. 9. (pp.1–51). New York: Academic Press, p.29.
of the Strange Situation. Bristol: Psychology Press, p.230, Table 27.

158 Mary Ainsworth
caregiver aversion to physical contact in B1 dyads was still lower than that for caregivers of
Group A dyads.257
Ainsworth sought to make sense of the profile of behaviours of Group A infants in the
Strange Situation and their experience at home. It appeared at first sight very surprising
that Group A infants were the most distressed and angry at home, and yet responded to the
Strange Situation with attentional and behavioural avoidance of the caregiver on reunion.
Yet this discrepancy made sense in light of findings that infants in Group A received rela-
tively less sensitive care. What appeared to Ainsworth to be taking place was visible and
predictable evidence of a psychological inhibition: the attachment behavioural system was
being activated by the Strange Situation, but its expression was being suppressed. In a letter
from 1967 with some early speculations, Ainsworth wrote to Bowlby:

This makes me think that at the same time that the baby is developing attachment behav-
iour (and I think of this as being developed through a feed-back rather than a reinforce-
ment model) he may well also be developing defensive reactions to bring into play when he
doesn’t get the feedback he has come to expect. So a baby at one-year who seems relatively
little attached may merely be one who has already built up a primitive but fairly effective
defensive system. The whole thing makes considerable sense to me—especially because the
babies of A1 nearly all had some fragments of attachment behaviour.258
At a time when speculations about motivation and inner processes were still anathema in
much of American psychology, in the context of a backlash against the speculative and un-
testable mechanisms posited by psychoanalysis, this was an electrifying finding. The pri-
ority given to scientific prediction, feedback, and passively learnt behaviour in behaviourism
was achieved precisely through the characteristic concerns of psychoanalysis on family
context, internal conflict, the active defensive strategies of the individual, and the role of
invisible motivations in prompting and inhibiting behaviour. Ainsworth’s distinction be-
tween Groups A and B integrated apparently irreconcilable trends within psychology: she
was studying family context and defensive strategies precisely through a laboratory-based,
replicatable observational study. This made her work especially eye-catching to contempor-
aries. Ainsworth later recalled that ‘the fact that the Strange Situation was not in the home
environment, that it was in the lab, really helped. I only did it as an adjunct to my naturalistic
research, but it was the thing that everyone could accept somehow. It was so demonstrable.’259
With new doctoral students entering the research group in the early 1970s, Ainsworth’s
team became interested in why infants from Group A dyads concentrated their attention
on toys just at the moment when other infants showed distress. The quality of this attention
to the toys was poor, ‘showing no investigative interest in the objects that they were either
manipulating or moving toward, but rather banging them about repeatedly or throwing and
retrieving them repeatedly’.260 It seemed that the toys were being used as a distraction to
avoid attending to the caregiver and other cues for the activation of the attachment system.
Group A was therefore termed ‘avoidant’ by Ainsworth.
257 Ibid. p.232 and Table 28.

258 Ainsworth, M. (1967) Letter to John Bowlby, 17 October 1967. PP/Bow/K.4/12.
259 Interview with Mary Ainsworth by Robert Karen cited in Karen, R. (1998) Becoming Attached: First
Relationships and How They Shape Our Capacity to Love. Oxford: Oxford University Press, p.163.

Influenced by Main’s emerging ideas (Chapter 3), Ainsworth came to believe that the
avoidant behaviour ‘protects the baby from experiencing the rebuff that he has come to
expect when he seeks close contact with his mother. It thus somewhat lowers his level of
anxiety (arousal). It also leads him to turn to the neutral world of things.’261 Avoidance
allows the infant to remain alongside the caregiver, even if closeness is not achieved,
whilst also escaping the contradiction between a desire to approach the caregiver and
concern about what has happened in the past when physical closeness has been sought.
The attention and orientation to the toys are aspects of what from the 1980s Ainsworth
would term, following Bowlby, the ‘defensive exclusion’ of the attachment behavioural
system: ‘It is suggested that Pattern A babies under stress systematically exclude from
perception (i.e. from highest-level processing) information that might intensely acti-
vate attachment behaviour. Thus they tend not to be distressed when the mother leaves

the room in the separation episodes of the strange situation and when she returns to the
reunion episodes, resorting instead to diversionary activity, which in this situation com-
monly consists of what appears to be exploratory behaviour.’262 Both distress and the
desire for comfort are excluded, resulting, Ainsworth speculated, in the maintenance of
equilibrium.
Subsequent researchers have taken from Ainsworth’s findings and theory the conclusion
that avoidant attachment results from caregiver insensitivity in the form of rebuff of attach-
ment signals. Though showing weaker effects than Ainsworth, later research generally repli-
cated Ainsworth’s findings that caregivers of avoidant dyads show lower levels of sensitivity
than securely attached dyads,263 and that caretakers in avoidant dyads specifically display
higher levels of intrusive behaviour towards their infants.264 Until recently, there has been
remarkably little discussion of alternative pathways to avoidant attachment besides rebuff
of attachment signals.265 This is despite the fact that researchers have, over the years, em-
pirically documented that other caregiver behaviours in naturalistic settings are associated
with avoidance in the Strange Situation. Main’s own analysis of the Ainsworth home obser-
vation data revealed that not only maternal rejecting behaviours but also maternal angry
261 Ibid. p.312.

262 Ainsworth, M. (1984) Attachment. In N.S Endler & J. McVicker Hunt (eds) Personality and the Behavioral
263 E.g. Grossmann, K., Grossmann, K.E., Spangler, G., Suess, G., & Unzner, L. (1985) Maternal sensitivity and
newborns’ orientation responses as related to quality of attachment in northern Germany. Monographs of the
Society for Research in Child Development, 50(1–2), 233–56; Koren-Karie, N., Oppenheim, D., Dolev, S., Sher, E., &
Etzion-Carasso, A. (2002) Mothers’ insightfulness regarding their infants’ internal experience: relations with ma-
ternal sensitivity and infant attachment. Developmental Psychology, 38(4), 534.
264 E.g. Smith, P.B. & Pederson, D.R. (1988) Maternal sensitivity and patterns of infant–mother attachment.
Child Development, 59(4), 1097–101; Isabella, R. & Belsky, J. (1991) Interactional synchrony and the origins of
infant–mother attachment. Child Development, 62, 373–84.
265 Early exceptions include Lamb, M., Thompson, R.A., Gardner, W., & Charnov, E.L. (1985) Infant–Mother
Attachment: The Origins and Developmental Significance of Individual Differences in the Strange Situation. Hillsdale,
NJ: Lawrence Erlbaum, p.96; Grossmann, K.E., Grossmann, K., & Schwan, A. (1986) Capturing the wider view of
attachment: a reanalysis of Ainsworth's Strange Situation. In C.E. Izard & P.B. Read (eds) Measuring Emotions
in Infants and Children, Vol. 2 (pp.124–71). New York: Cambridge University Press. More recently see Sroufe,
L.A. (2013) The promise of developmental psychopathology: past and present. Development & Psychopathology,
25(4.2), 1215–24. ‘We need studies that unpack the heterogeneity in current categories by examining differential
antecedents and pathways’ (1216); Waters, T.E.A. & Facompré, C.R. (in press) Measuring secure base script know-
ledge in the Adult Attachment Interview. In E. Waters, B.E. Vaughn, & H.S. Waters (eds) Measuring Attachment.
New York: Guilford: ‘Simply saying a child is avoidant tells us little about what specific parenting behaviors lead
that child to lose trust in their caregiver.’
160 Mary Ainsworth
behaviours predicted infant avoidance in the Strange Situation.266 And in her analysis of
videos of mother–toddler free play from her dissertation sample, Main reported:
Mothers of mother-avoidant infants mocked their infants, or spoke sarcastically to or about

them; some stared them down. One expressed irritation when the infant spilled imaginary
tea. Our ratings showed a strong association between avoidance and maternal anger.267
Grossmann and colleagues found that North German mothers’ expectation of self-reliance
from their infants meant that half their sample received an avoidant attachment classifica-
tion in the Strange Situation. This was despite the fact that, for a proportion of these infants,
the expectation of self-reliance was not linked to other forms of maternal rejection, or to
the later outcomes measured by the Grossmanns.268 In contrast, both maternal and paternal

sensitivity to the infant as assessed in home observations made sizeable contributions to later
outcomes. Such findings suggest that one pathway to an avoidant attachment classification
lies in cultural values around the suppression of distress signals, but that this pathway is not
necessarily one with the same implications as when avoidance stems from caregiver insensi-
tive and rejecting care.269 (The issue of the contribution of cultural differences in care to the
distribution of Strange Situation classifications is discussed further in the section ‘Cross-cul-
tural applicability of the Strange Situation’.)
Ambivalent/resistant attachment
Comparing the Strange Situation behaviour of infants from Group C dyads to the home ob-
servation data on the sample, Ainsworth and colleagues found that these infants displayed
more distress at home than the secure infants, and especially on occasions when the caregiver
attempted to leave the room.270 However, few other differences were noted in the children’s
behaviour at home. Ainsworth reported the qualitative impression of a mismatch between
266 Main, M. (1978) Avoidance of the attachment figure under stress: ontogeny, function and immediate caus-
ation. PP/Bow/J.4/1; Main, M. & Stadtman, J. (1981) Infant response to rejection of physical contact by the mother.
Journal of the American Academy of Child Psychiatry, 20(2), 292–307.
267 Main, M. (1980) Avoidance of attachment figures: index of disturbance. PP/Bow/J.4/1.
268 Grossmann, K., Grossmann, K.E., Spangler, G., Suess, G., & Unzner, L. (1985) Maternal sensitivity and
newborns’ orientation responses as related to quality of attachment in northern Germany. Monographs of the
Society for Research in Child Development, 50(1–2), 233–56; Fremmer-Bombik, F. & Grossmann, K.E. (1993)
Über die lebenslange Bedeutung früher Bindungserfahrungen. In: H. Petzold (ed.) Frühe Schädigungen—späte
Folgen? Psychotherapie und Babyforschung, pp.83–110. Paderborn: Jungfermann; Grossmann, K., Grossmann,
K.E., & Kindler, H. (2005) Early care and the roots of attachment and partnership representation. The Bielefeld
and Regensburg Longitudinal studies. In K.E. Grossmann, K. Grossmann, and E. Waters (eds) Attachment from
Infancy to Adulthood: The Major Longitudinal Studies (pp.98–136). New York: Guilford.
269 For aligned qualitative observations in a different context see Otto, H. (2014) Don’t show your emotions!
Emotion regulation and attachment in the Cameroonian Nso. In H. Otto & H. Keller (eds) Different Faces of
Attachment: Cultural Variations of a Universal Human Need (pp.215–29). Cambridge: Cambridge University Press.
Another pathway to an avoidant attachment classification in the Strange Situation may be the mind-mindedness
of the caregiver, distinct from their rejection of physical contact. Elizabeth Meins and colleagues found that
caregivers in dyads classified as avoidantly attached made fewer comments relevant to their child’s inner states
during free play. Meins, E., Fernyhough, C., de Rosnay, M., Arnott, B., Leekam, S.R., & Turner, M. (2012) Mind-
mindedness as a multidimensional construct: appropriate and nonattuned mind-related comments independ-
ently predict infant–mother attachment in a socially diverse sample. Infancy, 17(4), 393–415, Table 1.

infant signals and caregiver response, which contributed to low sensitivity. Even in the face
of queries regarding the integrity of the C classification, Ainsworth felt there was a defining
characteristic that linked the otherwise diverse behaviour of C1 and C2 infants. This was the
resistance and displays of ambivalence when the caregiver offered comfort, across contexts:
The ambivalence of Pattern C babies, both at home and in the strange situation, is easily
understood. Their mothers, who were very inconsistent in their responsiveness to signals,
often failed to pick the baby up when he most wanted contact, and often put him down
again long before he was ready to be put down. Consequently, when attachment behaviour
is intensely activated the baby has no confident expectation that his mother will respond
to his need for close contact. Having been frustrated in such situations often enough in the
past, his desire for close contact is intermingled with anger, because he rather expects his

mother to be unresponsive. He wants contact and is angry if his mother does not respond,
or if she tries some other mode of interaction, and yet he is still angry if she picks him up
and is difficult to soothe; indeed, he may struggle to be put down only to protest and seek
to be picked up again.271
The term ‘inconsistent’ was a common one in Ainsworth’s lexicon, and regularly used with
a variety of non-overlapping meanings. For instance, ‘inconsistent sensitivity’ is actually the
technical label for the mid-point (5) on her sensitivity scale; caregivers from ambivalent/
resistant dyads had scores on sensitivity well below this.272 Yet from the mid-1970s, ‘incon-
sistency’ was a term she frequently used to describe the caregivers of ambivalent/resistant
dyads, and was subsequently picked up by later attachment researchers as the defining cause
of ambivalent/resistant attachment in Ainsworth’s account. On this basis, later researchers
assumed a model in which inconsistent care creates a lack of contingency for the infant: in-
fants know that they can have their attachment signals heeded, but it is not clear when. As
such, the threshold for the activation of the attachment behavioural system is lowered, and
the threshold for termination raised. When the infant’s intensified attachment behaviours
and distress are accurately interpreted by the caregiver, this reinforces the strategy.273 Much
about this pathway is plausible, but the idea that it is ‘inconsistency’, specifically, that is the key
ingredient remains unevidenced.274 In fact, Ainsworth offers no data to suggest that the care-
givers of ambivalent/resistant infants are unpredictably sensitive: in Patterns of Attachment,
the specific behaviours that distinguished these mothers were in fact ‘delay in responding
271 Ainsworth. M.D.S. (1983) Patterns of infant–mother attachment as related to maternal care: their early
history and their contribution to continuity. In: D. Magnusson & V.L. Allen (eds) Human Development: An

273 Ainsworth, M.D.S., Bell, S.M., & Stayton, D.J. (1974) Infant–mother attachment and social development: ‘so-
cialisation’ as a product of reciprocal responsiveness to signals. In: M.J.M. Richards (ed.) The Integration of a Child
into a Social World (pp.9–135). Cambridge: Cambridge University Press; Cassidy, J. & Berlin, L.J. (1994) The in-
secure/ambivalent pattern of attachment: theory and research. Child Development, 65(4), 971–91; Mayseless, O.
(1996) Attachment patterns and their outcomes. Human Development, 39(4), 206–23; Crittenden, P.M. (1999)
Danger and development: the organisation of self-protective strategies. In J.I. Vondra & D. Barnett (eds) Atypical
Attachment in Infancy and Early Childhood Among Children at Developmental Risk (pp.145–71). Oxford: Blackwell.
274 The predictive power of contextual indicators of caregiver unpredictability have been examined by Belsky
and colleagues in the NICHD sample, but not considered in relation to infant attachment patterns. Belsky, J.,
Schlomer, G.L., & Ellis, B.J. (2012) Beyond cumulative risk: distinguishing harshness and unpredictability as deter-
minants of parenting and early life history strategy. Developmental Psychology, 48(3), 662–73.
162 Mary Ainsworth
to cry signals and occupying the time when holding the baby with routines’.275 Their mean
scores for sensitivity were the same as those of avoidant dyads. Though Ainsworth and col-
leagues found that mothers in ambivalent/resistant dyads displayed somewhat fewer re-
jecting behaviours towards their babies, the difference was not marked.276
However, there appears to be a second model in Ainsworth’s writings regarding caregiving
in ambivalent/resistant dyads. The quantitative findings reported in Patterns of Attachment
(1978) were of (i) delay in responding to cry signals and (ii) occupying the time when
holding the baby with routines. Main later described unpublished data from Ainsworth’s
Baltimore study, showing that ‘in the first 3 months of life, the mothers of the infants who
would later be resistant were extraordinarily inept in holding their infants (inept in 41% of
holding episodes) and almost never “tender and careful” (tender and careful in 2% of epi-
sodes). Their face-to-face interactions with their infants were marked by the absence of con-

tingent pacing.’277 This suggests a caregiver who is not adverse to closeness with their baby
but who finds it difficult to hold them fully in mind—leading to a delay in responding and
less full, satisfying, and well-judged responses when they do occur.
Indeed, Ainsworth herself specifically claimed that ‘the mother of a Pattern C baby is likely
to enjoy contact with him even though she is often imperceptive of his need for it’.278 The idea
that the caregivers of ambivalent/resistant dyads are less effective at tracking their infant’s
inner states is supported by findings that they make comments relevant to their child’s inner
states during free play, but these appear to observers to have little relationship with the child’s
actual states.279 It would be this imperceptiveness that would lead to a lack of confidence on
the part of the infant in the attentional availability of the caregiver, the reduced threshold for
activation of the attachment behavioural system, and the intensification of attachment sig-
nals. Ainsworth’s final word on the classification made no mention of ‘inconsistent’ care, and
instead foregrounded any caregiver behaviour that would lead a child to be uncertain about
availability.280
Overall, however, the relative infrequency of ambivalent/ resistant dyads in most
American and European samples, the absence of a scale for coding the passivity character-
istic especially of the C2 infants, and the lack of a sharp theory of caregiving antecedents
have all contributed to distinterest in the ambivalent/resistant classification over the dec-
ades. Even when researchers found distinct correlates of the ambivalent/resistant classifi-
cation, there has been little support for interpreting them from a network of other findings

276 Ibid. p.230, Table 27.
277 Main, M. (1995) Recent studies in attachment: overview, with selected implications for clinical work. In S.
Goldberg, R. Muir, & J. Kerr (eds) Attachment Theory: Social, Developmental and Clinical Perspectives (pp.407–70).
Hillsdale, NJ: Analytic Press, p.417.
278 Ainsworth, M.D.S. (1983) Patterns of infant–mother attachment as related to maternal care: their early
history and their contribution to continuity. In D. Magnusson & V.L. Allen (eds) Human Development: An
279 Meins, E., Fernyhough, C., de Rosnay, M., Arnott, B., Leekam, S.R., & Turner, M. (2012) Mind-mindedness
as a multidimensional construct: appropriate and nonattuned mind-related comments independently predict

infant–mother attachment in a socially diverse sample. Infancy, 17(4), 393–415, Table 1. See also Kelly, K., Slade,
A., & Grienenberger, J.F. (2005) Maternal reflective functioning, mother–infant affective communication, and in-
fant attachment: exploring the link between mental states and observed caregiving behavior in the intergenera-
tional transmission of attachment. Attachment & Human Development, 7(3), 299–311.
280 Ainsworth, M.D.S. & Eichberg, C.G. (1991) Effects on infant–mother attachment of mother’s experience re-
lated to loss of an attachment figure. In: C.M. Parkes, J. Stevenson-Hinde, & P. Marris (eds) Attachment Across the
Life Cycle (pp.160–83). London: Routledge, p.162.
and hypothesis-generating discussions.281 Ambivalence/resistance retains a place in the

system—17% of dyads received this classification according to a recent meta-analysis282—
but without researchers finding much need to give it discussion, with exceptions such as
Crittenden and Mayseless.283 The weak network of theory and empirical findings around
the ambivalent/resistant classification has led to significant problems for the field in pinning
down the relationship between the C and D classifications (Chapter 4).
Conflict behaviours in the Strange Situation
Ainsworth was generally of the view that the classificatory system she had developed was
open-ended. In Patterns of Attachment, she wrote ‘it is inconceivable that any system based

on a relatively small sample could comfortably accommodate all patterns represented in
the total population’.284 She wanted to ensure space to accommodate the unforeseen. One
strategy Ainsworth used to keep her system open was to have an explicit residual category, to
serve as a worksite for identifying further patterns. At the start of her work with the Strange
Situation, she used Group C to encompass infants who did not fit the other two groups, es-
pecially on the basis of seeming ‘disturbed’. This was still part of the function of the classi-
fication as late as Patterns of Attachment, though the classification came increasingly to be
defined by resistance (C1) and passivity (C2)—and then, after Ainsworth, generally just by
resistance. During her sabbatical at Stanford, Ainsworth wrote to Bell of the value of having
an intense and critical relationship between theory and observation, where each could
change a researcher’s perception of the other:
Even the most beautiful of theories is never as beautiful as truth or fact. To have destroyed
a theory is therefore an excellent thing. It is a step forward. And one need not tremble lest
a fact destroy a theory, even one’s own. One must seek it. Underneath lies a discovery.285
One unexpected discovery occurred when Ainsworth attempted to assess the stability of her
categories by bringing the dyads from one wave of her sample back two weeks later for a
second Strange Situation. Whilst the response of the infants in dyads classified as secure was
very largely the same, the response of the seven infants in avoidantly attached dyads was
281 Exceptions include Luijk, M.P., Velders, F.P., Tharner, A., et al. (2010) FKBP5 and resistant attachment pre-
dict cortisol reactivity in infants: gene–environment interaction. Psychoneuroendocrinology, 35(10), 1454–61;

Beebe, B. & Lachmann, F.M. (2014) Future resistant dyads. In The Origins of Attachment: Infant Research and
Adult Treatment (pp.104–114). London: Routledge. However, it is probable that these latter findings will not be
replicated by other laboratories: the coding system of Beebe and colleagues is not given in sufficient detail in their
published works to permit replication, and no training is available for other laboratories to learn the system.
282 Verhage, M.L., Schuengel, C., Madigan, S., et al. (2016) Narrowing the transmission gap: a synthesis of three
decades of research on intergenerational transmission of attachment. Psychological Bulletin, 142(4), 337–66,

Table 4.
283 Among the few sustained discussions of the classification are Cassidy, J. & Berlin, L.J. (1994) The insecure/
ambivalent pattern of attachment: theory and research. Child Development, 65(4), 971–91; Crittenden, P.M. (1999)
Danger and development: the organisation of self-protective strategies. In J.I. Vondra & D. Barnett (eds) Atypical
Attachment in Infancy and Early Childhood Among Children at Developmental Risk (pp.145–71). Oxford: Blackwell;
Scher, A. & Mayseless, O. (2000) Mothers of anxious/ambivalent infants: maternal characteristics and child-care
context. Child Development, 71(6), 1629–39.

285 Ainsworth, M. (1968) Letter to Sylvia Bell, 2 January 1968. Mary Ainsworth Archive, Box M3169, Folder 6.
164 Mary Ainsworth
quite different. Every one now approached the caregiver on reunion. But they did so whilst
showing fragments of avoidant behaviour. Ainsworth concluded that the infants had been
overstressed.286 The approach to the caregiver shown by these infants did not reflect their
expectation that the caregiver would be available, like the infants in secure dyads. Instead it
reflected the extent of their distress and fear.287 In this context, the avoidance was breaking
down into the kinds of conflict behaviour that had been described by Hinde (Chapter 1),
displaying wishes to both approach and avoid.
Between 1968 and 1973, Main undertook her doctoral research with Ainsworth. As well
as the measures required for her doctoral research, Main instructed her coders ‘to note each
time that the toddler did anything which seemed odd to them’; this included ‘hand-flapping;
echolalia; inappropriate affect; and other behaviours appearing out of context’.288 Five out
of the forty-nine infants were found to display such odd behaviour that it was difficult to

fit them within the Ainsworth classifications: two of these infants were force-classified as
secure, whereas three ‘were informally termed A–C infants within the laboratory’ and classi-
fied either as A or C.289 Main noted that two of these ‘A–C’ infants showed reunion behaviour
that combined an attempt to approach the caregiver with signs of fear and avoidance. One
threw her hands in front of her face on reunion, whereas the other engaged in asymmetric
handslapping whilst creeping forward. Main ‘asked Mary Ainsworth as my dissertation ad-
visor what to do. Characteristically cautious, but certain these infants were insecure, she re-
commended that for the time being (until more samples were collected and studied) ‘we
place them in Group A’.290 Main wrote wryly in a footnote to her doctoral thesis that, though
this technique was pragmatically useful, ‘Linnaeus might not approve’.291
From the 1970s, Ainsworth also took an interest in stereotypic and tension behaviours,
though this interest was hampered by the fact that the set-up of her Strange Situations had
only given observers a ‘profile view (at best)’ rather than a frontal view of the baby on re-
union.292 Ainsworth was also reliant on detailed written observations, rather than the full
video recording used by the next generation of attachment researchers. Initially Ainsworth
286 This finding would later be replicated, inadvertently, by researchers at Uppsala University: Granqvist, P.,
Hesse, E., Fransson, M., Main, M., Hagekull, B., & Bohlin, G. (2016) Prior participation in the strange situation and
overstress jointly facilitate disorganized behaviours: implications for theory, research and practice. Attachment &

288 Main, M. (1977) Analysis of a peculiar form of reunion behaviour seen in some day-care children. In R.
Webb (ed.) Social Development in Childhood (pp.33–78). Baltimore: John Hopkins, pp.70–71.
289 Main, M. & Solomon, J. (1990) Procedures for identifying infants as disorganised/ disoriented during
the Ainsworth Strange Situation. In M.T. Greenberg, D. Cicchetti, & E.M. Cummings (eds) Attachment in the
Preschool Years (pp.121–60). Chicago: University of Chicago Press, p.126. On the A–C designation see Crittenden,
P.M. & Ainsworth, M. (1989) Child maltreatment and attachment theory. In D. Cicchetti & Y. Carlson (eds) Child
Maltreatment: Theory and Research on the Causes and Consequences of Child Abuse and Neglect (pp.432–63).
Cambridge: Cambridge University Press. Also Lamb, M., Thompson, R.A., Gardner, W., & Charnov, E.L. (1985)
Infant–Mother Attachment: The Origins and Developmental Significance of Individual Differences in the Strange
Situation. Hillsdale, NJ: Lawrence Erlbaum. In a cluster analysis performed on four samples, there were seven clus-
ters: one was ‘a relatively rare group of infants displaying both avoidance and resistance’ (217).
290 Mary Main, personal communication, August 2012. See Duschinsky, R. (2015) The emergence of the disor-
ganized/disoriented (d) attachment classification, 1979–1982. History of Psychology, 18(1), 32–46.

291 Main, M. (1973) Exploration, play and cognitive functioning. Unpublished doctoral thesis, Johns Hopkins
University, p.21.
of the Strange Situation. Bristol: Psychology Press: Observers in Ainsworth’s procedure ‘had a good view of a baby’s
face as he approached either the mother’s or stranger’s chair, a profile view (at best) of a baby oriented to the door or
to a person entering’ (34).
associated stereotypic and tension behaviours with the B4 subclassification. B4 was intro-
duced in 1969 to capture infants from Silvia Bell’s sample who were very upset by the separ-
ation episodes, but who did manage to use their caregiver as a source of support and calm
before the end of the Strange Situation.293 In Patterns of Attachment, one of the criteria used
to define B4 was that the subclassification included any ‘other signs of disturbance, such as
inappropriate, stereotyped, repetitive gestures or motions’.294 As a consequence, the B4 sub-
type was regarded by some as a rather unstable element within the system.295 The inclusion
of other signs of disturbance as one of its criteria was later formally retracted as a criterion
for B4 by Ainsworth, with the introduction of the D classification (Chapter 3).296
In Patterns of Attachment, Ainsworth and colleagues explicitly considered the discrepant
behaviour they were seeing in the Strange Situation in terms of Hinde’s concept of ‘conflict
behaviour’:

When two antithetical systems are activated simultaneously, they may be said to be in con-
flict . . . The other system may not become manifest in behaviour until either the overriding
behaviour is terminated (or becomes less strongly activated) or some shift in the situation
increases the activation of the system until it overrides the behaviour of the previously
stronger system. When two conflicting systems are more nearly equal in level of activation
there may be alternation of behaviours, ‘compromise’ behaviours in which behavioural
elements of both systems are combined, or intention movements or other fragmentary be-
havioural representatives of one or the other system. Furthermore, the behaviour activated
by one stimulus object may be redirected towards another that is not involved in the con-
flict . . . Finally, overt behaviour may be determined by a third system, which is also at a
moderate level of activation, although not as high a level as the two conflicting systems that
tend to block each other—a phenomenon the ethologists label ‘displacement behaviour’.297
In an unpublished paper from around the period of composition of Patterns of Attachment,

Ainsworth and colleagues argued that there was a ‘distress language’ to be deciphered in
these odd behaviours, and that this language would be important for comprehending in-
fant experience: ‘Once one has gotten to know the distress language, these subtle signals
make the distress of the child who is trying with might and main to control it all the more
poignant’.298 The final pages of Patterns of Attachment were dedicated to a strong call—which
has never been quoted by subsequent researchers—for study of the meaning of conflict
293 Ainsworth, M., Bell, S.M., & Stayton, D.J. (1969) Individual differences in strange-situational behaviour
of one-year-olds. http://eric.ed.gov/?id=ED056742. Later version: Ainsworth, M.D.S., Bell, S.M., & Stayton, D.J.
(1971) Individual differences in strange-situation behavior of one-year-olds. In H.R. Schaffer (ed.) The Origins of
Human Social Relations (pp.17–58). New York: Academic Press.

295 E.g. Van IJzendoorn, M.H., Tavecchio, L.W.C., Goossens, F.A., Vergeer, M.M., & Swaan, J. (1983) How B
is B4? Attachment and security of Dutch children in Ainsworth’s Strange Situation and at home. Psychological
Reports, 52(3), 683–91.
296 Personal communication from Mary Ainsworth to Mary Main, January 1985, cited in Main, M. & Solomon,
J. (1990) Procedures for identifying infants as disorganised/disoriented during the Ainsworth Strange Situation.
In M.T. Greenberg, D. Cicchetti, & E.M. Cummings (eds) Attachment in the Preschool Years (pp.121–60).
Chicago: University of Chicago Press, p.149.

298 Ainsworth, M. (1976) Attachment and Separation in Paediatric Settings. Unpublished manuscript. PP/Bow/
J.1/40.
166 Mary Ainsworth
behaviours seen in the Strange Situation, and their different potential causes. In the early
1980s, Ainsworth wanted to hold a conference for Strange Situation coders, dedicated to
exploration of anomalous behaviours. Ultimately, though, she decided that she lacked the
time and money to host this event. Meanwhile, however, she encouraged her graduate stu-
dents Mary Main and Patricia Crittenden in their study of infants who showed behaviours
discrepant with her categories for coding the Strange Situation.299 And she anticipated that
research with clinical samples would ultimately lead to additions to her coding system, with
particular relevance for coding clinical samples. After all, her three categories had been de-
veloped on the basis of a sample of participants recruited according to demographic char-
acteristics that would reduce adversities.300 In the meantime, the anomalies did not appear
to threaten use of the Strange Situation as the basis for building a research programme.
Bowlby’s position seems to have been aligned with Ainsworth’s. Bowlby was not reconciled

to the idea that the three Ainsworth classifications were sufficient for work with clinical sam-
ples.301 However, he did not pursue the matter in print, except in his remarks on the work
of Ainsworth’s students. The issue would ultimately be left for the next generation of attach-
ment researchers (Chapter 3).
Everett Waters
Stability
Part of Ainsworth’s legacy for the field of attachment research was her mentorship of a re-
markable cohort of students, many of whom became leaders of the second generation of at-
tachment research. Among the undergraduates she taught who would become colleagues in
attachment-related research were Mark Cummings, Mark Greenberg, Robert Marvin, David
Olds, and Everett Waters. Her graduate students at Johns Hopkins were Silvia Bell, Mary
Blehar, Inge Bretherton, Alicia Lieberman, Mary Main, and Sally Wall. Michael Lamb was a
student with her for a year at Johns Hopkins, before Ainsworth left for Virginia. Ainsworth’s
graduate students and mentees at Virginia included Jude Cassidy, Deborah Cohn, Virginia
Colin, Patricia Crittenden, Rogers Kobak, Carolyn Eichberg, and Ulrike Wartner.302 In later
life, Ainsworth wrote fondly of her students as her ‘academic family’, a phrase that continues
to be used by attachment researchers in the developmental tradition to describe their com-
munity.303 The phrase has perhaps recurred in part because it captures the warmth, care, and
299 Crittenden, P.M. & Ainsworth, M.D. (1989) Child maltreatment and attachment theory. In D. Cicchetti & V.
Carlson (eds) Child Maltreatment: Theory and Research on the Causes and Consequences of Child Abuse and Neglect
(pp.432–61). Cambridge: Cambridge University Press; Ainsworth, M. (1991) Past and future trends in attachment
research. Film of the presentation made available by Avi Sagi-Schwartz (Chair), International Society for the Study
of Behavioral Development, Minneapolis, July 1991.
300 Ainsworth, M. (1985) Patterns of mother–infant attachments. Bulletin of the New York Academy of Medicine
61, 792–812, p.788.

301 Solomon, J., Duschinsky, R., Bakkum, L., & Schuengel, C. (2017) Toward an architecture of attachment disor-
ganization: John Bowlby’s published and unpublished reflections. Clinical Child Psychology and Psychiatry, 22(4),
539–60. See also Bowlby, J. (1981) Clinical applications: material for lectures. PP/Bow/F.3/103: ‘Patterns of attach-
ment: Secure attachment; Anxious attachment; Compulsive self-reliance; Compulsive care-giving; Psychopathic
detachment.’
302 Bretherton, I. (2000) Mary Dinsmore Salter Ainsworth (1913–1999). American Psychologist, 55, 1148–9.
303 E.g. Ainsworth, M. (1983, 2013) An autobiographical sketch. Attachment & Human Development, 15(5–6),
448–59, p.458.
Everett Waters 167
loyalty of the field of attachment research, which matches well the common idea of family;
and in part because it captures the conflict, compassion, and compromises of the field of at-
tachment research, which matches the all-too-human, fumbling-at-times actuality of family.
One of Ainsworth’s students who was important in shaping her legacy for developmental
psychology was Everett Waters. Waters worked with Ainsworth for almost two years, taking
courses and graduate seminars and acting as a research assistant. In addition, he joined
graduate students Mary Blehar and Sally Wall helping Ainsworth prepare the book length
report of her Baltimore project, Patterns of Attachment. Waters’ roles included preparing
the report on multivariate analysis of the ABC classification system and editing draft chap-
ters.304 In 1972, on the strength of Ainsworth’s recommendation, Waters was admitted to
the University of Minnesota’s Institute of Child Development. Classmates soon directed
him toward assistant professor Alan Sroufe and his work on the development of smiling and

laughter. Sroufe’s interests inclined toward a general theory of emotional development and
he was eager to tap Waters’ first-hand experience with attachment research. Mentoring led to
collaboration and soon a fast friendship, and a series of influential papers.
An early paper by Sroufe and Waters was a turning point for acceptance of Ainsworth’s
Strange Situation classifications within the wider developmental science community.305
In this paper ‘Attachment as an organizational construct’, published in 1977, Sroufe and
Waters summarised the cornerstones and significance of the theoretical work of Bowlby and
Ainsworth and discussed recent empirical results in a pointed response to several criticism
rooted in temperament and learning theory paradigms.
A first criticism that Sroufe and Waters sought to combat was that Ainsworth’s secure
vs insecure distinction could be explained as temperamental differences in sensitivity to
the stress of separation or in terms of how often mothers reinforced crying and depend-
ency. The mistake was that the secure vs insecure distinction did not map onto crying vs
non-crying in the separation episodes. In fact, half of secure one-year-olds protested separ-
ation in the Strange Situation (primarily B3 and B4) and half did not (B1, B2, and some B3).
Thus, as Sroufe and Waters pointed out, classifications depended not on separation protest
but on responses to mother’s return. Moreover, Sroufe and Waters pointed to data showing
that minimal separation protest in A infants was not simply a matter of temperament or re-
inforcement cycles. These explanations assumed that non-crying infants were indifferent to
separation. Certainly research with American parents indicated that many regarded the in-
fants who showed avoidant behaviour in the Strange Situation as the most competent, since
they seemed to not be distressed by the separations.306 In a piece of swift ingenuity, Sroufe
and Waters reported evidence from concurrent heart-rate recordings that ‘avoidant infants
showed sustained heart-rate acceleration on reunion (in the absence of vigorous motor
activity), suggesting clear affective response rather than indifference’.307 Infants in dyads
304 In 2015, Waters initiated a reissue of the now classic book with additional appendices and a new preface in
which he, Inge Bretherton, and Brian Vaughn discussed the book’s impact and its significance for current work.
305 Sroufe, L.A. & Waters, E. (1977) Attachment as an organizational construct. Child Development, 1184–99.
Goldberg, S. (2000) Attachment and Development. London: Routledge: ‘Ainsworth’s work was also subjected to
heated criticism and, until the publication of Sroufe and Waters’ 1977 paper on “attachment as an organisational
construct” and the emergence of supporting data from other laboratories, it was neither understood nor accepted
by the larger community of developmental psychologists’ (236).
306 E.g. Hubbs-Tait, L., Gray, D., Wierzbicki, M., & Englehart, R. (1994) Perceptions of infant boys’ behavior and
mental health: relation to infant attachment. Infant Mental Health Journal, 15(3), 307–15.
307 Sroufe, L.A. & Waters, E. (1977) Attachment as an organizational construct. Child Development, 1184–99,
p.1191. These findings are detailed further in Sroufe, L. & Waters, E. (1977) Heart-rate as a convergent measure
in clinical and developmental research. Merrill-Palmer Quarterly, 23, 3–27. In later research using heart-rate
168 Mary Ainsworth
classified as avoidant were interpreted as inhibiting a display of attachment behaviour and

keeping their distress secret from the caregiver, rather than having less desire for comfort
than other infants in the Strange Situation.
A second criticism was that infant attachment behaviours do not intercorrelate highly,
show weak stability over time, and are strongly influenced by context. This was taken by
critics as an invalidation of the construct of infant attachment.308 Of particular importance
was a paper published in Psychological Bulletin by Masters and Wellman from 1974, which
offered a detailed critique of attachment research, including a report that ‘little stability of
attachment behaviors was found if the intervening time was three minutes, one day, three
months, four months, or longer’.309 This was a potentially decisive criticism in light of the
central role early experience plays in attachment theory. Ainsworth and her small group
of collaborators experienced these critiques as a serious threat to the emerging paradigm.

There was ‘a real sense of emergency’ in the small attachment research community.310 One
response to Masters and Wellman was theoretical counterargument and clarification. Sroufe
and Waters argued that attachment is a behavioural system, not the particular behaviours
selected to achieve the set-goal. Particular behaviours are of primary relevance only to the
inferred inner organisation they are understood to express. If reaching works quickly to
achieve contact with the caregiver, there might be no need to also engage crawling towards
the caregiver. If fussing works, there might be no need to use crying. The selection of behav-
iours will be sensitive to context.311
A second response to Masters and Wellman was empirical. Waters undertook to demon-
strate that the results Masters and Wellman had reported regarding the absence of stability
stemmed from their focus on individual behaviours, rather than the organisation of behav-
iour. He collected Strange Situation assessments with 50 middle-class dyads at 12 months
and 18 months of age. There was almost no continuity of attachment behaviours displayed to-
wards the caregiver across the half-year period, as Masters and Wellman had predicted. Only
measures, there is some discrepancy between findings, suggesting that infants in avoidant dyads are more physio-
logically stressed than infants in secure dyads, and findings indicating comparability. Nonetheless, it has been
taken for granted by researchers since Sroufe and Waters that most infants in avoidant dyads are indeed distressed
by the separation, even after the widespread increase in daycare, and that their avoidant behaviour represents a
masking and inhibition of a desire to seek the caregiver’s availability, rather than an absence of this desire. On later
findings see e.g. Spangler, G. & Grossmann, K.E. (1993) Biobehavioral organization in securely and insecurely
attached infants. Child Development, 64(5), 1439–50; Hill-Soderlund, A.L., Mills-Koonce, W.R., Propper, C., et al.
(2008) Parasympathetic and sympathetic responses to the strange situation in infants and mothers from avoidant
and securely attached dyads. Developmental Psychobiology, 50(4), 361–76.
308 E.g. Cairns, R.B. (1972) Attachment and dependency: a psychobiological and social learning synthesis.
In J.L. Gewritz (ed.) Attachment and Dependency (pp.29–80). Washington, DC: Winston; Gewirtz, J. (1972) On
the selection and use of attachment and dependence indices. In J.L. Gewirtz (ed.) Attachment and Dependency
(pp.179–215). Washington, DC: Winston; Rosenthal, M. (1973) Attachment and mother–infant interaction: some
research impasses and a suggested change in orientation. Journal of Child Psychology and Psychiatry and Allied
Disciplines, 14, 201–207.
309 Masters, J. & Wellman, H. (1974) Human infant attachment: a procedural critique. Psychological Bulletin, 81,
218–37, p.224.
310 Everett Waters, personal communication, August 2019
311 Sroufe and Waters also showed that low correlations between individual attachment behaviours were also
partly an artifact of inadequate sampling of these low frequency behaviors. Specifically, because the discrete behav-
iors were typically samples for as little as 3-minutes, they did not provide a reliable (reproducible) estimate of an
infant’s typical behavior. Indeed, for many of these behaviors, Sroufe and Waters determined that it would require
hundreds of minutes of observation to obtain the reliable scores needed to correctly estimate stability. Waters then
showed that Ainsworth’s approach to scoring interactive behavior in terms of the organisation of multiple behav-
iors across Strange Situation episodes yielded more reliable scores on proximity seeking, contact, maintaining,
avoidance and resistance scores that could be quite stable across a full six month test-retest interval.
Everett Waters 169
smiling at the mother and touching the mother were stable across time; there was no con-
tinuity in other important behaviours such as proximity-seeking.312 Yet using Ainsworth’s
more broadly defined proximity-seeking, contact-maintaining, avoidance, and resistance
constructs, there was a consistent pattern of stability, particularly in the key reunion episodes
(for example, .62 for avoidance and .58 for resistance, both significant < .01). This empirical
result demonstrated more effectively than any narrative rebuttal the limitations of Masters
and Wellman’s critique as well as the strength and potential of the new attachment paradigm.
Forty-eight out of the fifty infants received the same classification at both 12 and 18 months.
This confirmed Waters’ expectation that more broadly defined and integrative assessments
would yield even higher stability than even individual scales. At the same time, this result is
much higher than in subsequent studies. Several factors could contribute to this. First, each
case was scored on the basis of a consensus among several experienced independent coders.

Moreover, the sample consisted of first-born infants from intact middle-class families. None of
the infants had experienced regular out-of-home care. It is also important to note a methodo-
logical limitation of the study. At the time, the only trained scorers for the Strange Situation were
Ainsworth and her students and the small Minnesota team. Consequently, the same experts
coded both time-points. Though coding the two time-points six months apart, the overlap in
coders may have unintentionally reduced the independence of their assessments and elevated
the stability of the clasifications.
There are times when having an extreme statement made allows the deep truth within it to be
released; and there are times when the extreme qualities end up skewing reception of a complex
truth in ways that are difficult to subsequently get heard. The sheer strength of stability of at-
tachment classifications over time reported by Waters reflects a little of both predicaments. This
was the first independent test of the stability of Ainsworth’s categories, and would prove critical
to their acceptance by the wider community of developmental scientists. Indeed, Waters’ study
appeared as empirical support for an image—suggested at times by Bowlby and Ainsworth—of
the results of early care as impervious to change, at least in the short term. This image was fur-
ther supported by the use of continuity from the Strange Situation by Main and colleagues in the
development and validation of new attachment measures (Chapter 3).313
Baldwin and Fehr suggest that a contributing factor to the birth of the social psychological
tradition of attachment research in the 1980s (Chapter 5) was the impression, given by the
reported findings of Waters, of individual differences in attachment as stable trait-like vari-
ables.314 As well as among many researchers, the idea of attachment patterns as set in early
childhood also became a pervasive part of the reception of Bowlby, Ainsworth, Waters, and
Main among clinical and child welfare readers.315 Aligned with Waters’ findings, several
312 Waters, E. (1978) The reliability and stability of individual differences in infant–mother attachment. Child
Development, 49(2), 483–94, p.488, Table 3.

313 For just one of many examples in which the Waters’ stability study and Main’s six-year follow-up are to-
gether taken to prove Bowlby’s claims about continuity from early childhood see Lipps, A.J. (2009) Review of
Klaus E. Grossman, Karin Grossman, and Everett Waters (eds): attachment from infancy to adulthood. Child &
Adolescent Social Work Journal, 26, 379–82.
314 Baldwin, M.W. & Fehr, B. (1995) On the instability of attachment style ratings. Personal Relationships, 2(3),
247–61.
315 See critical discussions of assumptions among practitioners about the fixedness of early attachment patterns
in Fraley, C.R. (2002) Attachment stability from infancy to adulthood: meta-analysis and dynamic modeling of de-
velopmental mechanisms. Personality and Social Psychology Review, 6(2), 123–51; Crittenden, P.M. (2016) Raising
Parents: Attachment, Representation, and Treatment, 2nd edn. London: Routledge; Granqvist, P., Sroufe, L.A.,
Dozier, M., et al. (2017) Disorganized attachment in infancy: a review of the phenomenon and its implications for
clinicians and policy-makers. Attachment & Human Development, 19(6), 534–58.
170 Mary Ainsworth
other samples in the 1980s similarly demonstrated high stability of the categories over
time.316 However, Waters was clearly concerned that such high levels of stability suggested a
lack of environmental responsiveness, which seemed implausible. He was also worried that
with only 50 infants in his original study, chance might have contributed to an overestimate
of stability. He therefore sought to conduct further research with high-risk dyads in order
to put boundaries on the previously reported stability before misapprehensions could arise
about attachment classifications being fixed for good in infancy.
Studying 100 high-risk dyads in the Strange Situation, Vaughn, Waters, and colleagues
found only 62 allocated to the same classification six months later. Discontinuities, however,
were often logical: dyads that changed from a secure to an insecure classification tended to
be those where the mother had experienced more stressful life-events in the meantime.317
Vaughn, Waters, and colleagues argued that this implied that patterns of attachment could

change if caregivers faced factors that would contribute to more or less sensitivity to their
child: ‘There is little room today for a construct that does not recognise the environmental
responsiveness of individual differences. The validity of the attachment construct is greatly
enhanced by our prediction and confirmation of stability in some cases and change in
others.’318 Ainsworth herself fully agreed that a child’s experiences of changes in interaction
with caregivers ‘predictably lead to changes in patterns of attachment’.319 She described her-
self as ‘delighted that researchers are increasingly turning to an examination of exceptions
to stability of attachment-pattern’,320 and encouraged her students to explore such excep-
tions.321 At the same time, Ainsworth cautiously anticipated that early experiences of secure
base/safe haven availability or unavailability may have relevance for later development, even
in the context of subsequent change (Chapter 4).322
It became a common finding that the Ainsworth classifications were generally less stable
in studies with participants drawn from high-risk samples. However, in the 1980s, studies
were also reported, even with low-risk samples, that found no short-term stability of the
categories. Sociologically, it may be noted that this coincides with a sharp rise in use of
day-care from the 1970s, which could be expected to loosen the association between ma-
ternal care and attachment classifications since less time would be spent together by mother
and child.323 The most influential study to show weak stability in the 1980s was a report
316 Main, M. & Weston, D.R. (1981) The quality of the toddler’s relationship to mother and to father: related to
conflict behavior and the readiness to establish new relationships. Child Development, 52(3), 932–40; Goossens,
F.A., van IJzendoorn, M.H., Tavecchio, L.W.C., & Kroonenberg, P.M. (1986) Stability of attachment across time
and context in a Dutch sample. Psychological Reports, 58(1), 23–32; Kermoian, R. & Leiderman, P.H. (1986) Infant
attachment to mother and child caretaker in an East African community. International Journal of Behavioral
317 Vaughn, B., Egeland, B., Sroufe, L.A., & Waters, E. (1979) Individual differences in infant–mother attachment
at twelve and eighteen months: stability and change in families under stress. Child Development, 50(4), 971–5.
318 Ibid. p.975.
319 Ainsworth, M. (1991) Past and future trends in attachment research. Film of the presentation made avail-
able by Avi Sagi-Schwartz (Chair), International Society for the Study of Behavioral Development, Minneapolis,
July 1991.
320 Ainsworth, M. (1985) Patterns of infant– mother attachments: antecedents and effects on development.
Bulletin of the New York Academy of Medicine, 61(9), 771–91, p.787.
321 Main, M. (1999) Mary D. Salter Ainsworth: tribute and portrait. Psychoanalytic Inquiry, 19(5), 682–736;
Bretherton, I. (2000) Mary Dinsmore Salter Ainsworth (1913– 1999). American Psychologist, 55, 1148– 9;
Crittenden, P.M. (2017) Gifts from Mary Ainsworth and John Bowlby. Clinical Child Psychology and Psychiatry,
22(3), 436–42.
322 Ainsworth, M. (1991) Past and future trends in attachment research. Film of the presentation made available
by Avi Sagi-Schwartz (Chair), International Society for the Study of Behavioral Development, Minneapolis, July
1991. https://www.youtube.com/watch?v=TWS1XBA7vmk.
323 https://www.census.gov/topics/families/child-care/data.html
Everett Waters 171
by Thompson, Lamb, and Estes.324 Like Vaughn and colleagues, they found that both con-
tinuity and discontinuity in attachment classifications was predictable in light of changes in
care: over half of infants changed attachment classification with their mother when she en-
tered full-time employment, and all the infants in the sample changed attachment classifica-
tion if they began receiving over 15 hours a week of non-parental care. However, the relative
instability of attachment classifications in the sample provided part of the staging platform
for Lamb and colleagues’ important early criticisms and qualifications of Ainsworth’s Strange
Situation procedure, which were published as a long article in Behavioral and Brain Sciences
in 1984. Though it does not seem to have been the authors’ intent, these criticisms were in-
terpreted as a rejection, rather than qualification, of the validity of the Strange Situation pro-
cedure. The article by Lamb and colleagues in the journal was followed by responses from
numerous researchers, mostly pouring scorn on the procedure and attachment theory in

general.325
Lamb and colleagues’ appraisal and qualification of every jutting aspect of the Strange
Situation, utilizing the detailed knowledge of an insider, offered excellent ammunition
to critics making general attacks on attachment as a research paradigm. Indeed, the ob-
servations of Lamb and colleagues are still cited today as clinching the ‘methodological
shortcomings and unsupported assumptions’ of Ainsworth’s research, and as making the
attachment paradigm as a whole untenable.326 Most importantly at the time, however,
Lamb and colleagues struck a nerve in headlining the lack of stability of classifications
of attachment in a low-risk sample, which appeared to contradict the claim by Bowlby
and Ainsworth that ‘this internalised something we call attachment’ is a ‘retentive inner
mechanism which serves as a kind of filter for the reception and interpretation of inter-
personal experience’, and in this way shapes later behaviour.327 Lamb and colleagues felt
that this offered an important qualification to the paradigm, but by no means intended
their stance as a rejection of attachment research. Nor did they intend their stance even as
a rejection of the Strange Situation, a form of measurement that they subsequently con-
tinued to use.
The Lamb and colleagues article brought home to the attachment community the extent
to which they were depending on tacit knowledge for organizing their research paradigm.
324 Thompson, R.A., Lamb, M.E., & Estes, D. (1982) Stability of infant–mother attachment and its relationship
to changing life circumstances in an unselected middle-class sample. Child Development, 53(1), 144–8.
325 Lamb, M.E., Thompson, R.A., Gardner, W.P., Charnov, E.L., & Estes, D. (1984) Security of infantile attach-
ment as assessed in the ‘strange situation’: its study and biological interpretation. Behavioral and Brain Sciences,
7(1), 127–47; Lamb, M.E., Gardner, W., Charnov, E.L., Thompson, R.A., & Estes, D. (1984) Studying the security of
infant–adult attachment: a reprise. Behavioral and Brain Sciences, 7, 163–71.
326 Vicedo, M. (2020) On the history, present, and future of attachment theory. Reply to Robbie Duschinsky,
Marinus van IJzendoorn, Sarah Foster, Sophie Reijman & Francesca Lionetti ‘attachment histories and futures’.
European Journal of Developmental Psychology, 17(1), 147–55, p.148.
327 Ainsworth, M. (1967) Infancy in Uganda. Baltimore: Johns Hopkins University Press, pp.429–30. In line with
Lamb’s concerns, researchers examined the respective contribution of infant attachment with mother and father
as assessed with the Strange Situation to later child attachment representations. The general finding was that nei-
ther makes an especially strong contribution, and that the extent of caregiver involvement with the child is an im-
portant moderator. Grossmann, K., Grossmann, K.E., & Kindler, H. (2005) Early care and the roots of attachment
and partnership representations. In K.E. Grossmann, K. Grossmann, & E. Waters (eds) Attachment from Infancy
to Adulthood: The Major Longitudinal Studies (pp.98–36). New York: Guilford; Steele, H. & Steele, M. (2005) The
construct of coherence as an indicator of attachment security in middle childhood. the friends and family inter-
view. In K.A. Kerns & R.A. Richardson (eds) Attachment in Middle Childhood (pp.137–60). New York: Guilford.
The continuity between infant–mother attachment in the Strange Situation and the six-year assessments in the
Berkeley sample is higher than most other studies. In part this is because the six-year samples were developed
semi-inductively from the infant data on the sample. See Chapter 3.
172 Mary Ainsworth
Most researchers at the time were using the Ainsworth scales in the absence of reliability
checks against existing research groups, or any demonstration of cross-group agreement.328
Ainsworth assumed that the scales and categories could be used without training or the con-
struction of a manual.329 The Thompson, Lamb, and Estes findings alerted Waters, Sroufe,
and other early attachment researchers in Ainsworth’s circle that the Strange Situation
coding protocols were not self-sufficient.330 Knowing how to code the Strange Situation
in the manner of Ainsworth was, in fact, partly an oral culture held by a small number of
researchers associated with Ainsworth’s group (then at Virginia), and the Berkeley and
Minnesota laboratories (Chapters 3 and 4). In response to the article by Lamb and col-
leagues, it was judged that inter-rater reliability between research groups and a formal pro-
cess of certification were needed to ensure valid use of the assessment. This prompted the
introduction of a training institute in coding the Strange Situation, which has been run

yearly at Minnesota by Alan Sroufe (and later Elizabeth Carlson and Robert Weigand) for
over three decades now.331
Curiously, the Ainsworth sensitivity scale was not subject to this formalisation of training
and certification.332 Presumably, since the scale was not published it was in less danger of
unrestricted use. Additionally, though central to the work of attachment researchers, this
scale has had less of a flagship status for the field’s critics and its defenders. By contrast, the
Strange Situation was treated as the foundation of attachment research as an empirical para-
digm, and so training to achieve scientific reliability was considered especially important.
However, the role of the Strange Situation coding training institutes extended well beyond
this, for instance in teaching relevant theory, providing opportunities for networking and
wider processes of enculturation. As van IJzendoorn observed at the time, these dynamics
of enculturation contributed to insularity in this community. In his view, a system of train-
ing institutes is ultimately necessary to ensure scientific reliability (though his laboratory,
like others, would use the Ainsworth sensitivity scale without any such training). However,
he expressed concern that training institutes may mean that researchers have to show
their submission to specific coding systems and their logic before that researcher’s voice is
328 Ainsworth expressed dismay in a letter to Bowlby that Michael Lamb had written to the editor of Child
Development to say that he had been a student of hers, as evidence of the validity of his coding. Though there
was no formal training process or certification, Ainsworth personally did not regard him as sufficiently trained
in coding the Strange Situation. Ainsworth, M. (1982) Letter to John Bowlby, 25 April 1982. PP/BOW/B.3/7.
However, van IJzendoorn and Kroonenberg would later empirically compare Lamb’s coding of a Swedish sample
to coding of Strange Situations in other contexts, and found minimal differences in how the coding protocols had
been interpreted. Van IJzendoorn, M.H. & Kroonenberg, P.M. (1990) Cross-cultural consistency of coding the
strange situation. Infant Behavior and Development, 13(4), 469–85.
329 Ainsworth, M. (1973) Letter to Georgette Marie Psarras, 9 July 1973. Mary Ainsworth Archive, Box M3176,
Folder 2.
330 One alternative to long training institutes was the development in the 1980s of an algorithm to turn scale
scores into category classifications by Richters, Waters, and Vaughn. The algorithm was used at times in the early
1990s, but quickly fell out of favour. The reasons for this are never discussed in print, but may have included the
introduction of the disorganized attachment classification (Chapter 3), and the social institutionalization of the
Minnesota Strange Situation training as authoritative. Richters, J.E., Waters, E., & Vaughn, B.E. (1988) Empirical
classification of infant– mother relationships from interactive behavior and crying during reunion. Child
331 Waters, E. (1983) The stability of individual differences in infant attachment: comments on the Thompson,
Lamb, and Estes contribution. Child Development, 54(2), 516–20. See https://attachment-training.com/.
332 This situation has only changed recently with training in the Ainsworth sensitivity scale offered by Marian
Bakermans-Kranenburg in 2019.
Everett Waters 173
acknowledged, leading to the exclusion of constructive disagreement and the insights that
can come from it.333
Ultimately, later meta-analytic research appeared to significantly qualify the strong
stability reported by the early studies—though whether this reflected the rise of daycare,
methodological differences between early and later studies, or simply the assumulation of
knowledge is difficult to unpick. In 2002, Fraley assembled findings from 23 papers using
measures from the developmental tradition of attachment research (i.e., not self-report
measures of attachment). Stability of a secure vs insecure attachment classification for as-
sessments of up to a year would prove much lower in Fraley’s meta-analysis than in the
Waters study: r = .32 (compared to r = .92 in Waters).334 Stability was higher in low-risk
(r = .48) than high-risk (r = .27) samples.
In the influential Minnesota Longitudinal Study of Risk and Adaptation (Chapter 4) there

was no continuity of secure vs insecure attachment classifications between infancy and the
Adult Attachment Interview. Nonetheless, as in the paper by Thompson and colleagues,
both continuity and discontinuity could be accounted for by a few specific factors, reflecting
stressors and support experienced by the caregiver. In the case of the Minnesota study these
were experiences of maltreatment after infancy, changes in family life stress, sensitivity in
parent–child relationships at 13, and quality of romantic relationships in adolescence.335
Such findings support the critics of Ainsworth and Waters: attachment classifications are
substantially less stable in the short term than originally thought, and more responsive to
context. However, it is critical to note that there is ‘lawful discontinuity’:336 the reasons for
the changes can be identified in factors that would certainly be expected to alter patterns of
attachment.
Furthermore, even if stability is lower than suggested by Waters, there remains substan-
tial evidence in support of the claim that attachment patterns have relevant continuity over
time. In a 2013 meta-analysis of 127 papers, and 21,072 dyads seen in the Strange Situation,
Pinquart and colleagues found equivalent findings to Fraley for short-term stability, and in
fact showed that stability remained moderate throughout the first five years. They were also
able to report that there were no significant differences between father–infant dyads and
mother–infant dyads in terms of stability. Pinquart and colleagues also reported the valuable
333 Van IJzendoorn, M.H. (1994) Process model of replication studies: on the relations between different types of
replication. In R. van der Veer, M.H. van IJzendoorn, & J. Valsiner (eds) On Reconstructing the Mind: Replicability
in Research on Human Development (pp.57–70). Norwood, NJ: Ablex: ‘Until such enculturation has taken place, a
researcher who cannot replicate a certain result may expect to be accused of being not competent enough to carry
out the experiment . . . A training of several weeks in one of the American research centers is considered necessary
for the reliable and valid coding of the observations or interviews, and therefore for a plausible and persuasive con-
tribution to the international discourse on attachment. When central theses of attachment theory are in danger
of being falsified by “untrained” researchers, this “incompetence” and lack of enculturation will be explicitly used
against the “dissident” (see, for an example, Waters, 1983)’ (62).
334 Fraley, C.R. (2002) Attachment stability from infancy to adulthood: meta-analysis and dynamic modeling of
developmental mechanisms. Personality and Social Psychology Review, 6(2), 123–51.

335 Weinfield, N.S., Sroufe, L.A., & Egeland, B. (2000) Attachment from infancy to early adulthood in a high-risk
sample: continuity, discontinuity, and their correlates. Child Development, 71(3), 695–702; Carlivati, J. & Collins,
W.A. (2007) Adolescent attachment representations and development in a risk sample. New Directions for Child and
Adolescent Development, (117), 91–106; Sroufe, L.A., Coffino, B., & Carlson, E.A. (2010) Conceptualizing the role of
early experience: lessons from the Minnesota Longitudinal Study. Developmental Review, 30(1), 36–51, pp.44–5.
336 Belsky, J. & Pensky, E. (1988) Developmental history, personality and family relationships: toward an
emergent family system. In R. Hinde & J. Stevenson-Hinde (eds) Relationships Within Families (pp.193–217).
Oxford: Clarendon Press.
174 Mary Ainsworth
discovery that secure attachments were much more likely to be stable than insecure attach-
ments (OR = 1.39), and that this effect was stronger in low-risk samples (OR = 1.73). Such
findings suggest that children are more likely to keep a sense of their caregiver’s availability
once they have found it than they are to retain a sense of their caregiver’s unavailability. They
found that this effect is stronger when there are fewer contextual risks facing the dyad.337
Pinquart and colleagues end on the optimistic note that the lower stability of insecure at-
tachments suggests that, against the public image that Bowlby’s popularizing writings helped
foster of attachment patterns as fixed for life, there is substantial room for change especially
among the families who may most need support. In these terms, there may be room for yet
more optimism. In 2018, Opie reported a meta-analysis from 56 studies, with stability of
r = .26. Accounting for differences between her findings and those of earlier studies, she had
access to many more published papers, and identified significant evidence of publication

bias, with studies reporting low stability less likely to have been published.338
Research on stability, initiated by Waters, overall raises two distinct points. First, con-
trary to the social learning theorists, that there is some stability suggests that the behaviour
reflects more than the pushes and pulls of immediate caregiver reinforcement. This finding
halted most, if not all, social learning critiques of attachment research, though the social
learning theorists did not join the attachment enterprise. Links between attachment and
social learning theory would have to be rediscovered by later researchers (see Chapter 6).
Second, the question of longer-term stability addresses the theoretical concern of the extent
to which early experience establishes a prototype for later interactions and/or are influenced
by later developments. The central conclusion drawn by researchers was that attachment is
influenced by later developments, and that this can be anticipated to be affected by changes
in caregiving.339 Use of language suggesting the fixedness of attachment patterns for most
people can still be seen in the writings of some major attachment researchers.340 However,
perhaps a majority of writings on attachment now emphasise predictable movement be-
tween patterns of attachment, with continuity in substantial part the result of continuity of
relationship qualities.341
337 Pinquart, M., Feußner, C., & Ahnert, L. (2013) Meta-analytic evidence for stability in attachments from in-
fancy to early adulthood. Attachment & Human Development, 15(2), 189–218. It should be mentioned, however,
that poor short-term stability was demonstrated on the large NICHD sample. Groh, A.M., Roisman, G.I., Booth-
LaForce, C., et al. (2014) IV. Stability of attachment security from infancy to late adolescence. Monographs of the
Society for Research in Child Development, 79(3), 51–66.
338 Opie, J. (2018) Attachment stability and change in early childhood and associated moderators. Unpublished
doctoral thesis, Deakin University, Melbourne.

339 An important work on this latter question, conducted by Waters and colleagues, was Waters, E., Merrick, S.,
Treboux, D., Crowell, J., & Albersheim, L. (2000) Attachment security in infancy and early adulthood: a twenty-
year longitudinal study. Child Development, 71(3), 684–9. This study is not detailed here since it would require ex-
planation of the Adult Attachment Interview, which is described in Chapter 3. In brief, the researchers found that
72% of the infants received a matching secure or insecure attachment classification in adulthood. However, 44%
of the infants whose mothers reported negative life events change attachment classifications by the time the Adult
Attachment Interview was conducted.
340 For a recent description of early attachment patterns as generally consigned for life in the absence of
intervention see Powell, B., Cooper, G., Hoffman, K., & Marvin, B. (2016) The Circle of Security Intervention.
341 Besides Lamb, another early advocate for integrating discontinuity into the tenants of attachment theory was
Crittenden, P.M. (1995) Attachment and psychopathology. In S. Goldberg, R. Muir, & J. Kerr (eds) John Bowlby’s
Attachment Theory: Historical, Clinical and Social Significance. New York: Analytical Press, pp.367–406. An em-
phasis on predicting discontinuity is also characteristic of the third generation of developmental researchers
(Chapter 6).
Everett Waters 175
Attachment Q-sort
As well as pursuing research on the stability of the Strange Situation classifications,

Waters was a fierce advocate for the need for full construct validation for attachment
measures. Following Cronbach and Meehl, he argued that any empirical findings take
their meaning from the network of other related results, which provide the context for
their interpretation.342 Waters and Ainsworth discussed their alignment with this pos-
ition in their correspondence.343 In the case of the Strange Situation procedure, Waters
and Deane expressed concern for a tendency within attachment research to search
for associations between the infant classifications and various outcomes without any
more general hypothesis than ‘all good things go together’.344 In their impression, with

each such study, attachment research becomes less incisive, and the concept of attach-
ment becomes increasingly debased into a general idea of close relationships in general
(Chapter 5).345
Waters argued for a return to focus on what he termed ‘the secure base phenomenon’,
which—confusingly for many readers—included for him both secure base and safe haven
dynamics. He also argued that too much attention had been paid to the ‘critical situations’
of distress and reunion, designed to prompt the activation of the attachment system, with
the Strange Situation providing visuals. Too little attention had been paid to the ordinary,
incremental moments that form the basis for trust or distrust in the availability of the care-
giver, especially contexts of non-distress, exploration, and quieter, ordinary feelings.346
Whilst overt attachment behaviour would be less frequently seen in these circumstances,
ordinary life in naturalistic settings was ultimately the substance that filled out and elabor-
ated the attachment system, and especially expectations about the caregiver’s availability as
a secure base. Like Bowlby and Ainsworth in their final years,347 and likely influenced by
Ainsworth’s stance directly, Sroufe and Waters were dismayed that attachment researchers
had leapt upon the Strange Situation procedure at the expense of the naturalistic research.
They felt that this direction of interest was due to a number of factor, including: poor gen-
eral understanding among researchers of Bowlby and Ainsworth’s specific characterisation
of the attachment behavioural system; the resource costs associated with detailed home ob-
servations; the unusual observational acuity of Ainsworth for seeing patterns, which not all
researchers possessed; and reification of the Group B classification as ‘security’ itself rather
342 Cronbach, L. & Meehl, P. (1955) Construct validity in psychological tests. Psychological Bulletin, 52(4), 281–
302; Waters, E. & Sroufe, L.A. (1983) Social competence as a developmental construct. Developmental Review 3,
79–97, p.80.
343 Ainsworth, M. (1973) Letter to Everett Waters, 7 December 1973. Mary Ainsworth Archive, Box M3176,
Folder 4.
344 Waters, E. & Deane, K.E. (1985) Defining and assessing individual differences in attachment relation-
ships: Q-methodology and the organization of behavior in infancy and early childhood. Monographs of the Society
for Research in Child Development, 50(1/2), 41–65, p.42.
345 Ibid. p.42.
346 The point is developed in Waters, E. (2008) Live long and prosper: a note on attachment and evolution.
http://www.psychology.sunysb.edu/attachment/gallery/live_long/live_long.html.
ality development. American Psychologist, 46(4), 333–41: The Strange Situation ‘procedure soon became widely
used, if not always wisely and well, and has quite overshadowed the findings of the research project that gave rise to
it’ (338).
176 Mary Ainsworth
than, as Ainsworth intended, a window into the child’s experiences of care by an attachment
figure and therefore a proxy for naturalistic observation.348
For any research paradigm truly loyal to Ainsworth’s principles, Sroufe and Waters argued
strongly that ‘the strange situation is used only because it can stand in place of attachment as
it would be observed in the home’.349 It struck them as dangerous that so much work subse-
quent to the Baltimore study was predicated (whether researchers knew it or not) on the reli-
ability of the link between Strange Situation assessments and secure base behaviour at home.
To offer a means to replicate and further explore this essential finding, Waters and colleagues
developed a new measure, the Attachment Q-Sort (AQS).350 Waters recalls that several col-
leagues in the attachment community were resistant to the creation of an alternative to the
Strange Situation.351 Yet many colleagues, including Sroufe, supported this development and
served as expert consultants on the development of the measure. Ainsworth stated her ap-

proval of the new measure.352
In the AQS, observers are asked to sort a set of behavioural descriptors of a child observed
for a few hours in a naturalistic setting like the home. These descriptors, printed on pieces of
card, are sifted into piles ranging from ‘most descriptive of this child’ to ‘least descriptive of
this child’. The set of descriptors covers a broad range of secure base and exploratory behav-
iour, affective response, social referencing, and other aspects of social cognition. Observers
can then be kept unaware of the constructs that will be scored from the data they provide,
whilst attachment phenomena can be picked out and distinguished from other aspects of be-
haviour. Like the Strange Situation, the measure is grounded in concretely observed behav-
iour. However, compared to the Strange Situation, the AQS can be used to assess ordinary
interaction between a child and caregiver in naturalistic settings; it therefore has greater
ecological validity than the dramatised prompts and interactions of the Strange Situation.
For instance, it could have better cross-cultural validity than the Strange Situation, since the
meaning of separation and reunion for young children likely differs by culture—though cul-
tural values may still inflect the descriptors on the AQS cards.
Ainsworth emphasised that the developmental changes from infancy to toddlerhood entail
nothing less than a ‘recasting of attachment relationship, so that they now include perspective
taking, include communication, negotiation and mutual plans’. Any adequate measure of at-
tachment after infancy, she felt, needs above all to consider the way that ‘seeing things from
348 Waters, E., Kondo-Ikemura, K., Posada, G., and Richters, J.E. (1991) Learning to love. In M.R. Gunnar & L.A.
Sroufe (eds) Self Processes and Development (pp.217–55). New York: Psychology Press, pp.241–2.
349 Sroufe, L.A. & Waters, E. (1982) Issues of temperament and attachment. American Journal of Orthopsychiatry,
52(4), 743–6, p.743.

350 On the history of Q methodology see Stephenson, W. (1953) The Study of Behavior: Q-Technique and its
Methodology. Chicago: University of Chicago Press; Good, J.M. (2010) Introduction to William Stephenson’s
quest for a science of subjectivity. Psychoanalysis and History, 12(2), 211–43. Though Waters and colleagues
cite Stephenson’s legacy, the more proximal examples of Q methodology research cited as a predecessor for the
Attachment Q-Sort were Baumrind, D. (1968) Manual for the preschool behavior Q-Sort. Berkeley: Institute of
Human Development, University of California; and Block, J. (1978) The Q-Sort Method in Personality Assessment
and Psychiatric Research. Palo Alto, CA: Consulting Psychologists Press.
352 Ainsworth, M. (1984) Attachment, adaptation and continuity. Paper presented at International Conference
on Infant Studies, April 1984. PP/Bow/J.1/57: ‘I am delighted with the extent to which this [Strange Situation]
procedure has proved useful in research, but have repeatedly stated that its success should stimulate rather than
discourage the development of other procedures . . . It is paradoxical that the search for new procedures comes not
from the critics of strange-situation research but from among those who have been most intimately involved in it.
I mention here particularly Mary Main and Everett Waters.’
Everett Waters 177
the partner’s point of view’ gets integrated with secure base/safe haven dynamics.353 Whilst
applications of the Strange Situation with toddlers and pre-schoolers have been developed,
such extensions have been troubled by the dual pull to respond to such maturational differ-
ences and, simultaneously, retain the semblance of Ainsworth’s infant system.354 A special ad-
vantage of the AQS is that, with minor adaptions, it can be used with children of different
ages.355 The AQS could readily incorporate expected changes in the way that children:
• use their caregiver as a secure base and safe haven in the context of more sophisticated
capacities for communication, negotiation, mutual planning, and perspective taking
• modulate the threshold for activation of the attachment behavioural system
• integrate cognition, affect, and behaviour as part of the system
• show attachment behaviour in interplay with other behavioural systems.

Q-methodology also has several psychometric advantages over Ainsworth’s categories.356
The primary output of the measure is the assessment of attachment security in terms of a
continuous variable, based on the extent to which children can use their caregiver as a secure
base and safe haven as needed within their everyday context. Waters did not, like Fraley and
Spieker, provoke controversy by publicly opposing the statistical and theoretical advantages
of a dimensional scale to a category-based approach. However, the way that the AQS was
constructed quietly downplayed the importance of the avoidant and resistant classifications,
focusing instead on how close or distant a case was from the paradigm of an infant using the
caregiver as a secure base and safe haven.357
A recent meta-analysis by Cadman and colleagues, on the basis of 245 studies and 32,426
child–caregiver dyads, reported substantial convergent validity for the AQS. When children
are assessed with the Ainsworth Strange Situation, and when the AQS is used for at least 180
353 Ainsworth, M. (1991) Past and future trends in attachment research. Film of the presentation made avail-
able by Avi Sagi-Schwartz (Chair), International Society for the Study of Behavioral Development, Minneapolis,
July 1991.
354 For discussions of, and different approaches to, this problem see Cassidy, J., Marvin, R., with the Attachment
Working Group of the John D. and Catherine T. MacArthur Network on the Transition from Infancy to Early
Childhood (1992) Attachment organisation in preschool children: procedures and coding manual. Unpublished
manual;Crittenden, P.M. (2017) Gifts from Mary Ainsworth and John Bowlby. Clinical Child Psychology
and Psychiatry, 22(3), 436–42; Waters, E., Vaughn, B., & Waters, H.S. (eds) (in press) Measuring Attachment.
New York: Guilford.
for Research in Child Development, 50(1/2), 41–65.
for Research in Child Development, 50(1/2), 41–65, p.53; Seifer, R. & Schiller, M. (1995) The role of parenting sensi-
tivity, infant temperament, and dyadic interaction in attachment theory and assessment. Monographs of the Society
for Research in Child Development, 60(2–3), 146–74.
357 A supplement to the Attachment Q-Sort to scale disorganized attachment (Chapter 3) was developed in
the 1990s by John Kirkland. However, the work was not published, or even discussed in print until recently. The
only study to date to have used this supplement is Handley, E.D., Michl-Petzing, L.C., Rogosch, F.A., Cicchetti, D.,
& Toth, S.L. (2017) Developmental cascade effects of interpersonal psychotherapy for depressed mothers: lon-
gitudinal associations with toddler attachment, temperament, and maternal parenting efficacy. Development &
Psychopathology, 29(2), 601–15. The authors report a strong association with disorganized attachment in the
Strange Situation, though on a study using only 10% of their sample of 125 toddlers. On disorganized attachment
and the Attachment Q-Sort measure see also the item analyses reported in Van Bakel, H.J. & Riksen-Walraven,
J.M. (2004) AQS security scores: what do they represent? A study in construct validation. Infant Mental Health
Journal, 25(3), 175–93.
178 Mary Ainsworth
minutes, agreement is substantial (r = .39), as are associations with measures of caregiver

sensitivity (r = .44). Agreement is, however, reduced when a shorter period of observation is
used (r = .25 with the Strange Situation; and .28 with caregiver sensitivity). The AQS assessed
by observers is also able to predict later socioemotional outcomes for children (r = .24). One
remarkable finding is that the AQS has an especially strong ability to predict later aggression
and conduct problems symptoms (d = .70). Some of this effect may be explained by the fact
that the AQS can be used with two-year-olds and this may be a developmental stage of par-
ticular importance for the development of conduct problems.358 At times, researchers have
also often asked parents themselves to complete the AQS. However, there are good reasons
to suspect that this approach is less valid, since associations with the Strange Situation and
sensitivity are much lower, and associations with caregiver report on the child temperament
are substantial (r = .33).359 It would seem that a child’s own caregivers are less effective than

unfamiliar observers at distinguishing a child’s attachment behaviours from a global im-
pression of their child’s temperament.360 Van IJzendoorn and colleagues praised the AQS,
claiming that the availability of a second well-validated measure of child attachment along-
side the Strange Situation helps ‘separate the concept of attachment in young children from
the way it is measured’, a major contribution to the field.361
A recent development of potential importance has been the development of a brief ver-
sion of the AQS by Fearon and colleagues.362 Ainsworth fully acknowledged that the Strange
Situation is far too laborious for screening purposes or regular use by clinicians and prac-
titioners, and urged methodological innovation in this area.363 To address this gap, the
Brief Attachment Scale (BAS-16) is a pared-down version of the full Attachment Q-Sort,
consisting of two scales: (i) harmonious interaction with the caregiver and (ii) proximity-
seeking behaviours. Validated against the full AQS on a portion of the large NICHD sample,
convergence between the measures was very good. The BAS-16 had associations with care-
giver sensitivity (r = .23) and child externalizing behaviours at 24 months (r = –.25) equiva-
lent to the full AQS, and there was no association with measures of infant temperament.
The NICHD sample also had classifications available for the Ainsworth Strange
Situation—though in this sample there was no association between the Ainsworth clas-
sifications and the full AQS. As a result, the BAS-16 had all but no link with the Strange
358 Fearon, R.P., Bakermans-Kranenburg, M.J., van IJzendoorn, M.H., Lapsley, A.M., & Roisman, G.I. (2010)
The significance of insecure attachment and disorganization in the development of children’s externalizing be-
havior: a meta-analytic study. Child Development, 81(2), 435–56, p.450.
359 Cadman, T., Diamond, P.R., & Fearon, P. (2018) Reassessing the validity of the attachment Q-sort: an up-
dated meta-analysis. Infant and Child Development, 27(1).

360 See also Tarabulsy, G., Provost, M.A., Larose, S., et al. (2008) Similarities and differences in mothers’ and ob-
servers’ ratings of infant security on the attachment Q-sort. Infant Behavior & Development, 31(1), 10–22.
361 Van IJzendoorn, M.H., Vereijken, C.M., Bakermans-Kranenburg, M.J., & Marianne Riksen-Walraven, J.
(2004) Assessing attachment security with the attachment Q sort: meta-analytic evidence for the validity of the
observer AQS. Child Development, 75(4), 1188–213, p.1207.
362 Cadman, T., Belsky, J., & Fearon, R.M.P. (2018) The Brief Attachment Scale (BAS-16): a short measure of
infant attachment. Child: Care, Health and Development, 44(5), 766–75. Independently of Fearon, Bakermans-
Kranenburg and colleagues also developed a brief version of the AQS about 15 years earlier. Bakermans-
Kranenburg, M.J., Willemsen-Swinkels, S.H.N., & van IJzendoorn, M.H. (2003) Brief Attachment Screening
Questionnaire. Unpublished manuscript, Leiden University, Centre for Child and Family Studies. This was used as
part of Rutgers, A.H., van IJzendoorn, M.H., Bakermans-Kranenburg, M.J., et al. (2007) Autism, attachment and
parenting: a comparison of children with autism spectrum disorder, mental retardation, language disorder, and
non-clinical children. Journal of Abnormal Child Psychology, 35(5), 859–70.
363 Ainsworth, M.D.S. (1980) Attachment and child abuse. In G. Gerbner, C.J. Ross, & E. Zigler (eds) Child
abuse: an agenda for action (pp.35–47). Oxford: Oxford University Press, p.45.
Everett Waters 179
Situation. Fearon and colleagues suggest that greater convergence may be anticipated when
the BAS-16 is used to assess responses to separation or some other mildly stressful event,
when the attachment behavioural system will be activated. If so, this would offer important
further validation of the new measure. Until then, what conclusions the field will draw re-
garding the usefulness of the BAS-16 as a brief measure of infant attachment are unclear, and
will likely depend on the extent to which developmental attachment research is willing to
rest weight on measures that do not agree with the Ainsworth Strange Situation.364 For the
second generation of attachment researchers, this would have been an anathema. For the
third-generation leaders of the field, with a greater concern to sustain a dialogue with rou-
tine clinical practice, it may not only be possible but perhaps expectable.365 The BAS-16 may
also help facilitate cross-cultural attachment research in a way that has not been possible for
the Strange Situation.

Secure base scripts
The Strange Situation, as a staged procedure requiring the semi-standardisation of caregiver

behaviour, offers little insight into what caregivers actually do to support attachment and
exploration in particular dyads. Yet Waters emphasised that the attachment behavioural
system requires support to achieve coherence. This includes, for example, ‘secure base/safe
haven games’ such as peek-a-boo, which dramatise the script of separation followed by re-
union, and obstruction resolved into relational closeness and accessibility.366 This teaching
and learning of the secure base script is evident also in adolescent and adult relationships,
when romantic relationships are invested with attachment components. In ‘Bowlby’s theory
grown up’, published in 1994, Crowell and Waters argued that ‘commitment in secure-base
terms is unlikely to be “hearts and flowers” responsiveness or “I want to marry you”; rather,
we suggest it may be more like “I’m here, I will be here, I’m interested in what you do and
what you think and feel, I will actively support your independent actions, I trust you and
you can trust me to be here if you need me, et cetera”.’367 All the small ways that adolescent
or adult dyads repeatedly, across situations, affirm their availability to one another form the
intricate little bones within the structure of the relationship and relationship-related expect-
ations, giving strength and stability to its capacity to serve as a secure base and safe haven.
Everett and Harriet Waters observed that Bowlby’s internal working model concept en-
compasses (i) how accessible or inaccessible attachment figures tend to be under ordinary
situations, (ii) whether attachment behaviours are regarded as acceptable or unacceptable,
and (iii) a forecast about how available and responsive these figures are when difficulties
364 See Ziv, Y. & Hotam, Y. (2015) Theory and measure in the psychological field: the case of attachment theory
and the strange situation procedure. Theory & Psychology, 25(3), 274–91.
365 Compare, for instance, attempts to pare down established assessments of caregiving behaviour to develop
a brief screening measure: Haltigan, J.D., Madigan, S., Bronfman, E., et al. (2019) Refining the assessment of dis-
rupted maternal communication: using item response models to identify central indicators of disrupted behavior.
Development & Psychopathology, 31(1), 261–77.
366 Waters, E., Kondo-Ikemura, K., Posada, G., and Richters, J.E. (1991) Learning to love. In M.R. Gunnar & L.A.
Sroufe (eds) Self Processes and Development (pp.217–55). New York: Psychology Press, p.236; Waters, E., Posada,
G., Crowell, J.A., & Lay, K.L. (1994) The development of attachment: from control system to working models.
Psychiatry, 57(1), 32–42, p.35.
367 Crowell, J.A. & Waters, E. (1994) Bowlby’s theory grown up: the role of attachment in adult love relation-
ships. Psychological Inquiry, 5(1), 31–4, p.32.

180 Mary Ainsworth
are faced (Chapter 1).368 Waters and Waters therefore argued that the idea of the internal
working model is too clumsy and general a concept for developing specific testable hy-
potheses. They advocated instead the idea of ‘scripts’, sedimented in procedural memory by
repetition and experience, which respond to particular cues with involuntary expectations
about what tends to happen next and predispositions to behave in certain ways. Attention to
attachment-related scripts therefore moves ‘toward explaining what exactly the development
of attachment representations is the development of ’.369 In fact, a similar point was proposed
to Bowlby in the 1980s by John Byng-Hall; Bowlby was highly sympathetic, describing the
specification of internal working models using the metaphor of a ‘script’ as ‘a most valuable
step’.370
Waters and Waters argued that at the heart of attachment theory are the secure base and
safe haven responses, and of all the different content included within an ‘internal working

model’ it is these that should be the focus of researchers’ efforts.371 This proposal has seen
increasing acceptance among attachment researchers in recent years.372 Such a script might
include some source of distress for one member of a dyad; this prompts a signal for help;
the signal is detected and help is offered; this help is accepted and proves effective; the inter-
action proves comforting; and matters are sufficiently resolved that other activities can be
recommenced. Waters and Waters proposed that having a secure base script organizing the
cognitive components of the attachment behavioural system is helpful for the expression
of secure behaviour when exploration is called for, and for requesting and making use of
support as needed if demands or threats arise. Waters and Waters developed methods for
assessing the secure base script, both using narrative methodologies appropriate for children
and in coding the Berkeley Adult Attachment Interview (Chapter 3). Such methods have
demonstrated that secure base scripts in adolescence and adulthood are predicted by early
experiences of sensitive care and early attachment patterns, are generally stable over time,
and in turn predict adult caregiving behaviour to children and other aspects of functioning,
including even adult physical health.373
368 Waters, H.S. & Waters, E. (2006) The attachment working models concept: among other things, we build
script-like representations of secure base experiences. Attachment & Human Development, 8(3), 185–97. They
were influenced by earlier work by Schank, R. & Abelson, R. (1977) Scripts, Plans, Goals, and Understanding: An
Inquiry into Human Knowledge Structures. New York: Lawrence Erlbaum; Bretherton, I. (1987) New perspectives
on attachment relations: security, communication, and internal working models. In J.D. Osofsky (ed.) Handbook of
Infant Development, 2nd edn (pp.1061–100). New York: Wiley. There was also other work in the 1990s that helped
set the stage for the proposal of secure base scripts, e.g. Kirsh, S. & Cassidy, J. (1997) Preschoolers’ attention to and
memory for attachment relevant information. Child Development, 68, 1143–53.
369 Waters, T.E.A. & Facompré, C.R (in press) Measuring secure base script knowledge in the Adult Attachment
Interview. In E. Waters, B.E. Vaughn, & H.S. Waters (eds) Measuring Attachment. New York: Guilford.
370 Bowlby, J. (1985) Letter to John Byng-Hall, 12 April 1985. PP/Bow/J.9/45, discussing Byng-Hall, J. (1985) The
family script: a useful bridge between theory and practice. Journal of Family Therapy, 7(3), 301–305.
371 Waters, H.S. & Waters, E. (2006) The attachment working models concept: among other things, we build
script-like representations of secure base experiences. Attachment & Human Development, 8(3), 185–97.
372 For instance, routine definitions of ‘internal working models’ have increasingly made appeal to the con-
cept of script. E.g. Bakermans-Kranenburg, M.J., Juffer, F., & van IJzendoorn, M.H. (2019) Reflections on the
mirror: on video-feedback to promote positive parenting and infant mental health. In C. Zeanah (ed.) Handbook of
Infant Mental Health, 4th edn (pp.527–42). New York: Guilford.
373 Waters, T.E., Brockmeyer, S.L., & Crowell, J.A. (2013) AAI coherence predicts caregiving and care seeking
behavior: secure base script knowledge helps explain why. Attachment & Human Development, 15(3), 316–31;
Waters, T.E., Ruiz, S.K., & Roisman, G.I. (2017) Origins of secure base script knowledge and the developmental
construction of attachment representations. Child Development, 88(1), 198–209; Farrell, A.K., Waters, T.E.A.,
Young, E.S., et al. (2019) Early maternal sensitivity, attachment security in young adulthood, and cardiometabolic
risk at midlife. Attachment & Human Development, 21(1), 70–86.
Cross-cultural applicability of the Strange Situation 181
Waters and Waters conceptualised security as a single dimension, running counter to

the central position of Ainsworth’s three patterns of attachment in the research imagination
of other attachment researchers. For them, the capacity to make effective use of others as
a secure base and safe haven is much like a skill, and the effectiveness of skills is generally
measured on continua. In support for this position, a taxometric study of secure base script
knowledge in late adolescence and adulthood found a dimensional latent structure at both
ages.374 More generally, Waters and Waters also hold that the distinction between avoidant
and ambivalent/resistant attachment is yet to have adequate substantion from home obser-
vations beyond those of Ainsworth, or empirical yield in decades of research. Nonetheless,
over the coming years it can be anticipated that the conceptualisation and operationalisa-
tion of secure base scripts will be asked to incorporate some attention to the different ways
that secure base scripts may be disrupted, given the characteristic interest in categories of

socioemotional problems among attachment researchers (Chapter 3) and in clinical dis-
courses (Chapter 6). This will no doubt include examination of associations with the dif-
ferent Strange Situation categories. Within the secure base script measures themselves, it
may be that differences will be articulated and explored between the narratives of individ-
uals who expect only instrumental support, those who expect insensitive care and partial
access to a secure base/safe haven, and those who expect little or no secure base/safe haven
availability at all.
Cross-cultural applicability of the Strange Situation
Two traditions of cross-cultural research
In evaluating the legacy of Ainsworth and the assessment measures she introduced, and im-
portant domain to consider is their application to cross-cultural research. Looking back in
2016 on the decades of research using Ainsworth’s methods, Mesman, van IJzendoorn, and
Sagi-Schwartz observed that ‘the current cross-cultural database is almost absurdly small
compared to the domain that should be covered’.375 A central reason for this seems to have
been that the early attachment researchers failed to secure an alliance with anthropology.
Margaret Mead had famously been an early critic of Bowlby. The nub of their disagreement,
from Bowlby’s perspective, was that Mead seemed to be arguing that an infant cared for by
interchangeable caregivers within a village would have the same prospects of healthy psy-
chological development as an infant cared for by a small number of very familiar and cher-
ished people. In direct discussion at the World Health Organisation in the 1950s, Mead put
to Bowlby that the child would do fine with twenty different caregivers. Bowlby replied that
in general he did not take the view that children would be harmed by having multiple care-
givers; however, there were limits. Roughly equal care by twenty different caregivers would
be unlikely to offer a young child the basis for discriminating, cherished relationships with
374 Waters, T.E., Fraley, R., Groh, A., et al. (2015) The latent structure of secure base script knowledge.

375 Mesman, J., van IJzendoorn, M.H., & Sagi-Schwartz, S. (2016) Cross-cultural patterns of attachment: uni-
versal and contextual dimensions. In J. Cassidy & P. Shaver (eds) Handbook of Attachment. New York: Guilford
(pp.790–815), p.809.
182 Mary Ainsworth
particular familiar figures.376 This agrees with later findings from the anthropological litera-
ture, summarised recently by the social anthropologist Sara Harkness as the conclusion that
‘even in contexts of multiple caregiving, infants generally do not form close relationships
with more than a few individuals’.377
Bowlby and Ainsworth held that care by more than one person was not anticipated to ne-
cessarily disrupt the quality of the attachment relationships formed with these people. For
example, ‘a child cared for by several caregivers can, and frequently does, form as secure an
attachment to one figure, his mother, as a child who has a more exclusive relationship with
one figure’.378 Unfortunately, however, Bowlby’s impression of what was meant when anthro-
pologists spoke of ‘multiple caregiving arrangements’ appears to have been frozen at Mead’s
characterisation of twenty interchangeable people. The result was a bizarre and quite specific
blindspot. Bowlby was the consummate interdisciplinary researcher, drawing in knowledge

from across disciplines including behavioural biology, cybernetics, linguistics, neurology,
and epidemiology. Furthermore, in fact he read anthropological research regarding grief
and mourning with great interest, and reported the benefit he had gained from reading an-
thropological studies such as those by Raymond Firth, Geoffrey Gorer, David Mandelbaum,
Phyllis Palgi, and Paul C. Rosenblatt. He also expressed appreciation for anthropological
theory, such as the ideas of Durkheim and Malinowski.379
However, he tended to treat anthropologists who raised questions about his work as
holding the same stance as Mead. So, for instance, when the Harvard anthropologists Sarah
and Robert Levine came to talk to him in London about their research on multiple care-
giving arrangements in Nigeria, he was apparently rude and dismissive.380 He systemat-
ically neglected discussion of the role of multiperson interactions in shaping attachment,
since these were conflated with multiple caregiving.381 With cross-cultural differences
neglected, many relevant issues in his theory remained unresolved. For instance, despite
reading much relevant ethnographic research, he left unaddressed in his writings the ques-
tion of whether, if all human infants have the capacity for use of the caregiver as a secure
base and safe haven, all cultures could be anticipated to scaffold and utilise this capacity.
Sadly, Bowlby seems to have experienced anthropologists as insufficiently uninterested in
the nuances of his work to make it worthwhile working out subtlties of his theory in dia-
logue with anthropology.
376 Tanner, J.M. & Inhelder, B. (eds) (1956) Discussions on Child Development: Proceedings of the WTO Study
Group of the Psychobiological Development of the Child, Vol. 2. London: Tavistock, p.90. See also Bakermans-
Kranenburg, M.J., Steele, H., Zeanah, C.H., et al. (2011) Attachment and emotional development in institutional
care: characteristics and catch up. Monographs of the Society for Research in Child Development, 76(4), 62–91.
377 Harkness, S. (2015) The Strange Situation of attachment research: a review of three books. Reviews in
Anthropology, 44(3), 178–97, p.196.

378 Ainsworth, M. (1963) The development of infant–mother interaction among the Ganda. In B.M. Foss (ed.)
Determinants of Infant Behaviour (pp.67–104), Vol. 2. London: Methuen, p.95. Bowlby’s annotations on this
chapter are in an edition held by Richard and Xenia Bowlby.
379 Bowlby, J. (1980) Loss. New York: Basic Books, p.126. W.H.R. Rivers was also an early influence. Van Dijken,
S., van der Veer, R., van IJzendoorn, M., & Kuipers, H.J. (1998) Bowlby before Bowlby: the sources of an intellec-
tual departure in psychoanalysis and psychology. Journal of the History of the Behavioral Sciences, 34(3), 247–69.
380 LeVine, R.A. (2014) Attachment theory as cultural ideology. In H. Otto & H. Keller (eds) Different Faces of
Attachment: Cultural Variations on a Universal Human Need (pp.50–65). Cambridge: Cambridge University Press.
381 Bowlby’s concerns fed Ainsworth’s focus on dyadic interaction. In turn, Bowlby’s antipathy to multi-person
interactions and Ainsworth’s focus on dyadic interaction together formed a serious obstacle to applications or
extensions of attachment research beyond the dyad, including small group research—though see Ein-Dor’s work
discussed in Chapter 5. Likewise, though there are exceptions, multi-person caregiving arrangements have been
comparatively underresearched by attachment researchers. Keller, H., Bard, K., Morelli, G., et al. (2018) The myth
of universal sensitive responsiveness: comment on Mesman et al. (2017). Child Development, 89(5), 1921–8.
By contrast, both Bowlby and Ainsworth were very encouraging of anthropological study
when they were confident that the researcher did not hold that twenty interchangeable carers
would offer the basis for secure attachments. When Ainsworth’s student Bob Marvin wrote
reporting from his collaboration with Sarah and Robert LeVine, Bowlby described the work
as ‘interesting’ and ‘very valuable’.382 No doubt a basis for Bowlby’s different stance was that
the LeVines took their observations of attachment behaviour shown by infants to multiple
caregivers as a falsification of attachment theory. By contrast, Marvin emphasised observa-
tions, from the very same fieldwork, that when children were distressed, they nonetheless
still generally sought their most familiar adult figure. For Ainsworth, such issues brought out
a fundamental difference between anthropology and psychology as research paradigms: psy-
chological research was grounded in the potential for quantitative assessment of inter-rater
reliability in the study of behaviour, whereas anthropological research was based on meticu-

lous ethnographic observations without attempts to achieve reliability with other observers.
She was a stanch advocate for anthropological and qualitative observational methods within
psychology.383
However, Ainsworth was also mindful that the standing of the emergent attachment para-
digm in the positivist context of American academic psychology depended on assembling
a body of quantitative findings. Immediacy’s grip led the field away from precisely the trad-
ition of mixed-methods research that had been fundamental to Ainsworth’s own intellectual
development until the mid-1960s.384 As the Strange Situation classification became increas-
ingly taken for granted, and developmental psychology moved towards an increased focus
on large numbers as the basis for validity, it is now very rare even to find studies that inves-
tigate and report on the specific qualities of cases that run against the overall association.385
Danziger referred to this phenomenon across academic psychology as the ‘triumph of the
aggregate’.386 Insofar as it has been reflected in the priorities of researchers after Ainsworth,
Klaus and Karin Grossmann described the triumph of the aggregate in the decline of mixed-
methods inquiry as an inestimable loss to attachment research.387
However, in addition to differences in epistemology, a further factor hindering the devel-
opment of an alliance with anthropology was the unattractiveness of Ainsworth’s Strange
Situation for ethnographic fieldwork.388 It was laborious to train to code the measure. Any
382 Bowlby, J. (1975) Letter to Robert Marvin, 5 November 1975. PP/Bow/J.9/132.

383 Ainsworth, M. (1998) Harold Stevenson—SRCD oral history interview. http://srcd.org/sites/default/files/
documents/ainsworth_mary_interview.pdf.
384 A plea for mixed methods research on attachment has also been made by Rothbaum, F. & Morelli, G.
(2005) Attachment and culture: bridging relativism and universalism. In W. Friedlmeier, P. Chakkarath, & B.
Schwarz (eds) Culture and Human Development: The Importance of Cross-Cultural Research to the Social Sciences.
Lisse: Swets & Zeitlinger; and by Keller, H. (2018) Parenting and socioemotional development in infancy and early
childhood. Developmental Review, 50, 31–41. For a recent example of use of mixed-methods see Suchman, N.,
Berg, A., Abrahams, L., et al. (2019) Mothering from the inside out: adapting an evidence-based intervention for
high-risk mothers in the Western Cape of South Africa. Development & Psychopathology, 32(1), 105–22.
385 Exemplary contrary cases include Hobson, R.P., Patrick, M., Crandell, L., Garcia-Perez, R., & Lee, A. (2005)
Personal relatedness and attachment in infants of mothers with borderline personality disorder. Development &
Psychopathology, 17(2), 329–47; Kozlowska, K. (2010) Family-of-origin issues and the generation of childhood
illness. Australian and New Zealand Journal of Family Therapy, 31, 73–91; Tharner, A., Verhage, M.L., Oosterman,
M., & Schuengel, C. (2019) The case of attachment non-transmission: zooming in on the pathways through par-
ental sensitivity. Paper presented at International Attachment Conference, Vancouver, 20 July 2019.
386 Danziger, K. (1990) Constructing the Subject: Historical Origins of Psychological Research.
Cambridge: Cambridge University Press, p.88.

387 Grossmann, K. & Grossmann, K. (2012) Bindungen— das Gefüge psychischer Sicherheit Gebundenes.
Stuttgart: Klett-Cotta.
388 For a discussion of some of the requisites for successful alliance and collaboration between psychology and
anthropology see Azuma, H. (1996) Cross-national research on child development: the Hess–Azuma collaboration
184 Mary Ainsworth
graduate students already needing to pursue fieldwork as part of their doctorate could quite
reasonably be concerned by the additional time and uncertainty of seeking to gain reliability
in Ainsworth’s coding system. Furthermore, Ainsworth’s tripartite division was regarded
as rather crude as an attempt to capture infant experience and care practices in their par-
ticularity. The measure may have relevance to understanding individual differences within
a culture, but this is only a minor goal within anthropology, which has generally been more
concerned with interpreting social practices and meanings.389 Anthropologists, often sus-
pended between two worlds as individuals or with their families, were also specially aware
from personal experience that separations and reunions have the potential for different
meanings in different cultural contexts. Furthermore, social and cultural anthropology es-
pecially, at least since the 1980s, has had a general antipathy to claims that appear universal-
istic, and all the more so when this universalism appears value-laden.390 To the extent that

attachment discourses seemed to prize security above insecurity, with stakes for how chil-
dren and families were evaluated, the paradigm appeared in conflict, quite fundamentally,
with a premise of social and cultural anthropology.391
Practically, it was also truly cumbersome to conduct the Strange Situation outside of the
laboratory.392 Furthermore, in some cultures, it was regarded as potentially transgressive or,
at least, quite rude for researchers to separate infants and caregivers.393 The Ainsworth sen-
sitivity scale might have been considered an alternative in the 1980s and 1990s. Ainsworth
had reported such a strong correlation between maternal sensitivity and infant security that
they could have been regarded as much the same construct. However, regrettably, the sen-
sitivity scale remained unpublished. It was by far eclipsed by the Strange Situation, which as
a standardised laboratory-based procedure was a more rhetorically useful source of cred-
ibility for the emergent attachment research paradigm than the sensitivity scale within the
psychological research community in North America and Europe, even if it had reduced
utility outside of it. With the sensitivity scale not even brought to market, researchers inter-
ested in examining the role of sensitivity in care across cultures were forced either to develop
in retrospect. In D.W. Shwalb & B.J. Shwalb (eds) Japanese Childrearing: Two Generations of Scholarship (pp.220–
40). New York: Guilford.
389 Keller, H. (2008) Attachment—past and present. But what about the future? Integrative Psychological and
Behavioral Science, 42, 406–15. A good illustrative case of the use of the idea of ‘attachment’, but not attachment re-
search, for interpreting social practices and meanings is Lowe, E.D. (2002) A widow, a child, and two lineages: ex-
ploring kinship and attachment in Chuuk. American Anthropologist, 104(1), 123–37. For reflection on the different
aims of interpretive and experimental forms of research see Reddy, W.M. (2014) Humanists and the experimental
study of emotion. In F. Biess & D.M. Gross (eds) Science and Emotions after 1945: A Transatlantic Perspective
(pp.41–66). Chicago: University of Chicago Press. The social psychological tradition of attachment research
(Chapter 5) has likewise seen little uptake within anthropology, sociology, human geography, or cognate social
sciences concerned primarily with practices. An exception is Quinlan, R.J. & Quinlan, M.B. (2007) Parenting and
cultures of risk: a comparative analysis of infidelity, aggression, and witchcraft. American Anthropologist, 109(1),
164–79. However, the social psychological tradition has also seen no critique from anthropologists. Presumably it
has fallen under the blanket of ‘developments after Ainsworth’, which anthropologists have tended to ignore.
390 Eriksen, T.H. & Nielsen, F.S. (2001) A History of Anthropology. London, Pluto Press.
391 See e.g. Keller, H. (2018) Universality claim of attachment theory: children’s socioemotional develop-
ment across cultures. Proceedings of the National Academy of Sciences, 115(45), 11414–19; Keller, H., Bard, K.,
Morelli, G., et al. (2018) The myth of universal sensitive responsiveness: comment on Mesman et al. (2017). Child
392 See e.g. Zevalkink, J., Riksen-Walraven, J.M., & Van Lieshout, C.F. (1999) Attachment in the Indonesian
caregiving context. Social Development, 8(1), 21–40; True, M., Pisani, L., & Oumar, F. (2001) Infant–mother at-
tachment among the Dogon of Mali. Child Development, 72(5), 1451–66.
393 Otto, H. (2014) Don’t show your emotions! Emotion regulation and attachment in the Cameroonian Nso. In
H. Otto & H. Keller (eds) Different Faces of Attachment: Cultural Variations of a Universal Human Need (pp.215–
29). Cambridge: Cambridge University Press.
their own scales, or to rely on dubious proxies for sensitivity such as household size. As a re-
sult, two research traditions developed in the 1980s and 1990s: developmental psychologists
using measures developed in America in other countries; and qualitative ethnographic stud-
ies that eschewed these measures. Each represented a segregated part of Ainsworth’s own
biographical journey, which had traversed both ethnography and laboratory science.
Cross-cultural studies in the 1980s and 1990s using standardised attachment measures
were generally conducted by attachment researchers or developmental psychologists with
some personal contact with Ainsworth or her students—with the signal exception of the
Japanese studies (discussed in the section ‘The Strange Situation “abroad” ’).394 An early
example was Kermoian and Leiderman (1986), two psychological researchers who con-
ducted a study of 26 Gusii infants from Kenya. The Strange Situation was adapted in several
thoughtful ways, such as by taking place outside the mother’s hut and by altering the reunion

so that infants would be greeted with, as they would expect, the ritual handshake of greeting.
The procedure was used both with the mother and with the person who most frequently
cared for the infant during the day (generally a sibling). The coding was conducted by re-
searchers trained by Main. Kermoian and Leiderman found that 61% of infants had a secure
attachment classification with their mother, and 54% with their daycarer. Whereas security
of attachment with mother was related to nutritional status, security of attachment with the
daycarer was related to infant cognitive and motor development.395 The researchers con-
cluded that the pervasive association between infant–mother attachment and infant func-
tioning which has been identified in American samples may be a reflection of the diversity
of activities in which American infants and mothers jointly engage, in contrast to the Gusii
where a high proportion of infant–mother interaction centres around feeding, and a high
proportion of infant–daycarer interaction centred around exploration and daily tasks.
A significant limitation of the cross-cultural studies conducted by psychological re-
searchers has been that only on rare occasions did these researchers accompany their re-
search with ethnography, and generally only when the Strange Situation had not proved
workable. A clear exception is Germán Posada and colleagues in their work on the sensitivity
construct and secure base behaviour, but this has proved unusual.396 Whilst attachment
researchers may personally have read ethnographic research in situating their study, their
write-up in psychology journals has rarely entered into dialogue with anthropologists. Nor
would this have likely been rewarded by psychology journals or their reviewers. Ultimately,
the ambition of these studies has been to demonstrate that the Strange Situation could tap
meaningful variation in individual differences across different contexts, and to examine
the role of culture in moderating the influence of caregiving on child attachment as repre-
sented by the distribution of Strange Situation classifications. The ambitions of attachment
researchers in using the Strange Situation cross-culturally have been therefore, at best, only
394 E.g. van IJzendoorn, M.H. & Kroonenberg, P.M. (1988) Cross- cultural patterns of attachment: a meta-
analysis of the strange situation. Child Development, 59, 147–56; Grossmann, K.E., Grossmann, K., & Keppler,
A. (2005) Universal and culture-specific aspects of human behavior: the case of attachment. In W. Friedlmeier, P.
Chakkarth, & B. Schwarz (eds) Culture and Human Development: The Importance of Cross-Cultural Research for the
Social Sciences (pp.75–97). New York: Psychology Press.
395 Kermoian, R. & Leiderman, P.H. (1986) Infant attachment to mother and child caretaker in an East African
community. International Journal of Behavioral Development, 9(4), 455–69.

396 Posada, G., Carbonell, O.A., Alzate, G., & Plata, S.J. (2004) Through Colombian lenses: ethnographic and
conventional analyses of maternal care and their associations with secure base behavior. Developmental Psychology,
40(4), 508–18; Posada, G. (2013) Piecing together the sensitivity construct: ethology and cross-cultural research.
Attachment & Human Development, 15(5–6), 637–56.
186 Mary Ainsworth
marginally aligned with the aim of anthropology to understand cultural practices. There
have been few discussions of how the Strange Situation and its coding might best be adapted
to account for cultural context and to offer insight into cultures of caregiving.397
As such, the accusation by the anthropologist Heidi Keller that ‘the only dimension that
attachment researchers have recognised as cultural is the distribution of the attachment qual-
ities’398 has an unfortunate degree of truth. Cross-cultural research was not especially well
aligned with developmental psychology as a disciplinary space in the 1980s and 1990s: there
are few funders who support cross-cultural psychology, and few rewards within the academic
community for the slower yield of publications this research strategy generally entails.399
Ethnographic work ahead of a quantitative study would be possible, but would risk being pen-
alised by developmental science journals, which look down on qualitative methods even for
exploratory work. Attachment researchers have largely had to furnish their own evidence of

cross-cultural relevance and, with few allies to take this work forward, the rate of publication
has been slow. Furthermore, attachment researchers are not trained to conduct qualitative
research or to draw on social and anthropological theories. Though there are certainly ex-
ceptions, this state of affairs contributes to a tendency for attachment researchers to use the
Strange Situation and related measures without the empirical or conceptual work (at least in
print) to examine whether or how these can be used in a valid way within a particular culture,
or to develop testable hypotheses about the role of culture for variations in the functioning of
the attachment and caregiving systems. This is precisely what collaboration with anthropology
could have facilitated, if there had been appetite, funds, and/or satisfaction of mutual interests.
On the other hand, a second research tradition developed of qualitative studies engaging
with issues of attachment. Sometimes this was conducted by developmental psycholo-
gists,400 but by the 1990s the tradition was sustained almost exclusively by professional an-
thropologists. This anthropological literature has had three particular markers. A first is
that it was cultural, not biological anthropology that took an interest in attachment. In fact,
with James Chisholm as an exception,401 there has been little sustained interest in attach-
ment among biological anthropologists.402 A second characteristic of this anthropological
397 Exceptions include Marvin, R.S., VanDevender, T.L., Iwanaga, M.I., LeVine, S., & LeVine, R.A. (1977) Infant–
caregiver attachment among the Hausa of Nigeria. In H.M. McGurk (ed.) Ecological Factors in Human Development
(pp.247–60). Amsterdam: North-Holland; Crittenden, P.M. & Claussen, A.H. (eds) (2000) The Organisation of
Attachment Relationships. Cambridge: Cambridge University Press; Thompson, R.A. (2017) Twenty-first century
attachment theory. In H. Keller & K. Bard (eds) The Cultural Nature of Attachment: Contextualizing Relationships
and Development (pp.301–19). Cambridge, MA: MIT Press.
398 Keller, H. (2013) Attachment and culture. Journal of Cross-Cultural Psychology, 44(2), 175–94, p.180.
399 The collaboration between Bob Marvin and the LeVines stands as an apparent exception to the division be-
tween these two traditions, even if it was not feasible to bridge the division completely through use of the Strange
Situation. However, it is striking also as a somewhat isolated case, at least until the past decade. It also illustrates
well the difficulties and lack of professional reward for such work. See Marvin, R.S., VanDevender, T.L., Iwanaga,
M.I., LeVine, S., & LeVine, R.A. (1977) Infant–caregiver attachment among the Hausa of Nigeria. In H.M. McGurk
(ed.) Ecological Factors in Human Development (pp.247–60). Amsterdam: North-Holland.
400 E.g. Brazelton, T.B. (1972) Implications of infant development among the Mayan Indians of Mexico. Human
Development, 15(2), 90–111; Tronick, E.Z., Morelli, G.A., & Ivey, P.K. (1992) The Efe forager infant and toddler’s
pattern of social relationships: multiple and simultaneous. Developmental Psychology, 28(4), 568.
401 Chisholm, J.S. (1996) The evolutionary ecology of attachment organization. Human Nature, 7(1), 1–
37; Chisholm, J. (2003) Uncertainty, contingency and attachment: a life history theory of theory of mind.
In K. Sterelny & J. Fitness (eds) From Mating to Mentality: Evaluating Evolutionary Psychology (pp.125–54).
Hove: Psychology Press.
402 A signal exception is work by Belsky and colleagues on early attachment experiences as priming regarding
the need for long-term or short-term focused reproductive strategies. This concept has generated substantial
interest among biological anthropologists. Belsky, J., Steinberg, L., & Draper, P. (1991) Childhood experience,
literature was that the quality of the ethnography is high, and publications such as those of
Nancy Scheper-Hughes have become classics of the anthropological literature in general.
A third was that researchers have by and large displayed little knowledge of developments
in attachment research since the 1980s; with some exceptions, their conversation has al-
most exclusively been with the ideas of Bowlby and Ainsworth, and especially Bowlby’s
writings.403 And like Bowlby’s earlier writings, they have tended—with exceptions404—to
assume that care relationships are attachment relationships, without consideration of the
attachment-specific qualities such as directional crying and preferential seeking suggested
by the theory.
Researchers after Bowlby and Ainsworth are at times treated as mute followers of these
founding figures by their anthropologist critics, rather than critical contributors to a living,
branching tradition of theory and research. The work of Main, for example her discussion of

conditional strategies, appears mostly unknown, or known only secondhand, in anthropo-
logical discussions of attachment.405 Germán Posada’s studies have likewise been ignored.
In Scheper-Hughes’ work, for example, ‘attachment’ was used to refer mostly to the care-
giving system, and Bowlby was interpreted in caricature as an instinct theorist in part as a
framing device through which the importance of economic and contextual factors can be
highlighted in shaping care practice.406 Bowlby was used rhetorically to represent ‘biology’,
against anthropology’s ‘culture’. As a consequence, differences between the disciplines and
their goals have combined with poor mutual impressions. Research psychologists seem to
hold an impression of anthropologists who discuss attachment as ill-informed and appar-
ently wilfully uninterested in contemporary attachment research—or as implacably hos-
tile, without openings for discussion of how to better conduct research in developmental
science.407
In a debate at the Leipzig Research Center for Early Child Development in November
2018, Ross Thompson argued that anthropologists have not recognised how the field has
evolved since Bowlby and Ainsworth. Keller replied that this is essentially irrelevant, since
interpersonal development, and reproductive strategy: an evolutionary theory of socialization. Child Development,
62, 647–70.
403 Though a critic of the tradition of attachment research located within developmental science, Heidi Keller
has been an important figure in seeking bridges between disciplines, and in updating the ‘working model’
held of attachment researchers by anthropologists. See e.g. Otto, H. & Keller, H. (eds) (2014) Different Faces of
Attachment: Cultural Variations on a Universal Human Need. Cambridge: Cambridge University Press. A hearten-
ing sign of a more serious interaction between developmental science and anthropology is the debate between Judy
Mesman and Heidi Keller and colleagues, though the tendency towards talking past one another remains in evi-
dence. See Keller, H., Bard, K., Morelli, G., et al. (2018) The myth of universal sensitive responsiveness: comment
on Mesman et al. (2017). Child Development, 89(5), 1921–8.
404 E.g. Meehan, C.L. & Hawks, S. (2015) Multiple attachments: allomothering, stranger anxiety, and intimacy.
In H. Otto & H. Keller (eds) Different Faces of Attachment: Cultural Variations on a Universal Human Need
405 See e.g. Quinn, N. & Mageo, J. (eds) (2013) Attachment Deconsidered: Cultural Perspectives on a Western
Theory. London: Palgrave; Morelli, G. (2015) The evolution of attachment theory and cultures of human attach-
ment in infancy and early childhood. In L.E. Jensen (ed.) The Oxford Handbook of Human Development and
Culture: An Interdisciplinary Perspective (pp.149–64). Oxford: Oxford University Press.
406 Scheper-Hughes, N. (1993) Death Without Weeping: The Violence of Everyday Life in Brazil.
Berkeley: University of California Press. Similar issues can be found in later anthropological works, even the other-
wise excellent Gottlieb, A. (2004) The Afterlife Is Where We Come From. Chicago: University of Chicago Press.
407 Though, unusually, collaboration and coauthorship between attachment researchers and anthropolo-
gists can be seen in Keller, H. (eds) (2017) The Cultural Nature of Attachment: Contextualizing Relationships and
Development. Cambridge, MA: MIT Press.
188 Mary Ainsworth
the version of attachment theory in wider circulation is that of Bowlby and Ainsworth, and
subsequent researchers have not raised their voices to correct this account:
The real problem I have is that attachment theory in the applied field is really causing,
to put it mildly, a lot of distress because children are evaluated according to attachment
theory . . . I know that you would never subscribe to such a view. But what I’m missing is
why attachment researchers don’t form a louder voice in order to distance themselves from
these appearances.408
Reflecting on a symposium on anthropology and attachment research held in Frankfurt the

previous year, Thompson expressed dismay that the only alternative hypothesis the anthro-
pologists seemed interested to present was that ‘cultures vary’. Anthropologists’ dismissal

of so much of attachment research methodology and theory had, in Thompson’s view, left
nothing but critique. Thompson expressed the sincere wish that anthropologists had sought
to be more constructive: ‘That cultures vary is not a hypothesis; it is a truism’. He urged that
critics seek to join the conversation with attachment researchers by making their criticisms
relevant to hypothesis-generation, for instance about when sensitivity is and is not so rele-
vant.409 In this way, attention to universal processes and culturally specific processes could
be brought together. Another example would be the Klaus and Karin Grossmann’s reflec-
tions on forms of attachment avoidance that have a basis more in cultural factors than in
insensitive care, and which may therefore have different correlates.
Keller, however, shot back at Thompson that attachment researchers only tend to
recognise hypotheses if they relate narrowly to the ‘gold standard’ quantitative meas-
ures.410 Furthermore, she alleged that hypotheses may not be recognised by attachment
researchers unless they are made by an insider. Writing in a book co-edited by Keller,
Weisner also observed that anthropologists feel deliberately ignored by developmental
attachment research, which seems implacable in response to their concerns about
Bowlby and Ainsworth’s ideas, and the Strange Situation procedure in particular, con-
tributing to a sense of frustration among anthropologists.411 Whether this frustration is
the anger of hope or the anger of despair is unclear. The overall result of interactions be-
tween attachment researchers and anthropologists from Bowlby to the present has been
weak common ground or basis for mutual curiosity, as neither tradition treats the work
of the other as engaging or valuable. Attachment researchers have mostly felt that they
had to go it alone in pursuing cross-cultural quantitative research without support from
anthropologists in adjusting or calibrating their measures, in the design of studies, or in
interpreting results.
Wundt Dialogue’, Leipzig University, 28 November 2018, hosted by the Leipzig Research Center for Early Child
409 Cf. Aviezer, O., Sagi-
Schwartz, A., & Koren-Karie, N. (2003) Ecological constraints on the formation
of infant–mother attachment relations: when maternal sensitivity becomes ineffective. Infant Behavior and
Wundt Dialogue’, Leipzig University, 28 November 2018, hosted by the Leipzig Research Center for Early Child
411 Weisner, T. (2014) The socialization of trust: plural caregiving and diverse pathways in human development
across cultures. In H. Otto & H. Keller (eds) Different Faces of Attachment (pp.263–77). Cambridge: Cambridge
University Press.
The Strange Situation ‘abroad’
The founding work of cross-cultural research conducted by attachment researchers was the
Bielefeld study by Klaus and Karin Grossmann, with results published in 1981. This study
sent shockwaves through the small community of attachment researchers, as well as the wider
community of developmental science researchers. The distribution of Strange Situation
classifications differed markedly from those of Ainsworth, with more avoidant than secure
dyads in the sample. This result was interpreted in terms of the aversion of German culture to
displays of distress and the importance placed on independence, reflected in childcare prac-
tices that promoted infant self-reliance such as separate sleeping, and that penalised com-
munication of anxiety by children.412 The study became a conventional reference point, cited

in reviews and textbooks, illustrating the limitations of the Strange Situation. In fact, how-
ever, subsequent studies in Germany have reported distributions that align well with both
Ainsworth’s distribution and other North American, European, and Australian samples. In
a second sample, from Regensburg, the Grossmanns found that 62% of the dyads received a
secure classification, 27.5% an avoidant classification, 5% an ambivalent/resistant classifica-
tion, and a further 5% that could not be classified into the Ainsworth categories.413 Another
study in Berlin found that 77.5% of dyads in the sample were classified as secure, 17.5% were
classified as avoidant, and 5% were classified as ambivalent/resistant.414
It has later been assumed that the Bielefeld findings could be explained by differences
in caregiving practices characteristic of northern Germany.415 Certainly, a subsequent ana-
lysis by the Grossmanns found that a proportion of the infants from avoidant dyads had
received otherwise highly sensitive care from parents who experienced cultural pressure to
encourage self-reliance in their children. In a later follow-up, these infants who had experi-
enced sensitive care had outcomes equivalent to those from securely attached dyads, and
unlike their fellow avoidantly attached dyads.416 Such later findings, however, have generally
412 Grossmann, K.E., Grossmann, K., Huber, F., & Watner, U. (1981) German children’s behavior towards their
mothers at 12 months and their fathers at 18 months in Ainsworth’s Strange Situation. International Journal of
Behavioral Development, 4(2), 157–81.
413 Wartner, U.G., Grossmann, K., Fremmer-Bombik, E., & Suess, G. (1994) Attachment patterns at age six in
south Germany: predictability from infancy and implications for preschool behavior. Child Development, 65(4),
1014–27. With the D classification included, the distributions were 50% B, 15% A, 4.5% C, and 30.5% D.
414 Beller, E.K., & Pohl, A. (1986) The Strange Situation revisited. Paper presented at 4th International
Conference on Infant Studies, Beverly Hills, April 1986. Distribution of attachment classifications reported in
van IJzendoorn, M.H. & Kroonenberg, P.M. (1988) Cross-cultural patterns of attachment: a meta-analysis of the
strange situation. Child Development, 59(1), 147–56. Even in a sample of German children known to social ser-
vices for potential child abuse and neglect, rates of secure attachment have been found to be high at 24 months.
Suess, G.J., Bohlen, U., Carlson, E.A., Spangler, G., & Frumentia Maier, M. (2016) Effectiveness of attachment
based STEEP™ intervention in a German high-risk sample. Attachment & Human Development, 18(5), 443–60.
415 However, empirical study of parents’ perceptions of the appropriate attachment behaviour of young chil-
dren showed few differences between north and south Germany: Scholmerich, A. (1996) Attachment security
and maternal concepts of ideal children in northern and southern Germany. International Journal of Behavioral
Development, 19(4), 725–38. An alternative explanation for the early independence encouraged in the Bielefeld
infants might be found in war trauma experienced by the parents in the Bielefeld sample. Bielefeld was heavily
bombed during World War II. By contrast, though the Messerschmitt aircraft factory and the oil refinery nearby
were attacked, the town of Regensburg itself received little bombing. On German war trauma and attachment pro-
cesses see the discussion in Kaiser, M., Kuwert, P., Braehler, E., & Glaesmer, H. (2018) Long-term effects on adult
attachment in German occupation children born after World War II in comparison with a birth-cohort-matched
representative sample of the German general population. Aging & Mental Health, 22(2), 197–207.
416 Grossmann, K.E., Grossmann, K., & Schwan, A. (1986) Capturing the wider view of attachment: a reanalysis
of Ainsworth’s Strange Situation. In C.E. Izard & P.B. Read (eds) Measuring Emotions in Infants and Children, Vol. 2
190 Mary Ainsworth
been ignored, except among attachment researchers. The salience of the early Bielefeld find-
ings, and their resonance with contemporary stereotypes about Germans as emotionally
suppressed but secretly insecure, have held the imagination: the ‘German’ tendency towards
avoidant attachment is still widely cited by both developmental psychologists417 and critics
of the attachment paradigm.418
The Grossmanns were trained to conduct the Strange Situation by Ainsworth, and were
given support in coding the procedure by both Ainsworth and Main. By contrast, applica-
tions of the Strange Situation by Japanese researchers were the first to be conducted by a
group without even distal ties to Ainsworth. A first study, published in 1984, was carried out
in Tokyo by Durrett and colleagues. Of their 39 infant–caregiver dyads, 61% were classified
as secure, 13% were classified as avoidant, 18% were classified as ambivalent/resistant, and
8% could not readily be classified into one of the Ainsworth classifications. The researchers

found that rates of security were higher among dyads where the parent experienced more
social support. These results generated little interest or discussion: they seemed merely to
confirm the status of secure attachment behaviour as the most common pattern, and that
it was associated with theoretically expectable antecedents. By contrast, a second Japanese
study by Takahashi was conducted in Sapporo and published in 1986; 68% of the sample
of 60 infant–mother dyads were classified as secure, 32% were classified as ambivalent/re-
sistant, and not a single infant was classified as avoidant.419
In interpreting these findings, Takahashi drew a contrast between the common occur-
rence of minor infant–mother separations in the American context, and the rarity of such
events in the lives of infants in traditional Japanese families, who generally experienced
co-sleeping, co-bathing, and being carried on their mother’s back. Takahashi emphasised
that a three-minute separation is not a standardised experience, but one shaped by culture.
For infants who have rarely, if ever, experienced separation from their mothers, the Strange
Situation may induce panic rather than serve as a mild stressor, and so fail to reflect experi-
ences in naturalistic settings. Considering these questions, Takahashi shared her cases with
Sroufe at Minnesota (Chapter 4). Takahashi and Sroufe agreed that the Strange Situation was
inappropriate for children who had so rarely experienced separations.420 They also agreed
417 E.g. Simonelli, A., De Palo, F., Moretti, M., Baratter, P.M., & Porreca, A. (2014) The Strange Situation pro-
cedure: the role of the attachment patterns in the Italian culture. American Journal of Applied Psychology,
3(3), 47–56.
418 E.g. LeVine, R.A. (2014) Attachment theory as cultural ideology. In H. Otto & H. Keller (eds) Different Faces
of Attachment: Cultural Variations on a Universal Human Need (pp.50–65). Cambridge: Cambridge University
Press. See also Grossmann, K.E. & Grossmann, K. (1999) Mary Ainsworth: our guide to attachment research.
Attachment and Human Development, 1, 224–8: ‘Our study, though, became known more because of the high
percentage of insecure-avoidantly attached infants (Grossmann, Grossmann, Huber, Wartner,1981) than for its
many confirmations of Ainsworth’s findings with a larger sample’ (224). Another important early cross-cultural
study of the 1980s was work by Sagi and Lamb exploring the attachment classifications of infants raised on Israeli
kibbutzim with their mother, father, and communal caretakers. This research was of particular importance in
highlighting the important contribution made by attachment figure overnight availability to security as assessed
in the Strange Situation. Sagi, A., Lamb, M.E., Lewkowicz, K.S., Shoham, R., Dvir, R., & Estes, D. (1985) Security
of infant–mother, –father, and –metapelet attachments among kibbutz-reared Israeli children. Monographs of the
Society for Research in Child Development, 50(1/2), 257–75; Sagi, A., van IJzendoorn, M.H., Aviezer, O., Donnell,
F., & Mayseless, O. (1994) Sleeping out of home in a kibbutz communal arrangement: it makes a difference for
infant–mother attachment. Child Development, 65(4), 992–1004.
419 Takahashi, K. (1986) Examining the Strange Situation procedure with Japanese mothers and 12-month-
old infants. Developmental Psychology, 22(2), 265–70. In a later report on the same sample, 75% were classified
as secure, 21% ambivalent/resistant, and 4% unclassifiable. Nakagawa, M., Lamb, M.E., & Miyake, K. (1992)
Antecedents and correlates of the strange situation behavior of Japanese infants. Journal of Cross-Cultural
Psychology, 23, 300–10.
420 Alan Sroufe, personal communication, January 2019.
that the apparently high rate of ambivalent/resistant infants did not reflect the predomin-
ance of this pattern of attachment, and was instead a misclassification of overdistressed in-
fants. It was noteworthy that the play of these infants was not poor prior to the separations,
as is the usual case for the anxious/resistant group. However, Sroufe states that Takahashi
was placed under institutional pressure to claim that the findings cast doubt on the cross-
cultural applicability of the Strange Situation in general. In her write-up she concluded that
the Ainsworth Strange Situation was a culturally specific artefact, with poor cross-cultural
applicability at least to traditional Japanese infant–caregiver dyads.421 This finding stirred
considerable attention. As Behrens subsequently observed, the Sapporo study findings res-
onated with a trend in social scientific research in the 1980s to emphasise the uniqueness
of Japan, and the lack of relevance of research paradigms developed on non-Japanese sam-
ples.422 Together with the Bielefeld study, the Sapporo study appeared to provide evidence of

vast differences in caregiving practices, or of the lack of cross-cultural validity of the Strange
Situation procedure, or both.
However, an additional option could be variance in how the Strange Situation procedure was
applied. Takahashi stated in the methodology for her paper that ‘as the original study (Ainsworth
et al. 1978) didn’t clearly describe when the episodes of distress were curtailed, we set the max-
imum duration at 2 min. Thus the durations of distress in this study were longer than in most
current studies in the United States.’423 This statement illustrates the distance of the Sapporo
researchers from Ainsworth: not only did they not consult with her, but also they had a weak
grasp of Patterns of Attachment, since the instruction to curtail episodes if the child becomes dis-
tressed is given clearly and repeatedly (on pages 35, 38, 39, 75, and 341). It is also mentioned in
Ainsworth’s other publications,424 and in publications by other American and European attach-
ment researchers in the 1970s.425 Furthermore, the coders of the Sapporo Strange Situations had
received no training or guidance from previous researchers who had used the coding protocols.
In acknowledgement of the methodological flaws in the study, Klaus and Karin Grossmann were
invited to Hokkaido University, Sapporo, and recoded the videos. Unfortunately, the results of
this reanalysis were printed only in the Annual Report of the Center for Child Development at
Hokkaido University, and so, in practice, were not in public circulation. The results of their re-
analysis are, however, interesting and important.
On review of the recordings, the Grossmanns affirmed that the separations had been con-
tinued too long. In line with Ainsworth’s instructions in print, no child was permitted to
cry for more than 30 seconds in either of the Grossmanns’ samples. In the Sapporo sample,
all infants were left to cry for at least 55 seconds and many for much longer: ‘some infants
cried as long as 4 minutes and 40 seconds in extreme despair’.426 Many infants could do little
421 Takahashi, K. (1990) Are the key assumptions of the ‘Strange Situation’ procedure universal? A view from
Japanese research. Human Development, 33(1), 23–30.

Online Readings in Psychology and Culture, 6(1), 7.

423 Takahashi, K. (1986) Examining the Strange Situation procedure with Japanese mothers and 12-month-old
infants. Developmental Psychology, 22(2), 265–70, p.266.

424 E.g. Ainsworth, M.D.S. & Bell, S.M. (1970) Attachment, exploration, and separation: illustrated by the be-
havior of one-year-olds in a strange situation. Child Development, 41(1), 49–67.

425 Serafica, F.C. & Cicchetti, D. (1976) Down’s syndrome children in a strange situation: attachment and explor-
ation behaviors. Merrill-Palmer, 22(2), 137–50; Smith, L. & Martinsen, H. (1977) The behavior of young children
in a strange situation. Scandinavian Journal of Psychology, 18(1), 43–52.
426 Grossmann, K.E. & Grossmann, K. (1989) Preliminary observation on Japanese infants’ behavior in
Ainsworth’s strange situation. Hokkaido University Annual Report, Research and Clinical Center for Child
192 Mary Ainsworth
but engage in exhausted sobbing through the second reunion rather than respond to the re-
appearance of their mother. This behaviour characterised 76% of the whole Japanese sample
left to cry intensely for more than two minutes.427 Klaus and Karin Grossmann identified
another factor that may have affected the distribution of classifications. Watching the videos,
they observed that the mothers were shy and formal in the laboratory setting, and barely
communicated with their infants while in the Strange Situation. A large minority rejected
their infant’s wish for contact, whereas this behaviour was shown by none of the German
parents, precisely contrary to stereotype.428 The Grossmanns observed that the Japanese in-
fants seemed surprised by the inaccessibility of their caregiver, and this may have encour-
aged the infants to intensify signals of distress and anger: ‘the infants showed through their
behaviour that they expected acceptance from their mothers’, implying that the withdrawn
behaviour was out of keeping with their usual expectations.429 The Grossmanns concluded

that the instructions given to the mothers had been interpreted by some as mandating with-
drawn behaviour, when this was in fact uncharacteristic of these dyads.
Whereas Takahashi reported 68% of the sample as secure and 32% as ambivalent/re-
sistant, the Grossmanns’ blind recoding of the sample yielded a distribution of 76% secure,
11% ambivalent/resistant, 2% avoidant, and 11% unclassifiable dyads. They also observed
conflict behaviours—what Main and Solomon would call indices of disorganised attach-
ment (Chapter 3)—in a very substantial proportion of both the Group C and the Group B in-
fants, but did not make a systematic report on this.430 They agreed with the original coding in
only 43% of cases originally coded as Group C. The Grossmanns suspected that the long sep-
arations had blown out the avoidant conditional strategy, which requires the redirection of
attention away from attachment-relevant stimuli. Such redirection is only possible at mod-
erate arousal, not at the high arousal of the infants in the Sapporo study.431 The first study
using the Strange Situation in Japan, by Durrett and colleagues, had a distribution far closer
to that of North American and European samples, though with somewhat fewer avoidant
dyads and somewhat more ambivalent/resistant dyads. This suggested to the Grossmanns
that cultural differences in caregiving might be playing a role in the distribution of condi-
tional strategies over and above overstress caused by the procedural issues. In line with this
assumption, they found that more Japanese Group C infants (50%) than German Group C
infants (10%) cried immediately as separation began. However, complicating the picture,
they also found that on average the Japanese sample cried less than the German sample as
separations began, suggesting that the sample as a whole was not necessarily overwhelmed
by separations per se.432
427 Grossmann, K., Fremmer-Bombik, E., & Freitag, M. (1991) German and Japanese infants in the Strange
Situation: are there differences in behavior beyond differences in the frequency of classes? Paper presented at
meeting of International Society for the Study of Behavioral Development, Minneapolis, July 1991.Copy shared by
Karin Grossmann.
428 Ibid.
429 Grossmann, K.E. & Grossmann, K. (1996) Kulturelle perspektiven der bindungsentwicklung Japan in und
Deutschland. In G. Trommsdorff & H.-J. Konrad (eds) Gesellschaftliche und individuelle Entwicklung Japan in und
Deutschland (pp.215–35). Konstanz: Universitätsverlag Konstanz.
430 Grossmann, K.E. & Grossmann, K. (1989) Preliminary observation on Japanese infants’ behavior in
Ainsworth’s strange situation. Hokkaido University Annual Report, Research and Clinical Center for Child
431 Grossmann, K., Fremmer-Bombik, E., & Grossmann, K.E. (1990) Familiar and unfamiliar patterns of attach-
ment of Japanese infants. Hokkaido University Annual Report, Research and Clinical Center for Child Development,
2, 30–39.
432 Grossmann, K., Fremmer-Bombik, E., & Freitag, M. (1991) German and Japanese infants in the Strange
Situation: are there differences in behavior beyond differences in the frequency of classes? Paper presented at
Debate about the Sapporo study went quiet for a decade, until the matter was revived
by Rothbaum, Miyake (one of the collaborators on the Sapporo study), and colleagues in
2000. In a high-profile paper in the American Psychologist they repeated Takahashi’s earlier
claims that the Sapporo data showed that the Strange Situation is not cross-culturally valid
as an assessment of individual differences.433 Like Takahashi, they pointed to the prolonged
skin-to-skin contact and the pre-emption of needs experienced by Japanese infants com-
pared to the distal interactions of western infants, which they supposed would make any
separations unbearable for most Japanese infants. Rothbaum and colleagues also drew on
the tradition of qualitative ethnographic research to propose that Japanese caregivers value
signs of infant dependency over displays of autonomy, and that this would account for the
higher numbers of ambivalent/resistant and the fewer avoidant dyads. They could point to
no quantitative findings in support of this claim, however, and the only direct study showed

the opposite: that Japanese caregivers value dependency far less than the infant using the
caregiver as a safe base from which to explore.434 Rothbaum and colleagues dismissed this
study, however, since it was from Tokyo and they presumed that the findings therefore re-
flected the western values of a metropolitan capital city.435
To appraise the claims of Rothbaum and colleagues, Kondo-Ikemura and colleagues con-
ducted another study in Sapporo in the 2010s. The Strange Situation procedure was carried
out in line with Ainsworth’s protocols. They found that 69% of dyads were classified as secure,
2% as avoidant, 16% as ambivalent/resistant, and 13% as disorganised. They found that the
infant attachment classifications were strongly predicted by Adult Attachment Interviews
with the parents, in line with research in other societies (Chapter 3).436 And a sizeable pro-
portion of mothers in the sample worked, suggesting that the infants were not unfamiliar
with separations. Nonetheless, the number of avoidant dyads was low. Kondo-Ikemura and
colleagues argued that the Strange Situation was generally a valid measure of individual dif-
ferences in the Japanese context, associated with expectable covariates. They qualified that
some aspect of Japanese childcare practice likely explains the low proportion of avoidant
dyads, but they did not speculate on what this might be.
In a recent chapter, Mary True surveyed discussions of the cross-cultural validity of the
Strange Situation.437 She makes two points of particular relevance to appraising Ainsworth’s
legacy to cross-cultural developmental research. She reports from a meta-analysis com-
paring cultures that conventionally include prolonged skin-to-skin contact between in-
fants and caregivers and cultures that conventionally use distal caregiving strategies. Rates
meeting of International Society for the Study of Behavioral Development, Minneapolis, July 1991.Copy shared by
Karin Grossmann. However, curiously, the Japanese infants also seemed much less wary of the stranger than the
German infants.
433 Rothbaum, F., Weisz, J., Pott, M., Miyake, K., & Morelli, G. (2000) Attachment and culture: security in the
United States and Japan. American Psychologist, 55(10), 1093–104.

434 Vereijken, C.J.J.L., Riksen-Walraven, J.M., & Van Lieshout, C.F.M. (1997) Mother–infant relationships in
Japan: attachment, dependency, and amae. Journal of Cross-Cultural Psychology, 28(4), 442–62
435 Rothbaum, F. & Kakinuma, M. (2004) Amae and attachment: security in cultural context. Human
Development, 47(1), 34–9. Cf. Posada, G., Lu, T., Trumbell, J., et al. (2013) Is the secure base phenomenon evident
here, there, and anywhere? A cross-cultural study of child behavior and experts’ definitions. Child Development,
84(6), 1896–905.
436 Kondo-ikemura, K., Behrens, K.Y., Umemura, T., & Nakano, S. (2018) Japanese mothers’ prebirth Adult
Attachment Interview predicts their infants’ response to the Strange Situation Procedure: the Strange Situation in
Japan revisited three decades later. Developmental Psychology, 54(11), 2007–2015.
437 True, M. (in press) Multiple pathways to infant disorganization: insights from an African dataset. In T.
Forslund & R. Duschinsky (eds) The Attachment Reader. Oxford: Blackwell.

194 Mary Ainsworth
of secure attachment are the same between these two contexts. By contrast, rates of avoid-
ance are lower and rates of ambivalent/resistance are higher when, as in the Japanese case,
prolonged skin-to-skin contact is part of caregiving norms.438 This accumulated evidence
indicates systematic variation in the kind of conditional strategy used in the context of differ-
ences in caregiving cultures. However, True argues that such accumulated findings represent
a denigration of Ainsworth’s true legacy, since without articulation of the specific processes
occurring within infant–caregiver interaction, any interpretation is recklessly speculative.
True states that attachment researchers have focused too much on supplying and then in-
terminably discussing distributions in Strange Situation patterns. A limitation of Ainsworth’s
approach was the prominence she gave to the Strange Situation, when in fact the generative
core of her insights came precisely from the combination of qualitative ethnography of gen-
eral processes and quantitative assessment of dyadic processes. True observes that this potent

combination has become fractured into the anthropological and developmental traditions of
cross-cultural research on attachment. This has stalled efforts to pin down and operation-
alise the role of cultural factors in contributing to both adverse forms of care and the role of
cultural protective factors, which may be of strong relevance to intervention science.439
Ainsworth’s focus on the Strange Situation made sense in the context of the priorities
and values of psychology as a discipline in the 1970s and 1980s. But it closed down con-
versation between psychology and anthropology, in a way that may well not have occurred
had she headlined the sensitivity scale. In recent decades, attachment researchers have ap-
plied Ainsworth’s sensitivity scale in measuring caregiving behaviour across various coun-
tries and contexts, where it has successfully predicted infant attachment behaviour and later
socioemotional development.440 It is certainly the case, as Röttger-Rössler among others has
shown, that Ainsworth’s language regarding ‘appropriate’ caregiver response is potentially
ambigious, depending heavily on training in order to clarify its meaning, and contributing
to the potential for coders to judge appropriateness by specific cultural standards.441 Much
depends here on the design of studies that are thoughtful about cultural context and val-
idity, and the work of coders to integrate the particularities of culture and childcare prac-
tices within the formal aspects of Ainsworth’s system. This work is barely mentioned in the
438 See also Mesman, J., van IJzendoorn, M.H., & Sagi-Schwartz, S. (2016) Cross-cultural patterns of attach-
ment: universal and contextual dimensions. In J. Cassidy & P. Shaver (eds) Handbook of Attachment (pp.790–815).
439 Exceptions include Posada, G., Carbonell, O.A., Alzate, G., & Plata, S.J. (2004) Through Colombian
lenses: ethnographic and conventional analyses of maternal care and their associations with secure base behavior.
Developmental Psychology, 40(4), 508–18; Howes, C. & Wishard Guerra, A.G. (2009) Networks of attachment re-
lationships in low-income children of Mexican heritage: infancy through preschool. Social Development, 18(4),
896–914; Fuertes, M., Ribeiro, C., Gonçalves, J.L., et al. (2020) Maternal perinatal representations and their asso-
ciations with mother–infant interaction and attachment: A longitudinal comparison of Portuguese and Brazilian
dyads. International Journal of Psychology, 55(2), 224–33.
440 Mesman, J., van IJzendoorn, M.H., & Sagi-Schwartz, A. (2016) Cross-cultural patterns of attachment. In
J. Cassidy & P.R. Shaver (eds) The Handbook of Attachment: Theory, Research, and Clinical Applications, 3rd edn
(pp.852–77). New York: Guilford; see also Posada, G. (2013) Piecing together the sensitivity construct: ethology
and cross-cultural research. Attachment & Human Development, 15, 637–56.
441 Röttger-Rössler, B. (2014) Bonding and belonging beyond WEIRD worlds: rethinking attachment theory on
the basis of cross-cultural anthropological data. In H. Otto & H. Keller (eds) Different Faces of Attachment: Cultural
Variations on a Universal Human Need (pp.141–68). Cambridge: Cambridge University Press. See also Carlson,
V.J. & Harwood, R.L. (2003) Attachment, culture, and the caregiving system: the cultural patterning of everyday
experiences among Anglo and Puerto Rican mother–infant pairs. Infant Mental Health Journal, 24(1), 53–73;
LeVine, R.A., Gielen, U.P., & Roopnarine, J. (2004) Challenging expert knowledge: findings from an African
study of infant care and development. In Childhood and Adolescence: Cross-cultural Perspectives and Applications
(pp.149–65). Westport, CT: Praeger/Greenwood.
published literature, which means that the principles and processes through which it has
been achieved are not transparent or available for discussion.442 This is an issue that would
likely have seen substantial resolution had anthropologists and developmental psychologists
been able to listen to and learn from one another on the basis of more common ground.
Yet evidence in favour of the cross-cultural relevance of Ainsworth’s construct is that dif-
ferences from western cultural norms are not, in themselves, generally associated with lower
scores for sensitivity. Rather, the majority of caregivers in most contexts around the world
are characterised as showing sensitivity, except where families are facing conditions of social,
economic, or political adversity, or where caregivers have themselves experienced trauma
or maltreatment. And even then, the effects of adversity and trauma may in some instances
be buffered by protective aspects of childcare practices.443 Nonetheless, the gulf between
attachment researchers and anthropologists has hindered the development of a global re-

search agenda to explore these processes, and in turn the take-up of attachment research
within the growing policy and research concern with child development and global public
mental health.444
The crying question
Ainsworth and Bell’s study of crying behaviour was the first analysis of the home observa-
tion data. They found that when mothers responded promptly to infants crying in the first
quarter, they cry less in the final quarter of the first year.445 This finding was a landmark report
at the time, offering a symbolic victory for attachment theory.446 It ran directly contrary to
the behaviourist theory that if crying for a parent is heeded and brings about a positive out-
come for the child, it will be repeated more often over time. Instead, the finding supported
Bowlby’s proposition that affection shown to children would not ‘spoil’ them and make them
dependent, but in fact would help them feel confident in the availability of their caregiver or
caregivers and less prone to distress.447 Furthermore, Ainsworth and colleagues soon after
442 Exceptions include Mesman, J., van IJzendoorn, M., Behrens, K., et al. (2016) Is the ideal mother a sensitive
mother? Beliefs about early childhood parenting in mothers across the globe. International Journal of Behavioral
Development, 40(5), 385–97; Posada, G., Trumbell, J., Noblega, M., et al. (2016) Maternal sensitivity and child
secure base use in early childhood: studies in different cultural contexts. Child Development, 87(1), 297–311;
Dawson, N.K. (2018) From Uganda to Baltimore to Alexandra Township: how far can Ainsworth’s theory stretch?
South African Journal of Psychiatry, 24, 8. Perhaps the only study to have examined the cross-cultural consistency
of coding the Strange Situation is Van IJzendoorn, M. H. & Kroonenberg, P.M. (1990) Cross-cultural consistency
of coding the strange situation. Infant Behavior and Development, 13(4), 469–85.
443 E.g. Fourment, K., Nóblega, M., Conde, G., del Prado, J.N., & Mesman, J. (2018) Maternal sensitivity in rural
Andean and Amazonian Peru. Attachment & Human Development, 27 March, 1–9.
444 Kieling, C., Baker-Henningham, H., Belfer, M., et al. (2011) Child and adolescent mental health world-
wide: evidence for action. The Lancet, 378(9801), 1515–25. However, see Bain, K. & Baradon, T. (2018) Interfacing
infant mental health knowledge: perspectives of South African supervisors supporting lay mother–infant home
visitors. Infant Mental Health Journal, 39(4), 371–84.
445 Bell, S.M. & Ainsworth, M.D.S. (1972) Infant crying and maternal responsiveness. Child Development, 43(4),
1171–90.
446 On the landmark status of this study see Lewis, M. (1997) Altering Fate: Why the Past Does Not Predict the
Future. New York: Guilford, p.145.

447 See the debate between attachment and behaviourist theorists over Ainsworth’s findings: Gewirtz, J.L.
& Boyd, E.F. (1977) Does maternal responding imply reduced infant crying? A critique of the 1972 Bell and
196 Mary Ainsworth
found that babies who reciprocated actively when held by their mother were less likely to
protest when put down, and more likely to immediately turn to independent exploration.448
Stayton and Ainsworth interpreted these findings by proposing that babies could accept ces-
sation of contact because they are confident that the caregiver will be accessible if needed.449
That infant crying was reduced by the end of the first year rather than increased by prompt
response by caregivers was a landmark finding, and one that Bowlby and Ainsworth continued
to mention until the end of their careers with a passionate, steady insistence. The finding neatly
encapsulated the empirical implications of attachment theory. Furthermore, in Patterns of
Attachment, Ainsworth and colleagues reported that ‘mothers who are promptly responsive to
crying signals in the early months have babies who later become securely attached’.450 The stat-
istical procedures by which Ainsworth and Bell came to these conclusions were contested by
her critics.451 However, without any other longitudinal data available to help answer the ques-

tion, the Baltimore findings were the only source of scientific information relevant to the heated
question of whether babies should be left to cry.452 In the 1980s, infant crying was also gaining in
importance in the context of proposals, for instance by Michael Lamb and colleagues, that infant
physical abuse may at times be triggered in parents as tiredness and emotional unrest and urgent
cues meld together as hot frustration.453
The topic drew the attention of the young Marinus van IJzendoorn, then in his late twen-
ties and newly appointed as full Professor at Leiden University. At the time, van IJzendoorn
was immersed in reading attachment theory, writing the first Dutch book offering a detailed
appraisal of Bowlby’s ideas.454 Van IJzendoorn had been asked to care for his six-month-old
godson whilst the child’s parents were away. He stood at the crib hour after hour, through the
night, trying helplessly to quiet the baby, who cried without ceasing. Days became endless,
and the nights even worse. By the time his godson’s parents returned, van IJzendoorn had
resolved to conduct a replication of Ainsworth and Bell’s research, to understand more about
how to reduce infant crying.455 He worked with a doctoral student, Frans Hubbard, to collect
Ainsworth report. Child Development, 48, 1200–1207; Ainsworth, M.D.S. & Bell, S.M. (1977) Infant crying and
maternal responsiveness: a rejoinder to Gewirtz and Boyd. Child Development, 48, 1208–16. For an example of
Bowlby using Ainsworth’s findings regarding crying as ammunition against social learning approaches see e.g.
Bowlby, J. (1973) Separation. London: Pimlico, p.358.
448 Ainsworth, M., Bell, S., & Stayton, D. (1972) Individual differences in the development of some attachment
behaviors. Merrill-Palmer, 18(2), 123–43, p.136.

449 Stayton, D. & Ainsworth, M. (1973) Individual differences in infant responses to brief, everyday separations
as related to other infant and maternal behaviours. Developmental Psychology 9(2), 226–35, p.233.

451 E.g. Lamb, M., Thompson, R.A., Gardner, W., & Charnov, E.L. (1985) Infant–Mother Attachment: The
Origins and Developmental Significance of Individual Differences in the Strange Situation. Hillsdale, NJ: Lawrence
Erlbaum: ‘A group difference in maternal unresponsiveness to crying in the first quarter led to the conclusion that
“Mothers who are promptly responsive to crying signals in the early months have babies who later become securely
attached” (Ainsworth et al. 1978, p.150). In fact, when the measure is expressed as a proportion (Maternal unre-
sponsiveness per hour/Infant crying per hour), the proportion of A and B infants are equivalent, and the deviant
group (C) contains only four dyads’ (65).
452 Newton, L.D. (1983) Helping parents cope with infant crying. Journal of Obstetric, Gynecologic, & Neonatal
Nursing, 12(3), 199–204.

453 Frodi, A.M. & Lamb, M.E. (1980) Child abusers’ responses to infant smiles and cries. Child Development,
51(1), 238–41.
454 Van IJzendoorn, M.H., Tavecchio, L.W.C., Goossens, F.A., & Vergeer, M.M. (1982) Opvoeden in
Geborgenheid. Een Kritische Analyse van Bowlby’s Attachment Theorie. Deventer: Van Loghum Slaterus.
455 Van IJzendoorn, M.H. (2004) Roos. In H. Procee, H. Meijer, P. Timmerman, & R. Tuinsma (eds) Bij die
Wereld wil ik Horen! Zevenendertig Columns & Drie Essays over de Vorming tot Academicus (pp.86–89).
Amsterdam: Boom.
data on what would be a very intensive study. Data collection began in 1983 and took four
years, using the new technology provided by an event recorder/FM audio registration unit.
The findings were clear-cut: when caregivers responded to fuss or cry signals, infants
cried more often by the end of the first year. Yet journals were not keen to publish the re-
sults, and it took until the 1990s before the paper was finally in print.456 Hubbard and van
IJzendoorn realised that ‘research on crying evolved into a “pièce de résistance” of attach-
ment theory . . . It constituted a cornerstone of attachment theory and therefore was not
really open to theoretical and empirical criticism.’457 However, Hubbard and van IJzendoorn
did not regard their results as representing an attack on attachment theory, but as criticism
of specific aspects of Ainsworth and Bell’s methodology. They argued that the Ainsworth
and Bell paper adopted an inappropriate statistical approach, a crude strategy of meas-
urement, and overconfidence in assertion of their findings on the basis of such a small

sample. Hubbard and van IJzendoorn proposed that a distinction needed to be drawn be-
tween low-level fussing and loud and prolonged cries. They speculated that whereas their
technology-enabled methodology allowed them to track the former, it is likely that only the
latter would have been recorded every time by Ainsworth’s live observers, even if fussing was
sometimes noted.
Low-level fussing is not, Hubbard and van IJzendoorn argued, an attachment signal.
Sensitive responsiveness to these behaviours may have little implication for informing the
expectations that organise the attachment behavioural system; instead, they may encourage
more fussing. This is in contrast to loud and prolonged cries by the infant, which are attach-
ment signals. The term ‘crying’ was inherited by developmental psychology from ordinary
language, but was not proving helpful since it hid this important distinction. Ainsworth her-
self would conclude: ‘Even the most responsive mothers did not respond to a little cry that
stopped spontaneously when a baby was put to sleep or a similar brief fuss when a baby was
trying to turn over and could not manage by itself, however, succeeding next. But rarely did
they fail to respond to a loud and prolonged cry.’458 The Hubbard and van IJzendoorn find-
ings have had little traction, however. Still today, the Ainsworth and Bell paper is widely cited
to prove that prompt response to infant crying reduces subsequent crying. For instance,
parents are taught about the finding to encourage nurturing care as part of Dozier and col-
leagues’ influential Attachment and Biobehavioral Catch-up parenting intervention.459
Coding individuals
North American developmental psychology had long felt the tension between a stance that
gave primacy to individuals and a stance that gave primacy to relationships.460 Ainsworth’s
456 Hubbard, F.O. & van IJzendoorn, M.H. (1991) Maternal unresponsiveness and infant crying across the first
9 months: a naturalistic longitudinal study. Infant Behavior and Development, 14(3), 299–312.
457 Hubbard, F.O.A. & van IJzendoorn, M.H. (1987) Maternal unresponsiveness and infant crying: a critical
replication of the Bell & Ainsworth study. In L.W.C. Tavecchio & M.H. van IJzendoorn (eds) Attachment in Social
Networks. Contributions to the Bowlby–Ainsworth Attachment Theory (pp.339–78). New York: Elsevier, p.344.
458 Personal communication to Marinus van IJzendoorn, cited in ibid, p.368.
459 Dozier, M. & Bernard, K. (2019) Coaching Parents of Vulnerable Infants: The Attachment and Biobehavioral
Catch-up Approach. New York: Guilford, p.11.

460 Sameroff, A. (2009) The transactional model. In A. Sameroff (ed.) The Transactional Model of
Development: How Children and Contexts Shape Each Other (pp.3–21). Washington, DC: American Psychological
Association. On the shift towards contextualism in developmental psychology in the 1970s see Lerner, R.M.,
198 Mary Ainsworth
group aligned themselves firmly with the latter perspective. The sensitivity measure is partly
an assessment of individual differences between caregivers—responsiveness to crying per-
haps more so, since it is a cruder measure. But in both cases there is no standardisation of
infant behaviour: the measure assesses the way that particular infant signals are noticed and
responded to by the caregiver, which is at least in part an assessment of the dyad in their
interactive dance. There was one assessment in which this dance was not observed, however,
and in which only individual behaviour was coded: the Strange Situation.
Ainsworth sought to semi-control the caregiver’s behaviour in the Strange Situation in
order to make the separations a standardised, ambiguous, but evocative stimulus. And the
written coding system for the Strange Situation that she created was for individual infant be-
haviour only, an approach generally extrapolated by her students to the coding of patterns
of attachment in separation–reunion procedures at later developmental stages.461 In itself

this is no problem. The behaviours of individual children in the Strange Situation are asso-
ciated with interactions of the dyad at home, and can predict caregiver behaviour towards
the child in other settings (Chapter 4). This is somewhat astonishing, profound even, and
suggests that a coding system for individual behaviours can serve as a workable window on
attachment as a dyadic property, since this behaviour reflects—even if imperfectly—infants’
expectations about their caregiver’s availability. Yet the methodological choice to code indi-
vidual behaviours was not fully owned: Ainsworth did not accompany this de facto focus of
the Strange Situation with any checklist to support its reliability as a measure of individual
differences, as might have been expected. For instance, researchers using the procedure
were not required to take note of whether a child was ill, was on relevant medication, or
even had received regular care by the parent.462 This is despite that fact that Ainsworth and
later researchers readily acknowledged these factors as relevant to the reliability of a Strange
Situation procedure. Klaus and Karin Grossmann believe that Ainsworth assumed that
the Strange Situation would usually be accompanied by naturalistic observation of dyads,
making a checklist for relevant individual differences superfluous.463 However, as discussed,
naturalistic observation fell away over time from attachment research, following the prior-
ities of the wider discipline of developmental science.
This potential limitation on reliability has been accepted quietly by subsequent re-
searchers. Perhaps it has been felt that the issue is minor, and that most infants in most
samples will nonetheless respond to separation and reunion with a caregiver in ways that
reflect to some extent the care they have received in that relationship. It may also have been
felt by second-generation attachment researchers, and all the more by third generation, that
it is now too late to add such reliability safeguards. Whereas such issues of reliability have
generally been ignored, some attachment researchers have explicitly wondered whether it is
Hultsch, D.F., & Dixon, R.A. (1983) Contextualism and the character of developmental psychology in the 1970s.
Annals of the New York Academy of Sciences, 412(1), 101–28.
461 Cassidy, J., Marvin, R., with the Attachment Working Group of the John D. and Catherine T. MacArthur
dren: procedures and coding manual. Unpublished manual; Main, M. & Cassidy, J. (1988) Categories of response
to reunion with the parent at age 6: predictable from infant attachment classifications and stable over a 1-month
period. Developmental Psychology, 24(3), 415–26.
462 Except when a sample has been specifically recruited with such factors in mind, e.g. Espinosa, M., Beckwith,
L., Howard, J., Tyler, R., & Swanson, K. (2001) Maternal psychopathology and attachment in toddlers of heavy
cocaine-using mothers. Infant Mental Health Journal, 22(3), 316–33.
463 Grossmann, K. & Grossmann, K. (2012) Bindungen— Das Gefüge psychischer Sicherheit Gebundenes.
Stuttgart: Klett-Cotta.
valid to assess attachment as a dyadic property with a focus on infant behaviours.464 A few
have sought to revise or create coding systems focused on dyadic interactions. The most
direct attempt has been that of Crittenden, who elaborated coding systems for the Strange
Situation and other assessment measures that explicitly assess caregiver–child interaction
rather than the individual behaviour of the child.465 One of the systems for coding behav-
iour at age six by the Berkeley group was the unpublished Strage and Main approach to
coding reunions of verbal children; this was also a dyadic coding system.466 And Lyons-
Ruth and colleagues developed a dyadic-based coding system called the Goal-Corrected
Partnership in Adolescence Coding System.467 Nonetheless, the predominant approach
to the assessment of child–caregiver attachment has certainly remained the coding of in-
dividual child behaviours following the protocol set out in Ainsworth and colleagues in
Patterns of Attachment.

Main attempted to title a paper ‘Security of attachment characterises relationships, not
infants’, with the running header of ‘Relationships, not infants’ (though the paper ended up
with a different title in its published version as a concession to gruelling rounds of peer-
review feedback).468 Bowlby put matters starkly in Attachment, Volume 1: ‘any statement
about a child of twelve months himself showing a characteristic pattern of attachment be-
haviour, distinct from the interactional pattern of the couple of which he is a partner, and
implying some degree of autonomous stability, is certainly mistaken’.469 Yet one consequence
of an individual-focused coding system for the Strange Situation has been that the predom-
inant language used to discuss the categories of the Strange Situation is of secure, avoidant,
and ambivalent/resistant infants. It is clear that a factor contributing to such language was
that the coding system assessed individual behaviours. However, an additional factor has
been that it is incredibly cumbersome to keep writing out ‘behaviour shown in the Strange
Situation by an infant in a dyad classified as avoidant, suggesting a certain history of infant–
caregiver interactions’; it is easier to refer to an avoidant or A infant. Such terminology
implied—or at the very least ceaselessly risked the implication—that attachment was a fixed
individual trait and ultimate explanation. This is the kind of implication that, once everyone
is asleep, creeps out and drinks the blood of a relationship-focused paradigm. Looking back
464 E.g. Fonagy, P. (2000) Attachment Theory and Psychoanalysis. New York: Other Press.
465 Crittenden, P.M. (1992) Preschool Assessment of Attachment. Miami: Family Relations Institute; Crittenden,
P.M. (2016) Raising Parents: Attachment, Representation, and Treatment, 2nd edn. London: Routledge. The
Cassidy/Marvin MacArthur preschool system has some elements of a dyadic focus, though these are not fore-
grounded. Cassidy, J., Marvin, R., with the ttachment Working Group of the John D. and Catherine T. MacArthur
dren: procedures and coding manual. Unpublished manual. Dyadic coding is more foregrounded in Marvin’s later
work: Britner, P.A., Marvin, R.S., & Pianta, R.C. (2005) Development and preliminary validation of the caregiving
behavior system: association with child attachment classification in the preschool Strange Situation. Attachment &
466 Main, M., Kaplan, N., & Cassidy, J. (1985) Security in infancy, childhood, and adulthood: a move to the level
of representation. Monographs of the Society for Research in Child Development, 50, 66–104; Strage, A. & Main,
M. (1985) Attachment and parent—child discourse patterns. Paper presented at Biennial meeting of Society for
Research in Child Development, Toronto. PP/Bow/J.4/4.
467 Obsuth, I., Hennighausen, K., Brumariu, L.E., & Lyons-Ruth, K. (2014) Disorganized behavior in adolescent–
parent interaction: relations to attachment state of mind, partner abuse, and psychopathology. Child Development,
85(1), 370–87.
468 Main, M. & Weston, D. (1981) The independence of infant–mother and infant–father attachment relation-
ships: security of attachment characterises relationships, not infants. PP/Bow/J.4/3. The paper was ultimately pub-
lished under the title ‘The quality of the toddler’s relationship to mother and to father: related to conflict behavior
and the readiness to establish new relationships’.
200 Mary Ainsworth
over three decades of research using the Strange Situation, and two decades of training cod-
ers, Sroufe acknowledges:
We readily slip into describing cause in terms of individual traits rather than develop-
mental systems. At the outset I want to adopt the curved finger of accusation and say that
attachment theorists, such as myself, are equally vulnerable to this problem. Frequently,
we slip into using terms such as ‘securely attached child’ when we know that attachment is
really a relationship term, and the proper description would be ‘a child with a history of a
secure relationship with the primary caregiver’. We don’t do it because this is unwieldy.470
Eagle proposed that this discourse has contributed to a focus on individual differences ra-
ther than relationships in attachment research more generally. He claimed that despite the

theoretical acknowledgement of attachment as a relational construct, in practice the fact of
the matter is that most attachment research is indistinguishable from a research programme
that imagines individual differences as fixed traits.471 There are certainly exceptions: an ex-
ample is the attention of Sroufe and Egeland to continuities in security and insecurity that
may be expressed as different forms of behaviour depending on the child’s stage of devel-
opment (Chapter 4). And Guy Bosmans and colleagues pursued innovative work exam-
ining attachment-relevant transitory states, contrasting them to more stable individual
differences in attachment.472 However, in general, Eagle’s observation does have purchase.
Fonagy and Campbell, taking the criticism further, recently argued that unless attachment
research can move away from a spiritless obsession with categories for individual differences
between humans, it will have no ‘intellectual vigour and relevance’, and likely no future (see
Chapter 6).473
Early experiences vs continuity of care
Ainsworth and colleagues found that caregiver sensitivity in the home observation data pre-
dicted infant attachment in the Strange Situation. And a generation of subsequent researchers
found a host of associations between the Strange Situation and later outcomes. However,
Ainsworth and her team did not have the data to make claims about the implications of
infant attachment for later development; and later researchers only very rarely undertook
extensive home observations. Those that did, such as Klaus and Karin Grossmann, con-
ducted their research with samples facing comparatively few adversities or sources of disrup-
tion. Given that caregiver sensitivity is quite stable over time unless specific changes intrude
which alter the resources available to the caregiving system, it remained entirely unclear
whether attachment as measured by the Strange Situation was functioning as an autonomous
471 Eagle, M. (2013) Attachment and Psychoanalysis. New York: Guilford, p.55. See also Kobak, R. & Bosmans, G.
(2018) Attachment and psychopathology: a dynamic model of the insecure cycle. Current Opinion in Psychology,
25, 76–80, p.76.
472 Bosmans, G., Van de Walle, M., Goossens, L., & Ceulemans, E. (2014) (In)variability of attachment in middle
childhood: secure base script evidence in diary data. Behavior Change, 31, 225–42.
473 Fonagy, P. & Campbell, C. (2015) Bad blood revisited: attachment and psychoanalysis. British Journal of
Psychotherapy, 31(2), 229–50, p.236.

predictive variable, or whether maternal sensitivity or other aspects of the caregiving context
were behind the scenes, doing the causal work. This question was raised by Michael Lamb
and colleagues in their controversial criticisms of the Strange Situation. They argued that
‘relationships between early experiences and later outcomes have been demonstrated only
when there is continuity in the circumstances’.474 It was not known, therefore, whether the
causal factor for these later outcomes was early experiences of care in early childhood, early
patterns of attachment, or experiences of care at the time of later outcomes. Or all three in-
dependently. Or an interaction. This question, left largely unexplored, has muddied uses of
attachment theory to inform clinical and preventative work.
Ainsworth acknowledged the problem head-on in a paper delivered to the International
Conference on Infant Studies in April 1984. She urged colleagues to accept that ‘stability of
patterns of attachment during infancy is influenced by the degree to which family interaction

is stable, while still not making what I think is the error of attributing continuity wholly to
such stability’.475 Likewise, the prediction from early attachment to later child outcomes was
argued to follow the same logic. Part of what was measured in the Strange Situation was
the consequences of the caregiving the child had received. Change the caregiving, and these
implications change. However, Ainsworth argued that the residue of experiences of care-
giving do, over time, come to organise the attachment behavioural system in relatively dur-
able ways. Yet this conference address by Ainsworth remained unpublished, and the position
generally attributed to her has been that infant attachment is of special importance because
of its major role in mediating early care and later development and mental health outcomes.
By the early 1990s, Everett Waters was expressing deep concern that this assumption had
come to function as ‘dogma and doctrine’.476 There were a few studies that examined the
contributions of infant attachment and later caregiving, generally finding that both made
a contribution to peer co-operativeness, language skills, school readiness, and behaviour
problems.477 In these, child outcomes were generally better when early insecure attachment
in the Strange Situation was followed by sensitive care than when early secure attachment
in the Strange Situation was followed by subsequent insensitive care. In 1998, Thompson
observed that ‘virtually all attachment theorists agree that the consequences of a secure or
insecure attachment arise from an interaction between the emergent internal representa-
tions and personality processes that attachment security may initially influence, and the con-
tinuing quality of parental care that fosters later sociopersonality growth’.478 However, no
studies were conducted to see whether the Strange Situation added to prediction beyond
early caregiver sensitivity.
474 Lamb, M., Thompson, R.A., Gardner, W., & Charnov, E.L. (1985) Infant–Mother Attachment: The Origins and
Developmental Significance of Individual Differences in the Strange Situation. Hillsdale, NJ: Lawrence Erlbaum, p.4.
on Infant Studies, April 1984. PP/Bow/J.1/57

477 E.g. Erickson, M., Sroufe, L.A., & Egeland, B. (1985) The relationship of quality of attachment and behavior
problems in preschool in a high risk sample. Monographs of the Society for Research in Child Development, 50
(1–2), 147–86; Belsky, J. & Fearon, R.P. (2002) Early attachment security, subsequent maternal sensitivity, and later
child development: does continuity in development depend upon continuity of caregiving? Attachment & Human
478 Thompson, R. (1998) Early sociopersonality development. In W. Damon & N. Eisenberg (eds) Handbook
of Child Psychology: Vol. 3. Social, Emotional, and Personality Development, 5th edn (pp 25–104). New York:
Wiley, p.58.
202 Mary Ainsworth
Sroufe and Egeland’s Minnesota group was one of the few laboratories that had lon-
gitudinal data on caregiver behaviour, infant attachment classifications, and later devel-
opmental outcomes. But they only made separate reports of the relationship between
sensitivity and the Strange Situation,479 and the Strange Situation and later outcomes,480
or folded caregiver sensitivity and infant Strange Situation classifications together into an
‘early caregiving experiences composite’.481 The position of Sroufe and Egeland appears to
have been that neither sensitivity nor the Strange Situation is in itself ‘attachment’, which
cannot be directly measured but must be inferred. As such, to Sroufe and Egeland, whether
it was the sensitivity scale or the Strange Situation that was doing the predicting was rather
irrelevant. Both measures were assumed to be only vantages on dyadic differences in the at-
tachment relationship, scientific proxies for detailed observations of infant–caregiver inter-
action in naturalistic settings. Sroufe and Egeland appeared to prefer the Strange Situation,

however, when they had specific hypotheses about the correlates of the avoidant or ambiva-
lent/resistant attachment classifications. Since the retirement of Sroufe and Egeland, the
Minnesota group have tended to stop reporting analyses on the longitudinal correlates of
the Strange Situation at all, and instead have generally reported longitudinal associations of
caregiver sensitivity.482
However, it should not be thought that Minnesota alone had the data to examine the
question of whether it is actually sensitivity that is behind associations between the Strange
Situation and other correlates. After Minnesota, there have subsequently been other labora-
tories with relevant data. One is the huge NICHD dataset, which has over a thousand dyads
with assessments using the Strange Situation and of sensitivity at multiple time-points,
including prior to the Strange Situation (though the measure of sensitivity had differences
from Ainsworth’s scale, since it included warmth and lack of intrusiveness).483 This dataset
has been available for three decades, and could have been used to see if the Strange Situation
added to prediction after inclusion of prior caregiver sensitivity. An obstacle to tests of medi-
ation, however, may have been norms and expertise in statistical analysis among the develop-
mental science community. Attachment researchers tended to rely in the 1980s on one-way
analyses of variance and other statistical tests that permitted examination of the implications
of caregiving on attachment, or attachment on some other variable. Mediational path ana-
lyses would only really enter into attachment research from the 1990s, despite earlier claims
made for its relevance.484 However, the issue was not simply methodological, but grounded
479 E.g. Egeland, B. & Farber, E.A. (1984) Infant–mother attachment: factors related to its development and
changes over time. Child Development, 55(3), 753–71 .

480 E.g. Sroufe, L.A. (1983) Infant–caregiver attachment and patterns of adaptation in preschool: the roots of
maladaptation and competence. In M. Perlmutter (ed.) Development and Policy Concerning Children with Special
Needs: The Minnesota Symposia on Child Psychology, Vol. 16 (pp.41–83). Hillsdale, NJ: Lawrence Erlbaum.
481 E.g. Jimerson, S., Egeland, B., Sroufe, L.A., & Carlson, B. (2000) A prospective longitudinal study of high
school dropouts examining multiple predictors across development. Journal of School Psychology, 38(6), 525–49.
482 See e.g. Farrell, A.K., Waters, T.E.A., Young, E.S., et al. (2019) Early maternal sensitivity, attachment security
in young adulthood, and cardiometabolic risk at midlife. Attachment & Human Development, 21(1), 70–86.
483 NICHD Early Child Care Research Network (1999) Child care and mother–child interaction in the first
three years of life. Developmental Psychology, 35, 1399–413. Belsky and Fearon conducted an analysis of the relative
contributions of attachment and sensitivity to later outcomes in this dataset, but only included measures of sensi-
tivity subsequent to the Strange Situation. Belsky, J. & Fearon, R.P. (2002) Early attachment security, subsequent
maternal sensitivity, and later child development: does continuity in development depend upon continuity of care-
giving? Attachment & Human Development, 4(3), 361–87.
484 Connell, J.P. (1987) Structural equation modeling and the study of child development: a question of good-
ness of fit. Child Development, 58, 167–75

in a theoretical commitment to attachment patterns as developmentally causal. This is sup-

ported by the fact that multiple regression was used with increasing frequency through
the 1980s by attachment researchers, in line with a broader trend in developmental psych-
ology, but attachment and sensitivity were not examined together in an attempt to identify
their respective contributions. When, in the 2000s, structural equation modelling was de-
ployed by the Minnesota group to analyse their data, a composite variable containing both
the Ainsworth infant attachment patterns and early caregiving experiences was used, rather
than separating these out.485
In 1998, Elizabeth Carlson from the Minnesota group published an article addressing
the disorganised attachment classification.486 Though Carlson’s work is addressed further
in Chapter 3 and 4, it is important to mention this article here, as it was the first published
work by the Minnesota group in which early caregiving and an infant attachment classi-

fication were entered into a mediational analysis. As an index of early caregiving, Carlson
used a composite which included observer-assessed caretaking skill of the mother, the
mother’s sensitivity, and information about whether the caregiver had been abusive or en-
gaged in suicide attempts. She found that the relationship between this early care composite
and adolescent mental health problems was fully mediated by the disorganised attachment
classification. She also found that avoidant attachment made an independent contribution
to adolescent mental health problems. However, she did not include avoidant attachment
within the mediational model.
The first full answer to Lamb’s question about the predictive significance of sensitivity
and attachment would wait for a study by Beijersbergen, Juffer, Bakermans-Kranenburg,
and van IJzendoorn, reported in 2012.487 Beijersbergen and colleagues conducted a follow-
up with 125 adolescents who had been adopted early in their childhoods. Maternal sen-
sitivity and infant attachment were assessed at 12 months of age. When the children were
14 years old, maternal sensitivity was assessed again during a discussion around difficult and
conflict-evoking topics. The Adult Attachment Interview (Chapter 3) was used to assess the
adolescents’ state of mind regarding attachment. The researchers found no continuity of at-
tachment between infant attachment in the Strange Situation and states of mind regarding
attachment in the Adult Attachment Interview. However, maternal sensitivity in infancy and
adolescence predicted continuity of secure attachment. Furthermore, a change from ma-
ternal insensitivity in infancy to sensitivity in adolescence predicted a change in attachment
patterns from insecure to secure. The researchers concluded that continuity of attachment
seems dependent on continuity of the caregiving context, as Lamb and Waters both pre-
dicted: ‘We therefore submit that attachment theory should be a theory of sensitive parent-
ing as much as it is a theory of attachment.’488
At age 23, the sample was asked to complete Waters and Waters’ attachment script assess-
ment. The same results appeared. There was no continuity of attachment from infancy to
young adulthood. Sensitive care during childhood, but not adolescence, was a good predictor
485 Carlson, E.A., Sroufe, L.A., & Egeland, B. (2004) The construction of experience: a longitudinal study of rep-
resentation and behavior. Child Development, 75(1), 66–83.

486 Carlson, E.A. (1998) A prospective longitudinal study of attachment disorganization/disorientation. Child
Development, 69(4), 1107–28.

487 Beijersbergen, M.D., Juffer, F., Bakermans-Kranenburg, M.J., & van IJzendoorn, M.H. (2012) Remaining or
becoming secure: parental sensitive support predicts attachment continuity from infancy to adolescence in a lon-
gitudinal adoption study. Developmental Psychology, 48(5), 1277–82.
488 Ibid. p.1281.
204 Mary Ainsworth
of secure base scripts at age 23.489 This was a controversial finding, given the extent of prior
emphasis on prediction from the Strange Situation. An informal network of quantitatively
oriented attachment researchers including Bakermans-Kranenburg and van IJzendoorn
made a call in 2014 for further work on this question. They asserted ‘an urgent need for
theory-driven studies that address mediating processes that account for such enduring ef-
fects, for example by addressing questions concerning whether such long-term continuities
are due to the ongoing supportive function of attachment relationships and/or the early ef-
fects of attachment experiences on the construction of stable psychological structures’.490
However, the emphasis on sensitivity by many researchers since Ainsworth has been
challenged by Ainsworth’s former student Jude Cassidy and her colleagues. Just as Waters
argued that amorphous measurements of ‘attachment’ should be superceded by precise at-
tention to the availability of a secure base/safe haven script, Cassidy and colleagues argued

that sensitivity should be superceded by the capacity of a caregiver to provide a secure base
and safe haven.491 Whilst in general this capacity will align with sensitivity, Cassidy and col-
leagues argued that they are entirely distinguishable. Sensitivity considers all of a caregiver’s
response to any infant cues, whereas secure base/safe haven provision is more specific about
which infant cues are relevant and which caregiver responses. Sensitivity is also concerned
with the caregiver’s promptness of response, which is not necessarily as central for secure
base/safe haven provision. These proposals have received initial support from a a recent
study by Woodhouse and colleagues of 174 mother–infant dyads facing socioeconomic ad-
versities.492 The researchers found that observed secure base/safe haven provision was a
much better predictor of Strange Situation classifications than observed sensitivity, and ac-
counted for 16% of variance. The association between the two observational measures of
caregiving was weak (r = .11). Woodhouse and colleagues also conclude that secure base/
safe haven provision likely has greater cross-cultural applicability than the Ainsworth sen-
sitivity construct.
Conclusion
See Table 2.2 in the Appendix to this chapter for consideration of key concepts discussed
here.
Ainsworth’s methodological innovations were a depth charge thrown into the water of
developmental psychology: bubbles from the explosion are still coming to the surface today.
The influence of Ainsworth’s research programme on subsequent attachment research has
been foundational, enduring, and profound. Ainsworth observed that ‘research methods
489 Schoenmaker, C., Juffer, F., van IJzendoorn, M.H., Linting, M., van der Voort, A., & Bakermans-Kranenburg,
M.J. (2015) From maternal sensitivity in infancy to adult attachment representations: a longitudinal adoption
study with secure base scripts. Attachment & Human Development, 17(3), 241–56.
490 Groh, A.M., Fearon, R.P., Bakermans-Kranenburg, M.J., van IJzendoorn, M.H., Steele, R.D., & Roisman, G.I.
(2014) The significance of attachment security for children’s social competence with peers: a meta-analytic study.
Attachment & Human Development, 16(2), 103–36, p.126.
491 Cassidy, J., Woodhouse, S.S., Cooper, G., Hoffman, K., Powell, B., & Rodenberg, M. (2005) Examination
of the precursors of infant attachment security: implications for early intervention and intervention research. In
L.J. Berlin, Y. Ziv, L. Amaya-Jackson, & M.T. Greenberg (eds) Enhancing Early Attachments: Theory, Research,
Intervention, and Policy (pp.34–60). New York: Guilford.
492 Woodhouse, S.S., Scott, J.R., Hepworth, A.D., & Cassidy, J. (2020) Secure base provision: a new approach to
examining links between maternal caregiving and infant attachment. Child Development, 91(1), 249–65.
Conclusion 205
influence the theoretical formulation associated with it. The reverse is also true.’493 The
Strange Situation is an extreme case. Rather than simply a tool for deployment in line with
researchers’ intentions, the Strange Situation extended the ambitions of an area of research,
whilst also shaping the kinds of action and thought that subsequently seemed feasible or
worthwhile.494 On the one hand, the Strange Situation has provided the basis for a cumula-
tive research paradigm for the study of early relationships for over 50 years—an astonishing
feat. Over 15,000 child–caregiver dyads have been seen in the Strange Situation in the course
of published research in this period. There will be many thousands more involved in unpub-
lished research and clinical and forensic practice. Without the development of the Strange
Situation, attachment theory would very likely have failed to take root within American de-
velopmental psychology.495 By the same token, Granqvist suggested that if a measure had
been developed for cognitive science or evolutionary psychology of equal importance as the

Strange Situation for developmental psychology, it is quite possible that attachment theory
would have put down roots in these disciplines too.496
On the other hand, to a degree that dismayed Ainsworth, the Strange Situation became
the essence of attachment research at the expense of mixed methods research and natur-
alistic observation: ‘I have been quite disappointed that so many attachment researchers
have gone on to do research with the Strange Situation rather than looking at what happens
in the home or in other natural settings—like I said before, it marks a turning away from
“fieldwork”, and I don’t think it’s wise.’497 Ainsworth was proud of the Strange Situation and
the way that it helped contribute to recognition for attachment theory. Nonetheless, Klaus
and Karin Grossmann report that ‘in our last encounter with Mary Ainsworth, she said that
she regretted the Strange Situation book [Patterns of Attachment], which had really eclipsed
what she had in mind. She said, sadly, that she rather would have written a book about sen-
sitivity.’498 The Strange Situation was constructed by Ainsworth as a heuristic, a means to
further explore Bowlby’s theory of behavioural systems. However, in general, the Baltimore
longitudinal study was part-misrepresented and part-misinterpreted as a hypothesis-testing
endeavour, giving findings the appearance of greater certainty and closure than they war-
ranted.499 And more specifically, the Strange Situation came to replace the interactions in the
home that it had been intended to capture and preserve. Individual differences in the form of
493 Ainsworth, M. (1969) Object relations, dependency and attachment. Child Development, 40, 969– 1025,
p.1003.
494 Cf. Sayes, E. (2014) Actor–network theory and methodology. Social Studies of Science, 44(1), 134–49.
495 For detailed argument on this point see van IJzendoorn, M.H., Tavecchio, L.W.C., Goossens, F.A., & Vergeer,
M.M. (1982) Opvoeden in Geborgenheid: Een Kritische Analyse van Bowlby’s Attachment Theorie. Amsterdam: Van
Loghum Slaterus. By way of comparison, for discussion of a movement in psychoanalytic theory in the same period
that failed to take root within academic psychology see McLaughlin, N.G. (1998) Why do schools of thought fail?
Neo-Freudianism as a case study in the sociology of knowledge. Journal of the History of the Behavioral Sciences,
34(2), 113–34.
496 Granqvist, P. (2020) Attachment, Religion, and Spirituality: A Wider View. New York: Guilford: ‘That the
placement of attachment theory and research within mainstream psychology is largely a historical coincidence,
resulting from some skilled developmental psychologists’ receptiveness to Bowlby’s ideas. They adopted the
theory, contributed to it, and made it a major part of psychology. However, attachment theory, originating in eth-
ology, was among the first fully developed theories within what was later to become evolutionary approaches to
human psychology. Similarly, as evidenced in Bowlby’s use of a cybernetic model and his borrowing of the internal
working model construct, attachment theory grew in the same soil that later produced cognitive science.’
Ainsworth. Monographs of the Society for Research in Child Development, 60(2/3), 2–21, p.12.
498 Klaus and Karin Grossmann, personal communication, November 2018.
499 The Sroufe and Egeland Minnesota group were an exception here; core to their mission was an attempt to
replicate and evaluate Ainsworth’s findings, to test out their scientific standing (Chapter 4).
206 Mary Ainsworth
attachment categories became the focus of attention, even as Bowlby’s behavioural systems
model became part of the backdrop rather than an active concern.
The second generation of attachment researchers inherited from Bowlby and Ainsworth
a theory with (i) apparently intuitive and accessible elements, including positions on norma-
tive child rearing, integrated with (ii) subtle, technical distinctions in the use of concepts and
theory (e.g. between attachment behaviour and the attachment behavioural system), and
(iii) complex observational measures. This facilitated the development of a division between
an inner core of specialised developmental researchers and a wide popular constituency
of practitioners, parents, and policy-makers interested in attachment theory and research,
with terms like ‘attachment’ and ‘security’ offering switchers and relays between these dif-
ferent domains. Frequently, these groups have talked right past one another with the same
words. Neither subtle, technical distinctions nor complex observational measures are easily

transmitted through a print medium. It is also hard to effectively debate and discuss them in
public, even when this can be acknowledged as a good use of time. One consequence is that
attachment ideas, for all their accessibility to diverse audiences, have been continually at risk
of being understood in ways that differ wildly from the understanding of researchers.
An important example has been the way that readers generally interpreted Ainsworth’s
construct of ‘sensitivity’ in line with the ordinary language connotations of the term. This is
hardly surprising, since the coding system for sensitivity remained unpublished (until the
2015 reissue of Patterns of Attachment by Waters) and the operationalisation of the construct
was little discussed in print. Chapter 1 describes how criticism of attachment theory has at
times mistaken the ideas popularised by Bowlby in the 1950s for the commitments of sub-
sequent researchers, leading attachment researchers to experience their paradigm as chron-
ically misunderstood and somewhat maligned, and critics to regard attachment research
as a heedless and arrogant enterprise.500 This has proven an obstacle to effective dialogue
between attachment research and anthropology. Furthermore, the gap between researchers
and their publics left open space for entrepreneurs to enter the market as authorities. At times
this has resulted in high-quality and well-informed popular texts and commercial trainings.
However, at times these entrepreneurs have profited from the circulation of accounts of at-
tachment theory and research that in important ways run contrary to available empirical
evidence (Chapter 6).
A second consequence of the role of subtle, technical distinctions and complex observa-
tional measures has been obstacles to the diffusion of expert knowledge, expertise, and au-
thority. It encouraged the development of an oral culture among developmental researchers
based on lines of mentorship and attendance at lengthy training institutes, to permit the de-
velopment and trained exercise of complex (and often implicit) skills of perception, thought,
appreciation, and valuation501—the development of an ‘attachment’ eye.502 Through the
training institutes, researchers could become socialised in tacit skills of observation, concep-
tualisation, and judgement that offered access to the practical intricacies of relevant theory,
500 See also White, S., Gibson, M., Wastell, D., & Walsh, P. (2019) Reassessing Attachment Theory in Child Welfare.
Bristol: Psychology Press. Collins has offered the term ‘alien science’ to describe the reception of research in cases
with such a sharp disparity. Collins, H.M. (1999) Tantalus and the aliens: publications, audiences and the search
for gravitational waves. Social Studies of Science, 29(2), 163–97.
501 Reijman, S., Foster, S., & Duschinsky, R. (2018) The infant disorganised attachment classification: ‘patterning
502 Cf. Bourdieu, P. (1989) The social genesis of the eye. In The Rules of Art, trans. S. Emanuel (pp.313–21).
Cambridge: Polity Press.

Conclusion 207
such as the secure base concept. However, the requirements of this socialisation within an
oral culture have made the reproduction of the field challenging, and the number of de-
velopmental attachment researchers has remained comparatively small. Reliance on an oral
culture has, at times, made it relatively difficult even for other developmental psychologists
to understand the details of the phenomena under discussion, for instance how the Strange
Situation categories are actually operationalised. For example, researchers speculate about
the meaning of differences in infant–caregiver behaviour in the Strange Situation in ways
largely cut loose from how it is coded.503
The role of an oral culture has also made it difficult for researchers not centrally con-
cerned with attachment to accurately take from and meaningfully contribute to the tradition
of research. An effect was that research groups where the oral culture could be accessed—
such as Berkeley (Mary Main; Chapter 3), Minnesota (Alan Sroufe and others; Chapter 4),

Regensburg (Klaus and Karin Grossmann), and SUNY (Everett Waters and Judith Crowell)
in the 1980s504—became centres of gravity for developmental attachment research and the
training of future researchers. The leaders of these research groups gained the status of ‘au-
thorities of delimitation’, with a certain capacity to set the terms of the developmental trad-
ition.505 These dynamics are not only apparent in retrospect. Even in the 1980s, attachment
researchers were themselves offering analysis and commentary on these sociological pro-
cesses in print, worrying that the result would be a dynamic of ‘insiderness’ among certain
research groups, to the exclusion of non-members.506
There were major advantages of having such centres of gravity for the developmental
tradition. One was the capacity to develop a centralised training and reliability test for the
Strange Situation, which has supported interlaboratory agreement as well as providing a site
with ritual as well as pragmatic functions, such as cultivating junior researchers and develop-
ing international solidarity and networks. Another advantage was the momentum that was
generated around high-investment, high-yield longitudinal studies such as the Minnesota
Longitudinal Study of Risk and Adaptation (Chapter 4). A third advantage was forms of col-
laboration and sharing of data for particular purposes, supported by strong interpersonal
relationships, for instance in appraising cross-cultural differences.507 The development of
the disorganised attachment classification was, as Chapter 3 will show, based in part on the
willingness of other laboratories to share their tapes of the Strange Situation with Main and
the Berkeley group.
503 E.g. Koós, O. & Gergely, G. (2001) A contingency-based approach to the etiology of ‘disorganized’ attach-
ment: the ‘flickering switch’ hypothesis. Bulletin of the Menninger Clinic, 65(3), 397–410.
504 By the 1990s, additional centres of gravity included—though were not limited to—Pennsylvania (Jay Belsky),
Harvard (Karlen Lyons-Ruth), Leiden (Marinus van IJzendoorn), Maryland (Jude Cassidy, Doug Teti), Haifa (Avi
Sagi-Schwartz), and London (Peter Fonagy, Miriam Steele, and Howard Steele). Berkeley and Minnesota espe-
cially, and Regesburg and SUNY to an extent, nonetheless appeared to retain particular significance and, in certain
regards, social priority. This dynamic can be seen in the pattern of citation by the research groups of one another.
505 For a discussion of the role of authorities of delimitation see Foucault, M. (1969, 1972) The Archaeology of
Knowledge, trans. A.M. Sheridan Smith. London: Routledge, Chapter 3.

506 E.g. Tavecchio, L.W. & van IJzendoorn, M.H. (eds) (1987) Attachment in Social Networks: Contributions to
the Bowlby–Ainsworth Attachment Theory. New York: Elsevier; Emde, R.N. & Fonagy, P. (1997) An emerging cul-
ture for psychoanalytic research? International Journal of Psychoanalysis, 78(4), 643–51: ‘Extensive training is often
needed for coding and observation, which can lead not only to isolation but to shared assumptions that are un-
specified among those doing the research’ (649).
507 E.g. van IJzendoorn, M.H. & Kroonenberg, P.M. (1990) Cross-cultural consistency of coding the strange situ-
ation. Infant Behavior and Development, 13(4), 469–85; Sagi, A., van IJzendoorn, M.H., & Koren-Karie, N. (1991)
Primary appraisal of the Strange Situation: a cross-cultural analysis of preseparation episodes. Developmental
Psychology, 27(4), 587–96.
208 Mary Ainsworth
However, one negative consequence of the manner in which the developmental tradition was
transmitted was that some aspects of Ainsworth’s approach were adopted with, in retrospect,
too little explicit discussion. One aspect, emphasised by Roisman and van IJzendoorn, was the
focus on intensive, small-scale work in distinct research groups.508 A huge advantage was that
this supported fidelity of measurement. In general, larger studies have tended to have poor asso-
ciations between the Strange Situation and measures of sensitivity, which has raised questions in
some quarters about whether coders rushed their task.509 Roisman and van IJzendoorn suggest
that an alternative might have been to build collaborative consortia, with data-sharing between
the many small groups. Pursuing just such a strategy, the willingness of researchers in the devel-
opmental tradition of attachment research to share and pool data has reached its apex in forms
of Individual Participant Data meta-analysis in recent years (Chapter 6).510 The strong, some-
what ‘familial’ interpersonal connections of the developmental tradition may be supposed to

have supported the development of such data-sharing initiatives.
Another less-advantageous aspect of Ainsworth’s approach that was followed by the
second generation, and which is interesting in retrospect, was her eschewal of popular
media. Bowlby was keen to use television, radio, magazine articles, and books published
by popular presses to influence clinicians, policy-makers, and the wider public.511 These
methods were not pursued by Ainsworth and they were not generally adopted by the subse-
quent generation of attachment researchers.512 In fact, Ainsworth’s correspondence time and
again reveals her deep discomfort when she was approached with requests for prescriptive
advice. The eschewal of popular media by Ainsworth and the vast majority of her confed-
erates stands out as unusual among their peers. Child development researchers, in general,
have been among the most active in public engagement among psychological researchers,
in part due to the ready public and practitioner audience for knowledge in this area, and the
growing policy problematisation of child development.513 It should be acknowledged that
508 See Roisman, G.I. & van IJzendoorn, M.H. (2018) Meta-analysis and individual participant data synthesis in
child development: introduction to the special section. Child Development, 89(6), 1939–42.
509 The particular object of contention here has been the NICHD sample: NICHD Early Child Care Research
Network (1997) The effects of infant child care on infant–mother attachment security: results of the NICHD Study
of Early Child Care NICHD Early Child Care Research Network. Child Development, 68(5), 860–79.
510 Verhage, M.L., Fearon, R.P., Schuengel, C., et al. (2018) Examining ecological constraints on the intergenera-
tional transmission of attachment via individual participant data meta-analysis. Child Development, 89(6), 2023–37.
511 The disjuncture between the willingness to engage with public media between Bowlby and Ainsworth was
already remarked upon by Karen, R. (1990) Becoming attached. The Atlantic, February 1990: ‘Ainsworth is all but
unknown to the public (and to many psychoanalysts and psychiatrists, who tend to be unfamiliar with trends in
developmental psychology), and yet her fame in the world of infant development exceeds that of John Bowlby
himself . . . Unlike Bowlby, who holds the light as if he were born to it, she doesn’t seem at home.’ Ainsworth, M.
(1990) Letter to John Bowlby, 24 1990, PP/BOW/B.3/8.
512 Some further limited exceptions can be identified. For instance, Sroufe was involved in writing a popular
textbook, and composed a few newspaper articles. Marti Erickson led the Children, Youth & Family Consortium
after finishing the STEEP intervention and the research, and worked directly in family policy. Generally, though,
the difference in scale between Bowlby’s public engagement activities and those of Ainsworth and the second
generation of attachment researchers is profound. As will be discussed further in Chapter 6, the lack of public
engagement by second-generation researchers stands in contrast to the extensive public-facing activities of third-
generation attachment researchers such as Sheri Madigan. Madigan, S. (2019) Beyond the academic silo: col-
laboration and community partnerships in attachment research. Paper presented at International Attachment
Conference, Vancouver, 20 July 2019: ‘The public are looking for information, and will get misinformation unless
we extend research findings to the public. I am taking it as a responsibility of mine to disseminate information to
those who want it, often parents and clinicians.’
513 See e.g. Britto, P.R., Lye, S.J., Proulx, K., et al., and the Early Childhood Development Interventions Review
Group, for the Lancet Early Childhood Development Series Steering Committee (2017) Nurturing care: pro-
moting early childhood development. The Lancet, 389(10064), 91–102; Leach, P. (ed.) (2017) Transforming Infant
Wellbeing: Research, Policy and Practice for the First 1001 Critical Days. London: Routledge.
Conclusion 209
there are exceptions. Patricia Crittenden and Peter Fonagy have been energetic in communi-
cating to public and professional constituencies about attachment theory.514 And the Circle
of Security graphic of the caregiver as a secure base and safe haven has circulated widely as
a visual representation of Ainsworth’s concepts.515 Nonetheless, as Goldberg observed, after
Bowlby ‘many attachment researchers (myself included) have been reluctant to take on this
responsibility’ of knowledge exchange with non-researchers.516
Beyond the model provided by Ainsworth, several further reasons might be identified for
the neglect of public engagement by the second generation of attachment researchers. One
would surely be the lack of incentives for public engagement within American academic life
in the 1980s and 1990s. The second generation also saw directly the problems and misunder-
standings caused by Bowlby’s popularizing texts; there may have been a sense of wanting to
seal themselves off from what they felt unable to mend. A further reason for the lack of public

engagement activities by Ainsworth and most of her successors is likely to have been the ur-
gent priority felt for the development of attachment as a distinct and differentiated scientific
paradigm, a ‘field’ for attachment research.517 This was understood to entail efforts to test and
secure the main aspects of what were judged Ainsworth’s legacy: the Strange Situation and its
expectable caregiving and behavioural correlates. In this regard, as we shall see in Chapter 3,
Main was both the quintessential second-generation researcher and a marked exception.
Few of the second generation were as deeply steeped in and committed to Ainsworth’s ideas
and methods, and most of Main’s work focused on correlates of the Strange Situation in one
way or another. Yet Main was also a radical innovator, introducing new theoretical ideas,
changing the coding of the Strange Situation procedure, and creating assessments of attach-
ment for later in the lifecourse. She also shaped interpretations of Ainsworth and Bowlby.
These developments, discussed in Chapter 3, would establish the methodological and the-
oretical mainstream for the second generation of attachment researchers through the 1990s
and 2000s.
514 See e.g. Spieker, S.J. & Crittenden, P.M. (2018) Can attachment inform decision-making in child protection
and forensic settings? Infant Mental Health Journal, 39(6), 625–41; Fonagy, P. & Higgitt, A. (2004) Early mental
health intervention and prevention: the implications for government and the wider community. In B. Sklarew, S.W.
Twemlow, & S.M. Wilkinson (eds) Analysts in the Trenches: Streets, Schools, War Zones (pp.257–309). Mahwah,
NJ: Analytic Press; Fonagy, P. (2018) Evidence submitted to the Evidence-Based Early-Years Intervention Inquiry.
Science and Technology Committee (Commons). http://data.parliament.uk/writtenevidence/committeeevi-
dence.svc/evidencedocument/science-and-technology-committee/evidencebased-early-years-intervention/
written/77644.pdf.
515 Powell, B., Cooper, G., Hoffman, K., & Marvin, R.S. (2009) The circle of security. In C.H. Zeanah (ed.)
Handbook of Infant Mental Health (pp.450–67). New York: Guilford.

516 Goldberg, S. (2000) Attachment and Development. London: Routledge, p.248. By way of comparison, con-
sider the use of popular forums even for scholarly debates between specialists in evolutionary psychology in the
1990s. Cassidy, A. (2005) Popular evolutionary psychology in the UK: an unusual case of science in the media?
Public Understanding of Science, 14(2), 115–41. The marked contrast between attachment research and evolu-
tionary psychology in this regard has been noted by Granqvist, P. (2020) Attachment, Religion, and Spirituality: A
Wider View. New York: Guilford.
517 Bourdieu, P. (1975) The specificity of the scientific field and the social conditions of the progress of reason.
Social Science Information, 14(6), 19–47; Bourdieu, P. (2004) The Science of Science and Reflexivity, trans. R. Nice.
Chicago: University of Chicago Press. For equivalent processes in another area of the social sciences see Turner,
S.P. (2012) De-intellectualizing American sociology: a history, of sorts. Journal of Sociology, 48(4), 346–63.
210 Mary Ainsworth
Appendix
Table 2.2 Some key concepts in Ainsworth’s writings
Security Personal wellbeing Ainsworth used the term ‘security’ to mean a person’s
and confidence confidence in their efficacy to access the people/resources
to have their needs met. When this confidence is available,
its source was described by Ainsworth as a ‘secure base’.
She also used the term as a category label for a group of
infants in her Strange Situation procedure. These infants
seek physical or distal contact, are readily comforted

following separations, and are able to return to play. This
implied to Ainsworth that they were confident in their
caregiver’s availability.
Attachment Mother An attachment figure is a familiar person who is sought
figure or wished for under conditions of alarm (i.e. when the
attachment system is activated).
Attachment in Continuation Ainsworth described separation anxiety and secure base
adulthood of individual and safe haven dynamics as aspects of relationships in
differences in both childhood and adulthood. And following Bowlby,
attachment from she anticipated some elements of continuity in how
early childhood individuals respond within relationships that have these
across the lifespan attachment components.
The role of these attachment components in adulthood
have been measured very differently, and more or less
directly, by different attachment researchers.
Strange Definitive test A structured laboratory-based observational procedure
Situation of individual developed by Ainsworth. The Strange Situation was
differences in intended to provide a window into a child’s expectations
attachment about their caregiver’s availability based on the history of
their prior interactions. It is a validated proxy for direct
observations of those interactions.
Sensitivity Warmth and Though used more expansively in her early work, the term
tenderness ‘sensitivity’ was a highly technical one for Ainsworth from
the late 1960s. By this she meant the ability of a caregiver
to (i) perceive and to (ii) interpret accurately the signals
and communications implicit in an infant’s behavior,
and given this understanding, to (iii) respond to them
appropriately and (iv) promptly. Ainsworth developed a
scale, unpublished until recently, for assessing caregiver
sensitivity.
Various other measures of sensitivity have subsequently
been developed by attachment researchers. Not all of them
measure sensitivity as technically defined by Ainsworth.
Illustrative Statement: ‘Ainsworth’s Strange Situation procedure showed that sensitive care was associated with
infant security with their attachment figure.’
Mistaken for: Ainsworth’s Strange Situation functions as a definitive test of the extent that individual infants de-
viate from a standard of security, representing their state of wellbeing. This state of wellbeing is associated with the
warmth and tenderness shown by their mother.
Technical meaning: Ainsworth’s Strange Situation is a validated proxy for naturalistic observations of dyads con-
taining an infant and someone who functions, at least to a material extent, as an attachment figure. It is an imper-
fect research instrument based on a small sampling of behaviour, and subject to measurement error. However,
findings using the procedure can be supported by convergence with those from other approaches, such as obser-
vations of caregiver sensitivity.
3
Mary Main, Erik Hesse, and the Berkeley
Social Development Study
Biographical sketch
Mary Main was one of Ainsworth’s first graduate students at the Johns Hopkins University.

However, her distinctive training prior to graduate work meant that she came to the study
of child development from an unusual angle. As an undergraduate she studied classics and
natural sciences at St John’s College, Maryland. This included four years of courses in lit-
erature and philosophy. Since Main had married one of her philosophy professors, Alvin
Main, if she wanted to continue to graduate school she would need to find a programme
nearby. Following her lifelong interest in language, Main applied to Hopkins to study psy-
cholinguistics.1 However, her application was accepted by Ainsworth for graduate work
on child–caregiver attachment. Main completed her doctorate under Ainsworth in 1973.
Her project was the third study to use the Strange Situation, after Ainsworth and Sylvia Bell,
and the results were part of the 106 cases analysed in Patterns of Attachment. In 1973 she
was appointed to a faculty position at the University of California, Berkeley. Settling in at
Berkeley was disrupted by the death of Alvin Main. However, during the 1970s she man-
aged to begin a longitudinal study in the Bay Area, conducting the Strange Situation with
both mothers and fathers. Already in her dissertation, Main’s interest in the work of Robert
Hinde had led her to study conflict behaviours observed in the Strange Situation. She pur-
sued this inquiry in her Bay Area sample, which led to the introduction of the disorganised
attachment classification. A subset of families from the sample were invited back to the la-
boratory when the child was aged six. Main and her group inductively identified associations
between the infant Strange Situation classifications and assessments of the six-year-old’s be-
haviour and family drawings. The Strange Situations also had associations with the form
taken by a parent’s autobiographical narrative. This latter finding led to the development
of the Adult Attachment Interview (AAI).2 The Berkeley study has been described as ‘the
most influential, in-depth, and complex study of intergenerational factors in attachment’ of
its era.3 Together with her second husband and collaborator Erik Hesse, Main introduced
the hypothesis that disorganised attachment may be caused by caregivers’ behaviour that
alarms their infant, and that this alarming behaviour can be predisposed by unresolved ex-
periences of loss or trauma. Building on the work of Bowlby and Ainsworth, contributions
1 Main, M., Hesse, E., & Kaplan, N. (2005) Predictability of attachment behavior and representational processes
at 1, 6, and 19 years of age: the Berkeley longitudinal study. In K.E. Grossmann, K. Grossmann, & E. Waters (eds)
Attachment from Infancy to Adulthood: The Major Longitudinal Studies (pp.245–304). New York: Guilford, p.248.
2 Hesse, E. (2016) The adult attachment interview: protocol, method of analysis, and empirical studies: 1985–
2015. In J. Cassidy & P.R. Shaver (eds) Handbook of Attachment: Theory, Research, and Clinical Applications, 3rd
edn (pp.553–97). New York: Guilford.
3 Karen, R. (1998) Becoming Attached: First Relationships and How They Shape Our Capacity to Love.

212 Mary Main and Erik Hesse
by the Berkeley group revolutionised and redefined the methodology and theory of attach-
ment research.
Introduction
Main was the first of Ainsworth’s graduate students to receive a faculty position and es-
tablish an independent research group. As a result, the most pressing item of business was
replication. Whereas Ainsworth’s original study in the mid-1960s had 26 infant–mother
dyads, and Patterns of Attachment could report 106 infant–mother dyads, Main submitted
a proposal in the mid-1970s to the William T. Grant Foundation for a much larger study,
including both mothers and fathers. The proposal was accepted and Main was able to hire

research assistance to support the recruitment of 189 families in 1976–77. For this replica-
tion and extension of Ainsworth’s study, Main’s sample was selected precisely to be low risk.
All the families were middle-or upper-middle class. No teenage mothers were included,
and none of the parental couples were divorced. Families were screened out if the infant
had been born prematurely or had a low birth weight, or if there had been problems with
the delivery. Families were also screened out if infants had experienced a separation of two
weeks or more from their attachment figures. And only Caucasian or Asian families were
included, to exclude as a confound the adversities black families faced with stigma and op-
pression. Families were also screened out if the father was unemployed, or if infants were in
more than 25 hours of daycare per week, though parents were not excluded on the grounds
of mental illness.
Deliberately unrepresentative of US families, Main’s sample was intended to be low risk,
as a point of comparison against which later studies addressing specific risks or particular
populations could be compared.4 However, it was also anticipated to be large enough for
further detailed study of the difference between secure and avoidant attachment, which was
Main’s central interest in the mid-1970s.5 Main’s inclusion of fathers in the Berkeley lon-
gitudinal study can be situated in the context of shifts in American public and academic
discourse in the 1970s that urged recognition of the significance of the father as a parent
and as an attachment figure.6 However, Ainsworth’s decision to focus on mothers was a prag-
matic one based on her limited resources and the childcare practice of the time in Baltimore.
A decade later, distal changes in discourses around the family combined with the better
at 1, 6, and 19 years of age: the Berkeley longitudinal study. In K.E. Grossmann, K. Grossmann, & E. Waters (eds)
Attachment from Infancy to Adulthood: The Major Longitudinal Studies (pp.245–304). New York: Guilford: ‘We be-
lieved that only after the sequelae “naturally” arising out of enduring differences in attachment relationships have
been delineated can researchers begin to trace—as P.T. Medawar put it in another context—the “variations which
depart” ’ (258).
ation. PP/Bow/J.4/3: ‘The principle aims of the project are a) to test the proposition that avoidance of attachment
figures in infancy predicts restricted affective responsiveness, aggression, and avoidance of adults other than the
parents in early childhood; b) to provide norms for infant behaviour with father in the Strange Situation and in
play settings; c) to compare and contrast mother–infant and father–infant relationships and the influence these
relationships have upon development; d) to relate the child’s behaviour toward peers and caregivers in nursery
school to “joint” classifications of relationships to parent in infancy and to conflict vs harmony in the parent–
parent relationship.’
6 Biller, H.B. (1974) Paternal Deprivation: Family, School, Sexuality, and Society. Lexington, MA: D.C. Heath;
Lamb, M.E. (1975) Fathers: forgotten contributors to child development. Human Development, 18, 245–66.
Introduction 213
resources of the Berkeley study, leading to an expanded lens on parenting as well as a larger
sample.
However, in fact Main did not ultimately have the funds to arrange rigorous coding for
all 378 Strange Situation recordings. A report was made by Main and Weston in 1981 from
a subset of 46 families who were invited to take part in both the Strange Situation at 12 and
18 months, and in an additional assessment of prosocial behaviour.7 Of the infants seen with
mother at 12 months, 68% of the dyads were classified as secure, 28% avoidant, and 4% am-
bivalent/resistant. For these same infants seen with father at 18 months, 59% of dyads were
classified as secure, 35% as avoidant, and 6% as ambivalent/resistant. The distribution was
therefore ‘highly compatible’ with the distributions from Ainsworth’s sample, and indeed are
well aligned with subsequent distributions.8 Main and Weston reported that the attachment
classification of an infant with one parent had a very weak, statistically insignificant associ-

ation with the attachment classification of an infant with the other parent.9 Furthermore,
there was no significant association on the proximity-seeking, contact-maintaining, avoid-
ance, or resistance scales.10
Main and Weston also reported that the attachment classification with each parent made
an independent contribution to prosocial behaviour. This was assessed in a rather unusual
laboratory-based procedure at 12 and 18 months. With a parent in the room, infants were
approached by a clown who sought to play with the child. The clown approached the child
with a mask, then removed the mask, called to the infant, and somersaulted about. The clown
was asked by an adult to leave, and began to protest wretchedly: ‘Extremely realistic sobbing.
The cry lasts 50 seconds. Its realism is so strong that many adults hearing this cry for the first
time are somewhat shaken.’11 Prosociality was rated if infants engaged positively with the
conflict behavior and the readiness to establish new relationships. Child Development, 52(3), 932–40.
8 Ibid. p.938. The rate of ambivalent/resistant was lower and the rate of avoidant attachment was higher than
the distributions reported in a recent international meta-analysis, but this has turned out to be quite usual for
American low-risk samples. Verhage, M.L., Schuengel, C., Madigan, S., et al. (2016) Narrowing the transmis-
sion gap: a synthesis of three decades of research on intergenerational transmission of attachment. Psychological
Bulletin, 142(4), 337–66, Table 4: 60.4% secure, 22.5% avoidant, 17.1% ambivalent/resistant. The meta-analysis
included only those studies that conducted both the Strange Situation and the AAI.
9 Later research debated possible reasons for the small positive association between infant attachment classifi-
cations with respective parents. Spangler emphasised the role of individual infant dispositions. Spangler, G. (2013)
Individual dispositions as precursors of differences in attachment quality: why maternal sensitivity is nevertheless
important. Attachment & Human Development, 15(5–6), 657–72. A recent meta-analysis found that interparental
conflict in non-clinical samples was inversely associated with attachment security (r = −.28). Interparental con-
flict may affect the care provided by both partners, or may directly impact the ambient sense of threat or caregiver
availability in the home for the baby in all relationships. Tan, E.S., McIntosh, J.E., Kothe, E.J., Opie, J.E., & Olsson,
C.A. (2018) Couple relationship quality and offspring attachment security: a systematic review with meta-analysis.
Attachment & Human Development, 20(4), 349–77. It should also be highlighted that in newer studies with pre-
schoolers—a period marked by an increase in father involvement, especially more recently—concordance is much
higher, e.g. Bureau, J.-F., Martin, J., Yurkowski, K., et al. (2017) Correlates of child–father and child–mother attach-
ment in the preschool years. Attachment & Human Development, 19, 130–50.
10 Relatively few longitudinal studies have conducted attachment measures with multiple caregivers. As a con-
sequence, little is known about the implications of convergent or divergent patterns of attachment with different
caregivers. The only reported data on this come from Fonagy and colleagues, who reported that—whilst there were
no effects for infancy—discrepant attachment classifications at age five in a modified Strange Situation were asso-
ciated with conduct problems. Fonagy, P., Target, M., Steele, M., et al. (1997) Morality, disruptive behavior, bor-
derline personality disorder, crime, and their relationships to security of attachment. In L. Atkinson & K.J. Zucker
(eds) Attachment and Psychopathology (pp.223–74). New York: Guilford, p.247. However, in subsequent decades
there has been no attempt to replicate these findings.
11 Main, M., Weston, D.R., & Wakeling, S. (1979) Concerned attention to the crying of an adult actor in infancy.
Paper presented at Society for Research in Child Development, San Francisco, March 1979. PP/Bow/J.4/3.
clown when invited to play and seemed sympathetic to the clown’s crying. Children with
two secure attachment relationships scored highest on prosociality; those with two inse-
cure attachment relationships scored lowest; and those with one secure and one insecure at-
tachment were rated in the middle.12 Children with avoidant attachment relationships were
found to often remain impassive in response to the crying clown, and their mothers were
more likely to show a derisive facial expression when the clown began crying.13 The Clown
procedure appeared to especially stress children in avoidantly attached dyads. Unable to seek
their caregiver directly for support, around half showed a variety of ‘conflict behaviours’ such
as ‘rocking back and forth whilst staring into space; assuming odd postures; engaging in
odd tension movements; sudden inappropriate or empty laughter; vocalising to the wall in
a “social” manner; an odd “frozen” facial expression; lying on the floor in foetal position
with eyes closed’.14 These behaviours were especially common in those who were avoidantly

attached with both parents. Main took this to signify that conflict behaviours were elicited
especially when the child experienced the relative unavailability of both caregivers as a safe
haven. In a later reflection, however, Main concluded that the association between avoidance
in the Strange Situation and conflict behaviours in the clown study may have been caused by
caregivers ‘who were harsh or frightened in addition to simply displaying aversion to phys-
ical contact’.15 The implication is that the close relationship Main initially would perceive
between avoidance and conflict behaviour was, in part, an artefact of a sample in which the
antecedents of avoidant attachment and the antecedents of conflict behaviours happened to
co-occur.
When participants were aged around six years old, in 1982, 40 families were invited back
to the laboratory, stratified by infant attachment classification.16 No families with parents
who had divorced in the intervening years were included.17 A small grant from the Society
for Research in Child Development helped cover the costs of the study, and Main’s lively
and intriguing classes attracted a talented group of Berkeley graduate students to work
on the project: Donna Weston, Carol George, Judith Solomon, Nancy Kaplan, and Ellen
Richardson. Each family was first visited at home by a member of the team, to take consent
and build rapport. On arrival in the laboratory, the family viewed excerpts together from
one of the films made by James Robertson (‘Thomas: ten days in fostercare’). Study of family
responses to the distressing, evocative film of a 14-day separation and reunions with father
12 For a review of developments and unanswered questions since Main and Weston’s report see Dagan, O. &
Sagi-Schwartz, A. (2018) Early attachment network with mother and father: an unsettled issue. Child Development
Perspectives, 12(2), 115–21.
ation. PP/Bow/J.4/3.
14 Main, M. & Weston, D. (1981) The independence of infant–mother and infant–father attachment relation-
ships: security of attachment characterises relationships, not infants. PP/Bow/J.4/3.

15 Main, M. (1999) Disorganized Attachment in Infancy, Childhood, and Adulthood: An Introduction to the
Phenomena. Unpublished manuscript, Mary Main & Erik Hesse personal archive.
16 There is a discrepancy regarding the number of families in the six-year follow-up. In 1985 Main states that
‘forty mothers, fathers, and their 6-year-old children (24 male, 19 firstborn) formed the sample of participants
in this 1982 study’: Main, M., Kaplan, N., & Cassidy, J. (1985) Security in infancy, childhood, and adulthood: a
move to the level of representation. Monographs of the Society for Research in Child Development, 50, 66–104, p.79.
However, Main stated on several later occasions that she and Ruth Goldwyn initially developed the AAI in 1982–
83 on 36 transcripts and then tested it on the remaining 66. If interviews were conducted with both mother and
father, then this would mean 51 families were called back. The likely resolution to this discrepancy is that further
data collection took place in 1983.
17 George, C. (1984) Individual differences in affective sensitivity: a study of five-year olds and their parents.
Unpublished doctoral dissertation, University of California, Berkeley.

Introduction 215
and mother was the basis of Carol George’s doctoral dissertation with Main.18 George found
that the parents from insecure dyads and children from insecure dyads were more likely to
show signs of anger, disgust, and sarcasm in response to viewing the child’s behaviour in the
film. Yet as well as a stimulus for George’s study, the film served as a prime for the attachment
behavioural system and caregiving behavioural systems ahead of subsequent assessments.19
The parents then left the room for individually administered life-history interviews (the
AAI). The children remained in the playroom with an examiner. For 15 minutes they were
asked to make a drawing of their choice, then one of their family. Next, the examiner asked
the child to respond to six pictures of child–caregiver separations, giving their thoughts on
what the separated child might feel or do (the Separation Anxiety Test).20 The child and par-
ents were reunited after about an hour’s separation; in half the cases the mother returned
first, and in half the cases the father. The reunion was filmed.

Main’s approach to the analysis of her six-year data was deeply influenced by the five
years she had spent as a graduate student within Ainsworth’s research group. Main had been
drenched to the bone in Ainsworth’s ideas at a point that attachment research as an empirical
paradigm was little more than this single laboratory at Baltimore. Ainsworth’s identification
of three patterns of infant attachment was felt by Main with incredible force. Why three?
Researchers outside of the little laboratory often simply drew the conclusion that the result
was arbitrary, primarily an artificial construction of diversity into pragmatic categories.21
But Main had conducted 50 Strange Situations as part of her dissertation research, and was
therefore in a position to see the Ainsworth system from the inside and to identify some-
thing remarkable: it was astonishingly un-arbitrary. Naturally there were marginal cases be-
tween the three groups. Yet the categories themselves seemed to Main to work unnervingly,
miraculously well in offering the three basic images to which all the infants approximated
in her dissertation sample. There really did seem to be, beyond any reasonable expectation,
three patterns of attachment displayed not just in the 26 middle-class Baltimore infant–
mother dyads recruited by Ainsworth for her initial study, but by the 106 middle-class
infant–mother dyads from the studies by Ainsworth, Bell, and from Main’s doctoral work.
The effectiveness of the Ainsworth classifications seemed to Main to imply something
about the fundamental reality of human emotional life. It is hardly unusual for a doctoral stu-
dent to look out at the whole world through the lens of her supervisor’s work. Chapter 2 dis-
cussed how Blatz’s work on security played a foundational role for Ainsworth in her thinking
about adult personality, the behaviour of infants in Uganda, and ultimately individual dif-
ferences in the Strange Situation. However, when Main started to look at the world through
18 Ibid. The role of an attachment prime (the video of separation and reunion of a child) prior to conducting the
AAI in the Berkeley study is not generally known and has not been discussed in the published literature. It could be
that this made no difference, but no study has addressed the question.
19 The question of whether priming is beneficial for or contaminates measures such as the AAI and the
Experiences in Close Relationships scale in their capacity to measure adult attachment remains an open one and
is rarely discussed. It is to be hoped that the growth of priming research will help bring this matter to light, and to
explicit examination (Chapter 6) No doubt part of the issue is that both the AAI and ECR have been interpreted as
measuring a ‘thing’ called adult attachment, when in fact matters are more complicated.
20 Hansburg, H.G. (1972) Adolescent Separation Anxiety: A Method for the Study of Adolescent Separation
Problems. Springfield: Thomas; Klagsbrun, M. & Bowlby, J. (1976) Responses to separation from parents: a clinical
test for children. British Journal of Projective Psychology, 21, 7–21; Kaplan, N. (1987) Individual differences in 6-
year olds’ thoughts about separation: predicted from attachment to mother at age 1. Unpublished doctoral disser-
tation, Department of Psychology, University of California, Berkeley, CA.
21 A significant recent example of this view is Gaskins, S. (2013) The puzzle of attachment. In N. Quinn & J.M.
Mageo (eds) Attachment Reconsidered: Cultural Perspectives on a Western Theory (pp.33–66). London: Palgrave.
the lens of the Ainsworth infant attachment classifications, two surprising things happened.
First, Main came to the exhilarating conclusion that the Ainsworth patterns represented the
three basic strategies used by all humans, whether infants or adults, for handling distress in
interpersonal contexts. Second, exceptions to these patterns therefore took on special sig-
nificance and interest. Main’s own powers as an observer of all scruffy particularities became
combined with a theoretical focus on exceptions, developing a fourth category for behaviour
suggesting a disruption in strategy.22 On this basis, Main developed an account of individual
differences in attachment as reflecting strategies for the direction of attention with respect to
attachment-related perceptions and memories.
These ideas are not widely understood, especially as several key texts by Main were ultim-
ately not published. However, they shaped the development of the AAI and other measures
developed by Main, Hesse, and collaborators in the Berkeley group in the 1980s and 1990s.

The theory and research of Main and colleagues has repeatedly been situated as a ‘revolu-
tionary shift’ in attachment research.23 Allen and Miga described how the work ‘permits
and indeed forces a new conceptualization of attachment, and opens up important avenues
for assessing the attachment system beyond childhood’.24 Holmes described it as the start of
attachment research ‘Phase 2’, and Toth observed that it brought about the interest in attach-
ment theory among the clinical community that Bowlby had sought but not managed on
his own.25
The work of Main and colleagues has offered innovative, persuasive methods and theory
with great heuristic power and value for researchers, especially in allowing them to extend
the applications of attachment theory beyond infancy. However, at times, the qualities of
these methods and theory have led them to enter into circulation without fine-grained dis-
tinctions being drawn, especially in the use of concepts and categories. This concern has
been widely recognised by developmental attachment researchers in recent years, including
by Main and her collaborators themselves.26 There are a variety of reasons for this, which are
discussed in this chapter. Of course, matters such as terminological precision and the exact
articulation of categories are not really the priorities of pragmatic researchers, for whom the
essential concern is whether measurement and prediction can be pursued more effectively or
persuasively.27 Yet, looking over the decades in a historical perspective, the pragmatic con-
cerns of individual researchers have at times led to collective problems. Misunderstandings
and miscommunication that are only minor irritants or not particular priorities for any in-
dividual researcher may, for a field, cause wide-ranging issues, played out incrementally over
22 Some of the indicators of disorganised attachment—such as asymmetric facial movements—are rarely, if ever
coded. They seem to be, at least in part, a product of Main’s unusual eye for observational detail.
23 Van IJzendoorn, M.H. & Bakermans-Kranenburg, M.J. (1997) Intergenerational transmission of attachment.
A move to the contextual level. In L. Atkinson & K.J. Zucker (eds) Attachment and Psychopathology (pp.135–70).
New York: Guilford, p.138; see also Durham-Fowler, J. (2013) An interview with Arietta Slade. DIVISION/Review,
7, 39–40.
24 Allen, J.P. & Miga, E.M. (2010) Attachment in adolescence: a move to the level of emotion regulation. Journal
of Social and Personal Relationships, 27(2), 181–90, p.182.

25 Holmes, J. (2009) Exploring in Security: Towards an Attachment- Informed Psychoanalytic Psychotherapy.
London: Routledge, p.4; Toth, S.L., Rogosch, F.A., & Cicchetti, D. (2008) Attachment-theory informed interven-
tion and reflective functioning in depressed mothers. In H. Steele & M. Steele (eds) The Adult Attachment Interview
in Clinical Context (pp.154–72). New York: Guilford, p.154.
26 E.g. Main, M., Hesse, E., & Hesse, S. (2011) Attachment theory and research: overview with suggested applica-
tions to child custody. Family Court Review, 49(3), 426–63.

27 Lakatos, I. (1970) Falsification and the methodology of scientific research programmes. In I. Lakatos & A.
Musgrave (eds) Criticism and the Growth of Knowledge (pp.91–196). Cambridge: Cambridge University Press.
Conditional strategies 217
decades, with costs mounting. In particular, researchers have tended to neglect or skip over
the underpinning logic of the contributions of the Berkeley group, to talk past one another
as different meanings are given to concepts, and to neglect questions with important conse-
quences for theory, research, and clinical practice.
Conditional strategies
The primacy of attention
During the years that Main was completing her doctorate in the Ainsworth laboratory, a fun-

damental question was the meaning of the behaviour shown by infants in Type A dyads. In
1969 Bowlby published Attachment, situating proximity with the caregiver as the set-goal of
the attachment system. Avoidance of the caregiver on reunion therefore seemed at first sight
to falsify Bowlby’s theory. In fact, Ainsworth’s home observation data were beginning to re-
veal a context for this behaviour, in the less-sensitive caregiving received by infants in avoid-
antly attached dyads. A year into her doctorate, at Ainsworth’s urging, in 1970 Main posted
a manuscript to Bowlby for feedback. In a wide-ranging theoretical text entitled ‘Infant play
and maternal sensitivity in primate evolution’, Main made a variety of gutsy points. One was
to question Bowlby’s assumption that the environment within which humans evolved would
reward a single ‘setting’ for the attachment system. Main proposed that the attachment
system would need to be capable of calibration to a variety of environments, favourable and
adverse. Sensitive caregiving is optimal, and the provision of a secure base would help a child
to explore and learn. However, less-sensitive caregiving could be expected to elicit responses
that would support survival even in adverse conditions. The vigilance of infants in ambiva-
lent/resistant dyads was readily explicable in this account, as it helped retain proximity to a
caregiver in potentially dangerous environments.28 Avoidant behaviour, however, remained
a puzzle for which Main, as yet, had no explanation.
Main’s 1973 doctoral research headlined results showing the higher scores on cognitive
development of the children from secure dyads.29 However, perhaps the most critical finding
for Main’s later thinking was the observation that infants in secure dyads in her sample had
the longest attention spans during play as toddlers. This suggested to her that children in
insecure dyads had other behavioural systems or responses active and sapping attention
from exploration. During her doctorate, Main also served as a coder for Berry Brazelton in a
micro-analysis of filmed interactions between five mother and infant dyads, which occurred
weekly over the first months of life. The focus of the analysis was on ‘cycles of looking and
non-looking, or attention and nonattention’.30 Though Brazelton had come into the study
28 Main, M. (1970) Infant play and maternal sensitivity in primate evolution. PP/Bow/J.4/1: ‘The strange situ-
ation behaviour of infants whose mothers are and are not sensitive suggests different probabilities of survival in
the environment of evolutionary adaptedness. In novel conditions, certainly insofar as the mother is present, the
infant who can explore is better off; in conditions of stress or alarm, the infant who seeks and maintains contact
with his mother [is better off].’
University.
30 Brazelton, T.B., Koslowski, B., & Main, M. (1974) The origins of reciprocity: the early mother–infant inter-
action. In M. Lewis & L. Rosenblum (eds) The Effect of the Infant on its Caregiver (pp.49–76). New York: Wiley, p.49.
with the assumption that the infant would display continuous attention within interactions,
the stop-frame analysis conducted by Main revealed a flexible movement between looking
at and away from the caregiver. The researchers concluded that ‘looking away behaviour re-
flects the need of each infant to maintain some control over the amount of stimulation he
can take in via the visual mode in such an intense period of interaction’.31 This looking away
strategy was adopted more frequently when the caregiver was insensitive in offering stimu-
lation that was not well aligned with the infant’s pacing. Drawing on contemporary ideas
from ethology, the researchers called this an ‘approach-withdrawal model’, a term that had
been used to describe the way that animals might flexibly use approach and withdrawal be-
haviours to modulate the intensity of stimulation, and in this way remain well regulated.32
This conclusion was supported by the finding that infant looking away behaviours were less
common when caregivers were alert to indications that the infant was becoming ‘upset or

disintegrated’ or otherwise having difficulty regulating:33
When the infant demonstrates unexpected random behaviour, such as the jerk of a leg or
an arm, the mother responds by stroking or holding that extremity, or by making a directed
use of that extremity to jog it gently up and down, thereby turning an interfering activity
into one that serves their interaction. In these ways she might be seen to teach the infant
how to suppress and channel his own behaviour into a communication system.34
The authors criticised Bowlby for focusing too strictly on behaviour, following his etho-
logical models, and his neglect of the role of attentional processes as a flexible resource for
maintaining regulation and coherently structured interaction. They speculated that the
interdependency of rhythms of attention and inattention within the relationship between
infant and caregiver ‘seemed to be at the root of their attachment’.35
After Main finished her doctorate in 1973, she continued reflecting on avoidant attach-
ment. Her assumption was that, in contravening the predicted set-goal of the attachment
system so directly, it would be ‘predictive of interactive and affective disturbance’.36 She de-
veloped scales for analysing caregiver–infant touch in the Ainsworth home observation
data. Main found that the mothers in dyads classified as avoidant in the Strange Situation re-
buffed physical contact with their babies much more frequently at home. They often rejected
their infants’ attachment behaviour, and more frequently displayed anger and flat affect in
response to displays of distress. Over time, this increased the frequency with which their in-
fants made tentative or circuitous approaches. In February 1974, Ainsworth wrote to Bowlby
that ‘we have found plenty of evidence that the mothers of A babies dislike physical con-
tact, and that it is through behaviour relevant to physical contact that they (at least in large
31 Ibid. p.60.
32 Ibid. The term ‘approach/withdrawal’ was drawn from Schnierla, T.C. (1965) Aspects of stimulation and or-
ganisation in approach/withdrawal processes underlying vertebrate behavioural development. In D. Lehrman, R.
Hinde, & E. Shaw (eds) Advances in the Study of Behaviour, Vol. 1 (pp.1–74). New York: Academic Press. On
Chance’s ethological work on gaze aversion see Kirk, R.G. (2009) Between the clinic and the laboratory: ethology
and pharmacology in the work of Michael Robin Alexander Chance, c. 1946–1964. Medical History, 53(4), 513–36.
33 Brazelton, T.B., Koslowski, B., & Main, M. (1974) The origins of reciprocity: the early mother–infant inter-
action. In M. Lewis & L. Rosenblum (eds) The Effect of the Infant on its Caregiver (pp.49–76). New York: Wiley, p.64.
34 Ibid. p.65.
35 Ibid. p.74.
University.
part) express rejection. Mary’s theory is that this puts babies in a double bind, for they are
programmed to want contact and yet are rebuffed (or at least have unpleasant experiences)
when they seek it.’37
Main’s undergraduate background in liberal arts at St John’s College made her aware that
novelists such as Hardy and Dostoevsky had ‘long been aware of the attraction irrationally
implicit in rejection’.38 Whilst at the College, she married her philosophy professor, Alvin
Nye Main. In the early 1970s, Alvin Main drew Mary Main’s notice to a potential evolu-
tionary basis for the attraction that may be prompted even in the context of rejection.39 In
The Voyage of the Beagle, Darwin noted that the amphibious sea lizard, when frightened,
would never flee towards the water. To test this further, Darwin repeatedly frightened sea
lizards, and then threw them into the water. They repeatedly swam right back to him on the
land, directly towards their ‘attacker’. Darwin interpreted this otherwise strange and coun-

terproductive behaviour in terms of the sea lizard’s evolutionary history:
Perhaps this singular piece of apparent stupidity may be accounted for by the circumstance,
that this reptile has no natural enemy whatever on shore, whereas at sea it must often fall
a prey to the numerous sharks. Hence, probably, urged by a fixed and hereditary instinct
that the shore is its place of safety, whatever the emergency may be, it there takes refuge.40
Alvin and Mary Main discussed the analogy with the infant of a caregiver disposed to re-
jecting or angry behaviour. For the sea lizard, ‘if the source of the attack is “the shore” itself,
the shore is nonetheless returned to as a haven of safety’.41 Similarly, according to Bowlby’s
theory of the attachment behavioural system, the infant would be disposed by an evolution-
arily channelled mechanism to seek proximity with the caregiver when alarmed as the haven
of safety, even if the caregiver is also experienced as rejecting or threatening.42—hence a
‘double bind’.
Alvin Main died of cancer in the summer of 1974, just following the end of Mary Main’s
first year as an assistant professor at Berkeley. In the grief of this loss, Mary Main experienced
a period of writer’s block that lasted around two years.43 One text written during the period
was a paper presented by Main and her friend Everett Waters, delivered in July 1975 to the
International Society for the Study of Behavioural Development at the University of Surrey.
Waters had earlier studied gaze aversion shown by infants in relation to the stranger. He pro-
posed that gaze aversion can best be understood as an attempt to ‘cut off ’ the stimulus from
37 Ainsworth, M. (1974) Letter to John Bowlby, 1 February 1974. PP/BOW/B.3/4. The ‘double bind’ would later
be redubbed a ‘paradoxical injunction’ by Main and Hesse in the 1990s.

38 Main, M. (1977) Analysis of a peculiar form of reunion behavior seen in some daycare children: its history
and sequelae in children who are home-reared. In R. Webb (Ed.), Social Development in Childhood (pp.33–78).
Baltimore: Johns Hopkins University Press, p.56.
39 That Alvin Main first drew Mary Main’s attention to the passage in Darwin is mentioned in the acknowledge-
ments to Hesse, E. & Main, M. (2006) Frightened, threatening, and dissociative parental behavior in low-risk sam-
ples: description, discussion, and interpretations. Development & Psychopathology, 18(2), 309–43.
40 Darwin, C. (1839, 1972) The Voyage of the Beagle. New York: Bantum, pp.334–5.
42 Cf. Ainsworth, M.D.S. (1977) Attachment theory and its utility in cross-cultural research. In P.H. Leiderman,
S.R. Tulkin, & A. Rosenfeld (eds) Culture and Infancy: Variations in the Human Experience (pp.49–67).
New York: Academic Press: ‘Infants may become attached to mothers who are rejecting, punitive, or actually
brutal’ (52–3).
43 Ainsworth, M. (1976) Letter to John Bowlby, 16 September 1976. PP/BOW/B.3/5.
perception, which serves to modulate tendencies that it might evoke.44 This agreed with the
conclusions of Main’s work with Brazelton. Main and Waters applied this idea to the be-
haviour of infants in Type A dyads seen in the Strange Situation.45 They speculated that, as
well as modulating arousal, visual cut-off allows the infant to remain close to a rejecting or
angry parent—without ‘falling into “all or nothing” response patterns which cannot be ter-
minated voluntarily such as approach for comfort, or crying’.46 In Ainsworth’s observations,
the children from dyads classified as Group A were precisely, and apparently paradoxically,
those who showed more distress and clinginess at home than other children (Chapter 2). In
Main’s only published work from this period, a book chapter in a volume dedicated to the
memory of her late husband, she proposed that the child who shows avoidant behaviour
turns to objects of attention in the environment that will not add to heartache, or that might
even provide the relief of distraction. In this way the child may retain control and flexibility

of response.47 Main offered her qualitative impression, however, that infants in avoidantly
attached dyads were still more disposed to show conflict behaviours in the home than infants
in secure dyads.48
Ainsworth had repeated the Strange Situation with the first wave of her sample two weeks
after the first administration (Chapter 2). All seven of the infants in avoidantly attached
dyads displayed conflict behaviour as they broke from their avoidance in panicked approach
to the caregiver. The intensified activation of the attachment system from yet more separ-
ations, Main supposed, made cut-off and therefore avoidance impossible. As a consequence,
these infants were no longer able to stay well regulated and near to the caregiver. Instead
they were thrown into the ‘all or nothing’ response of direct approach, despite their concerns
about the rebuff their experiences of their caregiver had led them to expect. The infants’ ap-
proach was mottled by conflict behaviours as markers of this concern.49 These reflections
were further developed when Main spent September 1977 to May 1978 as a Fellow at the
Centre for Interdisciplinary Research in Bielefeld, West Germany. The focus that year was on
the application of biological principles to observable behaviour, and the other Fellows rep-
resented future leaders across a wide variety of disciplines. Alongside Main, other attendees
included the biologists Richard Dawkins and John Maynard Smith, Robert Hinde’s student
and collaborator Patrick Bateson, Harry Harlow’s student and collaborator Stephen Suomi,
44 Waters, E., Matas, L., & Sroufe, L.A. (1975) Infant’s reactions to an approaching stranger: description, valid-
ation, and functional significance of wariness. Child Development, 46, 348–56.

45 They also reference the work of Daniel Stern as an influence on their thinking about gaze aversion and
regulation: Stern, D.N. (1974) Mother and infant at play: the dyadic interaction involving facial, vocal and
gaze behaviours. In M. Lewis & L.A. Rosenblum (eds) The Effect of the Infant on its Caregiver (pp.187–213).
New York: Wiley.
46 Main, M. & Waters, E. (1975) Autism and adaptation. Paper presented at International Society for the Study of
Behavioural Development, University of Surrey, July 1975. PP/Bow/J.4/1.

48 Main, M. (1977) Sicherheit und Wissen. In K.E. Grossmann (Ed.), Entwicklung der Lernfahigheit in der
Sozialen Umwel (pp.47–95). Munich: Kindler: ‘Behaviours indicative of conflict should be expected of such a
baby—such behaviours as in fact are found’ (68).
49 Reflecting decades later on the Ainsworth readministration of the Strange Situation, as well as later data
from the Uppsala Longitudinal Study, Main and colleagues concluded that ‘the magnitude of stress invoked by
the stressful re-encounter with the strange situation may be sufficient to break the pattern of defensive avoidance,
yielding an increased necessity of approach and the appearance of “disorganized” approach/avoidance behaviors
instead’. Granqvist, P., Hesse, E., Fransson, M., Main, M., Hagekull, B., & Bohlin, G. (2016) Prior participation in
the strange situation and overstress jointly facilitate disorganized behaviours: implications for theory, research and
practice. Attachment & Human Development, 18(3), 235–49, p.244.
and the child psychiatrist and psychoanalyst Robert Emde.50 The Fellowships were coord-
inated by the ethologist Klaus Immelmann and by Klaus and Karin Grossmann, who were
in the process of attempting the first replication of Ainsworth’s study outside of America.
Fellows engaged in structured and unstructured group conversations at least two or three
times a week with biologists and ethologists.
During her Fellowship in Bielefeld, Main applied ethological principles to the interpret-
ation of avoidant attachment behaviour as seen in the Strange Situation. Ethology asked four
questions of behavioural sequences (Chapter 1): first, the contribution of the behaviour for
species survival or reproduction; second, how the behaviour came about through natural
selection; third, the behaviour’s underpinning mechanisms; and fourth, how it develops
over the lifespan. Putting these questions to avoidant behaviour in the Strange Situation set
Main on a collision course with Bowlby. Bowlby, and following him Ainsworth, had pre-

sumed that the environment within which humans evolved would have predisposed infants
to seek their caregiver when alarmed, as their ultimate source of protection. Other behav-
iours were therefore abnormal and pathological. However, Maynard Smith and Dawkins
were in the process of introducing a new account of the evolutionary biology of mating and
reproductive behaviour, based on game theory. In game theory, the actions of individuals are
treated as moves in a game, responsive to the environment and its potential costs and bene-
fits. Applied to evolutionary processes, game theory suggested that species may develop a
repertoire of strategies that can be initiated in response to various expectable environments
and their costs and benefits.
In papers from 1979, Maynard Smith and Dawkins argued that the diversity of successful
mating strategies could be explained if it was assumed that natural selection had predisposed
a repertoire of behavioural patterns for achieving reproductive success in diverse circum-
stances.51 The direct achievement of mating would be sought when conditions were favour-
able. However, a range of more circuitous routes to mating success were also available when
a direct approach was not readily feasible. The latter were termed ‘conditional strategies’ by
Maynard Smith and Dawkins. The term is a technical one. It does not imply that the direct
achievement of a function or goal is ‘unconditional’; direct expression of a goal-oriented
response is usually dependent on a facilitative environment. Rather, the term ‘conditional’
was intended to refer to expectable behavioural sequences oriented towards at least partial
achievement of a function or goal that would be preferred if the direct strategy could be an-
ticipated to be unsuccessful. Main took this model of reproductive strategies and applied
it to the Strange Situation. Just as the environments for reproductive success varied for the
insects, reptiles, and birds of interest to Dawkins and Maynard Smith, so the caregiving en-
vironments of infants was also one with expectable variation in human evolutionary history.
From Bielefeld, Main sent Bowlby a manuscript entitled ‘Avoidance of the attachment
figure under stress: ontogeny, function and immediate causation’ towards the end of 1978.
She started from the ethological premise that ‘maternal investment varies. Some desert and
all must wean their infants. There must be a strategy for survival under these conditions.’52
50 A set of papers reporting reflections on the year was published as Immelmann, K., Barlow, G., Petrinovitch,
L., & Main, M. (1981) Behavioral Development: The Bielefeld Interdisciplinary Project. Cambridge: Cambridge
University Press.
51 Smith, J. M. (1979) Game theory and the evolution of behaviour. Proceedings of the Royal Society London,
B, 205(1161), 475–88; Brockmann, H.J. & Dawkins, R. (1979) Joint nesting in a digger wasp as an evolutionarily
stable preadaptation to social life. Behaviour, 71(3), 203–44.
She acknowledged that, at first sight, avoidant behaviour seen in the Strange Situation ap-
pears to disprove Bowlby’s position that proximity will be sought when the attachment
system is activated. However, the observation can be reconciled with the theory if it is as-
sumed that ‘avoidance of the attachment figure may function as a kind of conditional strategy
for proximity maintenance’. Conditional proximity supplies a stand-in for a relationship that
would offer genuine welcome. The infant can remain close by to a caregiver who might re-
buff them if they attempted a direct approach, a rebuff that would be emotionally painful and
might further reduce the caregiver’s availability. To set out the apparent paradox: the relin-
quishment of full proximity implied by avoidance is part of its effectiveness, part of its basis
for a qualified hope of a qualified caregiver availability.
Main later developed these ideas in two groundbreaking publications: a short note from
1979 entitled ‘The “ultimate” causation of some infant attachment phenomena’, and the 1981

chapter ‘Avoidance in the service of proximity: a working paper’ published in a book of pa-
pers by the Fellows at Bielefeld and co-edited by Main. In both texts, Main criticised Michael
Lamb and colleagues, who had stated that Bowlby’s position implies that infant avoidant be-
haviour is maladaptive since proximity with the caregiver is the basis for protection.53 Main
countered with an alternative reading of Bowlby. Bowlby himself had referred in passing
to ‘insecure attachment’ as a ‘strategy’ in Separation, though his discussion had been of the
clinginess of the ambivalent/resistant pattern.54 And Main highlighted Bowlby’s emphasis
on the evolutionary basis of behaviour. From an evolutionary perspective, avoidance could
be interpreted as a proactive response by the infant that ‘paradoxically permits whatever
proximity is possible under conditions of maternal rejection’.55 Evolutionary processes
would have selected for avoidance as one part of the infant repertoire for responding to care-
givers, since infants who are able to avoid antagonising their caregivers or making demands
that their caregiver will rebuff are more likely to have survived. Infants successfully utilising
an avoidant strategy maintain an indirect but real proximity to their caregiver, as well as the
regulatory control to continue to be responsive to the environment.56
Main also considered the ethological question of the behaviour’s underpinning mechan-
isms. She regarded the phenomenon as, essentially, ‘active visual, physical and communi-
cative avoidance’ of the caregiver in the context of an activation of the attachment system.57
53 Rajecki, D.W., Lamb, M.E., & Obmascher, P. (1978) Toward a general theory of infantile attachment: a com-
parative review of aspects of the social bond. Behavioral and Brain Sciences, 1(3), 417–36.
54 Bowlby, J. (1973) Separation. New York: Basic Books: ‘There are, however, persons of all ages who are prone
to show unusually frequent and urgent attachment behaviour and who do so both persistently and without there
being, apparently, any current conditions to account for it. When this propensity is present beyond a certain degree
it is usually regarded as neurotic. When we come to know a person of this sort it soon becomes evident that he has
no confidence that his attachment figures will be accessible and responsive to him when he wants them to be and
that he has adopted a strategy of remaining in close proximity to them in order so far as possible to ensure that they
will be available’ (165).
55 Main, M. (1979) The ‘ultimate’ causation of some infant attachment phenomena: further answers, further
phenomena, further questions. Behavioral and Brain Sciences, 2, 640–43, p.643.

56 Main, M. (1981) Avoidance in the service of proximity: a working paper. In K. Immelmann, B. Barlow, L.
Petrovich, & M. Main (eds) (1981) Behavioral Development: The Bielefeld Interdisciplinary Project (pp.694–9).
Cambridge: Cambridge University Press, p.686. Lamb would appear to later acknowledge the point: Lamb, M.,
Thompson, R.A., Gardner, W., & Charnov, E.L. (1985) Infant–Mother Attachment: The Origins and Developmental
Significance of Individual Differences in the Strange Situation. Hillsdale, NJ: Lawrence Erlbaum: It is ‘questionable
whether it is wise to view infantile behaviour as preadapted to an evolutionary niche that consists primarily of a
sensitively responsive adult . . . Evolution typically does not equip individuals with a single ideal pattern of behav-
iour, but rather with a repertoire of responses that may be selectively applied in different circumstances’ (49).
57 Main, M. & Weston, D. (1982) Avoidance of the attachment figure in infancy: descriptions and interpretations.
In C. Parkes & J. Stevenson-Hinde (eds) The Place of Attachment in Human Behavior, pp.31–59. New York: Basic
Books, p.31.
Yet, drawing on her work with Brazelton and with Waters, the paradigmatic form of avoid-
ance for Main was gaze aversion. In Bielefeld initially, and then to a greater extent in 1980
during a second visit to the Grossmanns after their move to Regensburg, Main gradually
came to the conclusion that gaze aversion was the most potent external marker of a fun-
damental internal process underpinning the visual, physical, and communicative qualities
alike. This internal process was the redirection of attention away from the caregiver. This
would inhibit the activating conditions of the attachment behavioural system. Gaze aversion
and other forms of avoidant behaviour were therefore not just keeping the wish to approach
from the caregiver’s notice. They were also keeping this wish from the baby itself.58 A ver-
sion of ‘Avoidance in the service of proximity’ was later published in a celebratory volume in
honour of Bowlby called The Place of Attachment in Human Behavior, edited by Colin Parkes
and Joan Stevenson-Hinde. Other than some sections cut for concision, the only substantive

change came in the conclusion. There it was argued that in infant avoidance of an attachment
figure, ‘we might better conceive of thought and behaviour as becoming actively reorganised
away from the parent and the memory of the parent’.59
However, in the reception of Main’s ideas from this period, the idea of attachment be-
haviour as reflecting attentional processes was little noticed. Instead, her application of evo-
lutionary game theory in conceptualising the Ainsworth classifications as ‘strategies’ took
centre stage, without bringing the attentional model with it from the wings. Given that both
ideas are emphasised across Main’s writings in the 1980s and 1990s, the period in which
the received image of her ideas was established, this reception is curious at first sight. Three
suggestions can be offered for the greater popularity of the idea of attachment strategies
to the exclusion of attentional processes in the reception of Main’s ideas. First, attention
seemed to already be encompassed by the concept of ‘internal working models’. Main’s em-
phasis on attention therefore did not seem salient or novel, even if, in practice, when in-
ternal working models were defined or operationalised the focus was rarely on attention.
Second, the concept of attachment strategies addressed a problem with Bowlby’s theory.
Bowlby’s presumption that the secure response was the one primed by evolution offered no
basis for understanding why avoidant and ambivalent/resistant behaviour could be seen
in sample after sample studied using the Strange Situation. By contrast, Main’s account of
58 Main, M. (1981) Avoidance in the service of proximity: a working paper. In In K. Immelmann, B. Barlow,
L. Petrovich, & M. Main (eds) (1981) Behavioral Development: The Bielefeld Interdisciplinary Project (pp.694–
9). Cambridge: Cambridge University Press: ‘The infant may hide its own needs, wishes, or behavioural ten-
dencies from its mother (as in the signal function of avoidance) or even (as in the cut-off function) from
itself ’ (687).
59 Main, M. & Weston, D. (1982) Avoidance of the attachment figure in infancy: descriptions and interpretations.
In C. Parkes & J. Stevenson-Hinde (eds) The Place of Attachment in Human Behavior (pp.31–59). New York: Basic
Books, p.56–7. Italics not in original but added at the request of Main in feedback on this chapter. The idea of avoid-
ance as a stable organisation of attention away from the caregiver when the attachment system was activated, as op-
posed to an artefact of the Strange Situation, was provisionally supported when a small study conducted by Main
found an association between avoidance shown in the Strange Situation and on reunion at daycare. Blanchard,
M. & Main, M. (1979) Avoidance of the attachment figure and social–emotional adjustment in day-care infants.
Developmental Psychology, 15(4), 445–6. However, this was not a presumption that she or the field of attachment
research would follow-up. An exception is Bick, J., Dozier, M., & Perkins, E. (2012) Convergence between attach-
ment classifications and natural reunion behavior among children and parents in a child care setting. Attachment
& Human Development, 14(1), 1–10. This paper, again, generated little further discussion. Though seemingly not
discussed by researchers, it is possible that one methodological complexity in using daycare reunions to assess at-
tachment with parents is that the daycare providers could also be attachment figures, leading to proximity-seeking
with them at times or brief conflict behaviour. Furthermore, naturalistic observations in daycare are messy from a
scientific perspective—including variability in the length of time children have been in daycare, and other occur-
rences in the environment. This will have made the setting less attractive to researchers.
individual differences in attachment as underpinned by attentional processes solved no ex-

isting problem. Additionally, methodologies for the study of attentional processes in young
children, such as eye-tracking technology, were still emerging.60 A third reason for the pre-
dominant emphasis on ‘strategies’ in interpretations of Main’s theory may have been the
timeliness of the concept. A sudden burst of appeals to the idea of ‘strategy’ occurred across
the human sciences in the late 1970s, especially in conceptualisations of the formation of
family units.61 It was a concept that allowed social scientists to model expectable, patterned
responses by individuals to their social environment, whilst keeping open the possibility for
change. The idea of ‘strategies’ permitted the explanation of patterns of social interaction
within and by families, as well as offering a lens on the techniques, including self-regulation,
used in order to maintain this stability.62
With her account of attentional processes underpinning attachment strategies, Main was

developing a model of individual differences seen in the Strange Situation. The infant in a
securely attached dyad could flexibly turn attention to the environment or the caregiver in a
manner responsive to the changing situation. Main regarded this strategy as reflecting a lack
of concern regarding access to proximity with the caregiver, permitting the child’s focus to be
responsive to the changing environment of the Strange Situation. The infant in an avoidantly
attached dyad, by contrast, was conceptualised by Main as attempting a partial reorganisa-
tion of attention away from the caregiver and towards the environment, in order to raise
the threshold for activating and lower the threshold for terminating the attachment system.
Gaze aversion, turning attention to the toys, and facing away from the caregiver all served to
keep the caregiver out of the infant’s focal awareness, thinning the influx of perceptions that
would otherwise activate the attachment system (and that might also activate anger). This
then raised the question of the evolutionary basis and underpinning mechanisms of ambiva-
lent/resistant behaviour in the Strange Situation.
Resistance and universality
As early as 1970 Main had come to conceptualise the proximity-seeking and clingy behav-
iour of infants in ambivalent/resistant dyads as explicable in an evolutionary sense: vigilance
regarding the availability of the caregiver might be expectable and helpful in some environ-
ments.63 Main’s further reflections on ambivalent/resistant attachment in the late 1970s were
informed by a groundbreaking paper by Robert Trivers.64 Whereas Bowlby had typically
treated the infant’s caregiver in terms of the caregiving behavioural system, Trivers empha-
sised that caregiving was only one priority for a parent. Other demands include the parent’s
60 Rayner, K. (1978) Eye movements in reading and information processing. Psychological Bulletin, 85, 618–60.
61 For a review see Viazzo, P.P. & Lynch, K.A. (2002) Anthropology, family history, and the concept of strategy.
International Review of Social History, 47(3), 423–52.
62 Influential cases include Bourdieu, P. (1976) Marriage strategies as strategies of social reproduction. In
R. Forster & O. Ranum (eds) Family and Society. Baltimore: Johns Hopkins University Press; Tilly, L.A. (1979)
Individual lives and family strategies in the French proletariat. Journal of Family History 4, 37–52; Becker, G. (1981)
A Treatise on the Family. Cambridge, MA: Harvard University Press. For discussion relevant to this wider shift in
the human sciences towards a concern with themes of strategy, family, emotion, and self-regulation see Reddy,
W.M. (1999) Emotional liberty: politics and history in the anthropology of emotions. Cultural Anthropology, 14,
256–88; Repo, J. (2018) Gary Becker’s economics of population: reproduction and neoliberal biopolitics. Economy
and Society, 47(2), 234–56.
63 Main, M. (1970) Infant play and maternal sensitivity in primate evolution. PP/Bow/J.4/1.
64 Trivers, R.L. (1974) Parent–offspring conflict. Integrative and Comparative Biology, 14(1), 249–64.
own survival, maintenance of sexual and social relationships, and the care of other offspring.
There was every potential for conflict between parents and offspring. In fact, Bowlby had al-
ready noted this in passing, in his observation of angry and distressed protest among human
toddlers during weaning or when their mother turns attention to a new baby.65 Bowlby’s re-
marks were influenced by the observations of his friend and colleague Robert Hinde, whose
research group had observed angry and distressed protest behaviour by young rhesus mon-
keys in response both to weaning and to short-term separations from their mother.66 Trivers
was fascinated by Hinde’s findings, and especially the observation that the more frequently
infants were pushed away by their mother prior to separation, the more distress, tantrums,
and clinging the infants showed on reunion. Trivers interpreted these responses as indicating
‘that the infant interprets its mother’s disappearance in relation to her predeparture behavior
in a logical way: the offspring should assume that a rejecting mother who temporarily disap-

pears needs more offspring surveillance and intervention than does a nonrejecting mother
who temporarily disappears’.67 The clinging, protesting behaviour observed by Hinde and
others was regarded by Trivers as serving to signal to the caregiver that the child needs more
investment before they would be able to survive alone. He regarded such signalling as part of
the primate behaviour repertoire, selected by evolutionary processes, since it may have con-
tributed to infant survival under conditions where caregiver investment in the child might
otherwise not be forthcoming.
Influenced by Trivers’ ideas, in a paper from 1979 Main identified that Trivers’ argu-
ment has strong relevance for thinking about ambivalent/resistant attachment.68 A lowered
threshold for activation of the attachment system, and the alternation of angry and dis-
tressed behaviours, could be regarded as a ‘conditional strategy’. It may well serve to retain
the attention of the caregiver and prompt the activation of the caregiving system. If the child
feels invisible to the caregiver, displays of distress, anger, and proximity-seeking/contact-
maintaining can be anticipated to increase the child’s visibility. However, it is a conditional
rather than a primary strategy, since it is unnecessary when the attachment system can be
satisfied directly through proximity-seeking. Furthermore, as Bowlby noted in Separation,
infant anger may coerce caregiver attention to attachment signals, but it also has the poten-
tial for backfiring by antagonising the caregiver.
In 1982, Jude Cassidy—then a doctoral student with Ainsworth—spent a year in Main’s
laboratory at Berkeley. Main and Cassidy reflected on what ambivalent/resistant attach-
ment would look like after infancy, and concluded that it would appear as behaviour that
attempted ‘to exaggerate intimacy with the parent, dependency on the parent, and his or her
relatively immature status’ with at least some ‘hostility or physical ambivalence’ evident.69
65 Bowlby, J. (1973) Separation. New York: Basic Books, p.213; Bowlby, J. (1984) Violence in the family as a dis-
order of the attachment and caregiving systems. American Journal of Psychoanalysis, 44, 9–27, p.11.
66 Hinde, R.A. & Spencer-Booth, Y. (1971) Effects of brief separation from mother on rhesus monkeys. Science,
173(3992), 111–18; Spencer-Booth, Y. & Hinde, R.A. (1971) Effects of brief separations from mothers during
infancy on behaviour of rhesus monkeys 6–24 months later. Journal of Child Psychology and Psychiatry, 12(3),
157–72.
67 Trivers, R.L. (1974) Parent–offspring conflict. Integrative and Comparative Biology, 14(1), 249–64, p.257.
68 Main, M. (1979) The ‘ultimate’ causation of some infant attachment phenomena: further answers, further
phenomena, further questions. Behavioral and Brain Sciences, 2, 640–43.

69 Main, M. & Cassidy, J. (1988) Categories of response to reunion with the parent at age 6: predictable from in-
fant attachment classifications and stable over a 1-month period. Developmental Psychology, 24(3), 415–26, p.418.
Though published in 1988, Cassidy’s visit to Berkeley and discussions with Main about the meaning of ambivalent/
resistant attachment after infancy occurred during 1982. The paper was submitted to Developmental Psychology in
mid-1986.
Whereas avoidant attachment represented a conditional strategy premised on the minimisa-

tion of attachment signals, ambivalent/resistant attachment included an intensification of at-
tachment signals as well as their punctuation by anger. This was an idea that Cassidy brought
back to Ainsworth’s group at Virginia, where it generated a great deal of discussion, and was
considered in print by Cassidy and Roger Kobak in 1988.70
In a paper published in 1994, Cassidy offered an interpretation of avoidance and am-
bivalence/resistance in terms of emotion regulation, as ‘minimising’ or ‘maximising’ the at-
tachment behavioural system itself (Chapter 5).71 Cassidy’s account aligned with Bowlby’s
discussion of defensive exclusion as inhibiting the activation of behavioural systems,
without specification of what exactly was being excluded (Chapter 1). It also aligned with
other voices, such as Schore, calling at the time for the reinterpretation of attachment as, in
general, the process through which self-regulation is achieved within close relationships.72 It

should be noted though that in contrast to Cassidy’s more diffuse position, Main emphasised
that these observable differences in strategy are underpinned by differences in attentional
processes.73 It is unclear the extent to which Cassidy realised that her description of min-
imising and maximising departed from Main in this regard. However, Kobak astutely ob-
served already at the time that two different models of ‘minimising and maximising’ seemed
to be in play. For his part, he strongly aligned himself with the centrality of attentional pro-
cesses, criticising Main for not doing more to highlight to readers the fundamental role of
attention in her conceptualisation of conditional strategies.74
For Main, avoidance was based on the direction of attention away from the caregiver,
as well as other perceptions and memories that may activate the attachment system.75 The
70 Cassidy, J. & Kobak, R. (1988) Avoidance and its relation to other defensive processes. In J. Belsky & T.
Neworski (eds) Clinical Implications of Attachment (pp.300–23). Hillsdale, NJ: Lawrence Erlbaum.
71 Cassidy, J. (1994) Emotion regulation: influences of attachment relationships. Monographs of the Society for
Research in Child Development, 59(2–3), 228–49; Cassidy, J. & Berlin, L.J. (1994) The insecure/ambivalent pattern
of attachment: theory and research. Child Development, 65(4), 971–91. The theory of ambivalence/resistance and
avoidance as representing maximising and minimising strategies has been attributed by Cassidy to both Main and
herself. See e.g. Berlin, L.J. & Cassidy, J. (2003) Mothers’ self-reported control of their preschool children’s emo-
tional expressiveness: a longitudinal study of associations with infant–mother attachment and children’s emotion
regulation. Social Development, 12(4), 477–95.
72 Cassidy drew the theme of emotion regulation from Thompson, R.A. (1994) Emotion regulation: a theme
in search of definition. Monographs of the Society for Research in Child Development, 59(2–3), 25–52. However,
Thompson did not address attachment strategies. Furthermore, in contrast to Cassidy, he was at greater pains to
emphasise that emotion regulation was not a single thing, and could be underpinned by a variety of processes: this
might include, but would not necessarily include, the manipulation of attention. An influential contemporary voice
emphasising emotion regulation was Schore, A.N. (1994) Affect Regulation and the Origin of the Self: Applications
to Affect Regulatory Phenomena. Mahwah, NJ: Lawrence Erlbaum Associates.
73 It is difficult to trace the lines of influence and the exact positions of Main, Cassidy, and Kobak in the 1980s
and early 1990s from the published record, as there was clearly a good deal of personal communication and sharing
of ideas between Berkeley and Virginia. Though a co-author on Cassidy’s 1988 chapter, it seems that it was Main’s
position, rather than Cassidy’s, that was subsequently adopted by Kobak.
74 Kobak, R.R., Cole, H.E., Ferenz-Gillies, R., Fleming, W.S., & Gamble, W. (1993) Attachment and emotion
regulation during mother–teen problem solving: a control theory analysis. Child Development, 64(1), 231–45.
75 The parameters of what Main intended by the idea of ‘attention’ fluctuate. At times it includes perception
and memory, to the degree that these processes are scaffolded by attention. However, at other times and foremost,
attention appears to mean the prioritisation of information within working memory, such that perception of the
environment and memory fall out of the definition of attention, even if they all work in the same direction, e.g.
‘Maintenance of “minimising” (avoidant) or “maximising” (resistant) behavioural strategy is therefore likely even-
tually not only to become dependent on the control or manipulation of attention but also eventually to necessitate
overriding or altering aspects of memory, emotion and awareness of surrounding conditions’: Main, M. (1995)
Recent studies in attachment: overview, with selected implications for clinical work. In S. Goldberg, R. Muir, & J.
Kerr (eds) Attachment Theory: Social, Developmental and Clinical Perspectives (pp.407–70). Hillsdale, NJ: Analytic
Press, p.451. On the question of whether ‘attention’ refers to a single process see e.g. Taylor, J.H. (2015) Against uni-
fying accounts of attention. Erkenntnis, 80(1), 39–56.
infant’s behaviour seemed to Main to be oriented to maintain the caregiver’s peripheral ra-
ther than focal attention, presumably since the caregiver tended actually to be most available
when not directly concerned with the baby: ‘if the mother picked them up, they turned away
and, as though attempting to distract her attention from themselves, pointed to toys and
other aspects of the environment’.76 In pointing to a toy on reunion, a behaviour that had
perplexed Ainsworth, Main saw a tactic through which the infant in an avoidant dyad could
manage the caregiver’s attention, so that the child remained in the caregiver’s awareness but
not directly the object of concern.77 By contrast, according to Main, distress and anger were
intensified for infants in ambivalent/resistant dyads by attentional vigilance towards the
availability of the caregiver, as well as other possible perceptions and memories that might
hold information relevant to the caregiver’s availability. In turn, this strategy centred the car-
egiver’s own attention on the child:

The A and C patterns of infant attachment organisation involve the overriding or manipu-
lation of the otherwise naturally occurring output of the attachment behavioural system.
Avoidance is conceived as a behavioural mechanism that permits the infant to shift atten-
tion away from conditions normally eliciting attachment behaviour, a shift that serves to
minimise the output of the attachment behavioural system . . . The focus upon the attach-
ment figure that is the essence of the type C strategy may be maintained by heightening
responsiveness to what would ordinarily be only minimally arousing cues to danger. Thus,
the type C infant may interpret a quiescent environment as threatening, in order to main-
tain the attention of the parent.78
The term ‘conditional strategy’ was drawn from ethological reflection on alternative forms
of adult sexual behaviour in non-human animals. Main similarly did not limit conditional
strategies to childhood in her theory. She proposed that the same strategies could be observ-
able in caregiving behaviour: either minimisation of attention to the child and reduced ac-
tivation of caregiving, or intensification of concern with the child.79 For instance, following
Trivers, she acknowledged that caregivers’ past experiences or present adversities may lead
them to minimise activation of the caregiving system, for instance to avoid overwhelming
feelings evoked by the child or to ensure resources remain available for other challenges.80
The minimising conditional caregiving strategy, Main suspected, may also serve to prime an
infant for an environment in which early independence might be beneficial. Conversely, the
maximising conditional strategy may serve to prime an infant for an environment in which
prolonged dependence on relationships would be especially salient and valuable. However,
Main was at pains to emphasise that the adoption of one caregiving strategy or another need
76 Main, M. & Goldwyn, R. (1983) Predicting rejection of an infant from mother’s representation of her own ex-
periences. National Conference on Infant Mental Health, Children’s Institute International, Los Angeles, February
1983. PP/Bow/J.4/4.
77 On infant declarative pointing as the direction of attention see also Tomasello, M., Carpenter, M., &
Liszkowski, U. (2007) A new look at infant pointing. Child Development, 78(3), 705–22.
78 Main, M. (1990) Cross-cultural studies of attachment organization: recent studies, changing methodologies,
and the concept of conditional strategies. Human Development, 33, 48–61, p.57.
80 This position has been elaborated in the recent anthropological literature on attachment, e.g. Seymour, S.C.
(2013) ‘It takes a village to raise a child’: attachment theory and multiple child care in Alor, Indonesia, and in North
India. In N. Quinn & J.M. Mageo (eds) Attachment Reconsidered (pp.115–39). London: Palgrave
not be conscious, and that it would be a category error to regard parents as motivated by the
evolutionary function of a strategy.81 Instead, the proximal cause will be the demands of the
present.
In a paper from 1990, ‘Cross-cultural studies of attachment organization’, Main argued
that ‘the maintenance of differing conditional strategies entails the utilisation of similar
cognitive mechanisms across individuals’.82 This made clear that she was playing for grand
stakes; her proposals amounted to no less than a global model of human emotional life.
Elaborating on the 1990 claim in a later work, Main asserted that ‘there exist species-wide
abilities that are not part of the attachment system itself, but can, within limits, manipu-
late (either inhibit or increase) attachment behavior in response to differing environments’.83
As described in Chapter 1, already in 1956 Bowlby, Ainsworth, and colleagues had written
that ‘the personality patterns of children who have experienced long separation tend to fall

into one or other of these two opposite classes’: either ‘over-dependent’ and ‘ambivalent’ or
‘mother-rejecting . . . having repressed their need for attachment’.84 This was already a mo-
mentous claim. However, whereas Bowlby and Ainsworth discussed the two classes as re-
sponses to separation, in her 1990 paper Main resituated these classes as the fundamental
alterations of the attachment system and caregiving system, or potentially any behavioural
system.
Main’s model of conditional strategies was not, therefore, merely a localised psychological
theory to account for the Strange Situation, but a philosophy of human experience in gen-
eral, with resonances of Plato and Kant. These resonances signal the important, if complex,
role of philosophy on the development of Main’s ideas. The only mention of Main’s parents in
her published works is the remark that ‘philosophy was much admired by my parents, who
had introduced me to Plato, Kant, and several Eastern philosophies by age 10’.85 As an under-
graduate, she re-encountered Plato and Kant at St John’s College in Maryland in the classes
of her future husband, Alvin Main.86 Now, in her 1990 paper, she was proposing that behind
81 Main can be seen here moving in the direction of what would later become the life-history model of attach-
ment, where the interaction between caregiving and attachment system is interpreted as a means through which
children receive signals about the strategies that will be best adapted for the degree of adversity that characterises
their environment, and in particular the relative need to prioritise immediate survival or long-term growth and
exploration. See Chisholm, J.S. (1996) The evolutionary ecology of attachment organization. Human Nature, 7(1),
1–37; Belsky, J. (1999) Modern evolutionary theory and patterns of attachment. In J. Cassidy & P.R. Shaver (eds)
Handbook of Attachment (pp.141–61). New York: Guilford. Main was cited and discussed by Chisholm and Belsky
in thinking about evolutionary trade-offs; the line of intellectual history is likely the influence of Trivers on all
three attachment theorists. Main’s discussions of the evolutionary basis of individual differences in attachment
essentially ended in 1990. By contrast, this would become a central focus in the work of Belsky over subsequent
decades.
at 1, 6, and 19 years of age: the Berkeley Longitudinal Study. In K.E. Grossmann, K. Grossmann, & E. Waters (eds)
84 Bowlby, J., Ainsworth, M., Boston, M., & Rosenbluth, D. (1956) The effects of mother–child separation: a
follow-up study. British Journal of Medical Psychology, 29, 211–47, p.238.

at 1, 6, and 19 years of age: the Berkeley Longitudinal Study. In K.E. Grossmann, K. Grossmann, & E. Waters (eds)
86 The curriculum at St John’s College used no textbooks and only provided students with primary texts. So
Main read Plato, Kant, and other philosophical texts in the original during her four years of study of the subject.
With wry humour, she recalled Ainsworth’s difficulties in trying to turn her new philosophically trained graduate
student into a proper developmentalist: ‘I heard that she said (unfortunately aptly, but I refused to consider the
truth-value of the statement at the time) that she dreaded sending me out on home visits to Baltimore mothers,
because I was virtually unable to engage in small-talk, and would probably ask them what they thought of Spinoza
the diversity of apparent infant behaviours there lies three essential forms for responding
to distressing and challenging situation (cf. Plato, Aristotle); the basis for these differences
stems from the structure of human experience, which shape human perception, language,
and behaviour (cf. Kant).
Though the three Ainsworth categories were identified using 23 middle-class infants in
the mid-1960s in Baltimore, for Main the categories exceeded this particularity to express
the three basic ways that humans can respond to distressing and challenging situations. In
the context of worries or other troubling feelings, there are three basic approaches: we can
communicate about our feelings to someone we anticipate or hope might help us; we can
keep our feelings to ourselves; or we can make our distress and frustration someone else’s
problem. As we saw in Chapter 2, Ainsworth took from Blatz the idea that the people we
may need to depend upon are independent of us in a deep sense, and this can be regarded as

a source of worry or as a source of reassurance, depending on what this freedom has implied
for us in the past. Main’s conditional strategies were responses to this predicament, based on
the selective exclusion of information that would risk reducing the availability of the care-
giver as a secure base and safe haven:
The conditional behavioral strategy (avoidance or preoccupation) is understood to be im-

posed on a still-active primary strategy, imposed on aspects of memory, and imposed on
awareness of surrounding conditions. Maintenance of the ‘minimizing’ (A) or ‘maximizing’
(C) behavioral strategy is then dependent on the control or manipulation of attention—
specifically, an organized shift of attention away from conditions activating attachment be-
havior in Group A infants, and a heightened vigilance maximizing responsiveness to even
minimal clues to danger in Group C infants. It seems inevitable that a continuing ‘mini-
mization or maximization of the display of attachment behavior relative to the naturally
occurring output of the behavioral system’ (Main, 1990) will eventually also involve the
defensive exclusion or defensive distortion of certain memories and perceptions.87
An infant in an ambivalent/resistant dyad directs attention away from potential information

that might suggest that the environment is unthreatening and that the caregiver is available.
This infant therefore sees clearly, indeed overclearly, the human predicament that attach-
ment figures are independent of us in worrying ways. Infants in an avoidant dyad direct
attention away from potential information that might elicit alarm, distress, or a tendency
to approach the caregiver for comfort. They see clearly, indeed with a reductive clarity, that
as humans we are partly independent of our attachment figures. Each of the conditional
or something.’ Main, M. (1999) Mary D. Salter Ainsworth: tribute and portrait. Psychoanalytic Inquiry, 19(5), 682–
736, p.690. The influence of Plato especially is visible throughout one of Main’s earliest papers, where her focus is
on how security can be achieved when humans cannot ultimately rest on knowledge gained from the apparent
world: Main, M. (1977) Sicherheit und Wissen. In K.E. Grossmann (Ed.), Entwicklung der Lernfahigheit in der
Sozialen Umwel (pp.47–95). Munich: Kindler. After the death of Alvin Main, it is not clear that Main continued
further reading in Plato and Kant, though her later work shows familiarity with philosophers such as Hans-Georg
Gadamer and her University of California colleague Paul Feyerabend. Kant’s noumenal–phenomenal distinc-
tion is drawn upon in Main, M. (1991) Metacognitive knowledge, metacognitive monitoring, and singular (co-
herent) vs. multiple (incoherent) models of attachment: some findings and some directions for future research. In
P. Marris, J. Stevenson-Hinde, & C. Parkes (eds) Attachment Across the Life Cycle (pp.127–59). London: Routledge.
87 Main, M. (1993) Discourse, prediction and recent studies in attachment: implications for psychoanalysis.
Journal of the American Psychoanalytic Association, 41, 209–44, p.233; Main, M. (1990) Cross-cultural studies
of attachment organization: recent studies, changing methodologies, and the concept of conditional strategies.
Human Development, 33, 48–61.
strategies is acutely and effectively attentive to some aspect of reality, and in this way enacts
a method of responding to alarm that may be of survival value under conditions where that
aspect is of special importance. This may provide important information for children about
how to calibrate the demands they make on the world in order to achieve what nurturance
and resources may be available. The conditional strategy may also be experienced as a kind
of ‘secondary felt security’, despite being held in place by anxiety; the reason for this is that
the conditional strategies nonetheless offer predictable and therefore reassuring access to
some sense of closeness and regulation.88 However, for Main, each conditional strategy must
also depend upon selective exclusion of another aspect of reality.89 There will be, she pre-
dicted, a price to be paid for the conditional strategy in the long run, to the degree that this
information about attachment relationships and about their own affective life remains lost or
relatively opaque to the individual.90

With Main’s model of Group A and Group C as conditional strategies, attachment theory
became a global account of human emotion and relationships.91 For Main, two conditional
strategies existed for the attachment and the caregiving systems, and potentially for other
behavioural systems. Its output could be minimised or intensified, and the basic architecture
for these strategies would lie in the direction of attentional processes.92 The magnitude of the
impact for subsequent attachment theory of the encompassing conditional strategy model
of individual differences in attachment cannot be overestimated. It has had more sway for
later attachment theory than even the theory of attachment as a behavioural system that it
ostensibly modified. Few developmental attachment researchers in the twenty-first century
have made more than token mention of behavioural systems, or developed specific hypoth-
eses from this concept (though see Chapter 5). Since the 1990s, the idea of attachment as
a ‘behavioural system’, grounded in human evolutionary history, has rather the status of a
memento that the developmental tradition of attachment research is pleased to have, and
90 Main has been criticised for this position at times by critics who argue that there may be ecological niches
where conditional strategies are simply superior. The implication that they represent a second-best option would
then be both overgeneralised and potentially ethnocentric: ‘Despite her recognition of alternative or “conditional”
attachment strategies, in referring to insecure attachment as a “secondary” strategy, Main (1990) may be clinging
to the view that because secure attachment is “primary” it must also be “normal”.’ Chisholm, J.S. (1996) The evolu-
tionary ecology of attachment organization. Human Nature, 7(1), 1–37, p.24. However, Main’s position has been
defended by Granqvist, who points to the disparity between markers of hidden distress and calm appearance char-
acteristic of attachment avoidance. Granqvist agrees that there will be ecological niches where this disparity may
be helpful. But he defends Main’s claim that this should be regarded as a back-up strategy, since it predictably
occurs primarily when direct communication about distress to the caregiver proves unsuccessful. Granqvist, P.
(2020) Attachment, Religion, and Spirituality: A Wider View. New York: Guilford.
91 Kobak, R.R., Cole, H.E., Ferenz-Gillies, R., Fleming, W.S., & Gamble, W. (1993) Attachment and emotion
regulation during mother–teen problem solving: a control theory analysis. Child Development, 64(1), 231–45.
92 Mary Main and Erik Hesse, personal communication, August 2019: ‘With developmental maturation, each
conditional strategy has greater variegation. Additionally, development allows humans to override a behavioural
system in other ways than the two conditional strategies, producing a wider variety of potential strategies than
those available to infants. These might well not be conditional strategies in the technical sense of being a behav-
ioural repertoire made available by human evolutionary history for solving problems of survival and reproduc-
tion. They could be described as “strategic” in the non-technical sense—but it depends on how the word is being
used.’ In fact, it is not clear the extent to which this argument holds for the controlling-caregiving and controlling-
punitive behavioural repertoires identified by Main and Cassidy in the six-year reunion system (see section ‘The
Main and Cassidy reunion system’). It could be imagined that these were ethological repertoires made available by
human evolutionary history. On ‘tertiary’ attachment strategies see Chapter 5.
without which they might feel bereft, but which never gets actually brought out for further
examination, except perhaps when teaching.93
By contrast, the image of human difference as strung out along axes of minimising and
maximising has been the beating heart of attachment research for three decades, and central
to hypothesis generation and the interpretation of empirical findings regarding individual
differences. It has formed the ‘grid of specification’ according to which kinds of attachment
behaviour in infancy and beyond have been divided, contrasted, related, grouped, clas-
sified, and derived in relation to one another.94 The idea of minimising and maximising
strategies is frequently cited as originating with Main, sometimes with Cassidy and Kobak,
and occasionally with Hinde.95 However, most often the image has been taken as simply a
timeless part of attachment research as a paradigm, implied by Ainsworth’s introduction of
three patterns of attachment. This interpretation has been supported by Main’s tendency

to attribute her own ideas to Ainsworth, where these grew out of the soil of Ainsworth’s
own thinking.96 It has likely also been supported by the elegant simplicity, amounting to
apparent obviousness, of the theory of individual differences divided into minimising and
maximising.
93 A few attachment theorists have, however, explicitly argued for the retirement or supersession of Bowlby’s
control systems theory, e.g. Cassidy, J., Ehrlich, K.B., & Sherman, L.J. (2013) Child–parent attachment and response
to threat: a move from the level of representation. In M. Mikulincer & P.R. Shaver (eds) Nature and Development
of Social Connections: From Brain to Group (pp.125–44). Washington, DC: American Psychological Association.
However, such arguments have not led to any sustained public discussion. Most attachment researchers impatient
with Bowlby’s theory of behavioural systems simply bypass it, and focus instead on individual differences in per-
ceptions of caregiver availability. One reason has been put forward by Granqvist, who observed that attachment
researchers have tended to avoid outright criticisms of Bowlby’s model of the function and workings of the at-
tachment system as a matter of courtesy, even if they know that Bowlby himself would have been dismayed and
scornful of any such nicety that held back theoretical development. Granqvist, P. (2020) Attachment, Religion, and
Spirituality: A Wider View. New York: Guilford. By contrast, in applying attachment theory to adult relationships,
the question of the nature of attachment at this later developmental stage became of critical concern to the social
psychological tradition of attachment research (Chapter 5).
94 Foucault, M. (1969, 1972) The Archaeology of Knowledge, trans. A.M. Sheridan Smith. London: Routledge,
Chapter 3.
95 Since the 1990s, the dominant position of Main’s interpretation of the categories of attachment has made
her work the standard reference point for the theory of conditional strategies. However, in the 1980s the concept
of attachment patterns as conditional strategies was actually more frequently ascribed to Hinde, who at the time
was the much more senior and well-known figure. This was despite the fact that Main was the first to present this
theory, in a short and quite obscure paper: Main, M. (1979) The ‘ultimate’ causation of some infant attachment
phenomena: further answers, further phenomena, further questions. Behavioral and Brain Sciences, 2, 640–43.
The concept of attachment patterns as conditional strategies was also offered by Hinde in a 1982 chapter without
reference to Main: Hinde, R.A. (1982) Attachment: some conceptual and biological issues. In J. Stevenson-Hinde
& C. Murray Parkes (eds) The Place of Attachment in Human Behavior (pp.60–78). New York: Basic Books. The
origins of Hinde’s use of the term are not fully clear. He had read Maynard Smith’s and Dawkins’ 1979 papers on
conditional strategies soon after they were published, and was soon citing them. Maynard Smith and Hinde were
also in direct discussion of these issues in 1982. See e.g. Hinde, R. (1982) Letter to Maynard Smith, 14 June 1982.
John Maynard Smith Archive, British Library, MS 86840/46. Hinde may also have been part of conversations about
conditional strategies with his student and collaborator Pat Bateson, who had been with Maynard Smith, Dawkins,
and Main at Bielefeld. In 1978, Main presented a paper on ‘detachment’ at a colloquium to Hinde’s research group,
and she may have mentioned her ideas about conditional strategies then. Given that Hinde was ordinarily quite
careful in attributing ideas that were not his own, it seems probable that all of these pathways of influence operated
simultaneously, giving Hinde the impression that the concept was simply a familiar one in contemporary etho-
logical discussions. Certainly already in 1979, Brockmann, Grafen, and Dawkins described the notion as ‘a fash-
ionable idea in modern ethology’: Brockmann, H.J., Grafen, A., & Dawkins, R. (1979) Evolutionarily stable nesting
strategy in a digger wasp. Journal of Theoretical Biology, 77(4), 473–96, p.473.
96 E.g. Main M. (1999) Mary D. Salter Ainsworth: tribute and portrait. Psychoanalytic Inquiry, 19, 682–
776: ‘Ainsworth, however, saw these rejected infants as responding to the increased stress imposed by the strange
situation by actively (although, of course, not necessarily consciously) shifting their attention so as to inhibit the
behavioral and emotional manifestations of attachment—notably, proximity-seeking, crying, and anger’ (719).
Ainsworth was, understandably, rather astonished that Main transformed her categories
for the Strange Situation into an encompassing philosophy of existence, applicable to all
behavioural systems. In a letter to John Bowlby in March 1984 (embarrassingly posted by
accident, in fact, to Main),97 Ainsworth expressed ‘unease’ with Main’s ambitious and uni-
versalising proposals:
She is convinced that I have discovered the three patterns of attachment—that she believes
to hold not only for one-year-olds but throughout the life span. This is very flattering. Also
I must confess I think that they are indeed the three major patterns. But . . . I cannot quite
believe that apart from the groups and subgroups I have identified there are [not] other
less frequent occurring patterns that may be impossible to comprehend within these three
major groups (A/B/C). To say nothing of cross-cultural variations.98

However, as she learned about Main’s theories over the course of 1985 and became increas-
ingly impressed with the promise of Main’s AAI, Ainsworth remarked to Bowlby: ‘You
were right that I am in a sense a student of Mary Main’s.’99 Ainsworth’s support played a
critical role in establishing Main’s account of conditional strategies as the central image of
individual differences within attachment theory, as a paradigm primarily concerned with
individual differences. In particular, Ainsworth’s late paper with Eichberg reporting an ex-
ceptionally strong relationship between the infant classifications and the new AAI was of
great symbolic as well as scientific importance.100 Ainsworth supported and encouraged
her different graduate students; Inge Bretherton, Everett Waters, Patricia Crittenden, and
Bob Marvin were all making both methodological and theoretical innovations at the turn
of the 1990s that built on Ainsworth’s contribution.101 However, by then the Ainsworth and
Eichberg paper was already in print, and quickly came to be widely interpreted as confirm-
ation of Main’s extension of the Ainsworth categories across the lifespan, passing the baton
of Ainsworth’s role as the field’s method-giver.102 In addition to Ainsworth’s support, Main’s
characterisation of avoidance and ambivalence/resistance as conditional strategies was seen
97 Main, M. (1999) Mary D. Salter Ainsworth: tribute and portrait. Psychoanalytic Inquiry, 19, 682–776.
98 Ainsworth, M. (1984) Letter to John Bowlby, 10 March 1984. PP/BOW/B.3/7.
99 Ainsworth, M. (1985) Letter to John Bowlby, 23 December 1985. PP/BOW/B.3/8.
100 Ainsworth, M. & Eichberg, C. (1991) Effects on infant–mother attachment of mother’s unresolved loss of an
attachment figure, or other traumatic experience. In C. Parkes, J. Stevenson-Hinde, & P. Marris (eds) Attachment
Across the Lifespan (pp.160–83). London: Routledge.
101 E.g. Bretherton, I., Prentiss, C., & Ridgeway, D. (1990) Family relationships as represented in a story-
completion task at thirty-seven and fifty-four months of age. New Directions for Child and Adolescent Development,
48, 85–105; Vaughn, B.E. & Waters, E. (1990) Attachment behavior at home and in the laboratory: Q-sort ob-
servations and strange situation classifications of one-year-olds. Child Development, 61(6), 1965–73; Crittenden,
P.M. (1992) Quality of attachment in the preschool years. Development & Psychopathology, 4(2), 209–41; Cassidy,
J., Marvin, R., with the Attachment Working Group of the John D. and Catherine T. MacArthur Network on the
Transition from Infancy to Early Childhood (1992) Attachment organisation in preschool children: procedures
and coding manual. Unpublished manual.
102 See the discussions of Ainsworth and Main for instance in Fonagy, P., Steele, H., & Steele, M. (1991) Maternal
representations of attachment during pregnancy predict the organization of infant–mother attachment at one
year of age. Child Development, 62(5), 891–905; Bakermans-Kranenburg, M.J. & van IJzendoorn, M.H. (1993)
A psychometric study of the Adult Attachment Interview: reliability and discriminant validity. Developmental
Psychology, 29(5), 870–79; Bacciagaluppi, M. (1994) The relevance of attachment research to psychoanalysis and
analytic social psychology. Journal of the American Academy of Psychoanalysis, 22(3), 465–79. On the role of conse-
cration of an heir at the end of life, and the balance between preserving the integrity of an inheritance and retaining
the loyalty to the family of non-selected heirs see Goody, J. (1973) Strategies of heirship. Comparative Studies in
Society and History, 15(1), 3–20; Bourdieu, P (2002, 2008) The Bachelors’ Ball: The Crisis of Peasant Society in
Béarn. Chicago: University of Chicago Press.
as a powerful explanatory tool. And the AAI, the first trainings in which by Main were co-
taught with Ainsworth, was regarded as an exciting methodological development. In her
final writings, Ainsworth urged her successors to retain an open-ended system, and not
to close either theory or method around her categories.103 However, a central goal of the
second generation of attachment researchers was the construction of a cumulative empirical
research paradigm of replicated results, and for this a settled coding system, rather than fur-
ther exploration, was treated as desirable.
Convincing Bowlby
The conventional reference for the theory of conditional strategies over the decades has

been Main’s 1990 paper ‘Cross-cultural studies of attachment organization’.104 It was more
elaborate than ‘The “ultimate” causation of some infant attachment phenomena’ from 1979,
which is more of a note than a full paper. Her 1981 book chapter ‘Avoidance in the service
of proximity’ was, by its own admission, meandering. It was also published in an edited
volume in a context in which journal articles, across the social sciences, were coming to re-
ceive comparatively greater prominence. However, as a consequence, readers have generally
missed the role of attention as the architecture underpinning Main’s conceptualisation of
conditional strategies. This was mentioned to a greater or lesser extent in all her work of the
period—except the 1990 paper.
Main observed that the conflict between anger and attachment could flood out and inter-
rupt the smooth expression of the attachment system. However, most ambivalent/resistant
infants used the oscillation between distress and anger in quite a smooth way to retain the
attention of their caregiver: when they were put down they cried to be picked up; when they
were picked up, they squirmed away. Just as avoidance was paradoxically in the service of
proximity for Group A infants, the distress and conflict of Group C infants appeared to in the
service of organisation via proximity. In Main’s perspective, the ambivalent/resistant clas-
sification seemed ‘much the least well organised’ of the Ainsworth classifications.105 And
in some work under Main’s supervision, the ambivalent/resistant and unclassifiable infants
were grouped together for analysis on the basis that both displayed conflict.106 Nonetheless,
ultimately, Main concluded that ‘like an avoidant baby, a resistant baby may be said to be “or-
ganized” in the sense of having a singular attentional focus’, i.e. retaining the attention of their
caregiver.107
103 Ainsworth M. (1990) Epilogue. In M.T. Greenberg, D. Cicchetti, & E.M. Cummings (eds) Attachment in the
Preschool Years (pp.463–88). Chicago: Chicago University Press.

and the concept of conditional strategies. Human Development, 33, 48–61.

105 Main, M. & Hesse, E. (1990) Parents’ unresolved traumatic experiences are related to infant disorganized
attachment status: is frightened/frightening parental behavior the linking mechanism? In M. Greenberg, D.

Cicchetti, & M. Cummings (eds) Attachment in the Preschool Years (pp.161–82). Chicago: University of Chicago
Press, p.179.
106 E.g. Weston, D.R. (1983) Implications of mother’s personality for the infant–mother attachment relation-
ship. Unpublished doctoral dissertation, University of California, Berkeley: ‘The C’s mix attachment behaviours
with resistance; the U’s [i.e. unclassifiable cases] mix attachment behaviours with resistance and avoidance. These
mixed patterns support the rationale for combining these two groups.’
Hillsdale, NJ: Analytic Press, p.420. Italics added. Main also offered another characterisation of the ambivalent/
Main went to visit Bowlby in March 1978, and they discussed these ideas. In the course of
these discussions Bowlby was persuaded that the attachment system may have evolved with
a repertoire of strategies.108 They discussed early draft material towards his book Loss, and
in the final version he cited Main’s perspective with approval. Bowlby had already from 1977
taken an interest in caregiving and compliant behaviour shown by children towards parents
when they might not receive adequate care (Chapter 1). Main’s proposal seemed to fit with
this, since avoidant, caregiving, and compliant behaviour had in common the substitution
of direct attempts to seek the caregiver’s support for an alternative strategy that would be
more effective given parental behaviour.109 The exclusion of information that might activate
the attachment system for children in this situation would have the predictable outcome
of increased caregiver responsiveness and so survival value within human evolutionary
history.110 In Loss, Bowlby therefore situated the avoidant conditional strategy within the

broader observation that ‘as children know in their bones, when mother is prone to be re-
jecting it may be better to placate her than to risk alienating her altogether’. However, he also
acknowledged Main’s emphasis on attention as the underpinning infrastructure of avoid-
ance as a conditional strategy:
An infant of this sort, instead of showing attachment behaviour as infants of responsive

mothers do, turns away from his mother and busies himself with a toy. In so doing he is
effectively excluding any sensory inflow that would elicit his attachment behaviour and
is thus avoiding any risk of being rebuffed and . . . in addition he is avoiding any risk of
eliciting hostile behaviour from his mother. Yet he remains in her vicinity. This type of
resistant infant: ‘Like the avoidant baby, and unlike the secure baby, however, her attention is not fluid, and she fo-
cuses upon only one aspect of her surround’ (420).
108 Bowlby, J. (1978) Notes following discussion with Mary Main in March 1978 about the draft of Vol. 3 Loss.
PP/Bow/H.78: Main found a ‘correlation between violent screaming and hitting mother in home and avoidance in
strange situation. The more attachment behaviour is aroused the more likely is avoidance to be exhibited. The less
attachment behaviour is aroused the more likely he is to show angry (e.g. hit) behaviour & also attachment behav-
iour. A conditional strategy.’
109 Bowlby’s emphasis in Loss on avoidance, caregiving, and compliant behaviour as reflecting different
forms of a common strategy in which a child masked distress would be influential for Patricia Crittenden,
who was pursuing a doctorate with Ainsworth at the time. Crittenden, P.M. (1983) Mother and infant pat-
terns of attachment. Unpublished doctoral dissertation, University of Virginia, Charlottesville; Crittenden, P.M.
(1988) Relationships at risk. In J. Belsky & T. Nezworski (eds) Clinical Implications of Attachment (pp.136–74).
Hillsdale, NJ: Lawrence Erlbaum. Main and Solomon later distinguished child caregiving behaviour towards
the parent from disorganisation; as a result it does not feature in the Main and Solomon (1990) indices. Main,
M. & Solomon, J. (1990) Procedures for identifying infants as disorganised/disoriented during the Ainsworth
Strange Situation. In M.T. Greenberg, D. Cicchetti, & E.M. Cummings (eds) Attachment in the Preschool Years
(pp.121–60). Chicago: University of Chicago Press, p.147. However, it was not given its own place in their infant
coding protocols, which became the dominant system not just for reporting studies of the Strange Situation
but also for measuring conceptualising attachment in general across childhood and the lifespan. Despite the
heavy emphasis on the importance of child caregiving behaviour by Bowlby in his later writings, this topic has
therefore largely disappeared from attachment research after his death, except in the work of Moss, Bureau, and
colleagues, e.g. Moss, E., Bureau, J.-F., Cyr, C., Mongeau, C., & St-Laurent, D. (2004) Correlates of attachment
at age 3: construct validity of the preschool attachment classification system. Developmental Psychology, 40,
323–34; Meier, M., Martin, J., Bureau, J.-F., Speedy, M., Levesque, C., & Lafontaine, M.-F. (2014) Psychometric
properties of the mother and father compulsive caregiving scales: a brief measure of current young adult care-
giving behaviours toward parents. Attachment and Human Development, 16(2), 174–91. Controlling-caregiving
behaviour is part of the Main and Cassidy six-year reunion system, and in this context is often mentioned in ex-
positions of attachment theory for clinical audiences. But there have been no trainings available in this measure
for decades.
110 Bowlby, J. (1980) Loss. New York: Basic Books: ‘Given certain adverse circumstances during childhood, the
selective exclusion of information of certain sorts may be adaptive’ (45).

response, Main suggests, may represent a strategy for survival alternative to seeking close
proximity to mother.111
Against his earlier position and also the common image of his work still in circulation today
(Chapter 1), Bowlby can be seen here in Loss following Main in turning away from an image
of the display of attachment behaviour as natural and best, towards acknowledgement of
the evolutionary function and the at least short-or medium-term benefits of diverse forms
of infant behaviour. And Bowlby would continue to refer to avoidance as a ‘strategy’ in his
subsequent writing, acknowledging both short-term advantages and potential survival value
and the longer-term contribution the strategy might make to mental illness.112 Such treat-
ment of avoidance as a conditional strategy fitted with Bowlby’s wider transition in his late
work from regarding proximity as the set-goal of the attachment system to seeing this as the

physical and attentional availability of the caregiver.
However, though recognised by Bowlby, fundamental aspects of Main’s account of con-
ditional strategies have been missed by the subsequent attachment literature. Klaus and
Karin Grossmann suggested that a key reason for this is that attachment researchers have
generally regarded Main’s introduction of the ‘disorganised attachment classification’ as
superseding any early reflections.113 As a result, very few attachment researchers have read
any of Main’s papers from prior to 1990. Attachment researchers repeat—and repeat—the
idea of ‘minimising’ and ‘maximising’ strategies in summaries of the paradigm and in inter-
preting empirical results. Yet the fact that the original idea was of minimising or maxim-
ising attentional processes, specifically, is essentially unknown. Main made several efforts
to clarify the centrality of attention to her account of individual differences in attachment.
The most strenuous was her decision to title a paper from 2000 ‘The organised categories of
infant child and adult attachment: flexible vs inflexible attention under attachment-related
stress’.114 Yet besides her most immediate collaborators such as Erik Hesse and Marinus van
IJzendoorn,115 and a few others with personal links to Main at one period or another such
as Roger Kobak, Pehr Granqvist, and Carlo Schuengel,116 researchers generally interpret
Main as describing the minimisation or maximisation of either attachment behaviour or of
distress (or both), rather than recognising her proposal of a causal account with attentional
111 Bowlby, J. (1980) Loss. New York: Basic Books, p.73.

112 E.g. Bowlby, J. (1989) Foreword. In E. Gut, Productive and Unproductive Depression (pp.xiii–xv).
London: Routledge: ‘Emmy Gut examines why people grow up unable to cope with painful problems and instead
become locked in some fruitless strategy of avoidance which, however successful it may be in the short term, leads
them in the longer term to become prone to depressive moods with no useful outcome’ (xiv).
113 Klaus and Karin Grossmann, personal communication, August 2012.
114 Main, M. (2000) The organized categories of infant, child, and adult attachment: flexible vs. inflexible atten-
tion under attachment-related stress. Journal of the American Psychoanalytic Association, 48(4), 1055–96.
115 E.g. Hesse, E. (1996) Discourse, memory, and the Adult Attachment Interview: a note with emphasis on the
emerging cannot classify category. Infant Mental Health, 17(1), 4–11.

116 Kobak’s writings on the AAI in the 1990s suggest he felt that Main had not gone far enough in emphasising
the role of attention, since she also headlined the concept of internal working model. The idea of the internal
working model connoted to Kobak a representational rather than attentional basis for individual differences.
Kobak, R.R., Cole, H.E., Ferenz-Gillies, R., Fleming, W.S., & Gamble, W. (1993) Attachment and emotion regula-
tion during mother–teen problem solving: a control theory analysis. Child Development, 64(1), 231–45. See also
Schuengel, C., de Schipper, J.C., Sterkenburg, P.S., & Kef, S. (2013) Attachment, intellectual disabilities and mental
health: research, assessment and intervention. Journal of Applied Research in Intellectual Disabilities, 26(1), 34–46;
Granqvist, P. (2020) Attachment, Religion, and Spirituality: A Wider View. New York: Guilford. Another exception
is Waters, T.E., Brockmeyer, S.L., & Crowell, J.A. (2013) AAI coherence predicts caregiving and care seeking be-
havior: secure base script knowledge helps explain why. Attachment & Human Development, 15(3), 316–31.
processes at the centre. This may have been facilitated by a lack of clarity regarding the rela-
tive overlap or differences between Main’s specific concern with attentional process and
Bowlby’s descriptive concept of defensive exclusion, since in Loss Bowlby treated the two
accounts as aligned (Chapter 1). Difficulties in identifying and understanding Main’s pos-
ition have also been facilitated by Cassidy’s influential reinterpretation of minimising and
maximising strategies in terms of ‘emotion regulation’ in the 1990s.117 In any case, eclipse
of the role of attentional processes in individual differences for Main has made for much
more general and less parsimonious hypothesis-generation. It has additionally obscured
the nature of the links Main perceived between infant behaviour and individual differences
in later development.
The infant disorganised classification

Conflict behaviour
In her doctoral dissertation Main had documented various forms of conflict behaviour in the
Strange Situation, including ‘stereotypies; hand-flapping; echolalia; inappropriate affect (inex-
plicable fears, inappropriate laughter) and other behaviours appearing out of context’.118 She had
also been impressed by an incident during her doctoral research. She had been meeting with an
infant, Sara, and her mother:
During the office interview, a thunderstorm took place and a bolt of lightning struck very
near the building. The event was frightening even for the adults present—and Sara, though
equidistant between both, dashed whimpering to the unfamiliar interviewer rather than her
mother.119
Sara received the highest score for avoidance in the sample in the Strange Situation, and her
mother had the highest score for aversion to physical contact with her infant. Main con-
cluded that Sara’s approach-avoidance conflict caused by the lightning was too intense to
permit an organised avoidant response. The result was a redirection of attachment behaviour
towards Main herself. For a time, interested to see if this effect would be replicable, Main
planned an empirical study in which ten infants avoidantly attached with both parents and a
comparison group of infants would be placed equidistant between the experimenter and the
117 Cassidy, J. (1994) Emotion regulation: influences of attachment relationships. Monographs of the Society
for Research in Child Development, 59(2–3), 228–49. For the rise of emotion regulation approaches to attach-
ment, important predecessors to Cassidy include Tronick, E. (1989) Emotions and emotional communication
in infants. American Psychologist, 44(2), 112–19, and Stern, D.N. (1985) The Interpersonal World of the Infant.
New York: Basic Books. Hofer’s work, discussed in Chapter 1, was also an important influence. Another significant
article of the period, somewhat later than Cassidy’s, was Lyons-Ruth, K. (1999) The two-person unconscious: inter-
subjective dialogue, enactive relational representation, and the emergence of new forms of relational organization.
Psychoanalytic Inquiry, 19(4), 576–617.
119 Main, M. (1980) Abusive and rejecting infants. In N. Frude (Ed.), The Understanding and Prevention of Child
Abuse: Psychological Approaches (pp.19–38). London: Concord Press, p.27.

The infant disorganised classification 237
mother and exposed to a sudden, loud crashing noise and flickering lights.120 Fortunately for
the infants, this study was not carried out.121
One reason may have been that it was not necessary: Main found that she could reliably
observe conflict behaviour in other contexts. One was among young children who had ex-
perienced abuse. Main’s graduate student Carol George proposed that conflict behaviour
was expectable when young children experienced physical maltreatment by their caregiver.
George observed ten physically abused toddlers and ten matched controls in a San Francisco
daycare. She conducted careful qualitative observations, making detailed narrative notes.
Based on analysis of these notes, George and Main reported that ‘all of the abused chil-
dren but none of the controls were observed to respond to peer affiliations with approach-
avoidance behaviour,’ such as approaching a professional carer but with their head averted.122
Additionally, Main found that she could reliably elicit conflict behaviour in the labora-

tory. Conflict behaviour was frequent, especially among infants in avoidantly attached
dyads, in response to the unnerving crying clown both in Main’s Berkeley sample and in
the Grossmanns’ Regensburg sample where the Clown procedure was also applied. In 1981,
Main, working together with her graduate student Donna Weston, formulated an unpub-
lished ‘Scale for Disordered/Disoriented Infant Behaviour’ for the Clown study. The ori-
ginal version appears to be lost, but a version of the scale was typed up by Main for Karin
Grossmann in June 1982.123 The manuscript shared with Grossmann is a nine-point scale,
indexing behaviours including ‘stereotypies, episodes of immobilisation, disoriented behav-
iour, misdirected behaviour, sudden disordered outbursts of activity, and sudden uninter-
pretable noises or movements’. In the 1982 manuscript, behaviour is identified by Main as
‘disordered’ based on the ‘extent to which such behaviour may be indicative of difficulties in
functioning’, for instance by virtue of lacking either ‘orientation’ or ‘purpose’.
Main was startled and intrigued by the fact that so many of these odd behaviours in human
infants resembled the conflict behaviours described by Hinde in animals (Chapter 1): simul-
taneous or sequential contradictions in approaching the caregiver, misdirected approaches,
poorly coordinated combinations of movements, freezing in place, or signs of confusion,
out-of-context anger, or tension on reunion with the caregiver. In some cases, a child’s be-
haviour in the Strange Situation seemed interrupted or misshapen by conflict behaviours.
However, in other cases, the infant’s response on reunion with their caregiver was dominated
by other behaviours, making an Ainsworth classification impossible. Hinde was in Berkeley
for the year 1979–80 as the Hitchcock Professor at the University of California, so Main
121 Later in her career, another relevant study designed but not conducted by Main was to administer the Strange
Situation without any toys. Since avoidance is a redirection of attention from the caregiver to the environment,
Main suspected that the absence of toys would foil the avoidant strategy, and result instead in intense distress.
Main, M. (1999) Epilogue. Attachment theory: eighteen points with suggestions for future studies. In J. Cassidy
& P. Shaver (eds) Handbook of Attachment (pp.845–87). New York: Guilford, p.858. Madigan and colleagues sub-
sequently showed that parents are more likely to show anomalous or alarming behaviours towards their children
when asked to play with them without toys, a more challenging demand. Madigan, S., Moran, G., & Pederson, D.
(2006) Unresolved states of mind, disorganized attachment relationships, and disrupted interactions of adolescent
mothers and their infants. Developmental Psychology, 42(2), 293–304.
122 George, C. & Main, M. (1980) Abused children: their rejection of peers and caregivers. In T. Field, S.
Goldberg, D. Stern, & A. Sostek (eds) High-Risk Infants and Children: Adult and Peer Interactions (pp.293–312).
New York: Academic Press, p.304.
123 Main, M. (1982) Scale for Disordered, Disoriented and Undirected Behaviors—Developed for Clown Session.
Unpublished manuscript, made available by Klaus and Karin Grossmann.

was able to discuss her observations and ideas with him; she later described the influence of
Hinde on her thinking about conflict behaviours shown in the Strange Situation as ‘strong
and direct’.124
A new category
In a paper from 1981, Main and Weston published an interim report on 46 families where
the Strange Situation had been coded for both infant–mother and infant–father interaction.
They reported that 12.5%125 could not be classified using the Ainsworth system: ‘infants
were not judged unclassifiable if they merely showed conflicted behavior or behaved oddly
during the strange situation: one infant stared, talked to the wall, and indeed seemed almost

to hallucinate, but was nonetheless regarded as secure in relation to the parent. Infants were
judged unclassifiable only if their social and emotional behavior toward the parent could not
be encompassed by the present Ainsworth system.’126 However, in fact the large majority of
the unclassifiable infants did show conflicted, confused, or apprehensive behaviours.
Yet Main and Weston’s report on ‘unclassifiable’ cases was only on the Strange Situation
procedures they had coded by that point. This was a fraction of the hundreds of Strange
Situations conducted by the research group. The others still needed to be coded. In 1977,
Judith Solomon joined Main’s lab following a graduate focus on ethology and compara-
tive psychology, which included attention to conflict behaviours in animals. Yet Main left
soon after Solomon’s arrival for Bielefeld, returning to Berkeley only in 1978. In her absence,
Main instructed Solomon to learn how to classify the Strange Situation guided by Donna
Weston and by feedback that Main sent by post.127 Solomon began to compile detailed notes
on cases she found difficult to classify, for discussion with Main on her return. At the same
time, Solomon began to study a sample of maltreated infants in the Strange Situation with
Carol George—a research project that was ultimately not completed or published. In coding
these two samples, Solomon observed a variety of behaviours discrepant with the Ainsworth
coding protocols, though they were particularly common in the maltreated sample: apparent
signs of depression in infants; indications that an infant was attempting to muster a coherent
124 Bahm, N.I.G., Main, M., & Hesse, E. (2017) Unresolved/disorganized responses to the death of important
persons: relations to frightening parental behavior and infant disorganization. In S. Gojman de Millan, C.
Herreman, & L.A. Sroufe (eds) Attachment Across Clinical and Cultural Perspectives: A Relational Psychoanalytic
Approach (pp.53–74). London: Routledge, p.56. The chapter also mentions that ‘Mary Main discussed conflict be-
haviour during two visits to Niko Tinbergen’, one of the founders of ethology and of the empirical study of conflict
behaviours: p.71.
125 The 12.5% figure is rather mysterious. Table 1 in Main and Weston has 12 Strange Situations unclassifiable
with either father or mother, out of 61 Strange Situations with each of father or mother. However, this would give a
proportion of 9.7% unclassifiable. The researchers’ comparison of classification with mother and father was made
on 46 infants, i.e. 92 strange situations; 12 of 92 is 13% which is much closer to the 12.5% figure, but then does
not agree with Table 1. Main, M. & Weston, D.R. (1981) The quality of the toddler’s relationship to mother and to
father: related to conflict behavior and the readiness to establish new relationships. Child Development, 52(3), 932–
40. In 1986, Main reported that ‘152 strange situations were reviewed; 19 of the strange situations (12.5%) were
judged unclassifiable’. Yet 19 of 152 is 8%, not 12.5%. Main, M. & Solomon, J. (1986) Discovery of a new, insecure-
disorganized/disoriented attachment pattern. In M. Yogman & T.B. Brazelton (eds) Affective Development in
Infancy (pp.95–124). Norwood, NJ: Ablex, p.103.
conflict behavior and the readiness to establish new relationships. Child Development, 52(3), 932–40, p.934
127 Duschinsky, R. (2015) The emergence of the disorganized/disoriented (D) attachment classification, 1979–
1982. History of Psychology, 18(1), 32–46.

strategy of approach or avoidance but failing to achieve this; infants initially approaching
the caregiver but then veering off; and disoriented behaviours (e.g. the child leaves its arm
hanging in the air).
The first use of a ‘D’ classification in the Berkeley study occurred in August 1979.128 This
was a case coded by Solomon. The infant’s behaviour on reunion displayed a whole variety
of conflict behaviours. After attempting at length to work out a best-fit Ainsworth classifica-
tion, Solomon eventually gave up: ‘Well, it is not A and it is not B and it is not C. I’m going
to call it D.’ Solomon’s first use of the ‘D’ label was in pique. However, from 1979 Solomon
and Main began thinking about and discussing a ‘D’ category for the Strange Situation. It
was in line with Hinde’s emphasis on conflict behaviours and on Main and Weston’s existing
work on cases unclassifiable by the Ainsworth coding protocols. Additionally, the fact that
such behaviours were more common in the maltreatment sample Solomon was also coding

spurred interest in their meaning.
Though they encased their observations within a category, Main and Solomon did not
assume that the various forms of discrepant behaviour all meant the same thing. They were
thinking in terms of the work by ethologists on conflict behaviour (Chapter 1):
Our continuing focus on conflict behavior as it related to disorganized attachment status

was derived from the work of the ethologists Niko Tinbergen, Konrad Lorenz and Robert
Hinde. Hinde (1966) noted that conflicting behavioral tendencies are present under many
conditions, with the most frequent outcome being that only one of the tendencies is actu-
ally observed. In some situations, however, both of the conflicting tendencies do appear in
an animal’s behavior, for example, in the alternating exhibition of the opposing behavioral
patterns. In other situations, a third, apparently unrelated behavior pattern appears, such
as preening in birds caught between feeding and flight. There are, additionally, conflict
situations in which an animal is observed to freeze, to exhibit abnormal postures, or to en-
gage in anomalous movements.129
Based on observations across various species, Hinde articulated very clearly the different
conditions that would lead to different forms of conflict behaviour. It would undoubtedly
have been possible for Main and Solomon to develop scales for conflicted, confused, or ap-
prehensive infant behaviours. The use of differentiated scales within an overarching cat-
egory was precisely the solution that Main’s research group later adopted in the 1990s when
studying parental behaviour.130 That some infants scored on more than one scale would
hardly have been a problem. Yet in the 1980s, in a field of empirical inquiry grounded on
Ainsworth’s Strange Situation and her patterns of attachment, categories were currency; to
a large extent, categories formed a horizon of how data could readily be conceptualised and
discussed at the time by attachment researchers. The publication of the influential DSM-III
in 1983 (Chapter 1), as a clinical system based on categories, may also have played a distal
role in supporting category-based thinking.
128 Judith Solomon, personal communication, September 2012.

129 Main, M. (1999) Disorganized Attachment in Infancy, Childhood, and Adulthood: An Introduction to the
Phenomena. Unpublished manuscript, Mary Main & Erik Hesse personal archive; Hinde, R. (1966, 1970) Animal
Behavior, 2nd edn. New York: McGraw.
130 Abrams, K.Y., Rifkin, A., & Hesse, E. (2006) Examining the role of parental frightened/frightening subtypes
in predicting disorganized attachment within a brief observational procedure. Development & Psychopathology,
18(2), 345–61.
In characterising conflict behaviours seen in the Strange Situation as a category, Main

and Solomon sought to support their attempts to win attention to a phenomenon—or inter-
related phenomena—that they saw as important, despite obstacles to such recognition.
The obstacles were substantial. There was intense resistance to making any change to the
Ainsworth coding system for the Strange Situation. Additionally, many of the behaviours
lasted only a few seconds, and could easily be discounted as simply babies doing odd things
because they are figuring out motor coordination. To gather more data and to support a
claim to wider relevance, Main’s laboratory began to collect unclassifiable tapes from other
researchers working with high-risk samples such as Mary J. O’Connor, Elizabeth Carlson,
Leila Beckwith, and Susan Spieker. Drawing on an analysis of both the Berkeley Strange
Situations and these additional recordings from other samples, in the winter of 1982 Main
and Solomon began work on what would be their 1986 chapter announcing ‘discovery of a

new, insecure-disorganised/disoriented attachment pattern’. This represented a shift in ter-
minology. In initial discussions with Solomon, Main used the terms ‘disordered’ and ‘dis-
oriented’ as overarching labels for the forms of conflict behaviour, following on from the use
of these terms in the coding system for the Clown study. However, the term ‘disordered’ was
judged to sound pejorative. It was not clear what the different behaviours meant, and so it
was regarded as premature to describe them as a mental disorder.
In looking for a more descriptive word, the term ‘disorganised’ was available to Main and
Solomon. The term had entered common use in neurology in the 1940s and 1950s to refer to
the potential for strong feelings to be experienced as overwhelming. Affects such as anxiety,
anger, awe, and ecstasy could be so intense and absorbing they make a person lost in the affect
and unable to respond to the cues of the situation they are in.131 The neurological literature
used ‘disorganised’ to refer to the state following such intense and absorbing affects. It essen-
tially meant ‘overcome’ by the feeling. Bowlby introduced the term into attachment theory
in his 1960 paper on ‘Separation anxiety’, in order to propose that, whether in childhood or
adulthood, grief could also be a disorganising affect. Having emphasised the value of the
concept of ‘disorganization’, Bowlby then promised that ‘this is a concept to which we shall
be returning in a paper to follow’.132 The promise was left unfulfilled, eliciting letters from
readers requesting more detail about this idea of ‘disorganisation’ and why Bowlby thought
it so important.133 In Attachment, Volume 1, Bowlby returned to the issue of disorganisation.
He restated his earlier claim that behaviour can become uncoordinated in the context of cer-
tain intense emotions: ‘Above a certain level, however, efficiency may be diminished; and,
when in an experimental situation total stimulation is very greatly increased, behaviour be-
comes completely disorganised.’134 Here he added, however, that such overwhelming inten-
sity is specifically expected in the context of conflicts between strong motivational systems,
and ‘in some cases, indeed, the behaviour that results when two incompatible behavioural
systems are active simultaneously is of a kind that suggests pathology’.135 He then reviewed
Hinde’s discussion of the various forms of ‘conflict behaviour’ seen in animals when two
competing behavioural tendencies were activated. These included rapid transition between
131 Leeper, R.W. (1948) A motivational theory of emotion to replace ‘emotion as disorganized response’.
Psychological Review, 55(1), 5–21; Goldstein, K. (1951) On emotions. Journal of Psychology, 31, 37–59.
133 E.g. Bastiaans, J. (1963) Letter to John Bowlby on behalf of the Dutch Psychoanalytic Society, 22 January 1963.
PP/BOW/B.3/20.
134 Bowlby, J. (1969) Attachment. London: Penguin, p.96–7.
135 Ibid.
one tendency and the other, poorly coordinated combinations, freezing in place, misdir-
ected movements, or signs of confusion or tension.136
Main’s colleagues at Berkeley, Block and Block, also drew on this term ‘disorganisation’
from neurology. They used the word to mean ‘immobilised, rigidly repetitive or behaviour-
ally diffuse’ flooding behaviours, which could be expected when a child was experiencing
‘a difficulty in recouping’ in the face of behavioural conflict and distress.137 The concept of
disorganisation may have also appealed to Main and Solomon in the context of a technical
use of the term ‘organisation’ that had sprung up following Ainsworth, Sroufe, and Waters
(Chapters 2 and 4), who had cut the term ‘organisation’ loose from its ordinary language
meaning and given it a technical sense: behaviour coordinated to achieving the set-goal of
the attachment system. Behaviour that did not seem oriented to the achievement of this set-
goal was therefore described by Ainsworth as ‘disorganised’ as early as 1968 in correspond-

ence138 and 1972 in print.139 These were, importantly, the exact years in which Main was
Ainsworth’s doctoral student. What the different forms of conflict behaviour observed by
Main and Solomon had in common was that they did not seem like components of a co-
herent and coordinated attempt to achieve proximity. Avoidance could be interpreted as ‘in
the service of proximity’, and therefore oriented to the achievement of the set-goal of the
attachment system. By contrast, conflicted, confused, or apprehensive behaviours did not
seem ‘organised’ to Main, in the technical sense of the word.140 It may well have seemed lo-
gical, therefore, to call them disorganised.
At a four-day workshop in 1985 at the University of Washington, Ainsworth sat on the
floor to be as close as possible to the screen as Main showed her tapes coded with the new
disorganised classification. At the end of the event, Ainsworth wrote to Bowlby that she and
‘everyone there was most impressed with the need for adding a new “D” or disorganised
category to the classification system’.141 Bowlby also became convinced that the behaviours
identified by Main and Solomon were likely ‘of great clinical concern’, and he expressed pride
in their work.142 However, in the margins of his personal copy of Main and Solomon’s 1986
chapter ‘Discovery of an insecure-disorganized/disoriented attachment pattern’, he wrote
136 Hinde, R. (1970) Animal Behavior, 2nd edn. New York: McGraw-Hill, Chapter 13.
137 Block, J. & Block, J. (1980) The role of ego-control and ego-resiliency in the organisation of behaviour.
In W.A. Collins (ed.) Development of Cognition, Affect, and Social Relations: The Minnesota Symposia on Child
Psychology, Vol. 13 (pp.39–101). Hillsdale, NJ: Lawrence Erlbaum, p.48.
138 Ainsworth, M. (1968) Letter to J.L. Gewirtz, 5 August 1968. PP/Bow/K.4/12: ‘I do agree that there are varied
indices of attachment, and my data suggest that these are not necessarily highly correlated. I also tend to agree that
the approach behaviours are more stable indices of attachment than are the “disorganization” responses—perhaps
because there may be more diverse determiners of disorganization behaviour than there are for approach behav-
iour to specific persons. I think it will require much more research to ascertain how “disorganization” responses
relate to the more “positive” components of attachment.’
139 Ainsworth, M. (1972) Attachment and dependency: a comparison. In J. Gewirtz (Ed.), Attachment and
Dependency (pp.97–137). Washington, DC: Winston: ‘Gewirtz and Cairns (both in this volume) have also dis-
tinguished the “positive” indices from other indices of attachment. They characterise the behaviour activated by
separation as disorganised, whether because of the emotional component contingent upon the frustration of sep-
aration or because of the disruption of other ongoing behavioural sequences’ (114).
140 Main, M. & Solomon, J. (1986) Discovery of a new, insecure-disorganized/disoriented attachment pattern.
In M. Yogman & T.B. Brazelton (eds) Affective Development in Infancy (pp.95–124.) Norwood, NJ: Ablex: ‘Infants
who cannot be classified within the present “A, B, C” system do not appear to us to resemble one another in strange
situation in coherent, organised ways’ (97).
141 Ainsworth, M. (1985) Letter to John Bowlby, 14 February 1985. PP/BOW/B.3/8.
142 Bowlby, J. (1988) The role of attachment in personality development. In A Secure Base (pp.134–54).
London: Routledge Press, p.124; Bowlby, J. (1990, 2015) John Bowlby: an interview by Virginia Hunter. Attachment,
9, 138–57: ‘Mary Main, Inge Bretherton . . . very admirable, able people. So the field is now being explored by first
class scientists doing first class research of high clinical relevance. That I’m very, very proud of ’ (151).
that the authors would have done better to call it a ‘status’ because the unitary term ‘pattern’
may result in confusion if readers interpret it in the Ainsworth sense. In this marginalia, he
observed that Main would likely agree with this reasoning, since she had indicated to him in
a discussion in March 1986 that, in her view, ‘trauma to the attachment system causes disor-
ganisation of behaviour but does not create a new category’.143
However, one key advantage of making a new classification was that it cleaned up the
existing categories: many of the children who could now be classified as disorganised in at-
risk samples had previously been classified as ‘secure’ because, despite manifest displays of
conflict or confusion, they had protested the departure of their caregiver and been com-
forted on reunion. This was a particular problem, for instance, in the influential Minnesota
Longitudinal Study of Risk and Adaptation (Chapter 4) and in clinical samples.144 Ainsworth
had also observed conflict behaviours in the samples reported in Patterns of Attachment,

and had even entered them into the coding protocol as characteristic of the B4 subtype
(Chapter 2). Main was able to demonstrate in Washington that conflicted, confused, or
apprehensive behaviours were not limited to this subtype, but could be found across the
Ainsworth patterns of attachment. In fact, Main could show examples of Strange Situation
procedures where forms of conflict, confusion, or apprehension so disrupted reunions that it
was the dominant response. As she wrote in a paper composed in 1986, the extent of conflict
behaviour was in theory on a spectrum. In a small number of cases, though, the conflict was
so pervasive that it drowned out other responses: ‘disorganization operates as a category only
in extreme cases, being otherwise . . . a dimension’.145
The meaning of disorganisation
By the time of their 1990 chapter offering protocols for coding the new category, Main and
Solomon had closely analysed 100 recordings of infants from low-risk samples and 100 re-
cordings from high-risk samples.146 They proposed certain infant behaviours to be indica-
tive of a ‘disorganised’ attachment response. They clustered the identified behaviours into
seven indices based on how they appeared:
I. Sequential displays of contradictory behaviour;

II. Simultaneous display of contradictory behaviour;
III. Undirected, misdirected or incomplete movements;
IV. Stereotypies, mistimed movements and anomalous postures;
V. Freezing or stilling;
143 Bowlby, J. (1986) Marginalia on Main and Solomon’s ‘Discovery of an insecure-disorganized/disoriented at-
tachment pattern’. PP/BOW/J.7/6; see Reisz, S., Duschinsky, R., & Siegel, D.J. (2018) Disorganized attachment and
defense: exploring John Bowlby’s unpublished reflections. Attachment & Human Development, 20(2), 107–34.
144 E.g. Gaensbauer, T.J. & Harmon, R.J. (1982) Attachment behavior in abused/neglected and premature in-
fants. In R.N. Emde & R.J. Harmon (eds) The Development of Attachment and Affiliative Systems (pp.263–89).
145 Main, M. & Cassidy, J. (1988) Categories of response to reunion with the parent at age 6: predictable from
infant attachment classifications and stable over a 1-month period. Developmental Psychology, 24(3), 415–26,
pp.423–4.
146 Main, M. & Solomon, J. (1990) Procedures for identifying infants as disorganized/ disoriented during
Preschool Years (pp.121–60). Chicago: University of Chicago Press.
VI. Display of apprehension of the caregiver;

VII. Overt signs of disorientation.
To facilitate coding, Main and Solomon presented general guidelines and a nine-point
scale.147 This scale is partly a measure of the extent of inferred disorganisation of the attach-
ment system, and partly a ranking of how certain a coder is that they are seeing behaviour
suggesting disruption of the attachment behavioural system. A score above 5 is sufficient for
placement of the dyad into a D classification.148
As Main and Solomon acknowledged, behaviours pertaining to the first five indices were
already discussed by Hinde and Bowlby as classic ‘conflict behaviours’ (Chapter 1). Main
and Solomon introduced two further kinds of behaviour, based on their analysis of the re-
cordings: apprehension directed towards the caregiver, and disorientation or confusion on

reunion or in proximity with the caregiver. They knew that apprehension did not really fit
under the term ‘disorganisation’, since the behaviour could well be behaviourally coherent
and coordinated. However, at a more abstract level of analysis, apprehension of the care-
giver nonetheless seemed in conflict with proximity, considered as the set-goal of the attach-
ment system.149 Unlike the Ainsworth patterns which tended to be more discrete, Main and
Solomon found that infants who showed one kind of conflict, apprehension, or disorien-
tation also regularly showed some other kind. This led them to regard the phenomena as
highly related, even if not necessarily all of a piece. As a result, the different behaviours were
grouped together as a ‘disorganised’ category.
Yet their use of the term ‘disorganised’ differed from the dictionary definition. And they
did not clarify this for the reader at the time, something both authors now regret. The dic-
tionary, everyday meaning of the term ‘disorganization’ suggests randomness and a lack of
predictable responsiveness to contingencies: ‘to destroy the system or order of; to throw
into confusion or disorder’.150 According to the connotations of the word in ordinary lan-
guage, what Main and Solomon appeared to be proposing was a category of undifferentiated
chaos. The scientific concept was taken hostage by its ordinary language connotations.151 In
Spangler and Schieche, for instance, the authors wrote that ‘as disorganized infants, by defin-
ition, do not have any coherent strategies, behavioral regulation is restricted or even not pos-
sible at all’.152 Likewise, Wanaza and colleagues wrote that ‘disorganization is defined as the
147 Ibid. p.133.

148 Technically, a score of 5 was given as the mid-point, and could be the basis for a D classification on the judge-
ment of the coder. However, subsequent coders have tended to mark such cases as 5.5.
149 Duschinsky, R. (2015) The emergence of the disorganized/disoriented (D) attachment classification, 1979–
1982. History of Psychology, 18(1), 32–46.

150 Oxford English Dictionary (1990) ‘disorganize’. Oxford: Oxford University Press.
151 Exceptions include Waters, E. & Crowell, J.A. (1999) Atypical attachment in atypical circumstances.
Monographs of the Society for Research in Child Development, 64(3), 213–20; Rauh, H., Ziegenhain, U., Muller,
B., & Wijnroks, L. (2000) Stability and change of infant–mother attachment in the second year of life: relations
to parenting quality and varying degrees of daycare experience. In P.K. Crittenden & A.H. Claussen (eds) The
Organization of Attachment Relationships: Maturation, Culture and Context (pp.251–76). Cambridge: Cambridge
University Press; Slade, A. (2014) Imagining fear: attachment, threat, and psychic experience. Psychoanalytic
Dialogues, 24(3), 253–66; Solomon, J. & George, C. (2016) The measurement of attachment security and related
constructs in infancy and early childhood. In J. Cassidy & P.R. Shaver (eds) The Handbook of Attachment, 3rd edn
152 Spangler, G. & Schieche, M. (1998) Emotional and adrenocortical responses of infants to the Strange
Situation. International Journal of Behavioral Development, 22(4), 681–706, p.700. A further factor subse-
quently contributing to the image of disorganisation as chaos was findings in the early 1990s by Spangler and by
Hertsgaard of elevated hypothalamic–pituitary–adrenal (HPA) reactivity following the Strange Situation among
young children who received a disorganised attachment classification. Spangler and Grossmann had 9 D dyads,
collapse of attachment strategy under conditions of stress; under such conditions, disorgan-
ized individuals select a set of behaviors that are irrelevant to their need for downregulation
of discomfort’.153 Commentators at the research–practice interface have criticised the idea of
attachment disorganisation as clinically evocative but unhelpful, since the concept means in-
comprehensibility and therefore offers no clues to clinicians about how to proceed.154
In fact, the term ‘disorganisation’ was used in three ways in Main and Solomon’s 1986 and
1990 chapters—not one of which aligns well with the dictionary definition. The term was
used by Main and Solomon to describe (i) observable behaviour that seemed to lack order
or relevance to achieving proximity with the caregiver; (ii) a disruption of the attachment
behavioural system, caused by past experiences of child–caregiver interaction and inferable
from observed behaviour; (iii) and the label given to the category used for coding the Strange
Situation.155 In using the same term ‘disorganised’ to refer to both behaviour and/or psycho-

logical process and/or the overall category, Main and Solomon had a specific aim, though
it was not well articulated at the time. ‘Disorganisation’ was used as a conceptual tool for
picking out ‘an observed contradiction in movement pattern, corresponding to an inferred
contradiction in intention or plan’.156 The theoretical stakes of using the term ‘disorganized’
to mean both behaviour and psychological process was the claim that the diverse behav-
iours picked up by the Main and Solomon indices could well have different antecedents and
sequelae, but what they had in common was that they suggested disruption or breakdown
at the level of the attachment system. This was the rationale for the ‘radical notion that the
many, highly diverse indices of disorganization and disorientation can be placed under one
heading’.157
In the early 1990s, Main created a revised version of the Main and Solomon coding proto-
cols, though she did not publish it. This version is distributed to trainees at the Minnesota
Strange Situation training institute, run by Alan Sroufe and Betty Carlson.158 Main’s revi-
sions predominantly entailed adding further forms of behaviour under the seven indices.
For instance ‘overbright greeting’ was added to Direct Indices of Apprehension (Index VI),
presumably as this was interpreted as an early form of caregiving or compliance by the child
and Hertsgaard had 11. However, in a later study by Luijk and colleagues with 57 D dyads, the earlier findings
failed to replicate. Furthermore, in 1999, Spangler and Grossmann later acknowledged that the overwhelming
majority of the association between D and HPA reactivity in their study was attributable to Index VII behaviour
(direct indices of disorientation), and there was no association at all for conflict behaviours where confusion or ap-
prehension were not also present. Spangler, G. & Grossmann, K.E. (1993) Biobehavioral organization in securely
and insecurely attached infants. Child Development, 64, 1439–50. Hertsgaard, L., Gunnar, M., Erickson, M.F., &
Nachmias, M. (1995) Adrenocortical responses to the strange situation in infants with disorganized/disoriented
attachment relationships. Child Development, 66(4), 1100–106; Spangler, G. & Grossmann, K.E. (1999) Individual
and physiological correlates of attachment disorganization in infancy. In J. Solomon & C. George (eds) Attachment
Disorganization (pp.95–124). New York: Guilford; Luijk, M.P., Velders, F.P., Tharner, A., et al. (2010) FKBP5 and re-
sistant attachment predict cortisol reactivity in infants: gene–environment interaction. Psychoneuroendocrinology,
35(10), 1454–61.
153 Wazana, A., Moss, E., Jolicoeur-Martineau, A., et al. (2015) The interplay of birthweight, dopamine re-
ceptor D4 gene (DRD4), and early maternal care in the prediction of disorganized attachment at 36 months of age.
Development & Psychopathology, 27, 1145–61, p.1157.
154 E.g. Baim, C. & Morrison, T. (2014) Attachment-Based Practice with Adults. Brighton: Pavilion.
155 Duschinsky, R. & Solomon, J. (2017) Infant disorganized attachment: clarifying levels of analysis. Clinical
Child Psychology and Psychiatry, 22(4), 524–38.

156 Main, M. & Solomon, J. (1990) Procedures for identifying infants as disorganized/ disoriented during
Preschool Years (pp.121–60). Chicago: University of Chicago Press, p.133.
157 Ibid. p.151.
158 https://attachment-training.com/training/#1.
towards the caregiver. Main also added guidance to coders on ‘Major Considerations’ for
coding disorganised attachment. The primary consideration was ‘Is the behaviour inexplic-
able (no evidence of immediate goal or rationale) OR is the behaviour explicable only if we
presume: a) The baby is afraid of the parent; b) The baby is inhibited from approach without
being able to shift attention to the environment?’159 Here we can see that disorganisation is
operationalised as behaviours that seem confused, apprehensive, or conflicted. Also visible
is the technical distinction for Main between avoidance and disorganised attachment, hin-
ging on the fact that the former can direct attention to the environment in order to maintain
regulation, whereas in the latter this conditional strategy of the use of attention to circum-
vent approach/avoidance conflict is not feasible. However, this manuscript remained un-
published, and generally available only to junior researchers who are tasked with coding.
With exceptions, many senior researchers in the field of attachment research have there-

fore not known how the category is actually coded, for instance that only some infants will
look apprehensive—others will look confused or conflicted (or a combination of the three).
Senior researchers have tended to be unaware of the significant role coders must give to in-
ferences about how the infant is directing attention in the Strange Situation.
In putting confused, apprehensive, and conflicted behaviours together, Main and
Solomon did not intend to imply that they all meant the same thing. In fact, they stated in
print that ‘our discovery of the D category of infant Strange Situation behaviour rested on
an unwillingness to adopt the “essentialist” or “realist” position regarding the classification
of human relationships’.160 Their epistemological position, like Bowlby’s, treated categories
as ‘provisional, albeit nonarbitrary approximations’,161 and Main and Solomon end their
1990 chapter with an extended criticism of essentialist approaches to categorisation in the
history of biology. Yet these remarks by Main and Solomon on the dangers of essentialism
have never been cited or discussed. In general, the ensuing literature respectfully cites but
gives little evidence of having directly read the book chapters introducing the classification.
Main and Solomon were widely interpreted as introducing an exhaustive addition to the
Ainsworth system. They have therefore been accused of being bent upon ‘reducing complex
human experience to typologies’.162 Likewise, it has been assumed that their category aimed
simply to soak up possible variation in human behaviour beyond the Ainsworth patterns
and treat it all as evidence of dysfunction.
For instance, Gaskin argued that ‘the category is really just a residual one’: rather than
itself designating any meaningful phenomena, the existence of the classification ‘might be
seen more productively as evidence of the inadequacy of the three attachment classifica-
tions’ introduced by Ainsworth on the basis of her Baltimore middle-class sample.163 This
159 The other major considerations were: “2) Timing of the appearance of disorganized behaviour: a) stronger
index of disorganisation if occurs at first moments of reunion; b) however, even D-like behaviour appearing only
in Episode 3 may yield a D classification. 3) Consider what the baby does next, namely, if the baby goes to the par-
ent as though for comfort after a bit of disorganisation (i.e. stereotypies then comforted). If they become organised
quickly, discount the D behaviour). Main, M. (undated) Disorganised/Disoriented Classification Scheme: Major
Considerations. Unpublished manuscript, received from Elizabeth Carlson, and cited with her permission.
of representation. Monographs of the Society for Research in Child Development, 50, 66–104, p.99.
162 O’Shaughnessy, R. & Dallos, R. (2009) Attachment research and eating disorders: a review of the literature.
Clinical Child Psychology & Psychiatry, 14(4), 559–74, p.559.

163 Gaskins, S. (2013) The puzzle of attachment. In N. Quinn & J.M. Mageo (eds) Attachment Reconsidered
(pp.33–66). London: Palgrave, p.39.

misunderstanding has a history that goes back to the very introduction of the classification.
Mark Cummings was one of the editors of the volume within which Main and Solomon’s
1990 chapter was published. In his contribution to the volume, he argued that ‘deviations
from expected sequences do not constitute a sufficient criterion for classification’.164 Against
what he took to be Main and Solomon’s perspective, he proposed that D behaviours could
not all be expected to reflect the same process of breakdown of ‘general functioning’, and that
therefore the category lacked coherence and meaning.165 A better criterion for disorganisa-
tion, Cummings argued, would be behaviours that do not appear to function to achieve ‘felt
security’, which Sroufe and Waters had proposed, as the set-goal of the infant’s attachment
system (Chapter 4).166
Psychological constructs are usually subjected to statistical analysis to see which elem-
ents cluster together and which among these clusters are especially involved in associations

with other variables of interest. By contrast, though it is not uncommon to hear praise of the
‘strong psychometric properties’ of the Strange Situation procedure and its classification as
a whole,167 in fact no psychometric study has ever been reported on the infant disorganised
attachment classification. Though perhaps less in social psychology than in developmental
psychology (Chapter 5), psychological scientists are generally satisfied to work with instru-
ments that have expectable correlates and fit with a conceptual framework, no matter that
psychometric work is incomplete. Furthermore, psychometric work receives little profes-
sional reward for psychological researchers. Another potential obstacle is that a sample of
adequate size for psychometric research is needed—though in the case of disorganised at-
tachment this cannot have been the only factor in play, since a number of such samples have
long been available now. Rather, it was simply assumed that disorganisation was best treated
as a single category.
No assessment was made, for instance, of whether the one-to-nine scale would be more
serviceable than the category. And no exploration was made of the component parts of dis-
organisation, for instance to economise the coding system. The underdeveloped and rather
complicated coding protocol made application of the classification time-consuming and, be-
yond this, very hard to learn. The result has been a significant bottleneck for research using
the Strange Situation due to the small number of reliable coders and the disincentives of
investing time and resource in learning the system. It has also limited the circulation of im-
portant practical knowledge, leading to frequent speculative discussions about disorganisa-
tion among attachment researchers that have been wholly cut loose from how it is actually
coded.168 Yet prediction of negative outcomes was treated by researchers—and subsequently
by clinicians and policy-makers—as validity. Indeed, the disorganised classification also
quickly bore fruit in predictive significance.
164 Cummings, E.M. (1990) Classification of attachment on a continuum of felt-security. In M.T. Greenberg, D.
Cicchetti, & E.M. Cummings (eds) Attachment in the Preschool Years (pp.311–38). Chicago: University of Chicago
Press, p.319.
165 Ibid. p.316.
166 Ibid. p.326.
167 E.g. Bernard, K., Dozier, M., Bick, J., Lewis-Morrarty, E., Lindhiem, O., & Carlson, E. (2012) Enhancing at-
tachment organization among maltreated children. Child Development, 83(2), 623–36, p.632.
within the disturbance of coherence’. Social Science & Medicine, 200, 52–8. See also van IJzendoorn, M.H. (1995)
Adult attachment representations, parental responsiveness, and infant attachment: a meta-analysis on the pre-
dictive validity of the Adult Attachment Interview. Psychological Bulletin, 117(3), 387–403: ‘Amount of training in
coding the disorganized/disoriented category also appeared to be strongly related to differences in effect sizes; less
training in the application of the complicated coding system was associated with smaller effect sizes’ (394).
The short-term test-retest stability of disorganised attachment as assessed twice using the
Strange Situation was r = .35. This was regarded as at least adequate to validate the measure,
and potentially rather high considering that the classification was, at times, made on the
basis of quite fleeting behaviours169—though stability between infancy and toddlerhood or
preschool has been found to be much lower.170 Incidence of the disorganised attachment
classification was also discovered to be more common in clinical samples, samples known to
social services, and in samples facing multiple adversities, suggesting that it reflected some
adverse experience or process. In community samples with relatively few adversities, around
15% of infants show a sufficiently high degree of confused, disoriented, or apprehensive be-
haviours towards their caregiver in the Strange Situation for a disorganised classification to
be assigned. However, this increases to a majority of infants from families drawn from sam-
ples facing multiple compounding adversities, or where maltreatment of the child has been

documented.171 Of particular importance for acceptance of the new classification, however,
was the prediction of conduct problems. An early and influential study by Lyons-Ruth and
colleagues, reported in 1991, found disorganised infant attachment associated with aggres-
sive behaviour in preschool in a sample of dyads known to child protection services.172
The link between the infant disorganised attachment classification and later aggressive
behaviour has been confirmed by numerous studies. By 2010, Fearon and colleagues could
report from a meta-analysis of 69 studies a modest but material association (d = .34) be-
tween disorganisation and various externalising behaviours (including attentional prob-
lems).173 The association was stronger in samples where families were known to clinical or
professional services (d = .43) or facing low socioeconomic status (d = .44). Indeed, for many
samples, disorganised infant attachment only predicted later externalising problems in con-
junction with other forms of adversity.174 For disorganised attachment, as for the Ainsworth
avoidant classification (Chapter 2), the Strange Situation conducted with two-year-olds
seemed to predict later externalising problems more effectively than when it was conducted
with infants, perhaps because one of the particular developmental challenges of this age is
169 Van IJzendoorn, M.H., Schuengel, C., & Bakermans-Kranenburg, M.J. (1999) Disorganized attachment in
early childhood: meta-analysis of precursors, concomitants, and sequelae. Development & Psychopathology, 11(2),
225–50.
170 The latest meta-analytic findings are reported by Opie, J. (2018) Attachment stability and change in early
childhood and associated moderators. Unpublished doctoral thesis, Deakin University, Melbourne. Extracting
results from 56 studies, stability from infancy to toddlerhood was r = .10, p = .09, df = 7.12. Stability from infancy to
preschool age was r = .19, p = .052, df = 6.81. Stability of disorganisation assessed using the Cassidy–Marvin system
was higher: r = .32, p = .02, df = 3.30.
171 Cyr, C., Euser, E.M., Bakermans-Kranenburg, M., & van IJzendoorn, M. (2010) Attachment security and dis-
organization in maltreating and high-risk families: a series of meta-analyses. Development & Psychopathology, 22,
87–108.
172 Lyons-Ruth, K., Repacholi, B., McLeod, S., & Silva, E. (1991) Disorganized attachment behavior in
infancy: short- term stability, maternal and infant correlates, and risk- related subtypes. Development &
Psychopathology, 3(4), 377–96. Another study of particular importance for acceptance of Main’s methodological
innovations as a whole was the finding by Fonagy, Steele, and Steele that four-way Strange Situation classifications
could be predicted by four-way AAIs conducted prenatally with mothers. Fonagy, P., Steele, H., & Steele, M. (1991)
Maternal representations of attachment during pregnancy predict the organization of infant–mother attachment
at one year of age. Child Development, 62(5), 891–905.
havior: a meta-analytic study. Child Development, 81(2), 435–56. On the interpretation of effect sizes in psych-
ology see Funder, D.C. & Ozer, D.J. (2019) Evaluating effect size in psychological research: sense and nonsense.
Advances in Methods and Practices in Psychological Science, 2(2), 156–68.
174 For a discussion see Tharner, A., Luijk, M.P., van IJzendoorn, M.H., et al. (2012) Infant attachment, parenting
stress, and child emotional and behavioral problems at age 3 years. Parenting, 12(4), 261–81.
learning to regulate frustration in relationships.175 Researchers have also documented other

sequelae, including small-to-moderate associations with social competence and friendships
(d = .25).176
A meta-analysis by Groh and colleagues found no association with later depression or
anxiety.177 But other mental health outcomes have been documented. For instance, a clas-
sification of disorganised/disoriented attachment in infancy predicts severity of later post-
traumatic stress disorder (PTSD) symptoms following a trauma—an association that was
found not to be attributable to the many co-occurring risk factors.178 In her landmark 1998
paper, Carlson also documented associations in the Minnesota Longitudinal Study of Risk
and Adaptation between infant disorganised attachment and general mental health in early
adulthood, with a particularly strong association for dissociative symptoms.179 The Carlson
paper helped secure perceptions of disorganisation as, in Sroufe’s words, an ‘incredibly

powerful construct, being among other things the strongest single predictor of later psy-
chopathology’.180 Yet Groh, Fearon, van IJzendoorn, Bakermans-Kranenburg, and Roisman
have presented meta-analytic findings inconsistent with Carlson’s apparent emphasis on the
implications of infant disorganised attachment for general psychopathology, finding instead
stronger links with externalising problems specifically.181 The Minnesota findings are dis-
cussed further in Chapter 4.
Fear and causality
The Main and Solomon 1990 chapter laid out protocols for coding the classification. Each
of the seven Index headings was followed by a variety of concrete examples of kinds of be-
haviour seen by Main and Solomon in their review. Some of the examples are italicised,
which means that behaviour of this sort may, on its own, be sufficient for a D classification.
Other examples are not italicised, which means that only when other kinds of conflicted,
confused, or apprehensive behaviour is present should a D classification be considered.
The rationale was not made explicit by Main and Solomon for the inclusion of particular
examples, or for which examples are italicised, or how to weight observations that differ
from the examples. This has contributed to the bottleneck in reliable coders. A careful
175 Fearon, R.P., Bakermans-Kranenburg, M.J., Van IJzendoorn, M.H., Lapsley, A.M., & Roisman, G.I. (2010)
havior: a meta-analytic study. Child Development, 81(2), 435–56, p.450.
176 Groh, A.M., Fearon, R.P., Bakermans-Kranenburg, M.J., Van IJzendoorn, M.H., Steele, R.D., & Roisman, G.I.
(2014) The significance of attachment security for children’s social competence with peers: a meta-analytic study.
177 Groh, A.M., Roisman, G.I., van IJzendoorn, M.H., Bakermans-Kranenburg, M.J., & Fearon, R.P. (2012) The
significance of insecure and disorganized attachment for children’s internalizing symptoms: a meta-analytic study.
Child Development, 83(2), 591–610.
178 MacDonald, H.Z., Beeghly, M., Grant-Knight, W., et al. (2008) Longitudinal association between infant dis-
organized attachment and childhood posttraumatic stress symptoms. Development & Psychopathology, 20(2),
493–508.
Development, 69(4), 1107–28.

180 Sroufe, L.A. (2003) Attachment categories as reflections of multiple dimensions. Developmental Psychology,
39(3), 413–16, p.414.

181 Groh, A.M., Fearon, R.M., IJzendoorn, M.H., Bakermans- Kranenburg, M.J., & Roisman, G.I. (2017)
Attachment in the early life course: meta-analytic evidence for its role in socioemotional development. Child
Development Perspectives, 11(1), 70–76, p.73.
retrospective examination of the protocol, conducted together with Solomon, reveals that
behaviours are included when they are conflicted, confused, or apprehensive. And, again
in retrospective examination, italicisation was based on six factors: (i) frequency of a be-
haviour, (ii) its pervasiveness or duration, (iii) its abruptness in behavioural sequence, (iv)
the extent to which it occurs either close to reunion or in physical proximity with the care-
giver, (v) whether it can be better explained as a reaction to the immediate environment,
and (vi) the extent to which the infants’ responses to their caregiver suggest the experience
of apprehension.182
This latter item in the protocol reflects a theory that Main had been developing with her
husband and collaborator Erik Hesse. Hesse entered undergraduate study in psychology at
Berkeley as an adult, after studying to become a professional musician. Following comple-
tion of his studies in 1981, he worked as a research collaborator with Main, for instance as

one of the coders for the Kaplan drawing system (discussed in the section ‘The Kaplan and
Main family drawing system’). At the start of her relationship with Hesse, Main had been
thinking intensively about the role of avoidant behaviour in the Strange Situation. She con-
cluded that a rejecting or angry parent would distress an infant, activating the attachment
system and a desire for proximity, but simultaneously would evoke a tendency to withdraw
from the parent. Avoidant behaviour would circumvent the ensuing approach–avoidance
conflict.183
Hesse’s ingenious proposal was that one cause of the conflicted, confused, and appre-
hensive behaviours that Main and Solomon were seeing in the Strange Situation would be
experiences of alarming behaviour by the caregiver, the haven of safety sought by the attach-
ment system. This would not permit the redirection of attention that, for Main, provided the
infrastructure for a conditional strategy: ‘the infant may at times be experiencing a fear or
distress too intense to be deactivated through a shift in attention (the Ainsworth A pattern),
yet at least momentarily cannot be ameliorated through approach to the attachment figure
(the Ainsworth B and C patterns)’.184 This predicament could be produced if ‘the fear the in-
fant experiences stems from the parent as its source’, due to the injunction of the attachment
system to approach the caregiver when alarmed.185 Alarming behaviour by an attachment
figure would cause an injunction to both approach and take flight, blocking the redirection
of attention upon or away from the caregiver that underpins the two forms of infant con-
ditional attachment strategy. Main and Hesse put forward this idea in a 1990 chapter that
immediately followed the Main and Solomon protocols, establishing it essentially and fun-
damentally within reception of the D classification. In their writings of the period, and in
personal communication with colleagues, they proposed that ‘it is the interjection of fear
into the caregiving experience that is essential to developing a disorganised/disoriented
attachment’.186
182 Duschinsky, R. & Solomon, J. (2017) Infant disorganized attachment: clarifying levels of analysis. Clinical
Child Psychology and Psychiatry, 22(4), 524–38.

183 Main, M. & Stadtman, J. (1981) Infant response to rejection of physical contact by the mother. Journal of the
American Academy of Child Psychiatry, 20(2), 292–307.

attachment status. In M.T. Greenberg, D. Cicchetti, & E.M. Cummings (eds) Attachment in the Preschool Years
(pp.161–81) Chicago: University of Chicago Press, p.163.
185 Ibid. p.182.
186 Cicchetti, D. & White, J. (1988) Emotional development and the affective disorders. In W. Damon (ed.) Child
Development: Today and Tomorrow (pp.177–98). San Francisco: Jossey-Bass, citing a personal communication
from Main, p.185.
Supporting evidence for association of the caregiver with fear as one pathway to disorgan-
ised infant attachment was already available by the time of Main and Hesse’s 1990 chapter,
from Dante Cicchetti and his research group (Chapter 4). The Harvard Child Maltreatment
Project was initiated in 1979 by Cicchetti and Aber, as a large longitudinal study, modelled on
the Minnesota Longitudinal Study of Risk and Adaptation.187 A small subsample were seen
in the Strange Situation and classified using the Ainsworth categories.188 Twenty-two dyads
were known to child protection services for abuse and/or neglect of the child or an older
sibling; 21 were a matched comparison group. However, a third of the sample were classi-
fied as securely attached according to the Ainsworth coding protocols. There were a few dif-
ferent proposals available at the time for additions to Ainsworth’s system for use with clinical
samples. One was that of Ainsworth’s student Patricia Crittenden, who observed that some
abused infants show both avoidant and resistant attachment (A–C).189 However, the Main

and Solomon disorganised attachment classification appealed to Cicchetti and colleagues
as more encompassing; it was assumed that finer-grained distinctions would be worked out
later. The disorganised classification also had emerging evidence of predictive validity.190
The data were recoded for disorganisation by Cicchetti and Barnett; Barnett was blind to the
children’s maltreatment status; 82% of the infant–caregiver dyads known to child protection
services were classified as showing disorganised attachment and 14% classified as secure; in
the comparison group, 19% were classified as disorganised and 52% classified as secure. At
18 months, 61% of dyads known to child protection services were classified as showing dis-
organised attachment, compared to 29% in the comparison group.191
In the empirical developmental attachment tradition initiated by Ainsworth, it was widely
held that home observations were needed to validate new measures or the adaptation of ex-
isting measures. By the early 1990s, Ainsworth was expressing her dismay that ‘so many at-
tachment researchers have gone on to do research with the Strange Situation rather than
looking at what happens in the home . . . in part it has to do with the “publish or perish”
realities of academic life’.192 And Sroufe and Waters agreed that ‘the strange situation is used
only because it can stand in place of attachment as it would be observed in the home’.193 As
such, ‘any procedure claiming to assess attachment (even variations or new applications of
187 Another early, important study of a sample with substantial rates of maltreatment was conducted by Mary Jo
Ward. Ward was also a former student of Sroufe’s, graduating a few years after Cicchetti: Ward, M.J. & Carlson, E.A.
(1995) Associations among adult attachment representations, maternal sensitivity, and infant–mother attachment
in a sample of adolescent mothers. Child Development, 66(1), 69–79.
188 Schneider-Rosen, K., Braunwald, K.G., Carlson, V., & Cicchetti, D. (1985) Current perspectives in attach-
ment theory: illustration from the study of maltreated infants. Monographs of the Society for Research in Child
Development, 50(1–2), 194–210.
189 Crittenden, P.M. (1988) Relationships at risk. In J. Belsky & T. Nezworski (eds) Clinical Implications of
Attachment Theory (pp.136–74). Hillsdale, NJ: Lawrence Erlbaum. See also Radke-Yarrow, M., Cummings, E.M.,
Kuczynski, L., & Chapman, M. (1985) Patterns of attachment in two-and three-year-olds in normal families and
families with parental depression. Child Development, 56, 884–93.
190 Carlson, V., Cicchetti, D., Barnett, D., & Braunwald, K. (1989) Finding order in disorganization: lessons
from research on maltreated infants’ attachments to their caregivers. In D. Cicchetti & V. Carlson (eds) Child
191 Barnett, D., Ganiban, J., & Cicchetti, D. (1999) Maltreatment, negative expressivity, and the development
of type D attachments from 12 to 24 months of age. Monographs of the Society for Research in Child Development,
64(3), 97–118.
Ainsworth. Monographs of the Society for Research in Child Development, 60(2–3), 2–21, p.12.
52(4), 743–6, p.743.

the Ainsworth procedure) must be anchored to observations of the attachment–exploration

balance in the natural environment. That is the crucial criterion.’194 Though the Strange
Situation is often described as the ‘gold standard’, Sroufe and Waters characterised detailed
naturalistic observations as the true ‘standard’ for assessing attachment. Cicchetti and col-
leagues revealed one kind of experience in the home environment that could substantially
increase the likelihood of disorganised attachment. Yet Main and colleagues regarded it as
implausible that in the low-risk Bay Area sample, all infants who received a disorganised at-
tachment classification had received maltreatment from their caregiver. And, as mentioned
in the section ‘A new category’, Main and Solomon found that apprehension of the caregiver
and disorientation were more common in the tapes they reviewed from maltreatment or
very high-risk samples, whereas these behaviours were uncommon in the infants classified
as disorganised in the Bay Area sample.

Conversations between Main and Hesse led to the conclusion that there could be other
forms of alarming caregiver behaviour besides abuse. Main’s written notes from home ob-
servations during her doctorate offered too little detail to be the basis for a published study,
and her group had not collected new observations of the child’s behaviour at home for the
Berkeley study. So she was not in a position to examine how disorganised attachment in the
Strange Situation related to a child’s behaviour at home in samples not specifically known
to child protection services. Ainsworth regarded a home observation study as ‘highly desir-
able’ for the validation of the classification.195 Likewise, van IJzendoorn alleged that ‘since
the D category has not been widely validated against home behavior, its meaning is not yet
fully clear’.196 To answer such concerns, Main, though she had no home observation record-
ings, did have observations of infant free-play sessions in the laboratory with both the child’s
mother and father. If alarming behaviour in playful and other more ordinary interactions
was shown predominantly by parents in dyads classified as disorganised in the Strange
Situation, then this would offer one form of ecological validation for the classification, and at
the same time help shed light on its cause or causes.
194 Sroufe, L.A. (1996) Emotional Development. Cambridge: Cambridge University Press, p.181.
195 Ainsworth, M. & Eichberg, C. (1988) Effects on Infant–Mother Attachment of Mother’s Experience Related to
Loss of an Attachment Figure or Other Traumatic Experience. Unpublished manuscript. PP/Bow/J.1/62: ‘It would
be highly desirable for future research to undertake intensive observation of parent–infant interaction in the nat-
ural environment of the home in the case of infants classified as D’ (40). Incredibly, in the three decades since the
introduction of the classification, this ‘crucial criterion’ has never been systematically pursued; the only work on
the question was the accidental observation of disorganised attachment behaviour at home in a few infants during
Carlo Schuengel’s doctoral study: Schuengel, C., van IJzendoorn M.H., Bakermans-kranenburg, M.J., & Blom,
M. (1998) Frightening maternal behaviour, unresolved loss, and disorganized infant attachment: a pilot-study.
Journal of Reproductive and Infant Psychology, 16(4), 277–83: ‘Unexpectedly, as it was not an aim of our study,
we observed disorganized attachment behaviour in two infants during the home observations. This behaviour
qualified, if it had been observed in the context of the Strange Situation, for a D-classification. Surprisingly, the
two infants who were disorganized at home were not disorganized in the Strange Situation’ (282). Another study
later examined the behaviour of dyads classified as disorganised, not at home, but in response to an injection at
the doctors’, yielding weak but interesting findings. Wolff, N.J., Darlington, A.S.E., Hunfeld, J.A., et al. (2011) The
influence of attachment and temperament on venipuncture distress in 14-month-old infants: the generation R
study. Infant Behavior and Development, 34(2), 293–302: ‘The current study showed that there was a trend for at-
tachment disorganization to predict higher levels of venepuncture distress in 14-month-old infants. Furthermore,
the interaction of disorganized attachment and fearful temperament was significantly associated with distress; fear
predicted an increase in distress only in infants with a disorganized attachment classification’ (299).
196 Van IJzendoorn, M.H., Goldberg, S., Kroonenberg, P.M., & Frenkel, O.J. (1992) The relative effects of ma-
ternal and child problems on the quality of attachment: a meta-analysis of attachment in clinical samples. Child
Development, 63(4), 840–58, p.854.
From the early 1990s, Hesse undertook doctoral research at Leiden University, supervised
by van IJzendoorn. As part of this research, Hesse and Main began to formulate a coding
system for alarming caregiver behaviours, based initially on theory and then on review with
their graduate students of 13 observations from the Berkeley sample of infant–mother or
infant–father interaction during unstructured play from before the Clown session.197 Many
of the potentially alarming behaviours they observed were fleeting; the coding system devel-
oped by Hesse and Main to identify and capture these moments is penetrating, subtle, and
extremely difficult to learn (even on the rare occasions on which training has been offered).
It was dubbed the ‘FR coding system’, named after two central components: frightening and
frightened caregiver behaviours. Examples of frightening behaviour provided by the coding
manual were directly threatening or predatory behaviours, such as bared teeth, or lunging or
looming at the infant without markers of play or prior warning. Merely angry behaviours by

the caregiver, or ordinary forms of ‘culturally sanctioned’ discipline, were not included.198
Frightened behaviours shown by the caregiver towards the infant were also identified, such
as sudden withdrawal from the baby or a strong startle response in response to the infant’s
approach. Such displays were anticipated to alarm the infant when they sought the care-
giver as a safe haven. A third kind of behaviour identified by Hesse and Main was apparently
dissociative behaviours by the caregiver, which were assumed to potentially frighten an in-
fant through the caregiver’s inexplicable attentional unavailability.199 The role of dissociative
caregiver behaviours was proposed to Main and Hesse by their friend Giovanni Liotti, whom
they had met at Bowlby’s 80th birthday party in 1987 and then visited in Rome in 1990.200
In the course of the 1990s, Hesse and Main continued to scope potentially alarming care-
giving behaviours through review of 62 further Berkeley recordings of caregiver–infant
interaction and also recordings drawn from four diverse samples where Strange Situation
data were also available. One was Karlen Lyons-Ruth’s sample of families known to social
services in Boston.201 Another was a doctoral project by their student Mary True, who had
recorded mother–infant interactions among the Dogon of Uganda.202 A third sample was
collected by Debby Jacobvitz in Texas; though a community sample, rates of alarming care-
giver behaviours and of disorganised attachment were unusually high.203 And a fourth
197 This was a pilot study for part of Kelly Abrams’ doctoral dissertation research: Abrams, K.Y. (2000) Pathways
to disorganization: a study concerning varying types of parental frightened and frightening behaviors as related to
infant disorganized attachment. Unpublished doctoral dissertation, University of California at Berkeley.
198 Curiously, the scale did explicitly not include angry verbal threats by the caregiver to a child. This might have
been simply an oversight, rather than a theoretically motived decision, given that verbal threats were included by
Main in the traumatic abuse scale for the AAI composed around the same time. Additionally it is likely that, in
practice, coders would code verbal threats as ‘threatening’ in any case.
199 Hesse, E. & Main, M. (1999) Second-generation effects of unresolved trauma in nonmaltreating parents: dis-
sociated, frightened, and threatening parental behavior. Psychoanalytic Inquiry, 19(4), 481–540.
200 Main, M. & Hesse, E. (1992) Attaccamento disorganizzato/ disorientato nell’infanzia e stati mentali dis-
sociati dei genitori. In M. Ammaniti & D. Stern (1992) Attaccamento e Psicoanalisi (pp.80–140). Rome: Gius,
Laterza & Figli; Liotti, G. (1992) Disorganized/disoriented attachment in the etiology of the dissociative disorders.
Dissociation: Progress in the Dissociative Disorders, 5, 196–204.
201 Lyons-Ruth, K., Bronfman, E., & Parsons, E. (1999) Chapter IV. Maternal frightened, frightening, or
atypical behavior and disorganized infant attachment patterns. Monographs of the Society for Research in Child
202 True, M., Pisani, L., & Oumar, F. (2001) Infant– mother attachment among the Dogon of Mali. Child
Development, 72(5), 1451–66.
203 Jacobvitz, D.B., Hazen, N.L., & Riggs, S. (1997) Disorganized mental processes in mothers, frightening/
frightened caregiving, and disoriented, disorganized behavior in infancy. Paper presented at Biennial meeting of
Society for Research in Child Development, Washington, DC; Jacobvitz, D., Leon, K., & Hazen, N. (2006) Does
expectant mothers’ unresolved trauma predict frightened/frightening maternal behavior? Risk and protective fac-
tors. Development & Psychopathology, 18(2), 363–79.
sample of bereaved mothers was collected by Carlo Schuengel at Leiden for his doctoral re-
search.204 Both study of the Berkeley sample205 and Schuengel’s Leiden sample206 revealed
that in fact dissociative caregiver behaviours were the most predictive of disorganised at-
tachment, more so even than directly frightening caregiver behaviours, though the two were
intercorrelated.207
By 1995, Hesse and Main’s work on the coding system had led to the addition of three fur-
ther kinds of behaviour.208 These were not in themselves assumed to be directly alarming, but
were held to index the caregiver’s potential for one of the three primary forms of alarming
behaviour. These were: (i) behaving in a timid or deferential way toward the child;209 (ii)
sexualised behaviours toward the infant, suggesting confusion between the caregiving and
sexual behavioural systems as well as a lack of capacity to monitor action; and (iii) behav-
iours by the caregiver that are coded in the Main and Solomon indices for infant disorgan-

isation, such as misdirected behaviours or briefly moving in a stiff, asymmetrical manner.
These latter behaviours were assumed to represent some significant disruption at the level
of the caregiving system and therefore serve as distal markers of alarming caregiving. In the
Berkeley sample, Hesse and colleagues found that these additional forms of FR caregiving
behaviour were not associated with infant disorganised attachment; however, they suspected
that this may have been a function of their low-risk cohort.210
204 Schuengel, C., Bakermans- Kranenburg, M.J., & Van IJzendoorn, M.H. (1999) Frightening maternal be-
havior linking unresolved loss and disorganized infant attachment. Journal of Consulting and Clinical Psychology,
67(1), 54–63.
205 Abrams, K.Y. (2000) Pathways to disorganization: a study concerning varying types of parental fright-
ened and frightening behaviors as related to infant disorganized attachment. Unpublished doctoral dissertation,
University of California at Berkeley.
206 This finding is discussed in Out, D., Bakermans-Kranenburg, M.J., & van IJzendoorn, M.H. (2009) The role
of disconnected and extremely insensitive parenting in the development of disorganized attachment: validation of
a new measure. Attachment & Human Development, 11(5), 419–43, p.422.
207 Neither study reported whether the range in the dissociation and frightening scales was equivalent. It was
therefore unclear whether dissociation was more predictive because it was more frequently coded, or because it
was indexing the more important cause. Anne Rifkin-Graboi, personal communication, July 2019.
208 Hesse, E., Main, M., Abrams, K.Y., & Rifkin, A. (2003) Unresolved states regarding loss or abuse can have
‘second-generation’ effects: disorganized, role-inversion and frightening ideation in the offspring of traumatized
non-maltreating parents. In D.J. Siegel & M.F. Solomon (eds) Healing Trauma: Attachment, Mind, Body and Brain
(pp.57–106). New York: Norton.
209 Hesse and colleagues identified that this behaviour was rare in the low-risk Berkeley sample and more fre-
quently observed in high-risk samples: Abrams, K.Y., Rifkin, A., & Hesse, E. (2006) Examining the role of parental
frightened/frightening subtypes in predicting disorganized attachment within a brief observational procedure.
Development & Psychopathology, 18(2), 345–61, p.357. See also Lyons-Ruth, K., Bureau, J.F., Easterbrooks, M.A.,
Obsuth, I., Hennighausen, K., & Vulliez-Coady, L. (2013) Parsing the construct of maternal insensitivity: distinct
longitudinal pathways associated with early maternal withdrawal. Attachment & Human Development, 15(5–6),
562–82.
210 Abrams, K.Y., Rifkin, A., & Hesse, E. (2006) Examining the role of parental frightened/frightening subtypes
in predicting disorganized attachment within a brief observational procedure. Development & Psychopathology,
18(2), 345–61. An expanded version of the FR coding system was developed by Lyons-Ruth, Bronfman, and
Parsons (the Atypical Maternal Behavior Instrument for Assessment and Classification (AMBIANCE)). This
assesses five dimensions of disrupted parental affective communication: negative-intrusive behaviour; role con-
fusion; disorientation; affective communication errors; and withdrawal from the child. AMBIANCE is a more
encompassing measure, with equivalent prediction of disorganised infant attachment to the FR system. Therefore,
on the grounds of parsimony, some attachment researchers have expressed a preference for FR, e.g. Out, D.,
Bakermans-Kranenburg, M.J., & Van IJzendoorn, M.H. (2009) The role of disconnected and extremely insensitive
parenting in the development of disorganized attachment: validation of a new measure. Attachment & Human
Development, 11(5), 419–43. However, training is available annually in AMBIANCE, unlike the FR system. Efforts
to economise AMBIANCE measure for use in routine clinical practice may also contribute to the popularity of the
measure in the future, e.g. Haltigan, J.D., Madigan, S., Bronfman, E., et al. (2019) Refining the assessment of dis-
rupted maternal communication: using item response models to identify central indicators of disrupted behavior.
Though only a tiny number of researchers have been trained by Main and Hesse to use it,
the FR coding system itself has been widely praised. For instance, Rothbaum and Morelli,
major critics of Ainsworth’s work, praised the FR system as less prone to unacknowledged
ethnocentric bias than Ainsworth’s sensitivity scale.211 Furthermore, empirical evidence
has accumulated in support of the predicted association between infant disorganised at-
tachment in the Strange Situation and caregiver behaviours coded with either the initial
or the expanded version of the FR coding system. Madigan, Bakermans-Kranenburg, van
IJzendoorn, and Moran reported from a meta-analysis that FR caregiver behaviour during
observations accounts for 13% of the variance in attachment disorganisation.212 Cassidy has
criticised Main and Hesse for overclaiming the importance of the frightening/frightened
pathway to disorganised attachment, since in fact ‘the relation between frightening/fright-
ened behaviour and infant disorganisation has been found to be relatively weak’.213 There

is certainly much unexplained variance, suggesting the role of factors besides frightening,
frightened, or dissociative caregiving. Nonetheless, these are clearly causally important: the
meta-analysis by Madigan and colleagues reveals that parents who displayed these behav-
iours were 3.7 times more likely to be part of dyads coded for disorganised attachment in the
Strange Situation (r = .34, N = 851).214
An expanded version of the FR coding system was developed by Lyons-Ruth, Bronfman,
and Parsons (the Atypical Maternal Behavior Instrument for Assessment and Classification
(AMBIANCE)). This assesses five dimensions of disrupted parental affective communica-
tion: negative-intrusive behaviour; role confusion; disorientation; affective communication
errors; and withdrawal from the child. AMBIANCE is a more encompassing measure, with
equivalent prediction of disorganised infant attachment to the FR system.215 Therefore, on
the grounds of parsimony, some attachment researchers have expressed a preference for
FR.216 However, training is available annually in AMBIANCE, unlike the FR system. Efforts
to economise AMBIANCE measure for use in routine clinical practice may contribute to the
further popularity of the measure in the future.217
An important meta-analytic finding by van IJzendoorn and colleagues was that disor-
ganised infant attachment was only weakly predicted by caregiver insensitive behaviours
in general. Van IJzendoorn and colleagues concluded that the FR pathway was therefore
211 Rothbaum, F. & Morelli, G. (2005) Attachment and culture: bridging relativism and universalism. In W.
Friedlmeier, P. Chakkarath, & B. Schwarz (eds) Culture and Human Development: The Importance of Cross-Cultural
Research to the Social Sciences. Lisse: Swets & Zeitlinger, p.110.
212 Madigan, S., Bakermans-Kranenburg, M.J., Van Ijzendoorn, M.H., Moran, G., Pederson, D.R., & Benoit, D.
(2006) Unresolved states of mind, anomalous parental behavior, and disorganized attachment: a review and meta-
analysis of a transmission gap. Attachment & Human Development, 8(2), 89–111, p.102.
213 Cassidy, J. & Mohr, J.J. (2001) Unsolvable fear, trauma, and psychopathology: theory, research, and clinical
considerations related to disorganized attachment across the life span. Clinical Psychology: Science and Practice,
8(3), 275–98, p.283.
214 Madigan, S., Bakermans-Kranenburg, M.J., Van IJzendoorn, M.H., Moran, G., Pederson, D.R., & Benoit, D.
analysis of a transmission gap. Attachment & Human Development, 8(2), 89–111.
215 Bronfman, E., Madigan, S., & Lyons- Ruth, K. (2009–2014) Disrupted Maternal Behavior Instrument for
Assessment and Classification (AMBIANCE): Manual for Coding Disrupted Affective Communication, 2nd edn.
Unpublished manuscript, Harvard University Medical School.
216 E.g. Out, D., Bakermans-Kranenburg, M.J., & Van IJzendoorn, M.H. (2009) The role of disconnected and
extremely insensitive parenting in the development of disorganized attachment: validation of a new measure.
217 Haltigan, J.D., Madigan, S., Bronfman, E., et al. (2019) Refining the assessment of disrupted maternal com-
munication: using item response models to identify central indicators of disrupted behavior. Development &
Psychopathology, 31(1), 261–77.
distinct from insensitivity.218 Subsequently, however, Bailey and colleagues as well as other
researchers provided evidence suggesting that the FR pathway and insensitivity overlap
in high-risk families.219 Proposals have also been put forward regarding the role of spe-
cific forms of insensitive care in disorganised attachment, including by Lyons-Ruth. Using
AMBIANCE, Lyons-Ruth and collaborators identified a specific association between care-
giver withdrawal from the child and disorganised attachment.220 Another pathway to dis-
organised attachment has been suggested by researchers such as Spangler, who have argued
that infant temperament or genetic factors may predispose at least some forms of disorgan-
ised attachment.221 Main and Hesse have by and large been sceptical of such temperamental
or genetic explanations and have pointed to the fact that classifications of disorganised at-
tachment with different caregivers have little association, and that studies of temperamental
or genetic antecedents have had a poor record of replication.222 However, they have been

intrigued by findings by van IJzendoorn and colleagues that suggest gene × environment
interactions in the prediction of disorganised attachment.223
Yet, just as the ordinary language connotations of ‘disorganisation’ misled readers of Main
and Solomon, so did the evocative connotations of the term ‘fear’ magnetise readers of Main
and Hesse’s chapter. Close examination of the Main and Hesse 1990 text from the vantage of
the present suggests that the term ‘fear’ was used in a way that was insufficiently specified. The
1990 chapter fell subject to a danger already identified by Bowlby: ‘unfortunately in collo-
quial English the word “fear” is used in many senses, often being synonymous with expectant
anxiety and sometimes with fright’.224 Main and Hesse meant to convey that one pathway to
disorganised attachment may be when children come to associate their caregiver, for what-
ever reason, with feelings of alarm.225 However, as a result of the way they used the term ‘fear’,
it has been widely assumed by readers that Main and Hesse were proposing that disorganised
225–50; Out, D., Bakermans-Kranenburg, M.J., & Van IJzendoorn, M.H. (2009) The role of disconnected and
extremely insensitive parenting in the development of disorganized attachment: validation of a new measure.
219 Bailey, H.N., Tarabulsy, G.M., Moran, G., Pederson, D.R., & Bento, S. (2017) New insight on intergenera-
tional attachment from a relationship-based analysis. Development & Psychopathology, 29(2), 433–48. See also
Gedaly, L.R. & Leerkes, E.M. (2016) The role of sociodemographic risk and maternal behavior in the prediction of
infant attachment disorganization. Attachment & Human Development, 18(6), 554–69.
220 Lyons-Ruth, K., Bronfman, E., & Parsons, E. (1999) Maternal frightened, frightening or disrupted be-
havior and disorganized infant attachment patterns. Monographs of the Society for Research in Child Development,
64(3), 67–96.
221 Spangler, G., Johann, M., Ronai, Z., & Zimmermann, P. (2009) Genetic and environmental influence on at-
tachment disorganization. Journal of Child Psychology and Psychiatry, 50(8), 952–61.

222 Main’s position was not, however, a full rejection of temperament-based explanations, e.g. ‘I am inclined,
however, to believe that researchers should still hold out the possibility of a small heritable component in disorgan-
isation (e.g. overall fearfulness). This is an empirical question . . . In my view, genetic differences may “get their in-
nings” with respect to attachment organisation late rather than early in life.’ Main, M. (1999) Epilogue. Attachment
theory: eighteen points with suggestions for future studies. In J. Cassidy & P. Shaver (eds) Handbook of Attachment
(pp.845–87). New York: Guilford, p.864–5. See also Groh, A.M., Narayan, A.J., Bakermans-Kranenburg, M.J., et al.
(2017) Attachment and temperament in the early life course: a meta-analytic review. Child Development, 88(3),
770–95.
223 Main, M., Hesse, E., & Hesse, S. (2011) Attachment theory and research: overview with suggested applica-
tions to child custody. Family Court Review, 49(3), 426–63, p.443; Bakermans-Kranenburg, M.J. & Van IJzendoorn,
M.H. (2007) Research review: genetic vulnerability or differential susceptibility in child development: the case of
attachment. Journal of Child Psychology and Psychiatry, 48(12), 1160–73.
225 Duschinsky, R. (2018) Disorganization, fear and attachment: working towards clarification. Infant Mental
Health Journal, 39(1), 17–29.

attachment represents a child afraid of their caregiver. It has been common to encounter
statements even by authorities such as van IJzendoorn and Bakermans-Kranenburg referen-
cing Main and Hesse as saying that, for instance, ‘the essence of disorganized attachment is
that the child is at times scared of the attachment figure’.226
To take further examples of the misapprehension of Main and Hesse’s position, Paetzold
and colleagues—the researchers in the social psychological tradition who have given most
notice to disorganised attachment—stated that, according to Main and Hesse, ‘infants in the
disorganized category develop a fear of their attachment figures because these figures dis-
play frightening behaviors in their daily interactions with their children’.227 Sinason wrote
that ‘disorganized attachment refers to grossly disorganized behavior on the part of the in-
fant or child: apprehension in the presence of the mother or primary caretaker’.228 And Rees
wrote to fellow paediatricians that ‘disorganized patterns arise if pervasive abuse leaves chil-

dren ineffective both in self-sufficiency and in using relationships, lacking understanding
of their own and others’ feelings. Safe independence is unlikely and criminality in adult-
hood common without recovery.’229 Reification of the fear–disorganisation relationship
also appeared in proposals for National Institute for Health and Care Excellence guidelines
on child attachment for clinicians in the UK.230 However, following feedback from Main,
Hesse, Lyons-Ruth, and others,231 these guidelines were amended in the final version to
avoid implying that frightening caregiver behaviour is the primary route to disorganised
attachment.232
One contributing factor to misapprehensions regarding the role of fear in disorganized
attachment may have been the illustrations in Main and Solomon’s 1990 chapter.233 This
comes into relief if the illustrations in the original manuscript and the published chapter
are compared. In both versions, pen drawings were presented of infants showing behaviour
listed in the coding protocols. The drawings were tracings of the film negatives of video
recordings of infants in the Strange Situation. In the original manuscript, the illustrations
are of a variety of the behaviours. In the published version, the only set of illustrations is
of explicit apprehension seen on reunion with the caregiver. There is a drawing of a child
covering her mouth on reunion with her caregiver, which might also suggest confusion,
but the drawing and Main and Solomon’s commentary suggest that the behaviour should
226 Van Rosmalen, L., van IJzendoorn, M.H., & Bakermans- Kranenburg, M.(2014) ABC+D of attach-
ment theory. In P. Holmes & S. Farnfield (eds) Routledge Handbook of Attachment: Theory (pp.11–30).
London: Routledge, p.21. See also Juffer, F., Struis, E., Werner, C., & Bakermans-Kranenburg, M.J. (2017) Effective
preventive interventions to support parents of young children: illustrations from the Video-feedback Intervention
to promote Positive Parenting and Sensitive Discipline (VIPP-SD). Journal of Prevention & Intervention in the
Community, 45(3), 202–14: ‘Disorganized attachment (Main & Solomon, 1990) is characterized by fear of the
parent’ (203).
227 Paetzold, R.L., Rholes, W.S., & Kohn, J.L. (2015) Disorganized attachment in adulthood: theory, measure-
ment, and implications for romantic relationships. Review of General Psychology, 19(2), 146–56, p.147.
228 Sinason, V. (2016) The seeming absence of children with DID. In E. Howell & S. Itzkowitz (eds) The
Dissociative Mind in Psychoanalysis (pp.221–8). London: Routledge, p.223.

229 Rees, C. (2011) Children’s attachments. Paediatrics and Child Health, 22(5), 186–92, p.187.
230 National Institute for Health and Care Excellence (2015) First draft of Children’s Attachment: Attachment in
Children and Young People Who Are Adopted from Care, in Care or at High Risk of Going into Care. London: NICE.
https://www.nice.org.uk/guidance/ng26/documents/childrens-attachment-full-guideline2.
231 https://www.nice.org.uk/guidance/ng26/documents/consultation-comments-and-responses.
232 NICE (2016) Children’s Attachment: Attachment in Children and Young People Who Are Adopted from Care,
in Care or at High Risk of Going into Care. London: NICE. https://www.ncbi.nlm.nih.gov/pubmed/26741018.

be interpreted as indicating apprehension. In general, the facial expressions in the draw-

ings convey terror and a crumpled misery. Though the drawings selected for inclusion in
the published version are somewhat higher-quality drawings (less blotchy, more human-
looking), their selection was also partly shaped by theory. They align much better with the
Main and Hesse hypothesis that fear is implicated in at least some forms of disorganised
attachment. Illustrations of visible fear of the caregiver served as a powerful encapsula-
tion and expression of the process theorised to be underlying the diversity of disorganised
behaviours. However, a disadvantage of privileging of illustrations of fear was that it was
unrepresentative. There are a variety of behaviours used to classify disorganised attach-
ment, including conflicted and confused behaviours without apparent apprehension of the
caregiver.234 And, in fact, visible apprehension of the caregiver in the manner of the draw-
ings is very uncommon in community samples, and relatively rare even in samples facing

significant adversities.235 Incidence of clear apprehension of the caregiver in maltreatment
samples has not been reported but is an outstanding question of the utmost significance,
not least in terms of understanding the potential for measurement variance across samples
among dyads who receive a disorganised attachment classification. As Padrón, Carlson,
and Sroufe indicated, if one group of children show predominantly outright apprehension
of the caregiver and another group predominantly non-frightened behaviours in the Main
and Solomon indices, both might receive a disorganised attachment classification but have
very different correlates.236
The choice of illustrations by Main and Solomon may well have had particular importance
for the imagination of the disorganised infant by the readers of their chapter.237 The image of
fear on reunion is central, for example, to explanations of the category by researchers in pa-
pers,238 discussion of disorganisation in textbooks for psychology students,239 and guidance
provided for social workers and other child-safeguarding practitioners in using disorgan-
ised attachment as an indicator of child maltreatment.240 Waters criticised Main and Hesse
for conveying an impression of disorganised attachment behaviours as always caused by
234 Additionally, Solomon reported that the process of tracing the film negative lent the resulting drawings an
overexpressed quality, which in her view conveyed even more of a sense of terror and misery than the films them-
selves. Personal communication, November 2016.
235 E.g. adolescent parents: Forbes, L.M., Cox, A., Moran, G., & Pederson, D. (2006) Exploring expressions of
disorganization in the Strange Situation in a high-risk sample. Poster presented at the World Association for Infant
Mental Health, Paris, July 2006. https://ir.lib.uwo.ca/psychologypres/6/.
236 Padrón, E., Carlson, E., & Sroufe, A. (2014) Frightened versus not frightened disorganized infant attach-
ment: newborn characteristics and maternal caregiving. American Journal of Orthopsychiatry, 84(2), 201–208.
237 The particular influence of visual representation for the interpretation of both new classificatory systems and
the emotional state of others is well documented by sociologists of science: Coopmans, C., Vertesi, J., Lynch, M.E.,
and Woolgar, S. (eds) (2014) Representation in Scientific Practice Revisited. Cambridge, MA: MIT Press.
238 E.g. Warren, S.L., Gunnar, M.R., Kagan, J., et al. (2003) Maternal panic disorder: infant temperament, neuro-
physiology, and parenting behaviors. Journal of the American Academy of Child & Adolescent Psychiatry, 42(7),
814–25: ‘Infants who showed unusual behaviors such as fear of the mother, freezing, hitting, or running from the
mother were classified as disorganized (group D)’ (818). The selection of behaviours is heavily tilted towards those
suggestive of fear, downplaying those suggestive of dissociation/confusion or conflict about approaching the care-
giver but without obvious fear.
239 Parke, R.D. & Clarke-Stewart, A. (2011) Social Development. New York: Wiley.
240 Shemmings, D. & Shemmings, D. (2011) Indicators of disorganised attachment in children. http://www.
communitycare.co.uk/2011/01/21/indicators-of-disorganised-attachment-in-children/; Wilkins, D. (2012)

Disorganised attachment indicates child maltreatment: how is this link useful for child protection social workers?
Journal of Social Work Practice, 26(1), 15–30. For critical appraisal of these statements see Granqvist, P., Sroufe,
L.A., Dozier, M., et al. (2017) Disorganized attachment in infancy: a review of the phenomenon and its implica-
tions for clinicians and policy-makers. Attachment & Human Development, 19(6), 534–58.
fear, and all in the same way, blocking attention to the potential diversity of aetiological fac-
tors.241 Lieberman criticised Main and Hesse on related grounds: for lack of clarity regarding
whether disorganised attachment is essentially an expression of PTSD in children or an in-
dependent (or overlapping) construct.242 And Bakermans-Kranenburg and van IJzendoorn
drew attention to the potential ‘heterogeneity of the mechanisms leading to disorganized at-
tachment’.243 In a 2006 article, Hesse and Main stated their wish that they had made it clearer
that they intended their emphasis on frightened or frightening caregiver behaviour as ‘one
highly specific and sufficient, but not necessary, pathway to D attachment status’.244 In this
paper, Hesse and Main acknowledged that how their earlier work was framed and argued ap-
pears to have misled readers.245
Six-year systems

A move to the level of representation
When participants in the Berkeley study were six years old, 45 families were invited back to
the laboratory, stratified by infant attachment classification. Whilst fathers and mothers went
to be interviewed, the children were asked to do a drawing of their family. They were then
asked to consider six pictured parent–child separations and offer their thoughts on what
the pictured child might feel and do. They played in a sandbox for a quarter of an hour, be-
fore their parents returned—counterbalanced so that some fathers returned first and some
mothers. In the early 1980s, the Berkeley group developed coding systems for these observa-
tional assessments. A coding system for the family drawings was developed by Main and her
doctoral student Nancy Kaplan.246 The response to the pictures of separation was analysed
with an adaptation of a coding system developed by Klagsbrun and Bowlby.247 And a coding
system was developed for the reunion episodes by Cassidy and Main.248 A coding system
241 Waters, E. & Valenzuela, M. (1999) Explaining disorganized attachment: clues from research on mild-to-
moderately undernourished children in Chile. In J. Solomon & C. George (eds) Attachment Disorganization
242 Lieberman, A. & Amaya-Jackson, L. (2005) Reciprocal influences of attachment and trauma. In L. Berlin, Y.
Ziv, L. Amaya-Jackson, & M. Greenberg (eds) Enhancing Early Attachments (pp.100–126). New York: Guilford.
There seems to be extreme diversity of opinion between researchers today on this issue.
243 Bakermans-Kranenburg, M.J. & van IJzendoorn, M.H. (2007) Research review: genetic vulnerability or dif-
ferential susceptibility in child development: the case of attachment. Journal of Child Psychology and Psychiatry,
48(12), 1160–73, p.1164. See also Granqvist, P. (2020) Attachment, Religion, and Spirituality: A Wider View.
New York: Guilford: ‘while “fright without solution” sounds ominous, I have not yet seen persuasive evidence that
the lion’s share of D behaviors do in fact reflect fear of the caregiver. Granted, behaviors fitting to index 6 (direct in-
dices of apprehension) yield converging evidence, but those behaviors tend to be quite rare in normal populations
of infants. Also, for behaviors fitting the remaining six indices, a central role of fear remains hypothetical.’
244 Hesse, E. & Main, M. (2006) Frightened, threatening, and dissociative parental behavior. Development &
Psychopathology, 18(2), 309–343, pp.310–11.

245 For further discussion of Main and Hesse’s position on the fear–disorganisation relationship see Duschinsky,
R. (2018) Disorganization, fear and attachment: working towards clarification. Infant Mental Health Journal,
39(1), 17–29.
246 Kaplan, N. (1987) Individual differences in six-year-olds’ thoughts about separation: predicted from attach-
ment to mother at age one. Doctoral dissertation, University of California at Berkeley.

247 Klagsbrun, M. & Bowlby, J. (1976) Responses to separation from parents: a clinical test for children. British
Journal of Projective Psychology, 21, 7–21.

248 Though Cassidy was Ainsworth’s student at Johns Hopkins, she completed her dissertation on Main’s data,
given the close links between the two research groups. Main, M. & Cassidy, J. (1988) Categories of response to
Six-year systems 259
specifically for the verbal interaction between child and caregiver on reunion was also devel-
oped by Amy Strage, a research associate in the group between 1983 and 1985.249
Findings from these analyses and those with the AAI were reported at a dedicated sym-
posium in 1985 at the Society for Research in Child Development held in Toronto, and a
summary published as the paper ‘Security in infancy, childhood, and adulthood: a move
to the level of representation’ in a monograph for the Society, edited by Everett Waters and
Inge Bretherton. New classifications for coding a mother’s autobiographical discourse in the
AAI had a strong correlation (r = .62) with the classification of that parent–child dyad in
the Strange Situation five years earlier, and a marked correlation also for fathers (r = .37).250
The prediction from infant attachment to responses at age six was an astonishing 68% to
85% in the case of the mother–child dyads. In the case of the father–child dyads, prediction
from the Strange Situation to reunion behaviour was equivalent. But prediction was weak

or non-significant for the other tasks. Main and colleagues supposed that, in contrast to a
direct reunion with the parent, these latter measures were less effective at priming mental
processes specific to the relationship with a particular parent. Their interpretation was that
since almost all of the children in the subsample had their mother as their most familiar and
primary caregiver, undifferentiated attachment primes such as a family drawing made chil-
dren think about their mother rather than, or more than, for father.
In the early 1980s, even close collaborators and friends of Bowlby acknowledged that
an inherent limitation of attachment research was that there was no way of measuring at-
tachment after infancy.251 Its relevance to other questions and concerns, such as therapeutic
work with adults, seemed limited. Main and colleagues appeared to conclusively demon-
strate that this was not so. Certainly the attachment behavioural system would change very
dramatically with maturation. However, having conceptualised attentional processes as the
underpinning architecture of individual differences in the Strange Situation, Main was in a
position to extrapolate upwards and examine how attention was directed in relation to at-
tachment figures in other contexts. For her, the Strange Situation procedure, children’s draw-
ings and discourse, and adult autobiographical discourse all had in common not any specific
behaviour but a question about the direction of attention in relation to attachment figures. In
‘A move to the level of representation’, Main stated that across the different measures:
In each of the insecure patterns of attachment, behavior and attention seem constricted
in readily identifiable ways. Throughout the Strange Situation, for example, the insecure-
avoidant infant attends to the environment and its features while actively directing atten-
tion away from the parent. The insecure-ambivalent infant, in contrast, seems unable to
direct attention to the environment, expresses strong and sometimes continual fear and
distress, and seems constantly directed toward the parent.252
reunion with the parent at age 6: predictable from infant attachment classifications and stable over a 1-month
period. Developmental Psychology, 24(3), 415–26.
249 Strage, A. & Main, M. (1985) Attachment and parent–child discourse patterns. Paper presented at Biennial
meeting of Society for Research in Child Development, Toronto. PP/Bow/J.4/4.

251 E.g. Heard, D.H. (1981) The relevance of attachment theory to child psychiatric practice. Journal of Child
Psychology & Psychiatry 22, 89–96.

252 Main, M., Kaplan, N, & Cassidy, J. (1985) Security in infancy, childhood, and adulthood: a move to the level
To situate and legitimate her ideas within the tradition of Bowlby’s theory, on the first page
of the ‘A move to the level of representation’ paper, Main described the individual differences
characterised by her measures as based on differences in ‘internal working models’.253 She
defined these, however, in a technical way to mean biases and filters in the processing of
attachment-relevant information: ‘We define the internal working model of attachment as
a set of conscious and/or unconscious rules for the organization of information relevant to
attachment and for obtaining or limiting access to that information, that is, to information
regarding attachment-related experiences, feelings, and ideation’.’254 In terms of positioning
within the academic field, claiming that the AAI assessed Bowlby’s classic concept of ‘in-
ternal working models’ had clear symbolic advantages for gaining recognition and accept-
ance of Main’s work.255 However, the ‘internal working model’ was a problematic concept.
Main did not mark for the reader that her technical definition departed from the way that the

term had been used in Bowlby’s work. There it could be used to mean almost anything cog-
nitive. In particular, it could sometimes refer to mental representations of specific caregivers
and sometimes to any cognitive process relevant to attachment (Chapter 1). One unfortu-
nate result was that readers were led away from the fundamental logic through which Main
extrapolated from the Strange Situation to coding systems for later in the lifespan, making
this leap seem mysterious, and making the development of precise hypotheses for testing
more difficult.256
In fact, for Main, the common thread linking the Strange Situation with her group’s new
measures was the ‘rules for the direction and organisation of attention’.257 Whereas in trad-
itional psychoanalytic theory, defensive processes operate on mental content, in Main’s
account, attentional processes are used to exclude information that might result in patterns
of attention and behaviour anticipated to be unhelpful based on representations of past ex-
periences. Whilst representations might guide the direction of attention, for Main it was
attentional processes that were ultimately what animated individual differences. So, for
instance:
In almost every assessment presented, children who had initially been judged in-
secure-avoidant [in infancy] showed an avoidant response pattern at 6 years of age.
They directed attention away from the parent on reunion, attending to toys or to activ-
ities; responded minimally (although politely) when addressed; and sometimes subtly
253 Zeanah and Barton reported that the concept of the internal working model, already by the end of the 1980s
and thus even within Bowlby’s lifetime, had all but become synonymous with Main’s AAI categories. Zeanah,
C.H. & Barton, M.L. (1989) Introduction: internal representations and parent–infant relationships. Infant Mental
Health Journal, 10(3), 135–41, p.139.
of representation. Monographs of the Society for Research in Child Development, 50, 66–104, p.66–7.
255 On academic positioning see Baert, P. (2012) Positioning theory and intellectual interventions. Journal for
the Theory of Social Behaviour, 42(3), 304–324.

256 Santarelli, M. (2017) Security as completeness: a Peircean semiotic reading of the psychology of attachment.
European Journal of Pragmatism & American Philosophy, 9(1). Attachment researchers such as Karlen Lyons-Ruth
and Howard and Miriam Steele have debated the extent to which Main’s ideas ‘were creative achievements in their
own right and were not contained in Bowlby’s work’. This debate has been hampered by the fact that, in legitimising
her methodological innovations, Main placed her work under the confused and confusing aegis of the ‘internal
working model’ concept. Lyons-Ruth, K. (1998) Commentary on Steele and Steele: lexicons, eyes, and videotape.
Social Development, 7(1), 127–31, p.128.
moved away from the parent. They seemed ill at ease in discussing feelings regarding
separation.258
The introduction of multiple new assessment methods was reported in the same chapter,
alongside new reflections on Bowlby’s concept of the ‘internal working model’. As a conse-
quence, all the explanations were highly compacted and their theoretical underpinnings are
only sketched. Details of the six-year assessments remained obscure—with some measures,
such as the family drawing, not reported at all. In the years after the ‘A move to the level of
representation’ paper, these further details were not forthcoming. Kaplan did not publish the
results from her dissertation. Strage did not write up her work. Cassidy published an analysis
of the results of her dissertation work259 but did not publish the coding system. The adapta-
tion of the Separation Anxiety Test by the Berkeley group was also left unpublished, though

it influenced work by Bretherton, Cassidy, George, and Solomon, and others on story-stem
and doll-play approaches to the assessment of attachment, which have since flourished.260 In
1986, Main completed a book manuscript with details of each of the coding systems, which
was due to be published by Cambridge University Press. However, Main’s sense that the sys-
tems were not yet adequately finished contributed to delays in the book’s publication. One
factor may have been that very few ambivalent/resistant dyads were available for the six-year
follow-up, so the ‘maximising’ aspects of the coding systems was relatively underdeveloped.
Main may also have had concerns that the coding system was ‘overfitted’ to her Berkeley
sample, and wanted to see replications and take the chance to amend the systems before
making them publicly available. Ultimately the book remained unpublished, though it circu-
lated to several colleagues in manuscript form.
Over the decades, only a handful of individuals were ever trained in any of the six-year
measures by Main.261 Main also did not delegate the matter of training to any colleagues—
until July 2019, when Naomi Gribneau Bahm, Kazuko Behrens, Anne Rifkin-Graboi,
and Deborah Jacobvitz received certification to deliver trainings in the Main and Cassidy
infant attachment classifications and stable over a 1-month period. Developmental Psychology, 24(3), 415–26.
260 Cassidy, J. (1988) Child–mother attachment and the self in six-year-olds. Child Development, 59, 121–34;
Bretherton, I., Ridgeway, D., & Cassidy, J. (1990) Assessing internal working models of the attachment relation-
ship: an attachment story completion task for 3-year-olds. In M. Greenberg, D. Cicchetti, & E.M. Cummings
(eds) Attachment in the Preschool Years: Theory, Research, and Intervention (pp.273–308). Chicago: University of
Chicago Press; Jacobsen, T., Edelstein, W., & Hofmann, V. (1994) A longitudinal study of the relation between rep-
resentations of attachment in childhood and cognitive functioning in childhood and adolescence. Developmental
Psychology, 30(1), 112–24; Emde, R.N., Wolf, D.P., & Oppenheim, D. (eds) (2003) Revealing the Inner Worlds of
Young Children. The MacArthur Story Stem Battery and Parent–Child Narratives. Oxford: Oxford University Press.
261 Ulrike Wartner received training from Main and achieved reliability with Main and with Ainsworth on the
Main and Cassidy system: Wartner, U.G., Grossmann, K., Fremmer-Bombik, E., & Suess, G. (1994) Attachment
patterns at age six in south Germany: predictability from infancy and implications for preschool behavior. Child
Development, 65(4), 1014–27. Cohn stated that her group received training on the Main and Cassidy system in
the 1980s, though no reliability test results were reported. Cohn, D.A. (1990) Child–mother attachment of six-
year-olds and social competence at school. Child Development, 61(1), 152–62. Nancy Kaplan also trained a few
colleagues to reliability: Pianta, R.C. & Longmaid, K. (1999) Attachment-based classifications of children’s family
drawings: psychometric properties and relations with children’s adjustment in kindergarten. Journal of Clinical
Child Psychology, 28(2), 244–55; Behrens, K.Y. & Kaplan, N. (2011) Japanese children’s family drawings and their
link to attachment. Attachment & Human Development, 13(5), 437–50; Granqvist, P., Forslund, T., Fransson, M.,
Springer, L., & Lindberg, L. (2014) Mothers with intellectual disability, their experiences of maltreatment, and
their children’s attachment representations: a small-group matched comparison study. Attachment & Human
reunion system. (The same researchers will also be trained and certified to deliver train-
ing in the FR coding system from summer 2020.)262 However, in the meantime, the lack of
trained coders produced a gap in the availability of consecrated measures of attachment. For
school-age children, alternatives such as the Manchester Child Attachment Story Task and
the Child Attachment Interview have been developed.263 Nonetheless, one consequence of
the de facto unavailability of the Berkeley six-year measures has been that, with few stud-
ies being conducted, over time publications that have used these measures have become
regarded as rather niche.264 A second consequence has been that what research has been
conducted with the Berkeley six-year measures has almost always occurred ‘off the books’,
without training or a standardised test to confirm reliability against other research groups.265
Researchers have been forced to proceed on the basis of the scanty information available in
the public domain. A rather sad case is the work of Dallaire and colleagues, who describe

their attempt to reconstruct a coding manual for the family drawing coding system, having
carefully scoured the little information made publicly available by Main’s research group.266
Finally, a third consequence has been that new measures to address this gap have emerged
only slowly, since they had to contend with the position of Main’s six-year system as the
orthodox and authoritative approach to assessment; as in the case of the Manchester Child
Attachment Story Task, often researchers sought validation of new assessments against the
Berkeley measures, treated as the gold standard.267 They also often emulated the Berkeley
four-category model. Lyons-Ruth, Bureau, and colleagues have been especially active in the
262 Mary Main, personal communication, July 2019.

263 Green, J., Stanley, C., Smith, V., & Goldwyn, R. (2000) A new method of evaluating attachment repre-
sentations in young school-age children: the Manchester Child Attachment Story Task. Attachment & Human
Development, 2(1), 48–70; Allen, B., Bendixsen, B., Babcock Fenerci, R. & Green, J. (2018) Assessing disorgan-
ized attachment representations: a systematic psychometric review and meta-analysis of the Manchester Child
Attachment Story Task. Attachment & Human Development, 20(6), 553–77. See also the ‘Security Scale’, Kerns,
K.A., Klepac, L., & Cole, A. (1996) Peer relationships and preadolescents’ perceptions of security in the child–
mother relationship. Developmental Psychology, 32, 457–66; the ‘Child Attachment Interview’, Target, M., Fonagy,
P., & Shmueli-Goetz, Y. (2003) Attachment representations in school-age children: the development of the Child
Attachment Interview (CAI). Journal of Child Psychotherapy, 29(2), 171–86; the ‘Friends and Family Interview’,
Steele, H. & Steele, M. (2005) The construct of coherence as an indicator of attachment security in middle child-
hood: the Friends and Family Interview. In K.A. Kerns & R.A. Richardson (eds) Attachment in Middle Childhood
(pp.137–60). New York: Guilford; the ‘School-Age Assessment of Attachment’, Crittenden, P.M., Kozlowska,
K., & Landini, A. (2010) Assessing attachment in school-age children. Child Clinical Psychology and Psychiatry,
14, 185–208; and the ‘School Attachment Monitor’, Vo, D.-B., Rooksby, M., Tayarani, M., et al. (2017) SAM: The
School Attachment Monitor. Paper presented at 2017 Conference on Interaction Design and Children (IDC ’17),
Stanford, CT, 27–30 June 2017, pp.671–4.
264 It is striking that a high proportion of subsequent studies using the six-year systems have been published in
Attachment & Human Development. The lack of formal interlaboratory reliability procedures may have contrib-
uted to the lack of wider interest in the results of findings using these measures, though this has not been a problem
for the Ainsworth sensitivity scale.
265 One route taken by some researchers has been to seek training in coding preschool Strange Situations from
Bob Marvin, and then extrapolate upwards. E.g. Moss, E., Cyr, C., & Dubois-Comtois, K. (2004) Attachment
at early school age and developmental risk: examining family contexts and behavior problems of controlling-
caregiving, controlling- punitive, and behaviorally disorganized children. Developmental Psychology, 40(4),
519–32.
266 Dallaire, D.H., Ciccone, A., & Wilson, L.C. (2012) The family drawings of at-risk children: concurrent rela-
tions with contact with incarcerated parents, caregiver behavior, and stress. Attachment & Human Development,
14(2), 161–83.
267 E.g. Green, J., Stanley, C., Smith, V., & Goldwyn, R. (2000) A new method of evaluating attachment repre-
sentations in young school-age children: the Manchester Child Attachment Story Task. Attachment & Human
Development, 2(1), 48–70; Target, M., Fonagy, P., & Shmueli-Goetz, Y. (2003) Attachment representations in
school-age children: the development of the Child Attachment Interview (CAI). Journal of Child Psychotherapy,
29(2), 171–86.
development of observational measures that offer equivalent categories to those identified

by Main.268 However, it is telling that they have tended to develop measures for alternative
ages, such as middle childhood and seven-to eight-year-olds. This would appear partly mo-
tivated by a desire to make assessments available for these age groups, but also, implicitly, to
avoid the appearance of competition with Main’s six-year measures, which remain widely
perceived as the ‘gold standard’ even if no training has been available and few studies con-
ducted over the past decades.
Access to two unpublished books by Main offers a chance to flesh out the brief description
of the reunion and family drawing measures in Main’s 1985 ‘A move to the level of represen-
tation’ paper. The first book is Behaviour and the Development of Representational Models of
Attachment from 1986. In one of Bowlby’s last pieces of professional correspondence, com-
posed less than two months before his death in October 1990, he wrote to his Portuguese

translator Sonia Monteiro de Barros: ‘I wonder if you are following Mary Main’s recent
work? . . . She has a book in press with Cambridge University Press which will be very remark-
able.’269 This book essentially comprises the coding manuals for the Strange Situation; the
AAI; the Strage and Main discourse system; the Kaplan family drawing system; and the Main
and Cassidy reunion system.270 The Kaplan adaptation to the Separation Anxiety Test was
not described. The second book is Four Patterns of Attachment Seen in Behaviour, Discourse
and Narrative: An Abstract for Psychoanalysis from 1995.271 This book was written in honour
of the psychoanalyst Joe Sandler; when he died just as the book was ready for submission,
Main chose not to submit it to the publisher. The 1995 book contains a description and com-
mentary on the measures from the 1986 book, together with a discussion of attachment and
psychoanalytic theory and of epistemological aspects of Main’s work. The six-year systems,
little discussed in print, help reveal the logic of Main’s approach to the more famous AAI,
as well as her thinking about classification and method in developmental science. Together
268 Bureau, J.-F., Easlerbrooks, M.A., & Lyons-Ruth, K. (2009) Attachment disorganization and control-
ling behavior in middle childhood: maternal and child precursors and correlates, Attachment & Human
Development, 11(3), 265–84; Brumariu, L.E., Giuseppone, K.R., Kerns, K.A, et al. (2018) Middle childhood
attachment strategies: validation of an observational measure. Attachment & Human Development, 20(5),
491–513
269 Bowlby, J. (1990) Letter to Sonia Monteiro de Barros, 6 August 1990. PP/ Bow/B.3/40;Main, M. (1986)
Behaviour and the Development of Representational Models of Attachment: Five Methods of Assessment. Unpublished
manuscript, Mary Main & Erik Hesse personal archive.
270 Though it was adapted by Kaplan and Main, the Separation Anxiety Test was an already-existing measure.
It was therefore not described in this book about new measurement systems. For reasons of space, and because
there are no more than a handful of reliable coders, the measure will not be discussed in detail below. The most
significant influence of the measure was on the rise of story-stem methods in attachment research: Bretherton,
I. & Oppenheim, D. (2003) The MacArthur story stem battery. In R.N. Emde, D.P. Wolf, & D. Oppenheim (eds)
Revealing the Inner Worlds of Young Children: The MacArthur Story Stem Battery and Parent–Child Narratives
(pp.55–80). Oxford: Oxford University Press. Nonetheless, some important studies using the Kaplan and Main
version of the Separation Anxiety Test include: Shouldice, A. & Stevenson-Hinde, J. (1992) Coping with security
distress: the separation anxiety test and attachment classification at 4.5 years. Journal of Child Psychology and
Psychiatry, 33(2), 331–48; Jacobsen, T., Edelstein, W., & Hofmann, V. (1994) A longitudinal study of the relation
between representations of attachment in childhood and cognitive functioning in childhood and adolescence.
Developmental Psychology, 30, 112–24; Granqvist, P., Forslund, T., Fransson, M., Springer, L., & Lindberg, L. (2014)
Mothers with intellectual disability, their experiences of maltreatment, and their children’s attachment represen-
tations: a small-group matched comparison study. Attachment & Human Development, 16(5), 417–36; Forslund,
T., Brocki, K.C., Bohlin, G., Granqvist, P., & Eninger, L. (2016) The heterogeneity of attention-deficit/hyperactivity
disorder symptoms and conduct problems: cognitive inhibition, emotion regulation, emotionality, and disorgan-
ized attachment. British Journal of Developmental Psychology, 34(3), 371–87.
271 Main, M. (1995) Four Patterns of Attachment Seen in Behaviour, Discourse and Narrative: An Abstract for
Psychoanalysis. Unpublished manuscript, Mary Main & Erik Hesse personal archive.
they offer deeper entry than the published record into Main’s approach to the measurement
of attachment and the theoretical commitments that organise this approach.
Guess and uncover
Central to Main’s work in the 1980s was a new, inductive approach to the development of
coding systems. Guess and uncover is a memory game for children. In the game, each player
secretly writes a number pattern that follows a rule. For instance, 3, 6, 10, 15, 21, 28. They
then cover all but two numbers from the right-hand-side. One player attempts to predict
the third number in their friend’s pattern. If they are correct (a ‘hit’), they get to continue
to guess the fourth number, and so on. If they do not guess correctly (a ‘miss’), the number

is revealed, but their opponent gets to take a turn. The game uses inductive processes in the
identification of a pattern, forming the basis for a deductive system that can support further
extrapolation. Main used this game as a methodology for developing new coding systems.
This is only mentioned in print in any detail in Main and Cassidy’s 1988 paper, and even
there it appears with sufficient brevity that it was not clear how someone else might use guess
and uncover in a scientific context.272 The approach has almost never been mentioned again
by subsequent attachment researchers,273 who instead tend to chalk up Main’s new coding
systems to her genius, solely and unreplicably.274 Yet there is no opposition between recog-
nition of Main’s creativity and consideration of the ‘guess and uncover’ methodology used in
the development of her coding systems. Furthermore, inattention to or lack of awareness of
Main’s approach has stopped later researchers from either using or evaluating it.
Main first applied this methodology in guiding the work of a new, young research as-
sistant in her group, Ruth Goldwyn. Edward Goldwyn had made a documentary film for
the BBC in the early 1970s, depicting the work of Ainsworth at Baltimore.275 During his
time working on the project, Goldwyn became friends with Mary Main. Some years later,
when he went to Berkeley to make another film, he and his 16-year-old daughter Ruth stayed
infant attachment classifications and stable over a 1-month period. Developmental Psychology, 24(3), 415–26: ‘A
judge who was well-informed with respect to infant attachment classification in general, but who was blind to the
infant attachment classifications for this sample, studied reunion responses across the sample as a whole and then
(one by one) studied each child’s reunion responses in an attempt to guess the probable infant attachment classifi-
cation with a given parent. The sixth-year system was gradually developed from this case-by-case study. We used
both rationales regarding a correct guess (match) and information regarding the actual infancy categories of sixth-
year misses (mismatches) in developing the sixth-year system’ (417).
273 An exception is Goldberg, S. (2000) Attachment and Development. London: Routledge, p.63. The semi-
dialectical method of scale development has, however, been used by former members of Main’s laboratory, e.g.
Solomon, J., George, C., & De Jong, A. (1995) Children classified as controlling at age six: evidence of disorganized
representational strategies and aggression at home and at school. Development & Psychopathology, 7(3), 447–63;
Cassidy, J., Marvin, R., with the Attachment Working Group of the John D. and Catherine T. MacArthur Network
on the Transition from Infancy to Early Childhood (1992) Attachment organisation in preschool children: pro-
cedures and coding manual. Unpublished manual.
274 E.g. Fonagy, P. & Campbell, C. (2015) Bad blood revisited: attachment and psychoanalysis. British Journal of
Psychotherapy, 31(2), 229–50. For discussions of the concept of ‘genius’ in the history of psychology and alterna-
tive approaches that nonetheless recognise individual qualities and their influence on science see Ball, L.C. (2012)
Genius without the ‘Great Man’: new possibilities for the historian of psychology. History of Psychology, 15(1),
72–83; Simonton, D.K. (2018) Creative genius as causal agent in history: William James’s 1880 theory revisited and
revitalized. Review of General Psychology, 22(4), 406–420.
275 If at First You Don’t Succeed . . . You Don’t Succeed, released August 1972. https://genome.ch.bbc.co.uk/
df74b910f3a64da5871107a126f61095.
with Main. By the time she finished secondary school, Ruth had no clear sense of her fu-
ture plans. Therefore, as a gap-year, in 1981 she went to Berkeley and joined Main’s group.
Goldwyn was invited to read books by Bowlby and Hinde, and Darwin’s book on the origin
of emotions, and she audited Main’s classes.276 She also worked with Kaplan in recruiting and
collecting data in the six-year follow-up, including conducting half of the interviews with
parents. Once the data were collected, however, Goldwyn was not immediately needed for
any other projects. Though study of events reported in the interview with the parents was
part of George’s doctoral project, George went on maternity leave after completing the inter-
view transcriptions.277
Main was intrigued by a particular transcript: ‘Although the speaker was clearly essentially
coherent and collaborative, Main noted that he went to slightly unusual lengths in describing
tender, emotionally affecting aspects of his life, lingering in somewhat lengthy descriptions of

his loss of a beloved family member . . . like a B4 infant, this speaker to a slightly unusual degree
seemed to be attempting to draw and maintain the interviewer’s attention, and (not untowardly)
to evoke a sympathetic response.’278 And, remarkably, this father was part of a dyad classified B4
five years earlier. Main set Goldwyn, her teenage research assistant, a ‘guess and uncover’ chal-
lenge. Could she match the transcripts of the parents’ autobiographical interviews to their child’s
Strange Situation classification from five years earlier?
Ultimately the result was the basis for the AAI coding system, which is discussed later in
the chapter in the section ‘Adult Attachment Interview’. The system was based on ‘guess and
uncover’ on a sample of 36 transcripts, and then applied to 66 further cases: ‘Developing the
system involved moving (blind) through each transcript in the development sample, and in
each instance using feedback (“correct” or “incorrect”) with respect to the infant’s attach-
ment classification to that adult) to refine and to further develop the rule system. This is
a slow-moving but highly profitable method of rule development, and it was used in the
creation of every succeeding system.’279 Part of what made the method so powerful and cre-
ative compared to the standard hypothesis-testing tradition in psychology was that both
‘hits’ and ‘misses’ fed fine-grained inductive information into a theory-driven deductive
system.280 It was therefore exceptionally well adapted for discerning unexpected links in
new data on the basis of a pre-existing frame of reference, in this case the infant attach-
ment classifications which Main had come to regard as reflecting the basic forms of human
emotional life. As Main put it, ‘hypothesis-testing procedures were combined with inductive
276 Ruth Goldwyn, personal communication, August 2013.

277 The research strategy for George’s dissertation had been to count incidents—such as losses—in the lives of
participants, rather than to examine participants’ interpretation of events: George, C. (1984) Individual differences
in affective sensitivity: a study of five-year olds and their parents. Unpublished doctoral dissertation, University of
California, Berkeley.
278 Hesse, E. (2016) The Adult Attachment Interview: protocol, method of analysis, and empirical studies: 1985–
2015. In J. Cassidy & P.R. Shaver (eds) Handbook of Attachment: Theory, Research, and Clinical Applications, 3rd
edn (pp.553–97). New York: Guilford, p.554.
279 Main, M. (1986) Behaviour and the Development of Representational Models of Attachment: Five Methods of
Assessment. Unpublished manuscript, Mary Main & Erik Hesse personal archive.
Psychoanalysis. Unpublished manuscript, Mary Main & Erik Hesse personal archive: ‘Our approach to the develop-
ment of methodologies for research in attachment has been at once dialectical and in keeping with the hypothesis-
testing canons of natural science . . . The reading of each transcript was informed by both our general knowledge of
the meaning implicit in differing kinds of discourse or narrative, and our more specialized knowledge of processes
influencing individual differences in attachment. On reading each transcript, we formed a hypothesis regarding
likely infant Strange Situation response to that parent as a speaker, utilizing the existing background rules.’
processes’.281 Assuming the accuracy of the infant classifications, the benefit would then be
that ‘this method maximises the likelihood of the discovery of new, attachment-associated
patterns’.282
In her unpublished 1995 book Four Patterns of Attachment Seen in Behaviour, Discourse
and Narrative, Main offered further reflection on the guess and uncover methodology.283
She linked it to the idea of the ‘hermeneutic circle’ in the work of hermeneutic philosophers
such as Heidegger, Jaspers, and Gadamer: ‘In this circle, the relations between the whole
and the part move forward with each repeatedly transforming and extending understanding
of the other. Thus, each part contributes to the understanding of the whole which, trans-
formed, then directs attention to new parts. These again alter the meaning of the whole.’284
For Heidegger and later philosophers in this tradition, the hermeneutic circle was a descrip-
tion of how all human understanding operates. Mortal and mutable, each of us interprets

new perceptions always in the context of our horizon of pre-existing assumptions, which in
turn can be modified by the new perception.
Main regarded the guess and uncover method as a sharpening and intensification of this
process of ‘dialectic’ between parts and wholes. The horizon of assumptions gets brought to
an explicit point every time the researcher dares to extrapolate from their existing rules in
guessing the next case, and the encounter with novelty is likewise sharpened as a source of
inductive learning by the repeated feedback of both ‘hits’ and ‘misses’ and attempt to then re-
vise existing rules. The process is then extended at a meta-level as the rules generated, ideally,
on a sub-sample are applied to the sample as a whole. The system is then fine-tuned and sub-
jected to attempts at replication on further samples. Through this process, the coding system
is likely to be changed by learning from the first few additional samples.285 Over time, the
rate of learning slows, as the system is stabilised by a system of rules that are effective enough
to deal with most novelty. Further refinements may occur, in principle. Without fail, every
one of Main’s coding systems identifies itself as unfinished and with future changes expected,
though none has seen revision now for decades.
Main offered a characterisation of this hermeneutic circle within Ruth Goldwyn’s ‘guess
and uncover’ challenge on the AAI transcripts:
Should a transcript predicted to belong to the parent of a secure infant then indeed be found
to ‘belong’ to the parent of a secure infant, we continued forward with the sample utilizing
the existing rules. A mis-match between our hypothesis regarding the way an infant would
respond to a given speaker in the Strange Situation (e.g., secure) and the infant’s actual
Strange Situation behavior toward the speaker (e.g., avoidant) was regarded as an instance
of real discordance if no theoretical explanation could be provided for the fact that the in-
fant had been judged avoidant. In this case the ‘rules’ for identifying the parents of secure
and avoidant infants would have been left unchanged. However, this same mis-match was
281 Ibid.
284 Ibid.
285 Main, M. (1987) Project proposal to the Guggenheim Memorial Foundation, November 1987. PP/Bow/B.3/
36/1: ‘It is an elegant property of these methods that within any sample they can be almost exhaustively refined
against the Strange Situation behaviour of the infant, so that even a review of this single Bay Area sample would no
doubt lead to modification and improvement.’
used to tentatively refine the systems for identifying both the parents of secure and the
parents of avoidant infants if a rationale for its appearance could be provided. In this case,
aspects of the previous rule-systems would have been eliminated, extended or altered, and
a tentative, modified system would have been devised. This modified system would then
be tested as further transcripts were read and infant Strange Situation status ‘uncovered’.
Should the modified system fail to correctly identify the parents of other avoidant infants,
it would likely be abandoned in favor of reversion to the earlier system, while success would
tend to instantiate the modifications made. This semi-dialectical method of system devel-
opment focuses initially upon the understanding of each individual dyad, and combines
inductive and deductive strategies. Note that this method recognizes and retains genuine
mis-matches between infant and parent.286

Reflecting on the ‘guess and uncover’ approach used by Main and colleagues in the develop-
ment of their measures, this approach has some quite clear limitations. Main was well aware
that the approach risked over-fitting coding systems to the data of the original sample. It also
took continuity as a methodological assumption: this is no problem when recognised as an
assumption. Yet with few details about ‘guess and uncover’ publicly available, attachment re-
searchers and the wider community of readers naturally interpreted Main’s identification of
continuities as a research finding. This contributed to an expectation of continuity in subse-
quent research, and to publication bias against reports of discontinuity.287
With ‘guess and uncover’ interpreted as proof of continuity rather than as a technique
for generating hypotheses, it was wholly to be expected that researchers in the 1990s would
be concerned when findings showed a ‘transmission gap’ between AAI classifications and
Strange Situation classifications.288 However, it was also wholly to be expected that later re-
searchers would face a growing weight of evidence suggesting much weaker continuity than
in Main’s findings, as they have done in the past few years (Chapter 6).289 In the meantime,
however, over decades the assumption of continuity, together with the dominance of the idea
of minimising and maximising strategies, has powerfully shaped the imagination of attach-
ment research. For instance, new measures developed for different age-groups have tended
to retain the same categories used by Main. Few attachment measures have had their cat-
egories developed inductively, based on theory about age-appropriate tasks and naturalistic
observation. They did not seem much need, since the relevant categories already seemed
to have been set out in advance by Main. The Minnesota group are among the exceptions
(Chapter 4), but their measures have had less theoretical influence, and uptake, than those of
the Berkeley group.
At the same time, for the purposes of making novel observations and links, the particular
strengths of ‘guess and uncover’ must be acknowledged. The integration of inductive and de-
ductive strategies within the approach was central to Main’s elaboration of her new methods,
287 Opie, J. (2018) Attachment stability and change in early childhood and associated moderators. Unpublished
doctoral thesis, Deakin University, Melbourne.

288 Van IJzendoorn, M. (1995) Adult attachment representations, parental responsiveness, and infant attach-
ment. Psychological Bulletin, 117(3), 387–403.

289 Verhage, M.L., Fearon, R.P., Schuengel, C., et al. (2018) Examining ecological constraints on the intergen-
erational transmission of attachment via individual participant data meta-analysis. Child Development, 89(6),
2023–37.
and to the novel hypotheses that they permitted. The hypothesis of a link between a dyad’s
attachment classification in the Strange Situation and qualities of the parent’s discourse was
perhaps the most remarkable, and is discussed further in the section ‘Adult Attachment
Interview’. However, ‘guess and uncover’ was also integral to Main and Cassidy’s identifica-
tion of the ‘controlling’ categories in the six-year reunion assessment.
The Main and Cassidy reunion system
Building on the success of Goldwyn’s use of ‘guess and uncover’ on the AAI transcripts, the
method was next applied by Cassidy during her visit to Berkeley in 1982. Several video-
tapes from the six-year reunion were stolen in a theft, and some tapes were unusable due to

technical difficulties. There were 33 children available with reunions with both parents for
Cassidy to analyse:
As the system was first being developed, a number of correct guesses (matches) were readily
made. For example, a child who craned her head around to smile at the mother, talked to
mother about her experiences of the last hour, and invited the mother to play was correctly
identified as having been secure with mother in infancy, and succeeding children who re-
sponded similarly to reunion with the parent were then correctly identified as having been
secure with that parent in infancy. We also used misses (mismatches) in system develop-
ment. For example, the first child seen to fail to speak in response to the parent’s increas-
ingly frantic conversational overtures was identified as a child who had been avoidant of
the parent as an infant. When this guess proved incorrect (the child had been disorgan-
ized/disoriented), we reviewed the reunion behaviour of children who had correctly been
identified as insecure-avoidant in infancy and discovered that avoidant children were at
least minimally responsive to the parent when pressed, giving brief answers (e.g., ‘What’s
that nice-looking set of toys you’ve been playing with?’ ‘Sandbox.’). Succeeding 6-year-olds
who confrontationally refused response to a particular parent (e.g., ‘Let me ask you again.
What’s that nice-looking set of toys you’ve been playing with?’ ‘ . . . ’) were considered more
likely to have been disorganized/disoriented with that parent in infancy. A new, control-
ling-punitive response category was gradually developed and separated from the insecure-
avoidant and secure response categories.290
The Main and Cassidy rule system was then given to an independent researcher to recode the
whole sample including, it would appear, the development set. The independent researcher
found 84% of child–mother reunions predictable from the infant attachment classification,
and 62% of child–father reunions.291
Main and Cassidy classified dyads as secure if, on reunion at age six, the child remained
calm and relaxed, but also expressed open pleasure on the parent’s return, and initiated con-
versation or interaction with the parent (or seemed ready to communicate if he or she did not
initiate interaction). Responsiveness to the parent was often displayed as a ready expansion
290 Main, M. & Cassidy, J. (1988) Categories of response to reunion with the parent at age 6: predictable from in-
fant attachment classifications and stable over a 1-month period. Developmental Psychology, 24(3), 415–26, p.417.
of the parent’s own remarks, continuing the conversation. In the Main and Cassidy system,
dyads were classified as avoidant if the child responded only minimally. The six-year-old
child did not turn away from the caregiver, or partially stonewall them, as might an avoidant
infant (especially the A1 subgroup). Instead, the child turned down the intensity of inter-
action to simmer whenever opportunities arose. For instance several seemed very interested
in examining the toys just as they saw their parent enter. Neither anger nor affection was
much in evidence. Dyads were classified as resistant if on reunion there was ‘exaggerative or
maximizing of relatedness to the parent’ by the child, for instance ‘putting the arm around
the parent and head-cocking while looking towards the camera (as though posing for a
mother–child portrait), speaking in a baby-like voice, or sitting on the parent’s lap.’ There
could be some signs of frustration with the parent, though this was much less in evidence
than in the infant category.

Main and Cassidy created two categories of disorganised attachment for the six-year re-
union system. In both cases the child’s behaviour towards the caregiver was controlling.
The first to be identified was behaviour that was punitive and directive: one previously dis-
organized child ordered the mother to ‘Sit down and shut up, and keep your eyes closed!
I said, keep them closed!’292 Another form of punitive behaviour was rejection of the parent
through a hostile and implacable silence, timed in such a way as to leave the parent humili-
ated or embarrassed. A further child from this group ‘pushed the door shut as his mother
tried to open it . . . another said immediately and angrily on his mother’s entrance, ‘Don’t
bother me’.’293 In the other category the control was solicitous and caregiving. In one such
dyad, the child asked the parent what she had been doing and how she was feeling, then care-
fully invited her to play, reassuring her that it would be fun for her. Another asked: ‘Are you
tired, Mommy? Would you like to sit down and I’ll bring you some [pretend] tea?’294 One
form of solicitous behaviour that particularly interested Main was ‘overbright’ behaviour,
in which the child showed a nervous, skipping cheerfulness on reunion. Main and Cassidy
observed that ‘brightness and caregiving of this kind do not merely alleviate the parent’s pre-
sent but temporary depression or anxiety, but ultimately are used to relieve or forestall the
experience of anxiety by the child’:
One child in the Berkeley sample had been in the playroom with the mother (with whom
she was secure) when she heard the father returning. She said to the mother, in an unmis-
takably depressed and resigned tone, ‘He’s coming’, but then immediately upon reunion
gave an exceptionally bright greeting and smile, (“DAD!”) and began immediately to at-
tempt to direct and guide her father’s attention (“See, this is a . . .”).295
292 Hesse, E. & Main, M. (2000) Disorganized infant, child, and adult attachment: collapse in behavioral and at-
tentional strategies. Journal of the American Psychoanalytic Association, 48(4), 1097–127, p.1107. That the punitive
category was identified before the caregiving category is revealed by a handwritten note sent by Main to Bowlby in
1982, identifying six categories from Cassidy’s work thus far: ‘1. Actively reestablishing relationship; 2. Responsive
and confident in relationship. 3. Avoidant. 4. Punitive/rejecting. 5. Hesitant/confused. 6. Not yet classified’: Main,
M. & Cassidy, J. (1982) Handwitten note attached to ‘Quality of attachment from infancy to early childhood: sta-
bility of classification, changes in behaviour’, submitted to SRCD. PP/Bow/J.4/4.
293 Main, M. & Cassidy, J. (1983) Secure attachment in infancy as a precursor of the ability to tolerate being left
alone briefly at five years. Paper presented at 2nd World Congress of Infant Psychiatry, Cannes, March 1983. PP/
Bow/J.4/4.
294 Hesse, E. & Main, M. (2000) Disorganized infant, child, and adult attachment: collapse in behavioral and at-
tentional strategies. Journal of the American Psychoanalytic Association, 48(4), 1097–127, p.1106.
In the case of controlling-punitive behaviour, the anger system appeared to have been re-
cruited in the service of attachment. Main and colleagues theorised that angry behaviour
permitted the child to regulate the relationship, and ensured a conditional kind of proximity,
even if it did not afford comfort.296 In the case of controlling-caregiving behaviour, the care-
giving system appeared to have been recruited in the service of attachment. As Bowlby had
described in Loss just a few years earlier, caregiving behaviour displayed by a child to a care-
giver might help keep the caregiver near and prop up the adult’s own capacity to offer care
(Chapter 1). The child was therefore also offered conditional access to proximity, even if this
proximity was achieved at the expense of acknowledgement of the child’s own attachment-
related feelings.
Though there were too few instances for Main and Cassidy to make a third category, the
researchers also noted that some children displayed sexual or romantic behaviours towards

their parent. They described this as an ‘emerging’ category, which would probably achieve
full category status after more cases had been seen.297 Again, this could be regarded as the
recruitment of components of what would later become the sexual system in adolescence,
in the service of maintaining the potential for closeness with the attachment figure.298 Some
further six-year-olds displayed conflict behaviours or disorientation on reunion with their
caregiver, with similar kinds of behaviours as those in the Main and Solomon infant system.
Initially, Main and Cassidy also had a category for ‘confused’ behaviour on reunion; how-
ever, this was not elaborated in later versions of the system.299
Ainsworth criticised Main and Cassidy for their use of the label ‘disorganised attachment’
to characterise controlling-punitive and controlling-caregiving behaviours in the six-year
system. She did not think it was an appropriate description for behaviours that clearly seem
environmentally responsive and coherently sequenced.300 Controlling-caregiving behaviour
(and perhaps controlling-punitive behaviour) could predictably offer access to conditional
proximity with the caregiver, making it ‘organised’ by Main’s own earlier definition. In the
controlling-punitive and controlling-caregiving categories, attention may also be coherent,
in being focused on the adult. Furthermore, the use of anger or caregiving (or sexuality)
in the service of attachment could be a behavioural repertoire channelled by human evo-
lutionary history, like the avoidant and ambivalent/resistant strategies. However, the term
‘disorganised’ was retained by Main and Cassidy to signal the longitudinal continuities with
the infant category. It also served to mark two ideas held by Main and Cassidy regarding
these children. A first was that the controlling behaviour suggested a chronic obstacle to the
296 An alternative/additional hypothesis has been put forward by Bureau and colleagues. In a chaotic and threat-
ening family context, it may not be wise to appear helpless; pre-emptive displays of threatening dominance may
save you from trouble. Bureau and colleagues found more hostility from mother at home in infancy to be associ-
ated with punitive behaviour at age 8 (r = .40). Punitive behaviour was also associated with severe physical abuse of
the participant by mother as reported by mother when participant was age 19 (r = .39). Bureau, J.-F., Easterbrooks,
A., & Lyons-Ruth, K. (2009) The association between middle childhood controlling and disorganized attachment
and family correlates in young adulthood. Society for Research in Child Development Biennial Meeting, Denver,
Colorado.
298 Main, M. & Cassidy, J. (1986) Categories of response to reunion with the parent at age six: predictable from
infant attachment classifications and stable over a one-month period. PP/Bow/J.4/4.

299 Main, M. & Cassidy, J. (1982) Handwritten note attached to ‘Quality of attachment from infancy to early
childhood: stability of classification, changes in behaviour’, submitted to SRCD. PP/Bow/J.4/4.

300 Ainsworth, M. (1991) Past and future trends in attachment research. International Society for the Study of
Behavioral Development, Minneapolis, July 1991. Film of the presentation made available by Avi Sagi-Schwartz
(Chair).
termination of the attachment system sometime earlier in the child’s experience. Anger or
caregiving would not have been pressed into service, they felt, if the child had received pre-
dictable access to proximity with their attachment figure in another way, whether through
directly communicating distress and seeking care or through an avoidant or ambivalent/re-
sistant conditional strategy. Secondly, a central aspect of the attachment system in childhood
is that children seek care from someone stronger and/or wiser when they are distressed.
However, controlling-punitive and controlling-caregiving behaviour involve an inversion of
power between child and parent. There is therefore assumed to be a symbolic ‘disorganisa-
tion’ of the attachment system, even if behaviour is coherent, in the chronic confusion of
roles.301
Controlling-punitive and controlling-caregiving behaviour may be smoothly sequenced
and ultimately achieve a conditional proximity with the caregiver. Both punitive and care-

giving behaviour allow children to approach, and their behaviour may help them retain the
attention of the caregiver—understood to be a requisite to any successful functioning of the
caregiving system. The fact that two main forms were found at age six, and that the evo-
lutionary function of the controlling-punitive and controlling-caregiving behaviour could
be inferred, suggests that they could be regarded as ‘conditional strategies’. However, Main
has never referred to the controlling-punitive and controlling-caregiving categories as
‘conditional strategies’—even though this has become common usage among close collab-
orators such as Cassidy and Liotti.302 She does not offer an explanation for this. The Main
and Cassidy six-year system was an extrapolation from the infant system, so it is quite pos-
sible that simply by definition for Main any behaviour labelled ‘disorganised’ could not be a
conditional strategy, with the connotations of the term itself shaping theory. Perhaps, how-
ever, it could also be that for Main a conditional strategy had to offer an alternative way of
achieving the goal of the behavioural system. Other attachment researchers defined the goal
of the attachment system more widely as the availability of the caregiver or the mainten-
ance of ‘organisation’, and so the controlling behaviours seem to be strategic in this sense.303
dren: procedures and coding manual. Unpublished manual: ‘Children who were formerly disorganized and/or
disoriented have become quite organized—organized in controlling the parent. (It may be that it is the relationship
that has become disorganized at age 6, in that the ordinary family structure, with the parent in control of the child,
has become reversed; see Marvin & Stewart, 1990.) It is as if a child who cannot count on the parent for structure
steps in and provides structure (takes control).’ Comparing the infant Main and Solomon classification to the six-
year controlling classifications, Cassidy and Marvin stated that ‘in one, the child’s behavior is disorganized; in the
other, it is the usual hierarchical structuring, with parent in control of child, which has become disorganized—or at
least seriously disordered’. This last qualification signals the slippage occurring as the term ‘disorganised’ is pressed
into diverse non-overlapping usages. Acknowledging that the term disorganisation is operating at multiple levels
of analysis, contributing to potential confusion, Moss and colleagues distinguished ‘behaviourally-disorganized’
forms of disorganisation, where apparent coherence is lacking, from controlling-caregiving and controlling-
punitive behaviours. O’Connor, E., Bureau, J.F., Mccartney, K., & Lyons-Ruth, K. (2011) Risks and outcomes as-
sociated with disorganized/controlling patterns of attachment at age three years in the National Institute of Child
Health & Human Development Study of Early Child Care and Youth Development. Infant Mental Health Journal,
32(4), 450–72.
302 E.g. Cassidy, J., Marvin, R., with the Attachment Working Group of the John D. and Catherine T. MacArthur
dren: procedures and coding manual. Unpublished manual: ‘the strategy of controlling the parent’s behavior ap-
pears to have evolved from a lack of any consistent strategy for dealing with the attachment figure in infancy’; Liotti,
G. (2006) A model of dissociation based on attachment theory and research. Journal of Trauma & Dissociation,
7(4), 55–73; Lyons-Ruth, K., Dutra, L., Schuder, M.R., & Bianchi, I. (2006) From infant attachment disorganization
to adult dissociation: Relational adaptations or traumatic experiences? Psychiatric Clinics, 29(1), 63–86.
303 See e.g Cassidy, J., Marvin, R., with the Attachment Working Group of the John D. and Catherine
T. MacArthur Network on the Transition from Infancy to Early Childhood (1992) Attachment organisation in
By contrast, Main’s model of the attachment system was formed by Bowlby’s early control
system model in which proximity terminated the system. The controlling-punitive and
controlling-caregiving six-year-olds, even if they do achieve proximity to hit or settle their
parent, do so primarily to the rhythms required for regulating the parent. Such behaviour
can be inferred to ultimately benefit the attachment system in offering some proximity, but
it is not clear that the system can be terminated if this rhythm is generally unrelated to the
child’s attachment needs.
Stability between infant attachment and the Main and Cassidy six-year system was high.
All 12 infants classified as secure with mother in infancy were classified as secure five years
later. Six out of the eight avoidant dyads received the same classification, as did eight out
of twelve disorganised dyads.304 This led Main and Cassidy to conclude that controlling-
punitive and controlling-caregiving behaviour can be regarded as an attempt to organise

and regulate the behaviour of a caregiver who might otherwise be alarming. For instance,
both behaviours might retain the attention of a caregiver struggling with trauma, who
might otherwise enter into episodes of abrupt unavailability. Cassidy recruited 50 dyads in
Charlottesville for a replication of the six-year reunion study. Intercoder reliability was 90%
(κ = .84) for the organised categories. However, eight of the twelve category disagreements in-
volved the D category. In a reapplication of the procedure after one month in Charlottesville,
stability for the disorganised classification was poor: only seven of fourteen (50%) children
classified as controlling in Session 1 were controlling in Session 2.305 This may be one reason
why Main regarded the system as unfinished and has only offered training on it to a few re-
searchers over the decades.306
An important later study using the Main and Cassidy system was conducted by Solomon
and George, former doctoral students in Main’s group. During the separation they had the
children in their sample complete a doll-play procedure featuring story stems regarding get-
ting hurt, a threatening monster, separation, and reunion; meanwhile, their mothers were
asked to complete the AAI and an interview about their experiences as a parent.307 The
six-year reunions were coded by Cassidy. Solomon and George found that the mothers of
preschool children: procedures and coding manual. Unpublished manual: ‘The concept of “strategies” refers to
each individual’s attempts to maintain a particular organization in relation to attachment.’
infant attachment classifications and stable over a 1-month period. Developmental Psychology, 24(3), 415–26.
305 Ibid. p.421.
306 Another replication of the Main and Cassidy study was conducted by Wartner and colleagues, who managed
to recall 40 of the Regensburg sample (92%) five years later. Wartner received training in the coding system from
Main, and a number of videos were second coded by Main and Ainsworth for reliability. Prediction from infancy
to the six-year reunion revealed 82% agreement (κ = .723). Wartner, U.G., Grossmann, K., Fremmer-Bombik, E.,
& Suess, G. (1994) Attachment patterns at age six in south Germany: predictability from infancy and implications
for preschool behavior. Child Development, 65(4), 1014–27. Kazuko Behrens’ doctoral thesis, supervised by Main
and Hesse, was intended as a further attempt to replicate Main and Cassidy in the Japanese context. Behrens found
a strong (r = .60) relationship between the AAI with parents and the six-year reunions. Behrens, K.Y., Hesse, E., &
Main, M. (2007) Mothers’ attachment status as determined by the Adult Attachment Interview predicts their 6-
year-olds’ reunion responses. Developmental Psychology, 43(6), 1553–67.
307 The procedure resembles the Kaplan and Main Separation Anxiety Test, another projective measure used
with the six-year participants in the Berkeley sample while the parents were taken for AAIs. The findings were
also strikingly similar. However, details of the findings from Kaplan’s Separation Anxiety Test were only briefly de-
scribed in print at the time, with most of Kaplan’s work remaining unpublished. For a subsequent summary of these
findings see the discussion of the Separation Anxiety Test in Main, M., Hesse, E., & Kaplan, N. (2005) Predictability
of attachment behavior and representational processes at 1, 6, and 19 years of age: the Berkeley Longitudinal
Study. In K.E. Grossmann, K. Grossmann, & E. Waters (eds) Attachment from Infancy to Adulthood: The Major
Longitudinal Studies (pp.245–304). New York: Guilford, p.256.
secure six-year-olds had a distinctive quality in offering a balanced and detailed provision
of helping the child cope with worries. The mothers of avoidant six-year-olds had a distinct-
ively rejecting quality both towards the child and towards the caregiver’s own difficulties,
with a pervasively negative attitude towards themselves and their child.308 And the mothers
of controlling six-year-olds described feeling helpless and out of control in their relation-
ships with the child: ‘Some children in the controlling group were described as wild or help-
less and vulnerable (e.g., locking the mother out of the house, persistent bed-wetting, wild
tantrums). Other children were described as precocious or powerful (e.g., comedian-like be-
havior, amazing acting abilities, caregiving skills, supernatural powers, special connections
with the deceased).’309 All of the caregivers classified as unresolved on the AAI were part of
dyads classified as controlling on reunion; indeed, there was a match between six-year re-
union and AAI classifications in 81% of cases (κ = .74).310

In their doll-play, the children classified as controlling demonstrated themes of fear and
danger, in contrast to the rest of the sample: ‘the narrative structure of these stories can best
be described as chaotic and flooded. Catastrophe, sometimes multiple catastrophes, often
arise without warning; dangerous people or events are vanquished, only to surface again and
again. Objects float and have magical, malignant powers; punishments are abusive and unre-
lenting’; ‘fantastic disasters frequently arose without warning. Further, children hastily gave
post hoc explanations for these as though they were surprised and disturbed by the direction
the story had taken.’311 Five of the seven children classified as controlling-caregiving seemed
‘extremely uncomfortable with the task and did not want to enact a story’. When given re-
peated prompts, their narratives were brimming with themes of ‘chaos and disintegration’.312
Solomon and George observed that for the controlling group as a whole, ‘fears disrupted
their doll-play by flooding the content in a chaotic and primitive way or were inflexibly con-
tained through a brittle strategy of inhibition of play’.313 Children from dyads classified as
controlling were more likely to be judged as having behavioural problems by their teachers
and by their parents, whereas there were no differences between the secure and organised-
insecure dyads on these measures.
After her work on the six-year reunion system, Cassidy undertook a large project with
another of Ainsworth’s former students, Robert Marvin, in the development of a pre-
school system for coding the Strange Situation. These efforts were supported by a working
group, including Ainsworth and Main, funded by the MacArthur Foundation.314 Unlike
the Ainsworth coding system, which developed inductively from naturalistic observation,
the Cassidy and Marvin coding protocol elaborated ‘up’ from the infant coding system and
308 George, C. & Solomon, J. (1996) Representational models of relationships: links between caregiving and at-
tachment. Infant Mental Health Journal, 17(3), 198–216, p.210.

309 Ibid. p.212.
310 Ibid. Table 4.
311 Solomon, J., George, C., & De Jong, A. (1995) Children classified as controlling at age six: evidence of disor-
ganized representational strategies and aggression at home and at school. Development & Psychopathology, 7(3),
447–63, p.454, 460.
312 Ibid. p.454.
313 Ibid.
314 The members of the working group who contributed to the Cassidy and Marvin system were Mary
Ainsworth, Kathryn Barnard, Leila Beckwith, Marjorie Beeghly, Jay Belsky, Janet Blacher, Inge Bretherton, Wanda
Bronson, Heather Carmichael-Olsen, Dante Cicchetti, Keith Crnic, Mark Cummings, Ann Easterbrooks, John
Gottman, Mark Greenberg, Robert Harmon, Lyn LaGasse, Mary Main, Colleen Morisset, Janet Purcell, Doreen
Ridgeway, Nancy Slough, Susan Spieker, Mathew Speltz, and Joan Stevenson-Hinde.
‘down’ from the six-year reunion system.315 Main’s ‘guess and uncover’ method was used
with a set of 300 Strange Situations drawn from various American (and one British) research
groups. The MacArthur preschool system is based in the first instance on common themes
between the two systems. First, ‘the pattern shown by infants classified insecure/avoidant
(pattern A) is strikingly similar to that shown by six-year-olds classified insecure/avoid-
ant: children at both ages avoid intimate interaction or contact, maintain an affective neu-
trality, and convey the impression that the parent’s return is of no particular importance to
them’. Second, ‘the pattern shown by infants classified securely attached (pattern B) is strik-
ingly similar to that shown by six-year-olds classified securely attached: children at both ages
show interest in proximity or at least interaction with the attachment figure’. Third, the am-
bivalent preschool classification incorporates the resistance of the Ainsworth infant system
and the intensified signalling of relatedness to the parent of the Main and Cassidy system.

Immature behaviour contributes to a subclassification of ambivalence, on the analogy of pas-
sivity in the Ainsworth infant system. Fourth, disorganised attachment is coded on the basis
of either Main and Solomon behaviours from the infant system or the controlling-caregiving
or controlling-punitive behaviours from the Main and Cassidy system. In 2002 Cassidy and
Marvin added some additional subclassifications, including a depressed subclassification
and one for children who seemed overtly fearful of their caregiver.316
The Cassidy and Marvin system has been used very widely and, with annual training
available, new studies using the approach continue to regularly appear. As such, in all likeli-
hood it represents the most important legacy of the Main and Cassidy system, where train-
ing has not been available. There has been a severe bottleneck for relevant research in middle
childhood over the decades. For instance, the only large randomised study of adoption after
institutionalisation, the English and Romanian Adoptee project, conducted a separation-
reunion procedure with their participants at age six. Due to the de facto lack of availability
of the Main and Cassidy method, the researchers chose to code their data using the Cassidy
and Marvin system. They found uninterpretable results.317 The researchers concluded that
the Cassidy and Marvin system was not appropriate for their study. Given the unique nature
and global importance for policy of the English and Romanian Adoptee project, this was a
wasted opportunity on a historic scale.
Given the lack of availability of training, the most—and potentially the only—especially
influential study since the 1990s to have used the Main and Cassidy system was a longitu-
dinal study by Ellen Moss and colleagues. With a community sample of 120 French Canadian
children, they coded the Strange Situation at 2.5 years using the Cassidy and Marvin system
and a separation-reunion procedure coded with the Main and Cassidy system at 5.5 years.
Though trained in the Cassidy and Marvin system, it is not clear that Moss and colleagues
were trained in the Main and Cassidy system, or sought (or could seek) interlaboratory re-
liability. The researchers found moderate stability of attachment classifications over the
preschool years (secure: 72%, z = 6.0; avoidant: 44%, z = 3.6; ambivalent: 62%, z = 5.0; disor-
ganized: 77%, z = 5.6). None of the dyads classified as disorganised at 2.5 years was classified
dren: procedures and coding manual. Unpublished manual.
316 Ibid. Update September 2002.
317 Kreppner, J., Rutter, M., Marvin, R., O’Connor, T., & Sonuga-Barke, E. (2011) Assessing the concept of
the ‘insecure-other’category in the Cassidy–Marvin scheme: changes between 4 and 6 years in the English and
Romanian adoptee study. Social Development, 20(1), 1–16.
as secure at 5.5 years. Changes from secure to disorganised attachment were associated with
changes in caregiver behaviour towards the child,318 lower marital satisfaction, and events
such as bereavement and parental hospitalisation. In addition to the separations associated
with hospitalisation, Moss and colleagues offered the speculation that ‘parental hospitaliza-
tion is also likely to compromise the child’s confidence in parents as a source of protection
and security’.319
In an expansion of the sample by Moss and colleagues to include a number of five-year-
olds, 68% of the children classified in the D group showed one of the controlling responses on
reunion; 32% showed Main and Solomon-style disorganisation. Maternal report of marital
conflict and unhappiness was associated with the display of Main and Solomon indices on
reunion, but not with the Main and Cassidy controlling behaviours: ‘The unpredictability
and overwhelming nature of the family environment, which is disrupted by severe marital

conflict, may compromise the possibility of the child’s forming an organized integrated
model of attachment, even one of a role-reversed controlling nature.’320 The researchers did
not include measures of exposure to violence, abuse, neglect, or other forms of adversity that
might have shed further light on the experiences of these children. However, contrary to
the researchers’ expectation, Moss and colleagues found no association between maternal
depression and any of the forms of disorganisation, including the controlling strategies.
Though given that this was a community sample, they by no means ruled out that such an
association would be seen in a population facing more adversities. Moss and colleagues did
find, though, that families of children who developed a controlling-caregiving pattern were
more likely to have experienced a major bereavement. Children who showed controlling-
punitive behaviours on reunion with their parent were rated by teachers as displaying more
conduct problems at school; children who showed controlling-caregiving behaviours were
rated by teachers as displaying more anxiety and depression at school.
The Kaplan and Main family drawing system
Another of the six-year systems developed by the Berkeley group was a system for coding
family drawings. The use of drawing tasks as a projective measure had roots in the work of
both Bowlby and Ainsworth.321 The children in the Berkeley six-year follow-up had been
318 The Moss and colleagues observational measure of caregiver– child interaction includes scales
for: Coordination, Communication, Partner Roles, Emotional Expression, Responsivity–Sensitivity, Tension,
Mood, and Enjoyment. See Moss, E., Rousseau, D., Parent, S., St-Laurent, D., & Saintonge, J. (1998) Correlates
of attachment at school age: maternal reported stress, mother–child interaction, and behavior problems. Child
319 Moss, E., Cyr, C., Bureau, J. F., Tarabulsy, G.M., & Dubois-comtois, K. (2005) Stability of attachment during
the preschool period. Developmental Psychology, 41(5), 773–83, p.781. Moss and colleagues also conducted the
Solomon and George doll-play measure with their sample when the children were aged eight. Agreement between
the Main and Cassidy system and the doll-play measure was 73% (κ = .45), Dubois-Comtois, K., Cyr, C., & Moss,
E. (2011) Attachment behavior and mother–child conversations as predictors of attachment representations in
middle childhood: a longitudinal study, Attachment & Human Development, 13(4), 335–57.
320 Moss, E., Cyr, C., & Dubois-comtois, K. (2004) Attachment at early school age and developmental risk: exam-
ining family contexts and behavior problems of controlling-caregiving, pontrolling-punitive, and behaviorally dis-
organized children. Developmental Psychology, 40(4), 519–32, p.529.
321 In his ‘Forty-four thieves’ paper, Bowlby reported on a case where he had used a drawing task to seek a sense
of the child’s inner life: ‘Lily had always been a miserable and frightened child . . . She spent most of her spare time
in the streets just mooning about. She was very slow and dreamy and took hours to do things. Her mother de-
scribed how she sometimes got “miles away” which made her feel “creepy” . . . When asked to do a drawing she pre-
ferred an abstract design to a picture.’ Bowlby, J. (1944) Forty-four juvenile thieves: their characters and home-life.
asked to complete a family drawing. Like Ainsworth’s use of projective measures (Chapter 2),
Kaplan and Main intended the family drawing task as a standardised, ambiguous, and evoca-
tive stimulus, to elicit individual differences based on past experiences in the family rela-
tionships. It was also well aligned with Kaplan’s background in arts and humanities from
her undergraduate studies at Sarah Lawrence College.322 The ‘guess and uncover’ method-
ology was, once again, utilised in the development of the Kaplan and Main drawing system.
However, rather than use a development sample and then apply the scheme to the rest of the
data, the number of family drawings was judged too small for this approach. Kaplan and
Main therefore performed no external test, but used the whole sample in the development
on the scheme: ‘Moving blind through the set of drawings while developing the rules for
identifying the classifications, the two original judges independently obtained hit-rate [with
the infant classifications] of close to 80%. The drawings and the rule-system were then given

to a third blind judge who had not participated in the development of the system: this judge
obtained an overall hit-rate of 76%.’323
Family drawings classified as secure by Kaplan and Main had the general quality of being
‘rational and realistic’. Kaplan and Main found that figures tended to be complete, with eyes,
nose, mouth, hair, hands, and feet. This was despite the fact that there was no relationship
between an independent assessment of drawing skill and attachment classification at age
six. Drawings classified as secure had figures that seemed well grounded, with legs planted
in a way that suggested that the floor beneath them, even if invisible, was adequately stable.
Figures appeared in the middle of the picture, and did not have their arms closed off from
one another or seem to be in over-rigid postures. When height differences were character-
ised, no figure loomed over anyone else. Objects and houses might appear in the picture,
but did not dominate the scene in a way suggesting that attention was being directed away
from the attachment figures. A curious aspect of the secure classification in the Kaplan and
Main system is that they specifically stated that the absence of the self from the picture is
not relevant to the classification. This was partly the result of their inductive methodology,
since the absence of the subject was seen in a number of pictures where the child had pre-
viously been classified secure in infancy. However, the conclusions also seems theoretically
motivated. Whereas ‘internal working models’ are generally described by other attachment
researchers as representations of the self and others (Chapters 1 and 5), the secure classi-
fication on the six-year drawing system shows that this was not Main’s primary concern.
Depictions of the self are not necessarily relevant: what mattered was the kind of attention
directed towards or away from attachment figures and attachment-relevant experiences
(Chapter 2).
International Journal of Psycho-Analysis, 25, 19–53, p.26. Bowlby, Ainsworth, and colleagues also asked children to
complete projective drawing tasks after they returned from hospitalisation. Ainsworth, M.D. & Boston, M. (1952)
Psychodiagnostic assessments of a child after prolonged separation in early childhood. British Journal of Medical
Psychology, 25, 169–201, p.175.

Assessment. Unpublished manuscript, Mary Main & Erik Hesse personal archive. No further information appears to
be available regarding the breakdown of associations between the infant attachment classifications and the six-year
drawing system, or the associations between the different six-year systems. The match between the infant classifica-
tions and the Kaplan drawing system for mothers was reported as 78% and as non-significant for fathers by Main,
M. (1987) Project proposal to the Guggenheim Memorial Foundation, November 1987. PP/Bow/B.3/36/1.
Family drawings classified as avoidant by Kaplan and Main had many merry figures with
little individuation. They also often lacked arms, or had their arms in postures not suitable
for physical contact with others. Kaplan and Main found this especially interesting, since
when asked to do a different drawing, of a bear, arms were then present and the apparently
unwelcoming posture was absent. The figures also seemed rather isolated from one another
on the page. The overriding characteristic was a picture of a family with attention directed
away from the potential for the members’ potential vulnerability to or intimacy with one
another. Family drawings classified as resistant tended to draw out-of-scale family figures,
placed very close together, sometimes even with overlapping arms or shoulders. Several
features of the drawings emphasised vulnerability. For instance, some of the children drew
themselves with round bellies; half of children in this category, and no other child in any
other classification, drew an unusual slant to the neck and head relative to the shoulders

and the rest of the body, a posture that suggested coy submission. However, there were rela-
tively few ambivalent/resistant children in the sample, and the classification was somewhat
unfinished.324
Drawings classified as disorganised had some disrupted or chaotic elements, such as the
child having to redo and redo the task or being unable to attempt it effectively. Kaplan and
Main also identified two elements that corresponded to the punitive and caregiving behav-
iours that had been observed by Cassidy in the reunion episodes. A first was the presence of
ominous elements in the picture, such as black clouds looming over the figures, dismem-
bered or floating body-parts, or rows of skeletons in the background.325 A second was over-
bright elements ‘which at first sight appear simply very sweet or very cheerful (e.g. a family
standing on a row of hearts with a small sun drawn directly over the mother’s head; a huge
sun with a smiling face dominates the figures in the picture)’. This was different to the add-
ition of hearts and flowers added by many other children, since here the cheerful elements
seemed to dominate the image and to have a role in emotionally propping up the characters
rather than expressing their inner state.
Though an unpublished system, the research community was intrigued by the Kaplan and
Main coding system, since it seemed to offer a special access to the child’s ‘internal world’.326
It was also the least resource-intensive measure to date produced by the developmental trad-
ition of attachment research. An attempt to replicate the Kaplan and Main study was con-
ducted by Fury, Carlson, and Sroufe using data from the Minnesota Longitudinal Study of
Risk and Adaptation. No training by Kaplan or Main or attempt to achieve cross-laboratory
324 Poor inter-rater reliability was reported by Pianta and Longmaid on the distinction between resistant and
disorganised family drawings: Pianta, R.C. & Longmaid, K. (1999) Attachment-based classifications of children’s
family drawings: psychometric properties and relations with children’s adjustment in kindergarten. Journal of
Clinical Child Psychology, 28(2), 244–55.
325 These ominous elements also seemed to resonate with Kaplan and Main’s insecure/disorganized-fearful clas-
sification for the Separation Anxiety Test. In response to pictures of child–caregiver separations, functioning as
story stems, children who received this classification showed behaviours suggestive of fear. These included extreme
voicelessness occurring only in response to the test stimuli, lapses in discourse disorganization suggestive of seg-
regated systems (e.g. ‘yes-no-yes-no-yes-no’), accounts of events with an eerie quality in which there is no human
cause, or worries that the caregiver might have died. Some children also became abruptly aggressive to the assessor
or the test materials. Kaplan, N. (1987) Individual differences in 6-years olds’ thoughts about separation: predicted
from attachment to mother at age 1. Unpublished doctoral dissertation, Department of Psychology, University of
California, Berkeley.
326 E.g. Warren, S.L., Emde, R.N., & Sroufe, L.A. (2000) Internal representations: predicting anxiety from chil-
dren’s play narratives. Journal of the American Academy of Child & Adolescent Psychiatry, 39(1), 100–107.
reliability was reported.327 The children in the study were eight-to nine-year-olds. Fury and
colleagues had concerns about some aspects of the system. First, ‘many of the Kaplan and
Main signs that required a large measure of subjectivity (e.g., “faint ominousness”; “pained
smile”) were excluded’.328 Second, some of the signs seemed overparticular, and a product
of overfitting of the coding system to the Berkeley sample. Indeed, they found that only
one of the indicators of avoidance from the Berkeley drawing study—the positioning of
figures’ arms—was associated with avoidant attachment in the infant Strange Situation in
the Minnesota sample. Fury also alleged that the semi-inductive ‘guess and uncover’ meth-
odology used by Kaplan and Main was overly scattershot in the identification of specific
markers, and that Kaplan and Main offered too little in the way of principles regarding ‘pre-
cisely how various descriptors were organised and distributed’.329 In the Fury study, none
of the individual markers of disorganised attachment in the drawing system was associated

with disorganised attachment in the infant Strange Situation. However, when signs were
combined into aggregate indices, the expectable associations appeared, though associations
were weak or modest (avoidance, r(168) = .15, p < .05; resistance, r(168) = .27, p < .001).330
The best association of the drawing system was not with the infant Strange Situation, but
with a global measure of mental health: a .32 association with resistance in girls (though no
association for boys) and a .36 association with avoidance in boys (though no association for
girls).331 Main was dismayed by these results, and in a 1995 chapter concluded that ‘children’s
drawings should not be used in the assessment of attachment’: at most, they should be ‘re-
garded only as correlates of early attachment’.332
The Minnesota findings were replicated by Madigan and colleagues with 127 dyads drawn
from a moderate-risk sample in Ontario, with a significant proportion of children diagnosed
with forms of heart disease. Again, no training by Kaplan or Main or attempt to achieve
cross-laboratory reliability was reported. These researchers found few links between in-
fant attachment and specific markers in the drawings made by the seven-year-olds in their
sample. However, they found that the match between the infant and the seven-year attach-
ment classifications was 80%, with most of the mismatches occurring for the avoidant in-
fant category: over half were classified as secure in the drawing system. The low stability in
the high-risk Minnesota study and the high stability in the lower-risk Berkeley and Ontario
samples is in line with findings regarding the stability of attachment classifications in general
(Chapter 2). Madigan and colleagues speculated that a higher proportion of disorganised
cases in the Minnesota study may also have played a role in the different results.333
327 Inter-rater reliability was instead reported between Fury and an undergraduate student in Fury, G.S. (1996)
The relation between infant attachment history and representations of relationships in school-aged family draw-
ings. Unpublished doctoral dissertation, University of Minnesota.
328 Fury, G., Carlson, E.A., & Sroufe, L.A. (1997) Children’s representations of attachment relationships in
family drawings. Child Development, 68(6), 1154–64, p.1115. A more detailed discussion of the excluded items is
available in Fury, G.S. (1996) The relation between infant attachment history and representations of relationships
in school-aged family drawings. Unpublished doctoral dissertation, University of Minnesota.
329 Ibid.
330 Fury, G., Carlson, E.A., & Sroufe, L.A. (1997) Children’s representations of attachment relationships in family
drawings. Child Development, 68(6), 1154–64, p.1161.

331 Ibid. Table 1.
Goldberg, R. Muir, & J Kerr (eds) Attachment Theory: Social, Developmental and Clinical Perspectives (pp.407–70).
333 Madigan, S., Ladd, M., & Goldberg, S. (2003) A picture is worth a thousand words: children’s representations
of family as indicators of early attachment. Attachment & Human Development, 5(1), 19–37.
Adult Attachment Interview 279
Running contrary to Main’s 1995 claim that the drawing system should no longer be used
as a measure of attachment, in a 2016 paper Gernhardt and colleagues observed that the
system had, in fact, become increasingly popular over the subsequent decades. Suspicious of
the measure, they wondered about its cross-cultural applicability. In their study, Gernhardt
and colleagues used the drawing system—though without the disorganised classification—
with 32 six-year-old children from Berlin and 21 six-year-old children from rural farming
villages around Kumbo in Cameroon. Once again, no training by Kaplan or Main or attempt
to achieve cross-laboratory reliability was reported. Their results showed that most of the
pictoral markers of insecurity were dramatically more common in the Kumbo sample. For
instance, neutral or negative facial affect characterised every drawing by the children from
Kumbo, compared to a tiny fraction of the children from Berlin. Likewise, lack of individu-
ation, arms downwards, and unusually small figures were common features in the Kumbo

pictures, whereas they were rare in the Berlin pictures. Nearly half of the Kumbo drawings
omitted the mother, whereas none of the Berlin drawings did so. Gernhardt concluded that
the Kaplan and Main system is essentially a cultural artefact, without any cross-cultural val-
idity. This conclusion is questionable since they did not examine whether the results of the
drawing measure correlate with expectable individual differences in the children’s lives or
behaviour. Their view was that to do so would have been culturally inappropriate and poten-
tially pathologising. Nonetheless, Gernhardt and colleagues do effectively make the point
that many markers of insecurity identified by Kaplan and Main on the Berkeley sample need
to be treated with caution when applied to contexts with different ways of thinking about,
representing, and experiencing family life.334 This caution is especially important since the
underpinning principles for the Kaplan and Main system, which could have been used by
researchers to responsively elaborate the specifics of the coding system to achieve greater
cross-cultural applicability, remain unpublished and, additionally, underdescribed even in
the unpublished texts.
Adult Attachment Interview
Semantic and episodic information
During the six-year follow-up at Berkeley, while their children were busy with tasks such as
completing the family drawing, mothers and fathers were interviewed separately about their
experiences relevant to attachment. Bowlby’s 1980 book Loss had been shared with Main
in draft a few years earlier. In this work Bowlby was interested in the variety of forms that
memory can take, and especially the work of Tulving. According to Tulving, ‘episodic’ in-
formation comprises temporally dated episodes or events and of relations between such epi-
sodes or events, as experienced by an embodied subject; by contrast, ‘semantic’ information
contains generalised propositions about the world.335 Bowlby argued that whereas episodic
information derives primarily from an individual’s own embodied experience, semantic in-
formation is more explicit, linguistically encoded knowledge, in particular what others have
334 Gernhardt, A., Keller, H., & Rübeling, H. (2016) Children’s family drawings as expressions of attachment rep-
resentations across cultures: possibilities and limitations. Child Development, 87(4), 1069–78.
335 Tulving, E. (1985) How many memory systems are there? American Psychologist, 40, 385–98.
told the individual, especially as a child.336 Kaplan, George, and Main therefore developed an
interview protocol to include prompts for both descriptions of key attachment relationships
and specific supportive memories.
The AAI began with a question, developed by Kaplan on the basis of her clinical inter-
ests,337 asking speakers to describe their relationship with their parents. Main added a
request that speakers choose five adjectives to describe each relationship, a prompt for se-
mantic memory.338 Speakers were then asked to explain what made them choose each adjec-
tive with reference to illustrative memories, a prompt for episodic memory. They were asked
what they did when they were upset in childhood, a verbal analogue in the AAI for the separ-
ations and reunions experienced by a child in the Strange Situation procedure. Speakers were
also asked whether they could remember being physically held by their parents for comfort
as a child. This question was likewise a verbal analogue, this time for Main’s ‘Aversion to

Physical Contact’ scale for infants, which had proven singularly effective in discriminating
caregivers in avoidant dyads in the Baltimore and Berkeley samples (Chapter 2). Another
question asked by the interview protocol concerned whether speakers had any experiences
of feeling rejected as a child, and if yes, why they now thought their parents had behaved as
they did. Additionally, participants were asked whether they had undergone any major sep-
arations from their parents in childhood, building on Bowlby’s classic concern (Chapter 1).
They were also asked about occasions when parents might have been threatening to them,
following on from Main’s interest in the role of threatening behaviour in the aetiology of
avoidant attachment. In line with a particular concern of Kaplan and Main’s, and the topic of
Bowlby’s most recent book on Loss, participants were questioned about experiences of loss
and bereavement. Participants were also asked about why their parents behaved as they did,
and whether their early experiences of care have affected them as an adult.339
Any one of these questions might be asked of a person in certain contexts in ordinary life.
However, the requirement to respond to this cumulative incision into speakers’ experience
of their attachment relationships was a demanding and potentially distressing challenge. It
cut away the familiar, conventionalised rhythms of interaction that usually offer us mooring
against memory. At the same time, a chance to talk to an interested person about personal
experiences without interruption, exasperation, or contempt is something quite special.
Amidst all its bustle, it is rare in ordinary life that we are freely given sustained and patient
attention by another. Furthermore, the overall tone of exemplary incidents from our child-
hood is something that we have little reason to share as adults, and may have few prompts
to consider in any structured way. There is a quality of matter-of-fact strangeness to the in-
timacy of delivering or receiving an AAI.
Applying the ‘guess and uncover’ approach, the first pattern identified by Goldwyn was
that some parents described their childhoods in flatly glowing terms. There were no prob-
lems; everything was fine. Yet, intriguingly, rather than belonging to the accounts of parents
from securely attached dyads, these accounts of perfect childhoods belonged to the parents
from avoidantly attached dyads. Main and Goldwyn therefore termed this ‘idealisation’.
In a conference paper giving the first public report from this research, Main described a
336 Bowlby, J. (1980) Loss. London: Pimlico, p.62–3.

337 Kaplan’s graduate work was part-funded by the National Center for Clinical Infancy Programs, and she sub-
sequently became a clinician.
338 George, C., Kaplan, N., Goldwyn, R., & Main, M. (1982–83) Attachment interview for parents. PP/Bow/J.4/4.
339 Ibid.
‘mother who stated initially that her mother “was a good one” and that they had “a fine re-
lationship” later in the interview told us—as though spontaneously—that she had painfully
broken her hand as a child. Although she had been in pain for weeks she had not told her
mother because her mother would have been angry. This incident was recounted in almost
the third person—i.e. “but one couldn’t tell her”. When this mother was seen in the Strange
Situation with her infant, the behaviour of the child led to the dyad receiving top scores for
avoidance.340 Another facet of the transcripts of parents from avoidantly attached dyads was
that though the parent asserted an ideal childhood, in fact they could supply few or no con-
crete memories of experiences of intimate and caring interactions. They often seemed to
show little regard for “the need to depend on others” or “recognition of missing and needing
others or being missed and needed by others”.’341
Furthermore, there were indications in the transcripts that at points speakers had experi-

enced rejection in their attachment relationships. Yet this did not seem to inform the speak-
er’s overall characterisation of the relationship, which was resolutely positive. Main ran an
initial test on the first 26 transcripts and found a substantial association between idealisation
of the parent and indications of early experiences of rejection (r = .51).342 She also found a
robust association between reports of early experiences of rejection in the AAI and displays
of aversion to physical contact with the infant in the Clown session: r = .56 for mothers and
r = .63 for fathers.343 However, Main had the qualitative impression that this association was
stronger for speakers where rejection was inferred by the coder, rather than consciously and
clearly reported by the speaker.344 Discussions between Goldwyn and Main identified what
appeared to be the common thread: ‘we may conceive of the AAI (like the Strange Situation)
as creating conditions that arouse and direct attention toward attachment’.345 Just as the in-
fant in an avoidant attachment relationship directed attention and behaviour away from the
caregiver on reunion in the Strange Situation, the parent in interview was directing attention
and discourse away from the actual events of their childhood and associated feelings.346 In
1983. PP/Bow/J.4/4.
1983. PP/Bow/J.4/4.
343 Main, M. (1990) Parental aversion to infant-initiated contact is correlated with the parent’s own rejection
during childhood: the effects of experience on signals of security with respect to attachment. In T.B. Brazelton & K.
Barnard (eds) Touch (pp.461–95). New York: International Universities Press, p.478.
344 Ibid. p.485.
345 This formulation is offered in unpublished texts from the 1980s, but eventually appeared in print in Main,
M. (1995) Recent studies in attachment: overview, with selected implications for clinical work. In S. Goldberg, R.
Muir, & J. Kerr (eds) Attachment Theory: Social, Developmental and Clinical Perspectives (pp.407–70). Hillsdale,
NJ: Analytic Press, p.452.
346 Main, M. & Goldwyn, R. (1984) Predicting rejection of her infant from mother’s representation of her
own experience: implications for the abused–abusing intergenerational cycle. Child Abuse & Neglect, 8(2), 203–
17: ‘Like the rejected infant, the rejected adult woman is expected to organize her attention away from attach-
ment experiences and her feelings regarding those experiences, in an effort to preserve a certain type of mental
organization’ (210–11). From an earlier draft: Main, M. & Goldwyn, R. (1983) Predicting rejection of an infant
from mother’s representation of her own experiences. National Conference on Infant Mental Health, Children’s
Institute International, Los Angeles, February 1983. PP/Bow/J.4/4: ‘In looking away from and ignoring the attach-
ment figure, the infant is excluding attachment-relevant information from further processing. The rejected infant
seems rather to resemble the rejected mother.’
both cases there is a swerve of the heart.347 Main theorised that ‘whether or not the move is
conscious and deliberate, the experience of rejection by the mother in childhood has led to
a shift of attention from attachment figures, experiences and feelings in adult life’.348 Such
experiences of rejection might include a parent’s aversion to physical contact, but could also
be produced by other forms of rebuff, incomprehension, or disparagement. In turn, infants
of these caregivers show avoidant behaviour in the Strange Situation, keeping potential dis-
tress away from their own and their caregiver’s attention. Main wrote to Bowlby, concluding
that ‘it appears that the infant does represent in its behaviour the parents’ life, unconscious
models and intentions’.349
During a visit to Berkeley in December 1982, Ainsworth looked at transcripts together
with Main. They came to the conclusion that the idealised narratives of parents from avoid-
ant dyads represented a reliance on semantic memory, at the expense of episodic memories

that might have qualified such undifferentiated accounts.350 This helped account for why
speakers often reported a lack of memory for specific occasions when asked to substantiate
the positive general account offered of their attachment relationships. Conversations with
Ainsworth in 1982 revealed her qualitative impression that in the Baltimore sample, too,
conversations with the mothers of avoidant dyads suggested idealising of their own child-
hoods.351 Main and colleagues initially termed the speech of these caregivers ‘detached’, since
its defining feature was a minimisation or disavowal of negative experiences, the feelings as-
sociated with these experiences, and of the value of relationships. However, this term from
Bowlby was ambiguous, and Ainsworth thought that it risked connoting a lack of attach-
ment (Chapter 1). Detached speech was therefore renamed ‘dismissing’ (labelled ‘Ds’).352
Another pattern identified by Goldwyn using ‘guess and uncover’ was that several speak-
ers seemed highly concerned with grievances; they remained angry and preoccupied with
their past and present relationships with attachment figures to such an extent that they often
took long conversational turns and lost track of the question. They seemed focused on the
vivid recollection of their childhood attachment relationships. Their goal at times seemed
347 In early support for this conclusion, Dozier and Kobak reported that dismissing speakers distinctively
showed increases in skin conductance levels from baseline in response to questions in the AAI about experiences
of separation, rejection, and threat from attachment figures. Dozier, M. & Kobak, R.R. (1992) Psychophysiology in
attachment interviews: converging evidence for deactivating strategies. Child Development, 63(6), 1473–80.
348 Ibid.
349 Main, M. (1983) Letter to John Bowlby, 15 January 1983. PP/Bow/J.4/4. Main’s correspondence with Robert
Hinde, John Crook, Mary Ainsworth, and John Bowlby is voluminous, as she was in continual contact with them
by letter. She and Hesse are still collating her copy of these letters, so they are not considered in this chapter but will
be the subject of a future article.
350 Main, M. (1982) Letter to John Bowlby, 9 December 1982. PP/Bow/ J.4/4: ‘If you would like some tran-
scripts to look over you would also be welcome. Mary Ainsworth had a wonderful time with them, and a splendid
idea re: semantic vs episodic memory: the A’s seem to have semantic memory.’ On the wider context of atten-
tion to contradictions between forms of memory in the 1980s see Hacking I. (1995) Rewriting the Soul: Multiple
Personality and the Sciences of Memory. Princeton, NJ: Princeton University Press; Danziger, K. (2008) Marking the
Mind: A History of Memory. Cambridge: Cambridge University Press, Chapter 6.
351 Reported as a personal communication in Main, M., Kaplan, N., & Cassidy, J. (1985) Security in infancy,
childhood, and adulthood: a move to the level of representation. Monographs of the Society for Research in Child
Development, 50, 66–104, p.97.
352 One ‘dismissing’ subclassification, Ds2, has been criticised by later researchers as a poor fit for the category.
Speakers in this classification derogate attachment relationships, though the classification can be made even on the
basis of quite brief passages of speech rather than characterising a whole transcript. The subclassification has been
found to fall with the preoccupied category in analyses on two large adult samples assessed using the AAI. Raby,
K.L., Labella, M.H., Martin, J., Carlson, E.A., & Roisman, G.I. (2017) Childhood abuse and neglect and insecure
attachment states of mind in adulthood: prospective, longitudinal evidence from a high-risk sample. Development
& Psychopathology, 29(2), 347–63.
to become that of making a case against their parents, rather than responding precisely to
the interviewer’s questions. In the terms that Main had developed for thinking about the
Strange Situation, these speakers were directing attention towards attachment-relevant cues,
even if this was at the expense of cooperation with the interview. The speech of these care-
givers was initially termed ‘enmeshed-conflicted’ (E); this was changed to ‘preoccupied’ as
the first term was ultimately considered stigmatising, though the category label ‘E’ was re-
tained. Preoccupied speech seemed to Main and Goldwyn to be analogous to the infant of
an ambivalent/resistant dyad in the Strange Situation, who is anxious at even the prospect of
separation, ‘unable to direct attention to the environment, expresses strong and sometimes
continual fear and distress, and seems constantly directed toward the parent’.353 And indeed,
some of the preoccupied speakers in the Berkeley sample had been part of such dyads five
years earlier.

Goldwyn also discerned qualities that seemed to distinguish the speech of parents who
had been part of securely attached dyads five years earlier. These speakers described both the
positives and negatives of their early relationships with good balance, even if these relation-
ships had been difficult. Conversations between Goldwyn, Main, and Hesse suggested that
a central property of the transcripts of secure infants was that the speaker appeared to have
ready access to both semantic and episodic memory and so could coordinate a response to
both kinds of prompts.354 Just like the secure infant in the Strange Situation could turn at-
tention to the toys or the caregiver in response to changing environmental cues, the speakers
seemed to be able to attend to both semantic and episodic memory, and positive and nega-
tive experiences, in a flexible way. In fact, the caregivers of dyads classified B3 in the Strange
Situation in the Berkeley sample displayed these features most prototypically.355 Main,
Goldwyn, and Hesse therefore termed the speech of these caregivers ‘secure-autonomous’,
not because the speaker seemed independently minded, but because the defining feature
of the transcript was the absence of restrictions in the discussion of experiences. The cat-
egory label given to these speakers was F, since they seemed ‘freed through experience or
thought to recognise the importance of attachment relationships, yet independent enough to
evaluate them’.356 This seemed to be in contrast to parents who had been part of avoidant and
resistant dyads, where there seemed to be ‘restrictions of varying types . . . placed on atten-
tion and the flow of information with respect to attachment’.357 In a handwritten annotation
353 Main, M., Kaplan, N., & Cassidy, J. (1985) Security in infancy, childhood, and adulthood: a move to the
level of representation. Monographs of the Society for Research in Child Development, 50, 66–104, p.74. Later re-
search found greater physiological arousal in preoccupied speakers in response to questions about separation and
threat: Beijersbergen, M.D., Bakermans-Kranenburg, M.J., van IJzendoorn, M.H., & Juffer, F. (2008) Stress regula-
tion in adolescents: physiological reactivity during the Adult Attachment Interview and conflict interaction. Child
Development, 79(6), 1707–20, p.1716.
1983. PP/Bow/J.4/4.
355 Main, M. (1991) Metacognitive knowledge, metacognitive monitoring, and singular (coherent)
vs. multiple (incoherent) models of attachment: some findings and some directions for future re-
search. In P. Marris, J. Stevenson-Hinde, & C. Parkes (eds) Attachment Across the Life Cycle (pp.127–59).
New York: Routledge, p.142.
356 George, C., Kaplan, N., Goldwyn, R., & Main, M. (1982–83) Attachment interview for parents. PP/Bow/J.4/
4. Since A, B, and C were the Ainsworth classifications, Main termed her AAI classifications D, E, and F. D stood
for “dismissing’. By the mid-1980s, though, this was confusing since D was also the term used for the new Strange
Situation classification, so dismissing discourse was relabelled ‘Ds’.
on a very early draft of the coding system, Main remarked that in the most prototypical of
secure-autonomous speakers, ‘there is a striking ability to integrate existing information’.358
This developing coding system was clearly unusual. Whereas most psychological meas-
ures for coding interviews focused on coding answers to particular questions, the devel-
oping Main, Goldwyn, and Hesse method for coding the AAI examined the transcript as a
whole, across the dance and drift of spoken discourse. The questions induced speech about
personal experiences without respite: ‘in clinical terms, we would say that the objective of
this interview is to “surprise the unconscious” with respect to attachment, through repeated,
insistent probing’ combined with a lack of reciprocity from the interviewer.359 However, the
most distinctive feature of the AAI as a scientific protocol, as well as perhaps its most auda-
cious and uncanny feature, was its focus on apparent restrictions on information, which put
the coder in the position of comparing ‘the subject’s own semantic categorisations of her ex-

periences’ with what ‘we could discern [of] actual experience’.360
As such, the coding focused neither on the subject’s stated opinions, nor on the individual’s
inferred actual history, since this is unknowable from an interview. The focus was instead on
the manner in which the speaker attended to and communicated about attachment-related
experiences. The coder was asked to consider both the internal consistency of the account
and the extent to which the speaker’s interpretation seemed plausible based on the episodic
information provided.361 In this, the coder had to consider the world from the speaker’s
point of view, and take a step back to evaluate the speaker’s account of the world and appraise
potential distortions of information. In this sense, the coding system had clear analogies to
the clinical interviewing of the period in the USA, placing the coder in the position of clin-
ician. Through the 1970s, under the influence of psychoanalysis, a dominant trend in clinical
interviewing in the context of mental health was to seek to identify psychological defences
or confusions expressed through speech and decipherable by the clinician.362 However, from
the 1980s, under the influence of the American Psychiatric Association’s DSM-III, clinical
interviewing increasingly sought to identify specifiable markers for categories of mental
pathology out of the particularities of what and how individuals report their experiences.363
The AAI had elements of both forms of clinical interview, which made it well suited and
timely for clinicians with a psychoanalytic background in providing a scientific basis and
359 George, C., Kaplan, N., & Main, M. (1985) Adult Attachment Interview, March 1985, 1st edn. PP/Bow/J.4/4.
1983. PP/Bow/J.4/3.
361 This was discussed by Main with reference to the debate in analytic philosophy between the coherence and
correspondence theories of truth. Main, M. (1991) Metacognitive knowledge, metacognitive monitoring, and
singular (coherent) vs. multiple (incoherent) models of attachment: some findings and some directions for fu-
ture research. In P. Marris, J. Stevenson-Hinde, & C. Parkes (eds) Attachment Across the Life Cycle (pp.127–59).
New York: Routledge, p.144; Main, M. (1993) Discourse, prediction and recent studies in attachment: implications
for psychoanalysis. Journal of the American Psycho-analytic Association, 41, 209–44, p.237.
362 E.g. Edinburg, G.M., Zinberg, N.E., & Kelman, W. (1975) Clinical Interviewing and Counselling: Principles
and Techniques. New York: Appleton-Century-Crofts.

363 Greco, M. (2016) What is the DSM? Diagnostic manual, cultural icon, political battleground: an overview
with suggestions for a critical research agenda. Psychology & Sexuality, 7(1), 6–22. A wider context can be given
also in the rise of assessment and classification of the speech of service-users as a mode of knowing and adminis-
trating within the professions from the 1970s to the 1980s, and with a particular concern with individuals able—or
not able—to know and regulate themselves. The DSM contributed to this shift, but was also part of a wider pro-
cess. Foucault, M. (2008) The Birth of Biopolitics: Lectures at the College de France, 1978–1979, trans. G. Burchell.
New York: Palgrave MacMillan.
classifications for blockages or disruptions in how individuals are able to know or regulate
themselves.
Main and colleagues felt that ultimately ‘we were attempting to trace what could happen
to information regarding negative or rejecting attachment experiences, other than their easy
and coherent recognition and evaluation’.364 On the one hand, it could be defensively ex-
cluded through lack of memory and/or idealisation. On the other hand, it could become a
preoccupying focus of attention in a way that hindered evaluation.365 Main described the
object of the interview as the adult’s state of mind with respect to attachment.366 However,
the phrase is quite misleading. Not least what is in question is not what psychological re-
search generally means by a ‘state’: a transitory experience responding to an external prompt
without the expectation of stability over time.367 Unfortunately, the pivotal article in which
Main set out and described the concept of ‘state of mind’ was accepted by the journal

Developmental Psychology but then withdrawn by Main as ultimately unsatisfactory.368 As
such, the meaning of the term has often been unclear to her readers, contributing to mis-
understanding of what the AAI actually measured, and fanned the flames of controversy
regarding other forms of assessment (Chapter 5). Fonagy and colleagues, for instance, were
left at a loss regarding the object of the AAI, wondering whether the coding system could
actually be best considered as an indirect assessment of the caregiving system, rather than
an assessment of anything attachment-related at all.369 Fonagy and Campbell subsequently
concluded that the AAI does assess something related to attachment, but that Main’s own
description of ‘state of mind regarding attachment’ in terms of internal working models has
been confusing and, ultimately, a ‘reductionist over-simplification’.370 However, Fonagy and
Campbell offered no further thoughts on what it is that the AAI does, in fact, measure.
In addressing this question, a first thing to note is that, despite its name, the AAI was
not an assessment of individual differences in a unitary entity called ‘attachment’.371 This is
despite the fact that, among developmental psychologists, ‘adult attachment’ has come to
mean ‘what the AAI assesses’. In fact, the name ‘Adult Attachment Interview’ appears to have
1983. PP/Bow/J.4/3.
365 George, C., Kaplan, N., Goldwyn, R., & Main, M. (1982–83) Attachment interview for parents. PP/Bow/
J.4/4: ‘The individual is not at all freed of the influence of early attachment relationships. He/she is unable to grow
beyond them, and either accepts this state passively, or actively struggles against it without success. Although the
individual may talk very readily and at length of the influence of early relationships, he/she is not independent
enough to evaluate them and his/her place within them.’
366 The underspecification of the concept of ‘state of mind regarding attachment’ might be seen as in part an ef-
fect of the importation of aspects of a clinical interview into the AAI methodology, but without the conventional
aim of the interview—which was changing in any case in the period, from the identification of defences to the
making of diagnoses.
367 On the concept of ‘state’ in psychology see e.g. Chaplin, W.F., John, O.P., & Goldberg, L.R. (1988) Conceptions
of states and traits: dimensional attributes with ideals as prototypes. Journal of Personality and Social Psychology,
54(4), 541–57.
368 Main, M. & Goldwyn, R. (1989) Interview Based Adult Attachment Classifications: Related to Infant–Mother
and Infant–Father Attachment. Unpublished manuscript, Developmental Psychology.

369 Stein, H., Jacobs, N.J., Ferguson, K.S., Allen, J.G., & Fonagy, P. (1998) What do adult attachment scales
measure? Bulletin of the Menninger Clinic, 62, 33–82, p.49. Worry about the object or objects of the AAI was also
expressed by Stern, who reviewed various possibilities, but acknowledged that ultimately this remained an un-
solved question. Stern, D.N. (1998) The Motherhood Constellation: A Unified View of Parent–Infant Psychotherapy.
London: Karnac, p.38.
370 Fonagy, P. & Campbell, C. (2015) Bad blood revisited: attachment and psychoanalysis. British Journal of
Psychotherapy, 31(2), 229–50, p.234.

371 Goldberg, S. (2000) Attachment and Development. London: Routledge, p.242.
originated from the interview’s concern with attachment experiences. True, the demand to
attend to and communicate about attachment-relevant experiences might well be alarming,
and might therefore also activate the attachment system. But this was not tested or even dis-
cussed. In fact, the central focus of the coding system developed by Main, Goldwyn, and
Hesse was on the capacity to attend to and communicate about attachment-relevant experi-
ences and the feelings they evoke. Main referred to this, somewhat loosely, as individual
differences in ‘adult attachment’, since this capacity seemed to her to have strong common-
alities with the Ainsworth classifications at the level of attentional processes. The term ‘adult
attachment’ was also a powerful assertion that the ideas of Ainsworth were relevant across
the lifespan. However, use of the term generated confusion, as well as acrimony between
developmental and social psychology researchers in which ownership over the capacity to
measure something called ‘attachment’ became a central stake (Chapter 5). However, Main

and Hesse’s basic position is that ‘there exist species-wide abilities that are not part of the
attachment system itself, but can, within limits, manipulate (either inhibit or increase) at-
tachment behavior in response to differing environments’.372 In Main and Hesse’s conceptu-
alisation of the AAI, what is measured is the deployment of such abilities as applied to adult
attention and communication regarding attachment-related matters, not the ‘attachment
system itself ’.
Nor was the AAI developed primarily as an assessment of ‘internal working models’, if by this
is meant representational content regarding attachment figures. (Admittedly, if ‘internal working
model’ is taken to mean any cognitive components recruitable by the attachment system, then
the AAI does tap these, but the claim does not tell us much.) Comparison of published with un-
published manuscripts suggests that Main’s use of the concept of ‘internal working model’ in the
mid-1980s in discussing the AAI was primarily a matter of tying her new methods to existing
theory, rather than actually the basis on which she was conceptualising the measure. It was also a
position she would drop already in the 1980s. At the meeting of the Society for Research in Child
Development in Baltimore in June 1987, Main spent time with Hinde. Hinde had huge respect
for Main’s work, as well as a good deal of personal affection. However, underspecified causal
claims about motivational processes seems to have been his personal bête noire. At the Baltimore
conference he passionately chided Main for her appeal to the ‘internal working model’ concept.
Main wrote to Bowlby to report that she could see Hinde’s point, and that she was plan-
ning to give the matter further thought.373 From 1987 onwards, she ceased equating states
of mind regarding attachment and internal working models. And in the late 1990s Main
acknowledged in print Hinde’s point that the characterisation of the AAI as an assessment
of internal working models was, ultimately, ‘misleading and unwarranted’.374 Fonagy and
Target agreed, arguing that Main’s description of the AAI as assessing internal working mod-
els had prompted much ‘futile’ research and generally ‘distracted attachment researchers’.375
at 1, 6, and 19 years of age: the Berkeley Longitudinal Study. In K.E. Grossmann, K.Grossmann, & E. Waters (eds)
373 Main, M. (1987) Letter to John Bowlby, 3 June 1987. PP/Bow/B.3/36/1.
374 Main, M. (1999) Epilogue. Attachment theory: eighteen points with suggestions for future studies. In J.
Cassidy & P. Shaver (eds) Handbook of Attachment (pp.845–87). New York: Guilford, p.877.
375 Fonagy, P. & Target, M. (2002) Early intervention and the development of self-regulation. Psychoanalytic
Inquiry, 22(3), 307–35, p.328. For an example of confusion caused by Main’s characterisation see, for instance, the
widely cited paper Johnson, S.C., Dweck, C.S., & Chen, F.S. (2007) Evidence for infants’ internal working models of
attachment. Psychological Science, 18(6), 501–502. The authors found a difference in attentional processes between
infants based on Strange Situation classifications. From this they conclude, in an unmonitored non sequitur, that
They were glad that she had altered her description of the AAI. However, this shift in Main’s
work has generally gone unnoticed and unheeded. Kobak and Esposito described how
Main’s appeal to internal working models inadvertently trained researchers’ attention on
personality trait-style qualities of individuals and their ‘models’ of others, rather than on at-
tentional or communicative processes.376 The characterisation of the AAI as an assessment
of ‘models’ of attachment relationships led to the misleading characterisation of the measure
as an assessment of ‘attachment representations’. The AAI was not primarily an assessment
of individual differences in representations held of particular parents, or even in representa-
tions of relationships in general, though both might influence classifications on the measure.
Yet thanks to its initial characterisation in terms of internal working models, the AAI is still
commonly described, including by trainers in the measure, as an assessment of ‘attachment
representations’ in adults. This occurs even when researchers’ own descriptions of what the

AAI assesses—for instance, the processing of attachment-relevant information—actually do
not agree with the usual meanings of the noun ‘representations’.377
The term ‘representations’ essentially functions as a placeholder. But its use has unsurpris-
ingly contributed to unwarranted assumptions about the measure. For instance, researchers
have often wondered why adults are assigned a single classification, rather than a classifi-
cation for each attachment relationship, as could be anticipated if the AAI was measuring
representations of relationships.378 The use of a single classification makes sense, however,
if Main’s focus on attentional and communicative processes in the years of its development
is recognised: in her work, the AAI has been conceptualised as an assessment of individual
differences prompted by the interview’s demand on speakers to attend to and communicate
regarding attachment-relevant experiences. The only researcher to have consistently high-
lighted the importance of Main’s attentional model of states of mind regarding attachment
has been van IJzendoorn. Van IJzendoorn had several advantages over other readers. For in-
stance, in a paper from 1992, and citing his personal copy of the unpublished Developmental
Psychology manuscript on the concept of ‘states of mind’, van IJzendoorn observed that gen-
erally in Main’s writings ‘the concepts of “internal representation,” “state of mind,” and “in-
ternal working model” of attachment are used interchangeably’. Nonetheless, he argued, the
‘concept of “state of mind” ’ is in fact less about representational content and more about ‘the
direction and organization of attention and memory’, as well as related affective and behav-
ioural aspects.379 Recognition of the importance assigned by Main to attention has gone on
this is proof of abstract mental representations of attachment figures: ‘Secure infants looked relatively longer at the
unresponsive outcome than the responsive outcome compared with the insecure infants. These results constitute
direct positive evidence that infants’ own personal attachment experiences are reflected in abstract mental repre-
sentations of social interactions’ (502).
376 Kobak, R. & Esposito, A. (2004) Levels of processing in parent–child relationships: implications for clin-
ical assessment and treatment. In L. Atkinson & S. Goldberg (eds) Attachment Issues in Psychopathology and
Interventions (pp.139–66). Mahwah, NJ: Lawrence Erlbaum, p.143.
377 Verhage, M.L., Fearon, R.P., C. Schuengel, et al. (2018) Examining ecological constraints on the intergenera-
tional transmission of attachment via individual participant data meta-analysis. Child Development, 89(6), 2023–
37; Messina, S., Reisz, S., Hazen, N., & Jacobvitz, D. (2019) Not just about food: attachments representations and
maternal feeding practices in infancy. Attachment & Human development, 23 April, 1–20.
378 It has sometimes been assumed that the AAI aimed to measure internal working models, but was faulty in
its execution of this aim. For a later attempt to construct and validate such a measure see Miljkovitch, R., Moss, E.,
Bernier, A., Pascuzzo, K., & Sander, E. (2015) Refining the assessment of internal working models: the Attachment
Multiple Model Interview. Attachment & Human Development, 17(5), 492–521.
379 Van IJzendoorn, M.H. (1992) Intergenerational transmission of parenting: a review of studies in non-
clinical populations. Developmental Review, 12(1), 76–99, p.80. Van IJzendoorn later even punned that the
‘Move to the Level of Representation’ was itself a ‘revolutionary shift in attention’ for the field of attachment
to guide van IJzendoorn’s research using the Strange Situation and AAI, for instance his con-
cern with genetic polymorphisms that influence the dopamine system, given the role of this
system in attention and reward processing.380
Coherence
From early on in their examination of the transcripts, a central concept used by Main and
Goldwyn to mark individual differences between the transcripts was ‘coherence’. As van
IJzendoorn and colleagues noted, however, the Main and Goldwyn usage was technical, and
differed in potentially confusing ways from everyday English.381 Indeed, a recent study by
Lind and colleagues found an association of only r = .37 between Main and Goldwyn’s scale

for ‘coherence’ and a conventional measure of narrative coherence.382 Main’s first use of the
term was in the first edition of the coding manual, in 1986, where they defined coherence
as ‘the extent to which the reader finds a unified, yet free-flowing picture of the subject’s
experiences, feelings and viewpoints within the interview, such that none of these require
the reader to make her own, differing interpretation. In addition, the interview transcript
is considered coherent when the subject seems able to easily point to the principle and ra-
tionale behind her responses; has thought or else seems ready to think about her past and its
influences.’383 In judging coherence, the manual urged the coder to look out for contradic-
tions, especially between episodic and semantic memory, logical contradictions, and factual
contradictions.
In the early 1990s, Main and Hesse continued to develop the concept of ‘coherence’ in re-
flecting on the question of what the coding system was capturing. In this, their thinking was
influenced by the ideas of a colleague at Berkeley, the philosopher of language Paul Grice.384
Grice had argued that cooperative discourse usually demands, except when permission is
given by the partner, adherence to four conversational maxims:
research, away from behaviour and towards the manner in which attachment-relevant experiences are com-
municated: van IJzendoorn, M.H. (1995) Adult attachment representations, parental responsiveness, and in-
fant attachment: a meta-analysis on the predictive validity of the Adult Attachment Interview. Psychological
Bulletin, 117(3), 387–403, p.388. Other researchers close to Main and with access to her unpublished works
have described ‘state of mind’ as referring to ‘the way adults process attachment-related thoughts, memories
and feelings’. Again, the difference from the ready connotations of the term ‘attachment representations’ is strik-
ing. Dozier, M. & Bates, B.C. (2004) Attachment state of mind and the treatment relationship. In L. Atkinson
& S. Goldberg (eds) Attachment Issues in Psychopathology and Intervention (pp.167–80). London: Lawrence
Erlbaum, p.167.
380 E.g. Van IJzendoorn, M.H. & Bakermans-Kranenburg, M.J. (2006) DRD4 7-repeat polymorphism moder-
ates the association between maternal unresolved loss or trauma and infant disorganization. Attachment & Human
381 Beijersbergen, M.D., Bakermans-Kranenburg, M.J., & van IJzendoorn, M.H. (2006) The concept of coher-
ence in attachment interviews: comparing attachment experts, linguists, and non-experts. Attachment & Human
Development, 8(4), 353–69. Morelli and Rothbaum also argued that the concept of coherence is defined in dif-
ferent ways in different cultures; it is not a cultural universal. Morelli, G. & Rothbaum, F. (2007) Situating the child
in context: attachment relationships and self-regulation in different cultures. In S. Kitayama & D. Cohen (eds)
Handbook of Cultural Psychology (pp.500–527). New York: Guilford.
382 Lind, M., Vanwoerden, S., Penner, F., & Sharp, C. (2019) Narrative coherence in adolescence: relations with
attachment, mentalization, and psychopathology. Journal of Personality Assessment, https://doi.org/10.1080/

00223891.2019.1574805.
384 Grice, H.P. (1989) Studies in the Way of Words. Cambridge, MA: Harvard University Press.
Quality—‘be truthful and have evidence for what you say’

Quantity—‘be succinct, and yet complete’
Relation—‘be relevant to the topic as presented’
Manner—‘be clear and orderly’
Main partly retained her earlier definition of ‘coherence’, but also partly redefined it in terms
of Grice’s maxims.385 Whereas initially the term had been defined as (i) the extent to which
the reader finds a unified, yet free-flowing picture by the speaker that agrees with the read-
er’s own account, the concept was tucked into (ii) the four dimensions identified by Grice.
Characteristic of a secure-autonomous (F) transcript was that there was good episodic evi-
dence for semantic generalisations; speakers could flexibly turn their attention to their past
and to the interviewer in being succinct and yet complete and relevant: ‘For these speakers,

the focus of attention appears to shift fluidly between the interviewer’s queries and the mem-
ories that are called upon. In this way, the speaker is able, regardless of the difficulty of the
subject matter, to remain both truthful (consistent) and collaborative.’386 Finally, the dis-
course of a secure-autonomous speaker was characteristically clear and orderly, often with
a fresh quality to the speech. These qualities require a capacity to both turn genuine and
curious attention to attachment-related memories and let go again in order to attend to the
interviewer. Main and colleagues developed subscales to help coders in making a judge-
ment about whether a transcript should be classified as secure, parallel to the function of
Ainsworth’s scales for the Strange Situation (Chapter 2). Of particular importance was a
scale for the coherence of the transcript.387
Dismissing (Ds) interviewees tended to have poor episodic evidence for their semantic
generalisations. They also tended to offer oversuccinct answers lacking the details of con-
crete experience, since they were not able to resource their responses through access to a
richly populated and textured inscape of episodic information. Presumably this information
had become defensively excluded. Relevance might also be weakly violated as the speaker
deflected the conversation away from sensitive topics. In general, however, what is said is
orderly. Preoccupied (E) interviewees tended to have poor alignment between episodic and
semantic memory, as they were focused on the negative aspects of episodic memories even
as they recounted events that could have other more balanced interpretations. However,
the primary violations of coherence in a preoccupied transcript were excessive quantity of
speech about topics that were not direct answers to the questions of the interview, often with
poor orderliness in the communication of experiences caused by a focus on the memories
themselves rather than on the comprehension of the interviewer. Their transcripts are a bit
wild, marked by ‘highly entangled, confusing, run-on sentences; failures to use past markers
in quoting conversations with the parents; rapid oscillations of viewpoint within or between
385 Main, M. (1991) Metacognitive knowledge, metacognitive monitoring, and singular (coherent) vs. multiple
(incoherent) models of attachment: some findings and some directions for future research. In P. Marris, J. Stevenson-
Hinde, & C. Parkes (eds) Attachment Across the Life Cycle (pp.127–59). New York: Routledge: ‘Coherence appears
both in an analysis based upon Grice’s maxims with respect to coherence of discourse (Grice 1975), and in terms of
overall plausibility’ (129).
386 Hesse, E. (1996) Discourse, memory, and the Adult Attachment Interview: a note with emphasis on the
emerging cannot classify category. Infant Mental Health Journal, 17(1), 4–11, p.6.
387 Secure-autonomous transcripts are also expected to demonstrate a range of feelings appropriate to the com-
plexity of the autobiography being related. This was not developed as a scale by Main and colleagues, but does
feature as a scale in a version of the AAI appropriate for young people: Steele, H., Steele, M., & Kriss, A. (2009) The
Friends and Family Interview (FFI) Coding Guidelines. Unpublished manuscript.
sentences; unfinished sentences; insertion of extremely general terms into sentence frames
(‘sort of thing’, ‘and this and that’), and use of nonsense words or trailers as sentence endings
(‘dada-dada-dada’).’388 Main and Hesse also created scales to help coders identify dismissing
and preoccupied transcripts. To help distinguish dismissing states of mind regarding attach-
ment, the researchers created scales including for idealisation of the parent and insistence
on lack of recall. To help distinguish preoccupied states of mind regarding attachment, they
created scales for involving/preoccupied anger and for passivity or vagueness of discourse.
Like the Ainsworth scales, the scales for the AAI were initially given the primary role of
supporting placement of cases within categories. As with the Strange Situation, coders were
enjoined to record their scale scores, resulting in an archive of largely unpublished findings.
These have become the target of great interest recently, especially in the context of Individual
Participant Data meta-analysis (Chapter 6). However, from the 1990s onwards, the coher-

ence score came to see use as a dimensional alternative, or complement, to the category-
based coding system.389
Hesse reflected upon the meaning of coherence at an interpersonal level.390 He concep-
tualised the AAI as constituting two tasks: to reflect on memories of attachment-relevant
experiences, and to communicate about these in a way that holds in mind the interviewer’s
specific questions. Transcripts classified as dismissing violate the first of these tasks; tran-
scripts classified as preoccupied violate the second. In the same period, Main was reflecting
on the meaning of coherence at a cognitive level. She proposed that the coherence of their
discourse suggests that secure-autonomous speakers have a relatively unitary model of their
experiences and how these influence their behaviour. She perceived that this integration of
different sources of information and the lack of need to divert attention to the implementa-
tion of a conditional strategy might support a capacity to retain balance in the evaluation of
388 Main, M. (1991) Metacognitive knowledge, metacognitive monitoring, and singular (coherent) vs. mul-
tiple (incoherent) models of attachment: some findings and some directions for future research. In P. Marris, J.
Stevenson-Hinde, & C. Parkes (eds) Attachment Across the Life Cycle (pp.127–59). New York: Routledge, p.144.
One group of transcripts showed substantial markers of splintering or incoherence. These speakers were preoccu-
pied not so much by anger towards their caregiver, as by ‘fearful attachment experiences—for example, experi-
ences of physical or sexual abuse, traumatic loss, psychosis in a parent, or simple cruelty. There is evidence within
the interview of active struggle with these experiences, but the subject is still implicitly fearful, confused or over-
whelmed. The subject is not yet objective, or able to gather these chaotic and fearful episodes of experiences into
a single abstract yet personally meaningful form.’ Main gave these their own subclassification: ‘fearful’ (labelled
E3). In some unpublished work, Main included E3 as a marker of unresolved/disorganised states of mind, e.g.
Main, M., van IJzendoorn, M.H., & Hesse, E. (1993) Unresolved/Unclassifiable Responses to the Adult Attachment
Interview: Predictable from Unresolved States and Anomalous Beliefs in the Berkeley–Leiden Adult Attachment
Questionnaire. Unpublished manuscript. https://openaccess.leidenuniv.nl/dspace/bitstream/1887/1464/1/168_
131.pdf. Several subsequent researchers criticised the inclusion of E3 as a form of preoccupation in the AAI
system, e.g. George, C. & West, M.L. (2012) The Adult Attachment Projective Picture System. New York: Guilford,
p.194. One interpretation would be that E3 speech, whilst not technically itself an unresolved/disorganised state
of mind, meets the conditions that especially produce lapses: (i) extensive speech and (ii) about traumatic experi-
ences, (iii) in a manner guided especially by the memories themselves rather than with relevance and order set by
the interviewer’s questions.
389 E.g. Fonagy, P., Steele, H., & Steele, M. (1991) Maternal representations of attachment during pregnancy
predict the organization of infant–mother attachment at one year of age. Child Development, 62, 891–905. A spur
to use of the coherence score as an alternative to the category-based system came with a discriminant function
analysis conducted by Crowell and colleagues, which found that the coherence score was of special importance
for secure/insecure discrimination. Crowell, J.A., Treboux, D., Gao, Y., Fyffe, C., Pan, H., & Waters, E. (2002)
Assessing secure base behavior in adulthood: development of a measure, links to adult attachment representa-
tions, and relations to couples’ communication and reports of relationships. Developmental Psychology, 38(5),
679–93.
390 Hesse, E. (1996) Discourse, memory, and the adult attachment interview: a note with emphasis on the emer-
ging cannot classify category. Infant Mental Health Journal, 17(1), 4–11.
relationships, to perceive how the same reality could be seen in different ways, and to allocate
attention as needed to different tasks.391
Main was also impressed by apparent differences in epistemology between speak-
ers: secure-autonomous speakers ‘adopted a more thoroughly constructivist view of their
own knowledge-base than less secure adults’, with more subtle forms of awareness of the
appearance-reality distinction. Secure-autonomous speakers may acknowledge that their
own perspective on an event may differ to that of a family member, or that their recall may
be distorted by their regrets about the occurrence.392 She believed that this constructivist
view of knowledge served to support a speaker’s capacity to examine their own experiences
(‘metacognition’) when prompted by the environment, since there is less segregated infor-
mation: ‘more epistemic ‘space’’ is available for such an individual ‘because her thinking pro-
cesses are not compartmentalised’.393 Though Main acknowledged that it could be, in theory,

that metacognitive capacities were facilitating secure-autonomous speech, or that there were
bidirectional effects, her view was that secure attachment was the cause of metacognitive
strengths such as clearsightedness in self-evaluation. A metacognitive monitoring scale was
added to the AAI coding manual in 1991 to support coders in making a classification of
secure-autonomous state of mind regarding attachment. (Nearly three decades later, how-
ever, the scale is still identified as unfinished in the AAI manual. The measurement of meta-
cognitive monitoring would not be pursued further by Main, and would instead be taken up
by Fonagy, Steele, and Steele in their work on reflective functioning and mentalisation.394
Main’s lack of further work on the metacognition scale has generally been interpreted as tacit
endorsement of the approach of Fonagy and colleagues.)
Main conceptualised the reduced coherence of dismissing and preoccupied speakers as
reflecting the conflict between segregated information, and thus conflict between different
interpretations by speakers of their own past. Everyone has multiple models of reality, Main
observed, but dismissing and preoccupied speakers seemed to have ‘implicitly contradictory
models of the same aspects of reality’, segregated from one another perhaps without the
speaker’s awareness, but available to someone reading the transcript of their interview.395 On
this basis, Main criticised Bowlby’s notion of the internal working model as too calm and set-
tled a metaphor, to the point of actually being ‘somewhat misleading’. Whilst perhaps applic-
able to secure-autonomous speakers, when asked to describe and evaluate their attachment
experiences and relations, dismissing and especially preoccupied speakers present an array
of ‘contradictory thoughts, feelings, and intentions which can only loosely be described as a
“model” ’.396 These speakers ‘evidence difficulties in obtaining access to attachment-related
391 Main reported unpublished data in support of this conclusion: in a study of 174 college students conducted
in collaboration with Waters, ‘self-reported difficulty dividing attention among several simultaneous tasks was
found associated with lack of memory for childhood, with descriptions of the subject’s mother as unforgiving,
and with uncertainty that the subject could turn to one or both parents in times of trouble’. Main, M. (1991)
Metacognitive knowledge, metacognitive monitoring, and singular (coherent) vs. multiple (incoherent) models of
attachment: some findings and some directions for future research. In P. Marris, J. Stevenson-Hinde, & C. Parkes
(eds) Attachment Across the Life Cycle (pp.127–59). New York: Routledge, p.155.
392 Ibid. p.153.
393 Ibid. pp.146–8 and Table 8.1.
394 Fonagy, P., Steele, M., Steele, H., Moran, G.S., & Higgitt, A.C. (1991) The capacity for understanding mental
states: the reflective self in parent and child and its significance for security of attachment. Infant Mental Health
Journal, 12(3), 201–18.
Stevenson-Hinde, & C. Parkes (eds) Attachment Across the Life Cycle (pp.127–59). New York: Routledge, p.132.
396 Ibid. p.132.
information; in maintaining organisation in attachment-related information; and in pre-

venting attachment-related information from undergoing distortion’.397 Whereas psycho-
analytic theory might conceptualise such effects as evidencing the repression of mental
contents relevant to past experiences, Main argued that what is ‘defensively excluded in this
conceptualisation is not the memory . . . but rather an alternative attentional and behavioural
patterning’.398
Just as Main conceptualised the infant showing avoidant behaviour in the Strange
Situation as primarily avoiding disorganisation, and only secondarily the attachment figure,
so dismissing speakers were conceptualised as primarily avoiding a breakdown in their con-
ditional strategy and its attentional and behavioural architecture. And like the ambivalent/
resistant infant, the preoccupied speaker was conceptualised as keeping attention and be-
haviour trained on attachment-relevant information, avoiding cues from the environment

that might suggest other priorities for attention and behaviour. Whether in childhood or
in adulthood, what conditional strategies have in common in Main and Hesse’s thinking is
that they ‘follow upon alterations in the focus of attention’.399 For it is attention that orches-
trates what attachment-relevant information and cues are received from the environment or
from memory. As such, these alterations in attachment-relevant information are topic spe-
cific. They were not anticipated to be carried over into the processing of matters irrelevant to
secure base/safe haven availability.400
Yet alterations in attention to these specific matters were nonetheless anticipated by Main
and colleagues to have wide-ranging implications for mental health. In line with this account,
both dismissing and preoccupied states of mind have been found to be transdiagnostic.
Though more prevalent in clinical samples, dismissing and preoccupied states of mind re-
garding attachment are not characteristic simply of any one form of mental illness (with the
partial exception of a link between preoccupied states of mind and personality disorders).
They are also remarkably stable during adulthood, though periods of developmental transi-
tion or changes in circumstances will often result in changes in states of mind regarding at-
tachment.401 However, contrary to Main’s expectation, there has only been evidence of weak
or moderate continuity from infant attachment classifications to AAI classifications, though
infant disorganised attachment does generally predict a later insecure classification on the
397 Ibid. p.143.

398 Main, M. (1993) Discourse, prediction and recent studies in attachment: implications for psychoanalysis.
Journal of the American Psycho-analytic Association, 41, 209–44, p.234.
399 Main, M. & Hesse, E. (1992) Attaccamento disorganizzato/disorientato nell’infanzia e stati mentali dissociati
dei genitori. In M. Ammaniti & D.N. Stern (eds) Attaccamento e Psicoanalisi (pp.80–140). Rome: Gius, Laterza &
Figli, p.86.
400 See Bakermans- Kranenburg, M.J. & van IJzendoorn, M.H. (1993) A psychometric study of the Adult
Attachment Interview: reliability and discriminant validity. Developmental Psychology, 29(5), 870–79; Sagi, A., van
IJzendoorn, M.H., Scharf, M., Koren-Karie, N., Joels, T., & Mayseless, O. (1994) Stability and discriminant validity
of the Adult Attachment Interview: a psychometric study in young Israeli adults. Developmental Psychology, 30(5),
771–7; Crowell, J.A., Waters, E., Treboux, D., et al. (1996) Discriminant validity of the Adult Attachment Interview.
Child Development, 67(5), 2584–99. The boundaries of ‘attachment-relevant information’ are not set out by Main
and colleagues.
401 Bakermans-Kranenburg, M.J. & Van IJzendoorn, M.H. (1993) A psychometric study of the Adult Attachment
Interview: reliability and discriminant validity. Developmental Psychology, 29(5), 870–79; Crowell, J.A. & Hauser,
S.T. (2008) AAIs in a high-risk sample: stability and relation to functioning from adolescence to 39 years. In H.
Steele & M. Steele (eds) Clinical Applications of the Adult Attachment Interview (pp.341–70). New York: Guilford.
There are ongoing discussions about the extent to which AAI classifications should be expected to be stable in the
context of psychotherapy, e.g. Daniel, S.I.F., Poulsen, S., & Lunn, S. (2016) Client attachment in a randomized clin-
ical trial of psychoanalytic and cognitive-behavioral psychotherapy for bulimia nervosa: outcome moderation and
change. Psychotherapy, 53(2), 174.
AAI.402 The question of the relationship between actual childhood experiences in attach-
ment relationships and retrospective discourse in the AAI is one that remains debated, as the
case of discussions of ‘earned security’ reveals especially clearly.
Earned security
In her use of ‘guess and uncover’ to identify forms of adult discourse associated with in-
fant secure attachment, Goldwyn identified the importance of the speaker’s ability, in the
present, ‘to take a balanced view of relationships’. However, the personal history recounted
by such speakers seemed to have one of two forms. A first was ‘a believable picture of one
or both parents serving as a secure base or haven of safety in childhood, a picture which

is not contradicted within the interview and which may even be illustrated by incidents of
parental giving of comfort or support’.403 Narrated in such accounts, it seemed to Main and
colleagues, was secure-autonomous speech as a consequence of relatively unbruised and
secure attachment relationships in childhood. Yet several transcripts in the development
sample revealed speech that was about difficult childhoods, but marked by flexible attention
to the good and the bad, to memories of the past and to communication with the interviewer
in the present, and to their own perspective and that of others. For example, one mother
who had experienced rejection by her family responded to the researchers’ initial query re-
garding the nature of her early relationships: ‘how many hours do you have? I have one of
those families that they should write a whole book about. Okay, well to start with, my mother
was not cheerful, and I could tell you right now the reason was that she was overworked.’404
Though this mother had experienced rejecting caregiving, she seemed to Goldwyn and Main
nonetheless able to acknowledge negative aspects of the relationship, with insight into both
her own and her parents’ experiences. When this mother and her infant had been seen in
the Strange Situation five years earlier, they were classified as B3, and the dyad received the
lowest possible score on the Ainsworth avoidance scale.405
In the first version of the coding system from 1982–83, Main and Goldwyn identified
three trajectories that seemed to be linked, despite a caregiver’s own difficult history of care,
to secure-autonomous speech and a secure attachment relationship in the infant–caregiver
relationship. A first was that some speakers seemed to have ‘actively engaged in a period of
rebellion and escape from the parents’, moving away from these relationships as an oppor-
tunity to reorganise how they respond to distress in the context of relationships.406 A second
trajectory seemed to be that some speakers had ‘forgiven the parents’, holding in mind both
their own perspective and feelings about events but also articulating a sense of the difficulties
402 Grossmann, K.E., Grossmann, K., & Waters, E. (eds) (2005) Attachment from Infancy to Adulthood: The
Major Longitudinal Studies. New York: Guilford; Pinquart, M., Feußner, C., & Ahnert, L. (2013) Meta-analytic
evidence for stability in attachments from infancy to early adulthood. Attachment & Human Development, 15(2),
189–218.
405 Main, M. & Goldwyn, R. (1984) Predicting rejection of her infant from mother’s representation of her own
experience: implications for the abused–abusing intergenerational cycle. Child Abuse & Neglect, 8(2), 203–17,
p.215–16.
their own parents had faced.407 A third trajectory, perhaps predisposed by the fact that
Main’s sample was from near the University of California, Berkeley campus, was that some
had ‘engaged in a period of study undertaken with a view to understanding child–parent
relationships and their influence’.408 What all three trajectories seemed to have in common
was the effortful achievement of a new perspective. By 1988, Main had come to refer to this
as ‘earned security’.
The first study to discuss earned security in print and to study it empirically was con-
ducted by Main and Hesse’s colleagues at the University of California, Carolyn and Philip
Cowan, and published in 1994. The Cowans were conducting a longitudinal study of the
transition to parenthood, and the AAI was administered to 40 adults when the first-born
children in the study were 42 months. In the sample, 10 speakers were classified as insecure,
10 as ‘continuous-secure’, and 20 as ‘earned secure’. Earned security was defined in prac-

tice by the Cowans as speakers who were classified as secure-autonomous in speech, but
whose transcripts also indicated a high score for neglecting and rejecting behaviour by ei-
ther caregiver, and a comparatively low score for loving behaviour by at least one caregiver.
The Cowans found that the half of their sample who were judged to be continuous-secure
speakers reported lower levels of depressive symptoms than the insecure or earned-secure
speakers. However, observations of child–caregiver interaction indicated that the earned-
secure speakers and continuous-secure speakers offered the same degree of warmth and
structure as their children played, in contrast to a lower degree of warmth and structure
offered by speakers who had been classified as insecure on the AAI.409 The image of individ-
uals able to ‘earn’ the benefits of secure early relationships through their own efforts in later
childhood, adolescence, or adulthood has naturally been one with great appeal, especially
in the American clinical community.410 And indeed, researchers found that earned-secure
speakers are more likely to have spent time in psychotherapy than continuous-secure, pre-
occupied, or dismissing speakers.411 The classification also offered attachment researchers
a powerful answer to accusations that attachment theory is concerned only with continuity
from infancy to adult relationships. Moreover, it supported the application of the AAI as an
outcome measure for evaluating psychotherapeutic interventions.412
Roisman and Sroufe, however, felt that there was a need to ‘be cautious about retrospective
reports’.413 They cited Main and Goldwyn who had observed that it cannot be ‘presumed
407 Ibid.
408 Ibid.
409 Pearson, J.L., Cohn, D.A., Cowan, P.A., & Cowan, C.P. (1994) Earned-and continuous-security in adult
attachment: relation to depressive symptomatology and parenting style. Development & Psychopathology, 6(2),
359–73.
410 E.g. Guina, J. (2016) The talking cure of avoidant personality disorder: remission through earned-secure at-
tachment. American Journal of Psychotherapy, 70(3), 233–50.

411 Caspers, K.M., Yucuis, R., Troutman, B., & Spinks, R. (2006) Attachment as an organizer of behavior: impli-
cations for substance abuse problems and willingness to seek treatment. Substance Abuse Treatment, Prevention,
and Policy, 1(1), 32; Saunders, R., Jacobvitz, D., Zaccagnino, M., Beverung, L.M., & Hazen, N. (2011) Pathways to
earned-security: the role of alternative support figures. Attachment & Human Development, 13(4), 403–20.
412 Levy, K.N., Meehan, K.B., Kelly, K.M., et al. (2006) Change in attachment patterns and reflective function in
a randomized control trial of transference-focused psychotherapy for borderline personality disorder. Journal of
Consulting and Clinical Psychology, 74(6), 1027–40.
413 Roisman, G.I., Padrón, E., Sroufe, L.A., & Egeland, B. (2002) Earned-secure attachment status in retrospect
and prospect. Child Development, 73(4), 1204–19, p.1205. Since the criticism of the earned-secure classification
has become identified with Roisman, it is worth highlighting that Roisman and Sroufe are jointly the corres-
ponding authors for the paper. The discussion is long, but the main section on p.1216 reads firmly as in Sroufe’s
voice and refers in the first person plural to other research not conducted by Roisman.
that these retrospective interviews can provide a veridical picture of early experience’.414 In a
paper published in 2002, Roisman and Sroufe set out to examine the earned-secure classifi-
cation prospectively, drawing on the Minnesota Longitudinal Study of Risk and Adaptation
(Chapter 4); 170 participants in the study completed the AAI at age 19. Since the Cowan’s
study, Main had also altered the coding manual to require that participants receive a score
of lower than 2.5 on the scale for inferred loving parental behaviour in order to qualify as
earned secure.415 Roisman and Sroufe noted, however, that only three of the participants
in their whole sample met this stringent standard, perhaps because they had conducted the
AAI at age 19 so there had been only scant opportunity for individuals to achieve secure-
autonomous status following such adverse care.416 Roisman and Sroufe adopted an approach
that resembled the Cowans, though with a few alterations based on methodological discus-
sions in the intervening years.417 On the basis of this approach, 24 participants were classi-

fied as earned secure.418
Roisman and Sroufe replicated the Cowans’ finding that earned-secure speakers re-
ported more current depressive symptoms. Prospective data also revealed that, a decade
and a half earlier, their mothers had reported more symptoms of anxiety and depression in
their children compared to both the continuous-secure speakers and those with a preoccu-
pied or dismissing AAI classification. Findings were in the same direction in adolescence.
Observations of the 19-year-old speakers in interactions with romantic partners revealed
that earned-secure speakers scored much the same as continuous-secure speakers, and
much more highly than dismissing and preoccupied speakers, on measures such as shared
positive affect, provision of the partner with a secure base, and conflict resolution.
Yet, looking back in time, data from the prospective cohort study also revealed that there
had been no difference between earned-and continuous-secure speakers in terms of their
infant attachment classifications with either fathers or mothers. Additionally, in an obser-
vational assessment with their mothers at 24 months, earned-secure speakers had received
the most supportive care of any of the children, more than the preoccupied, dismissing, and
continuous-secure speakers. At age 13, the mother–child interactions of both earned-secure
and continuous-secure speakers were much more supportive and positive than those of pre-
occupied and dismissing speakers. Roisman and Sroufe concluded that the ‘earned-secures
414 Main, M. & Goldwyn, R. (1998) Adult Attachment Scoring and Classification Systems, Version 6.3.
Unpublished manuscript.
415 This decision is discussed in Hesse, E. (2016) The Adult Attachment Interview: protocol, method of ana-
lysis, and selected empirical studies: 1985–2015. In J. Cassidy & P.R. Shaver (eds) Handbook of Attachment: Theory,
Research, and Clinical Applications, 3rd edn (pp.553–97). New York: Guilford.
and prospect. Child Development, 73(4), 1204–19, p.1209. None of the 19-year-old participants in the Berkeley
follow-up study was classified as ‘earned secure’ according to the stringent criteria either. Main, M., Hesse, E.,
& Kaplan, N. (2005) Predictability of attachment behavior and representational processes at 1, 6, and 19 years
of age: the Berkeley longitudinal study. In K.E. Grossmann, K. Grossmann, & E. Waters (eds) Attachment from
Infancy to Adulthood: The Major Longitudinal Studies (pp.245–304). New York: Guilford.
417 This followed slight changes to the operationalisation of earned security made by Phelps and colleagues,
and then by Paley and colleagues: Phelps, J.L., Belsky, J., & Crnic, K. (1997) Earned security, daily stress, and
parenting: a comparison of five alternative models. Development & Psychopathology, 10, 21–38; Paley, B., Cox,
M.J., Burchinal, M.R., & Payne, C.C. (1999) Attachment and marital functioning: comparison of spouses with
continuous-secure, earned-secure, dismissing, and preoccupied attachment stances. Journal of Family Psychology,
13(4), 580–97.
418 Examination of Table 1 in Roisman et al. (2002) reveals that distinguishing this group seems to have been
more on the basis of reported difficult childhood relationships with fathers than with mothers. Roisman, G.I.,
Padrón, E., Sroufe, L.A., & Egeland, B. (2002) Earned-secure attachment status in retrospect and prospect. Child
Development, 73(4), 1204–19.
were the beneficiaries of among the most supportive maternal care in a high-risk sample’
and, as a consequence, ‘we cannot rule out the possibility that self-described differences
in early experience between retrospectively defined earned-and continuous-secures were
primarily a function of positive and/or negative reporting biases (e.g. negative attentional
biases associated with depression)’.419
Such findings underline the focus of the AAI on current states of mind regarding attach-
ment: differences identified in the interview between subgroups of secure-autonomous
speakers did pick out prospective differences in histories of care and relationships, such as
more parent-reported symptoms of anxiety and depression in preschool among the earned-
secure speakers. However, the findings seemed in some regards to also put into question
the accounts of earned-secure speakers of adverse forms of care in childhood. Roisman and
Sroufe interpreted their findings as suggesting that ‘earned-secures did not rise above mal-

evolent parenting through sheer will; rather, their success was scaffolded by caring adults,
their security was a natural extension of a supportive (although not necessarily ideal) past’.420
They therefore recommended that an alternative to the term ‘earned security’ might be con-
sidered. Hesse and Main subsequently agreed that ‘the term earned-secure is not an ideal
taxonomic label’, and proposed the term ‘evolved secure’ as an alternative—though this pro-
posal seems to have been ignored so far, or come too late in the day to change the presiding
discourse.421 And indeed, attachment researchers have generally continued to depict the
childhoods of earned-evolved secure speakers as adverse, and secure-autonomous speech
in the AAI as reflecting an individual achievement, something ‘earned’, in the face of this
adversity.422
A further study by Roisman and colleagues has provided additional evidence relevant to
appraisal of the ‘earned-secure’ classification. They found that speakers classified as earned/
evolved secure, even on the basis of the stringent Main and Goldwyn criteria, experienced
levels of maternal sensitivity comparable to dismissing and preoccupied speakers at between
5 and 64 months.423 Their caregivers also had more financial difficulties and depressive
symptoms. However, on other measures, these speakers appeared to have had more posi-
tive care. From middle childhood they experienced care that was around the sample average
for sensitivity, and in adolescence observations of child–caregiver suggested that they re-
ceived more sensitive care than preoccupied and dismissing speakers. They also received
419 Ibid. p.1215. In support of attentional biases interpretation see Roisman, G.I., Fortuna, K., & Holland, A.
(2006) An experimental manipulation of retrospectively defined earned and continuous attachment security.
and prospect. Child Development, 73(4), 1204–19, p.1216. See also Roisman G. & Haydon K.C. (2011) Earned-
security in retrospect: emerging insights from longitudinal, experimental, and taxometric investigations. In D.
Cicchetti & G.I. Roisman (eds) The Origins and Organization of Adaptation and Maladaptation: The Minnesota
Symposia on Child Psychology, Vol. 36 (pp.109–54). New York: Wiley.
421 Hesse, E. (2016) The Adult Attachment Interview: protocol, method of analysis, and selected empirical
studies: 1985–2015. In J. Cassidy & P.R. Shaver (eds) Handbook of Attachment: Theory, Research, and Clinical
Applications, 3rd edn (pp.553–97). New York: Guilford, p.572. The term ‘evolved’ may also still carry over too many
connotations from eugenics discourses to feel comfortable on the tongues of developmental researchers.
422 E.g. Reiner, I. & Spangler, G. (2010) Adult attachment and gene polymorphisms of the dopamine D4 re-
ceptor and serotonin transporter (5-HTT). Attachment & Human Development, 12(3), 209–29. However, see re-
cently Iyengar, U., Rajhans, P., Fonagy, P., Strathearn, L., & Kim, S. (2019) Unresolved trauma and reorganization in
mothers: attachment and neuroscience perspectives. Frontiers in Psychology, 10, 110.
423 Roisman, G.I., Haltigan, J.D., Haydon, K.C., & Booth-LaForce, C. (2014) Earned-security in retrospect: de-
pressive symptoms, family stress, and maternal and paternal sensitivity from early childhood to mid-adolescence.
Monographs of the Society for Research in Child Development, 79(3), 85–107, p.105.
much more positive paternal care than the preoccupied and dismissing speakers, with no
differences from those classified continuous-secure. The study by Roisman and colleagues
has been subject to theoretical and methodological criticism from Hesse, who remains un-
convinced that earned-/evoked-secure speakers really did have more positive care.424 In the
context of such debates, the ‘earned/evolved secure’ classification will no doubt be subject
to further research over the coming years by Roisman and other third-generation attach-
ment researchers.425 Nonetheless, all parties to the debate about ‘earned security’ agree that
it should be emphasised that the AAI solicits a retrospective account: it is qualities in the
speaker’s discourse and reasoning about attachment-relevant experiences that form the basis
of classification, not the nature of the events described. This is why attachment researchers
have been especially intrigued by occasions when discourse or reasoning about attachment-
relevant experiences appears to be disrupted.

Unresolved loss
An early observation made by Goldwyn was that at points some narratives became ‘splin-
tered and incoherent, so that ideas were lost, superficially unconnected ideas invaded one an-
other, and the whole approach to the topic of attachment became disorganised’.426 Goldwyn
and Main documented occasions of varying degrees of such disrupted discourse—some ex-
tensive, some more momentary—across dismissing, preoccupied, and secure-autonomous
transcripts, though they seemed somewhat predominant among preoccupied and dismiss-
ing speakers.427 Goldwyn’s application of ‘guess and uncover’ revealed that these transcripts
frequently belonged to parents in dyads that had earlier been unclassifiable in the Strange
Situation according to the Ainsworth categories. Furthermore, this semi-inductive method
revealed that many of these speakers had experienced loss of attachment figures, especially
before adolescence.428 However, further examination of the transcripts in 1983 revealed that
424 This finding is highlighted and discussed in Hesse, E. (2016) The Adult Attachment Interview: protocol,
method of analysis, and selected empirical studies: 1985–2015. In J. Cassidy & P.R. Shaver (eds) Handbook of
Attachment: Theory, Research, and Clinical Applications, 3rd edn (pp.553–97). New York: Guilford.
425 A review of mood induction experiments to explore the meaning of earned/ evolved security is pre-
sented in Roisman, G.I. & Haydon, K.C. (2011) Earned-security in retrospect: emerging insights from lon-
gitudinal, experimental, and taxometric investigations. In D. Cicchetti & G.I. Roisman (eds) The Origins and
Organization of Adaptation and Maladaptation: The Minnesota Symposia on Child Psychology, Vol. 36 (pp.109–54).
New York: Wiley.
1983. PP/Bow/J.4/3.
427 In the early 1980s, Main still regarded conflict behaviours as especially characteristic of avoidant infants
under stress; likewise, in a conference presentation of 1983, Main proposed that splintered and incoherent elem-
ents would especially characterise discourse too, when a speaker adopting an avoidant strategy was faced with
a procedure, like the AAI (or the Separation Anxiety Test), which asked them to contradict their characteristic
conditional strategy and turn their attention to attachment-related experiences and feelings. However, by the mid-
1980s and the ‘Move to the level of representation’ paper, Main had come to regard disorganisation as varying
independently of the three Ainsworth classifications and their analogues in the AAI coding system. Main, M.
& Goldwyn, R. (1983) Predicting rejection of an infant from mother’s representation of her own experiences.
National Conference on Infant Mental Health, Children’s Institute International, Los Angeles, February 1983. PP/
Bow/J.4/3.
428 Main, M. (1982) Letter to John Bowlby, 9 December 1982. PP/Bow/J.4/4: ‘There is a second-generation effect
of early loss (through death) upon infant attachment behaviour. The infant whose parent lost a parent or other
attachment figure before maturity becomes unclassifiable as A, B or C in the Ainsworth infant system.’ Other
speakers had experienced bizarre forms of early care, such as a mother whose obsessional symptoms meant that
her children were regarded as too dirty to be allowed to touch her. By the end of 1982, Goldwyn and Main had
early loss in itself was not a good predictor of infant attachment classification. Rather, what
seemed critical was that speakers seemed to show disruptions in their discourse, to be in a
semantic sense ‘at a loss’, when discussing the dead attachment figure. The concept of ‘unre-
solved grief ’ or ‘unresolved mourning’ had been gaining prominence in the clinical litera-
ture of the 1970s.429 This development drew on earlier accounts by psychoanalysts, including
Bowlby, of the way in which acknowledging and accepting a loss could contribute to mental
health symptoms.430 Building from both Bowlby and the contemporary clinical literature,
Main and Hesse conceptualised the speakers as ‘unresolved with respect to the mourning
of an attachment figure’. These speakers were allocated an Unresolved/disorganised (U/d)
classification.
In the late 1980s, Main and Hesse developed a ‘Lack of Resolution for Mourning’ scale,
with support from a research assistant Anitra DeMoss. Most often markers of lack of reso-

lution occurred in response to the direct question in the AAI about losses, but they could also
be identified in other parts of the transcript that touch upon the relationship with the dead
person. Main, Hesse, and Demoss distinguished two species of lapses in the transcripts.431
Sometimes these co-occurred, especially in clinical or forensic samples, but mostly tran-
scripts displayed one or the other as a characteristic form. A first species was lapses in moni-
toring of reasoning without recognition by the speaker, leading to breaches in ‘coherence’
through interruptions in the plausibility of the speaker’s account.432 For instance, such lapses
were implied in a belief that the dead person remains alive and actively involved in the speak-
er’s life (e.g. Mrs Q, Chapter 1). This includes statements in interview such as ‘So I don’t now
what field I’m going to select, but my father [deceased 15 years ago in childhood] says that
I should choose law’. Main and colleagues commented on ‘this slip of the tongue to the pre-
sent tense regarding a lost attachment figure. It is taken as an indication of disbelief that the
person is dead and leads to U/d category placement because: a) the speaker is talking about
something going on in his immediate surroundings; b) which is currently of vital import to
him, and is c) actively bringing the long-deceased into a freshly constructed sentence; d) as
though he or she continued to have input. However, slips to the present tense sometimes do
not indicate serious disorientation/disorganisation. If, for example, a person says of a de-
ceased father ‘my father has been in banking for a number of years’ the slip may be minor.433
developed two categories for transcripts containing splintered discourse. A first category was ‘Lost: an attach-
ment figure lost through death and parent has not mourned sufficiently’. A second category was ‘Untouchable: the
parents’ parents were untouchable in a peculiar way, e.g. a mother whose mother always implied she was dirty, so
that the children must not touch her’. George, C., Kaplan, N., Goldwyn, R., & Main, M. (1982–83) Attachment
interview for parents. PP/Bow/J.4/4. This second category was not subsequently included in the coding system,
presumably in part because it turned out to be rare.
429 E.g. Lewis, E. (1979) Inhibition of mourning by pregnancy: psychopathology and management. British
Medical Journal, 2(6181), 27–8; Fulmer, R.H. (1983) A structural approach to unresolved mourning in single par-
ent family systems. Journal of Marital and Family Therapy, 9(3), 259–69.
430 Deutsch, H. (1937) Absence of grief. Psychoanalytic Quarterly, 6, 12– 22; Bowlby, J. (1963) Pathological
mourning and childhood mourning. Journal of the American Psychoanalytic Association, 11(3), 500–541. See
also Granek, L. (2010) Grief as pathology: the evolution of grief theory in psychology from Freud to the present.
History of Psychology, 13(1), 46–73.
431 That these were identified between 1987 and 1989 is suggested by the fact that the distinction is quite foreign
to an earlier draft of the scale, sent to Bowlby as Main, M. & Hesse, E. (1987) Lack of resolution of mourning, 15
November 1987. PP/Bow/B.3/36/1.
Stevenson-Hinde, & C. Parkes (eds) Attachment Across the Life Cycle (pp.127–59). New York: Routledge, p.144–5.
433 Main, M., Goldwyn, R., & Hesse, E. (2002) Adult Attachment Scoring and Classification System. Unpublished
manuscript, University of California at Berkeley, Department of Psychology.

The coding manual positioned lapses in reasoning about whether an attachment figure was
dead or not dead as the paradigmatic case of unresolved loss. Other rarer lapses in reasoning
around loss included confusion between the dead person and the speaker, and characterisations
of events in time or space that are not possible—such as being present at a family tragedy and
also, simultaneously, being absent from the event in another country. Lapses in reasoning were
ultimately characterised as ‘things which cannot be true in the external world’,434 though, antici-
pating problems with this definition, Main and colleagues advised coders as best they could to
exclude statements that are grounded in a self-aware and integrated religious or cultural view-
point.435 An additional set of relatively common lapses of reasoning was identified by Main and
Hesse as occurring when speakers stated that they have done something that cannot be true psy-
chologically, such as using willpower to erase experience of a past event. These are claims that, as
it were, cannot be true of the internal world.

A second type of lapse indicating unresolved states of mind was lapses in the monitoring of
discourse without recognition by the speaker. Examples included: invasions of remarks about
a death into apparently unrelated discussions; sudden use of eulogistic speech; exceptionally
long blank pauses that the speaker does not him-/h erself seem to notice; or apparent absorp-
tion in particular details surrounding the death.436 Though there was no explicit statement
in the AAI coding manual stating that lapses in reasoning are more important in making a
classification than lapses in discourse, it is notable that only a few of the examples given of
lapses in discourse are afforded a high score on the scale for unresolved loss, whereas many
of the lapses in reasoning are given a high score. There may be more ambiguity regarding the
cause of a lapse in the monitoring of discourse.437 Additionally or as a consequence, it may be
434 Ibid.
435 Behrens, K.Y., Hesse, E., & Main, M. (2007) Mothers’ attachment status as determined by the Adult
Attachment Interview predicts their 6-year-olds’ reunion responses: a study conducted in Japan. Developmental
Psychology, 43(6), 1553–67: ‘In Japan, it is a common, culturally polite practice in certain contexts to refer to de-
ceased persons in the present tense. From early on, Japanese children are often encouraged to talk as if a deceased
person is alive, as this is considered an act of respect for the deceased in a culture that traditionally has encour-
aged ancestral worship. Initially, speech usage of this kind was confusing for the AAI coder (Kazuko Y. Behrens)
when attempting to score unresolved status. This was because guidelines in the AAI manual stipulate that pre-
sent tense references to deceased persons can, when marked, imply a “lapse in reasoning” referred to as “dead/
not-dead” (Hesse, 1999, p. 405). In other words, in English, some present tense slippages suggest that a speaker
holds two incompatible belief systems, one in which the deceased person is understood to be dead, and a second
in which he or she is considered to be alive (in the physical, not religious or meta-physical, sense). After studying
a number of Japanese texts with present tense usage regarding deceased persons, however, it was possible (as it is
in English) to distinguish normative from nonnormative forms. Thus, for example, when Japanese mothers dis-
cussed both talking to and/or instructing their child to talk to a deceased grandmother in the present tense at a
portable shrine or altar before going to sleep, this could be considered analogous to a Western prayer. Hence, it is
culturally sanctioned, and as such does not imply the frightening ideation that seems to often accompany anom-
alous dead/not-dead usages in English (see Hesse & Main, 2006). In contrast, Japanese normally uses the past tense
when conveying factual information regarding deceased persons to a third party. Thus present tense usages in this
latter context would be considered as potential slippages or lapses in speech, which could, depending on intensity,
lead to a U placement’ (1559).
436 A third, rare form of lack of resolution was characterised as reports of extreme behavioural reactions to the
death, such as of episodes of uncharacteristic violence or suicide attempts, where the speaker does not appear
to realise in the present that the behaviour requires some remark to the interviewer to contextualise, explain it,
or situate the action in relation to the speaker’s present self. However, ‘both our own experience and those of
other investigators informally queried indicated that assignment to the unresolved-disorganised adult attachment
category on the basis of reports of extreme behavioural reactions is very rare’. Main, M. & Morgan, H. (1996)
Disorganization and disorientation in infant Strange Situation behavior: phenotypic resemblance to dissociative
states. In L. Michelson & W. Ray (eds) Handbook of Dissociation: Theoretical, Empirical and Clinical Perspectives
(pp.107–38). New York: Plenum Press, p.119.
437 Though described as ‘lapses’, it is worth noting that markers of unresolved/disorganised states of mind are
sometimes but not generally entirely out of the blue, as the term might suggest. There may often be a logic to their
inferred that lapses in discourse are less proximal markers than lapses in reasoning for what
Main and colleagues were seeking to capture.
In line with this supposition, by the early 1990s, Main’s work with Solomon on infant dis-
organisation had led to a re-evaluation of the concept of lack of resolution. In 1991, Main
and colleagues wrote that ‘as we have gained an increasing understanding of the nature of the
link between the adult’s and infant’s state, unresolved/disorganised/disoriented has come to
seem the best descriptor’.438 Main and colleagues concluded that there could be various ways
that a loss might be left unresolved over time but without contributing to a disorganised/
disoriented state of mind regarding the attachment-relevant cognition: ‘thus, for example,
effective dismissal of the import of a loss is certainly indicative of failure of resolution of
mourning (and is often referred to as “failed mourning”), but is not considered disorganised/
disoriented’.439 In such a case, attention has been effectively directed away from the poten-

tially disorganising loss, much as Main conceptualised the Group A pattern in the Strange
Situation as avoidance of conflict and loss of regulation.440 A smooth surface is put in play by
effective dismissal, against which grief and sorrow will ricochet.
Likewise, distressed pining after a lost attachment figure is often considered by clinicians
as indicating that the work of mourning is not finished. And the term ‘unresolved’ would
suggest that such cases would be included and coded. Indeed, several other researchers have
advocated for the extension of the unresolved/disorganised classification to include this kind
of response.441 Again, however, Main and Hesse claimed that this ‘distress is not considered
evidence for a disorganised/disoriented state of mind’, though they acknowledged that the
boundary is less clear in this case since weeping does entail ‘a very slight index of disorienta-
tion’.442 The distinction between preoccupation with a loss and disorganisation/disorienta-
tion is real but somewhat permeable, just like in Main’s thinking about C and D in the infant
classification system, essentially since distress can be dysregulating. This implies a subtle but
vital qualification. Though the category was named ‘unresolved for loss’, Main and colleagues
specified in the manual that coders were not actually being instructed to identify lack of
resolution. Rather, the coding protocols were situated as seeking to identify instances of the
‘disorganised/disoriented state of mind’ with regard to attachment that unresolved losses
often seemed to occasion.443 Discussions of an unresolved loss in the AAI seemed to be the
interruption of a state of mind regarding attachment. For instance, a dismissing speaker might close down to
clipped replies even more in a discussion leading up to or following a lapse, in a strategy of avoidance of disorgan-
isation; a preoccupied speaker might, derailed by a lapse, further lose track of the question and focus further on
their feelings of grievance.
438 Main, M., Demoss, A., & Hesse, E. (1991) Unresolved (disorganised/disoriented) states of mind with respect
to experiences of loss. In M. Main, R. Goldwyn, & E. Hesse (2002) Adult Attachment Scoring and Classification
System. Unpublished manuscript, University of California at Berkeley, Department of Psychology.
439 Ibid.: ‘One parent of a very secure child had been orphaned in traumatic circumstances. The parent de-
scribed these circumstances briefly, adding firmly “topic closed”. We did not consider this refusal indicative of an
unresolved/disorganised response to the loss under discussion.’
440 Ibid.: ‘Whether the alternative being avoided is sorrow, fear, anger or some unwonted behaviour pattern, in
failed mourning the subject avoids disorganisation by failing to focus on loss.’

441 By way of contrast, Sagi-Schwartz and colleagues, and George and West, later argued that dismissal of a loss
and distressed pining following a loss should have been regarded as unresolved states of mind. Sagi-Schwartz, A.,
Koren-Karie, N., & Joels, T. (2003) Failed mourning in the Adult Attachment Interview: the case of Holocaust
child survivors. Attachment & Human Development, 5(4), 398–409; George, C. & West, M.L. (2012) The Adult
Attachment Projective Picture System: Attachment Theory and Assessment in Adults. New York: Guilford.

443 Ibid.
frequent occasion of disruptions in attentional processes and the processing of information

relevant to attachment. In Main and Solomon’s work on the Strange Situation, the term ‘dis-
organised’ was used to refer to both behaviour and motivation. The term sought to laminate
conflicting, confused, or apprehensive observable behaviour with inferences about the func-
tioning of the (invisible) attachment system. So too in the AAI. The concept of ‘unresolved’
speech laminated observable lapses in reasoning or discourse in transcripts with inferences
about the (invisible) integration and coherence of attentional processes in the retrieval and
communication of information about attachment-relevant experiences.
Main and Hesse developed a nine-point scale for coding unresolved/disorganised/dis-
oriented states of mind regarding attachment, which was based partly on the inferred extent
of disruption to the retrieval and communication of information about attachment-relevant
experiences, and partly on how certain the coder was that lapses of reasoning or discourse

were what they were seeing in the transcript. As with the infant system, scores above 5 war-
ranted a classification. In most instances, unresolved/disorganised/disoriented states of
mind appeared as an interruption or splintering within the speaker’s characteristic forms of
reasoning and discourse when the topic of loss or abuse was raised, though in some cases it
could be more pervasive across the transcript. Main and Hesse conjectured that frightening
experiences or circumstances surrounding or following a loss or frightening ideas regarding
the relationship with the lost figure would be one pathway to disorganised/disoriented states
of mind. This idea was supported by the finding that unresolved states of mind regarding at-
tachment are moderately associated with frightening, frightened, or dissociative behaviours
by caregivers towards their infants (r = .28).444 It was also supported by work by Beverung
and Jacobvitz, which demonstrated that unresolved loss was twice as common when speak-
ers perceived the loss as sudden.445 However, to the researchers’ surprise, there was no asso-
ciation in their sample between U/d following bereavement and age at loss, the nature of the
relationship, the extent of emotional support, or the cause of death.
A fundamental study for the establishment of the U/d classification was conducted by
Ainsworth and Eichberg. This was Ainsworth’s final direct involvement in empirical re-
search. The AAI and the Strange Situation were conducted with a sample of 45 Charlottesville
mother–infant dyads. Ainsworth and Eichberg found that 30 dyads of the sample had ex-
perienced loss of an attachment figure, but 20 displayed few or no markers of unresolved/
disorganised/disoriented state of mind regarding attachment. A qualitative review of these
transcripts revealed that these speakers had experienced various forms of social connected-
ness following the bereavement, whether in terms of the availability of comfort from others
or in terms of their own responsibility for other family members. Only two of the moth-
ers classified as resolved for their loss were part of dyads classified as disorganised in the
444 Madigan, S., Bakermans-Kranenburg, M.J., Van IJzendoorn, M.H., Moran, G., Pederson, D.R., & Benoit, D.
analysis of a transmission gap. Attachment & Human Development, 8(2), 89–111, Table 2.
445 Beverung and Jacobvitz also speculated that the causal relationship may actually be reversed: U/ d may
subsequently make a loss feel like it took place more suddenly. However, their design was cross-sectional and a
prospective study would be needed to examine which way causality runs, or whether there is a bidirectional re-
lationship between U/d and perceived suddenness of the bereavement. Beverung, L.M. & Jacobvitz, D. (2016)
Women’s retrospective experiences of bereavement: predicting unresolved attachment. OMEGA-Journal of Death
and Dying, 73(2), 126–40. In another study, Lyons-Ruth and colleagues found that retrospective report of parental
death in childhood was only associated with unresolved loss at the level of a trend, and was not statistically signifi-
cant (r = 0.20). Lyons-Ruth, K., Yellin, C., Melnick, S., & Atwood, G. (2003) Childhood experiences of trauma and
loss have different relations to maternal unresolved and hostile-helpless states of mind on the AAI. Attachment &
Strange Situation, showing clearly that loss in itself was not a powerful predictor of infant
disorganised attachment.446 In the final count, ten bereaved mothers received a U/d classi-
fication, and of these all were members of dyads classified by blind coders as disorganised/
disoriented in the Strange Situation. The other five dyads who received a D classification had
experienced frightening occurrences. Many of these were quite recent, such as a near-death
experiences or a partner’s severe drug dependency.447 Half the mothers whose discourse re-
ceived a U/d classification had a secondary classification as preoccupied. Ainsworth cited
conversations with Main and Hesse that there can be a close relationship between preoccu-
pation and unresolved states of mind. For instance, ‘In the case of a mother who is preoccu-
pied with her early attachments, it is reasonable to suppose that the memory of her fear of
the parent might sometimes intrude into everyday life, and re-evoke the anxiety; as such
times the infant might find his mother’s behaviour especially frightening since there was

no apparent occasion for it.’448 Ainsworth and Eichberg also found a 90% match between
the other classifications: secure-autonomous state of mind in the AAI was highly associated
with infant secure attachment, and dismissing state of mind highly associated with avoidant
attachment. There was no association between preoccupied state of mind and ambivalent/
resistant attachment.
The strong association between the Strange Situation and the AAI, evidenced in a
study coded by no less than Ainsworth herself, provided a powerful consecration of the
later measure. However, the results for unresolved/disorganised states of mind from the
Ainsworth and Eichberg study have subsequently had to be excluded from meta-analyses,
since the study has been such an outlier in terms of the strength of the association between
the AAI and the Strange Situation procedure.449 Van IJzendoorn observed that, though gen-
erally regarded as a replication, the study should frankly be recognised as another explora-
tory work.450 In fact, this was in no way masked by the authors. Ainsworth clearly stated in
the paper that, after coding the sample a first time, she repeated her coding on the basis of
changes made by Main to the coding manual, which led to the perfect correlation between
unresolved loss and infant disorganised attachment.451 However, there were no changes
mentioned in the classification of secure-autonomous or dismissing states of mind regarding
446 These findings were soon after replicated by Marian Bakermans-Kranenburg as part of her doctoral re-
search: Bakermans-Kranenburg, M.J. & Van IJzendoorn, M.H. (1993) Gehechtheidsbiografie, verlieservaringen
en beleving van het ouderschap. In J.R.M. Gerris (ed.) Opvoeding, Specifieke Groepen en Minderheden (pp.33–54).
Lisse: Swets & Zeitlinger: ‘We examined in a group of 75 mothers with loss experiences in which mothers with un-
resolved loss were distinguished from the others. The number of loss experiences that were experienced turned out
to be of no importance to lack of resolution’ (33). The researchers also reported that ‘Unresolved loss is found in a
minority (17%) of mothers with loss experiences’ (48).
447 Ainsworth, M.D.S. & Eichberg, C.G. (1991) Effects on infant–mother attachment of mother’s experience re-
lated to loss of an attachment figure. In C.M. Parkes, J. Stevenson-Hinde, & P. Marris (eds) Attachment Across the
Life Cycle (pp.160–83). New York: Routledge, p.164.
448 Ibid. p.180.
449 van IJzendoorn, M.H. (1995) Adult attachment representations, parental responsiveness, and infant attach-
ment: a meta-analysis on the predictive validity of the Adult Attachment Interview. Psychological Bulletin, 117(3),
387–403; Verhage, M.L., Schuengel, C., Madigan, S., et al. (2016) Narrowing the transmission gap: a synthesis of
three decades of research on intergenerational transmission of attachment. Psychological Bulletin, 142(4), 337–66.
450 Van IJzendoorn, M.H. (2019) Replication crisis lost in translation? Paper presented at International

451 Ainsworth, M.D.S. & Eichberg, C.G. (1991) Effects on infant–mother attachment of mother’s experience
related to loss of an attachment figure. In C.M. Parkes, J. Stevenson-Hinde, & P. Marris (eds) Attachment Across
the Life cycle. New York: Routledge, p.164.This recoding and the discussions with Main that led to it are further de-
scribed in Ainsworth, M. (1990) Letter to John Bowlby, 17 January 1990. PP/BOW/B.3/8.
attachment. It is therefore likely that Ainsworth’s remarkable powers as an observer also con-
tributed to the strength of the association.
In contrast to the 100% agreement reported by Ainsworth and Eichberg, the latest meta-
analytic finding regarding the association between U/d parental discourse and the D classi-
fication in the Strange Situation based on three decades of research is r = .21. This is weaker
than the associations between the other paired categories: secure-autonomous (F) discourse
and secure (B) infant attachment have an association of r = .31; dismissing (Ds) discourse
and avoidant (A) infant attachment have an association of r = .29; and preoccupied (E) dis-
course and ambivalent/resistant (C) attachment have an association of r = .22.452 There has
been much discussion of factors that may ‘close the transmission gap’ by mediating between
a caregiver’s state of mind regarding attachment and the classification of the caregiver–infant
dyad in the Strange Situation.453 A variety of proposals have been made including: caregiver

perception of the child’s behaviour in terms of intentional mental states (mentalisation);
gene × environment interactions; caregivers’ support for their child’s exploration; caregivers’
use of effective limit-setting; and the nature of the repair offered following mismatches be-
tween children’s signals and their caregiver’s responses.454 However, a further proposal for
the source of the transmission gap has had to do with the operation of the U/d classification,
and especially in the way that unresolved traumatic experiences have been operationalised.
Unresolved traumatic abuse
Between 1987 and 1989, Main and Hesse ran three training institutes for the AAI: in
London (organised by John Bowlby and John Byng-Hall), in Virginia (organised by Mary
Ainsworth), and in Rome (organised by Nino Dazzi and Massimo Ammaniti).455 During
this time, Main and Hesse saw many new transcripts, especially those collected by clinical
colleagues. These new transcripts clearly showed that disruptions of discourse suggestive of
disorganised/disoriented states of mind regarding attachment could occur when speakers
discussed memories besides bereavement. Besides bereavement, the other frequent occasion
452 Verhage, M.L., Schuengel, C., Madigan, S., et al. (2016) Narrowing the transmission gap: a synthesis of
three decades of research on intergenerational transmission of attachment. Psychological Bulletin, 142(4), 337–
66. An additional recent meta-analytic finding work led by Madigan as part of the Collaboration on Attachment
Transmission Synthesis (see Chapter 6) has been that parents with U/d classifications on the AAI are more likely
to be part of dyads with disorganised attachment relationships, but not more or less likely to be part of dyads
with avoidant or resistant attachment relationships than any other parents. Madigan, S. and the Collaboration on
Attachment Transmission Synthesis (2019) An Examination of the Cross-Transmission of Parent–Child Attachment
Using an Individual Participant Data Meta-Analysis. Unpublished manuscript, cited with permission of Sheri
Madigan.
453 van IJzendoorn, M.H. (1995) Adult attachment representations, parental responsiveness, and infant attach-
ment: a meta-analysis on the predictive validity of the Adult Attachment Interview. Psychological Bulletin, 117(3),
387–403; IJzendoorn, M.H. & Bakermans-Kranenburg, M. (2019) Bridges across the intergenerational transmis-
sion of attachment gap. Current Opinion in Psychology, 25, 31–6.
454 Fonagy, P. & Target, M. (2005) Bridging the transmission gap: an end to an important mystery of attach-
ment research? Attachment & Human Development, 7(3), 333–43; Beebe, B. & Steele, M. (2013) How does micro-
analysis of mother–infant communication inform maternal sensitivity and infant attachment? Attachment &
Human Development, 15(5–6), 583–602; Bernier, A., Matte-Gagné, C., Bélanger, M.È., & Whipple, N. (2014)
Taking stock of two decades of attachment transmission gap: broadening the assessment of maternal behavior.
455 Hesse, E. (1999) The Adult Attachment Interview: historical and current perspectives. In J. Cassidy
& P. Shavers (eds) Handbook of Attachment: Theory, Research, and Clinical Applications (pp.395–433).
for lapses in discourse or reasoning seemed to be experiences of traumatic abuse. Main and
Hesse’s growing attention to traumatic abuse may also be placed in the context of the in-
creased prominence of this topic within academic psychology and wider American cultural
discourses by the late 1980s.456 Until that point, Main and Hesse had advised coders to ex-
tend use of the ‘unresolved loss’ classification to encompass disorganised/disoriented states
of mind about abuse: ‘researchers were advised to use the indices of disorganisation and
disorientation in thought processes during discussions of a loss in order to identify unre-
solved trauma of other kinds’.457 One problem with this extension was that the standard AAI
questions were not well adapted to exploring abuse, since the topic is only raised in quite a
general way, in contrast to loss experiences which are extensively probed.458 Nonetheless,
the interview questions regularly elicited lapses in monitoring of reasoning and discourse in
higher-risk samples, and occasionally in lower-risk samples too.459 By the end of the 1980s,

Main and Hesse had ‘now completed a draft of a separate scale for assessing unresolved ex-
periences of physical abuse’.460 That bereavement and traumatic abuse could have the same
kinds of consequences for attention and information-processing, and hence on discourse,
seemed intuitive to Main and Hesse in light of their idea that the intrusion of fear in relation
to an attachment figure might be a cause of both infant disorganised attachment and disor-
ganised/disoriented states of mind. Traumatic abuse is, definitionally, frightening.
In fact the traumatic abuse scale, completed at the start of the 1990s, was not just for phys-
ical abuse but also included any experience where an attachment figure may have trauma-
tised the child through frightening behaviour. Sexual abuse and threats to kill the child were
therefore also included. Main and Hesse excluded cases where a ‘parent was described as
hostile or mean at times, but without being clearly highly frightening in the view of the judge’.
They also advised that traumatic abuse should not be coded if the event was a one-off, even
if a child was for instance hit in the face by a parent or touched sexually by a drunk parent
through clothing, if this was not ‘overwhelmingly frightening’ or ‘expected to escalate’. The
coding manual suggested that an important factor in making this differentiation is the extent
to which the speaker did or ‘did not fear that the parent would go out of control’.461 Main and
colleagues advised that if the child was very frightened in the situation and/or afterwards,
then it should be coded as traumatic abuse, even if the speaker does not now regard the be-
haviour as abusive.
456 Hacking, I. (1995) Rewriting the Soul. Princeton, NJ: Princeton University Press; Fassin, D. & Rechtman, R.
(2009) The Empire of Trauma. Princeton, NJ: Princeton University Press.

(pp.161–81). Chicago: University of Chicago Press, p.177.
458 Bailey, H.N., Moran, G., & Pederson, D.R (2007) Childhood maltreatment, complex trauma symptoms, and
unresolved attachment in an at-risk sample of adolescent mothers. Attachment & Human Development, 9(2), 139–
61: ‘In contrast to the extensive probes around loss experiences, abuse (in particular, sexual abuse) experiences are
explored in less detail during the AAI in order to avoid distressing the participants’ (143).
459 The rarity of U/d for traumatic abuse in low-risk samples is discussed in Hesse, E. & van IJzendoorn, M.
(1999) Propensities towards absorption are related to lapses in the monitoring of reasoning or discourse during
the Adult Attachment Interview: a preliminary investigation. Attachment & Human Development, 1, 67–91. In a
group of 190 Berkeley college students reported by the authors, only three were classified as U/d on the basis of un-
resolved states of mind regarding traumatic abuse (p.76).
(pp.161–81). Chicago: University of Chicago Press, p.177.

Nonetheless, an unresolved/disorganised/disoriented (U/d) classification depended in

part on how events were spoken about. Sexual or physical abuse by a parent, spoken about
with ‘continuing pain and regret’ but without lapses in reasoning or discourse would not re-
ceive an unresolved/disorganised/disoriented classification.462 An important example given
in the coding system was ‘alternating clear report of abuse with denial that it was abuse’. This
had an important paradigmatic status in the manual, much like statements that imply that
a dead person is still alive in scale for unresolved loss, since it suggested direct conflict be-
tween segregated accounts of reality.463 Other markers of unresolved traumatic abuse given
in the coding system included: feelings of having personally deserved abusive treatment by
an attachment figure; confusion by speakers between themselves and the perpetrator; as well
as other signs similar to unresolved mourning such as invasion of the topic of abuse into dis-
cussion of other matters.

In both the case of loss and abuse, Main and Hesse specified that the classification cannot
be given if there is no specific and identifiable historical event mentioned in the transcript.464
Additionally, in the case of abuse but not loss, the abuse described must be established as
sufficiently severe and overwhelmingly frightening. Hence there may be experiences that,
despite appearances, may not be coded as unresolved even if relevant disturbances of rea-
soning or discourse are present. These include: a parent leaving and never being seen again
(but presumably remaining alive); anticipated bereavement in the case of a terminally ill at-
tachment figure; experiences of chronic neglect in childhood; exposure to an atmosphere
of emotional abuse between parents during childhood but without there being a stand-out
‘event’; and fear of an intimidating current partner or ex-partner.465 Similarly, it is not clear
462 Ibid. However, the boundary can sometimes be unclear. For instance, in the AAI a traumatic event or terrible
loss might be spoken about coherently, but then, shortly after, another subsequent event might be discussed with
lapses in reasoning and/or discourse. It can be assumed that the demand to discuss the first traumatic event has
depleted the attentional resources of the speaker, so that the second picks up some of the distress and confusion
that had been held at bay. However, the result is that markers of U/d become attached to events that are neither be-
reavements nor traumatic. There seems to be a diversity of practice regarding how coders deal with such cases.
463 Despite their paradigmatic status in the manual, alternations of reporting and denial of abuse suggestive of
segregation may sometimes be difficult for coders to identify sharply in practice, given that cultural discourses on
abuse are themselves quite confused and contradictory. What appears as alternation may simply be the implemen-
tation of a dismissing strategy as arousal increases in the course of the interview.
464 Bakermans-Kranenburg and colleagues observed that part of what is at stake is the ‘resolution’ of trauma
or loss, something that the AAI coding protocol only assesses, at best, implicitly: ‘the classification system for un-
resolved loss or trauma identifies only positive markers for an unresolved state of mind. Markers for successful
resolution of loss are not evaluated, so the classification system does not include a “resolved” category.’ Bakermans-
Kranenburg, M.J., Schuengel, C., & van IJzendoorn, M.H. (1999) Unresolved loss due to miscarriage: an addition
to the Adult Attachment Interview. Attachment & Human Development, 1(2), 157–70, p.162. Cf. Iyengar, U., Kim,
S., Martinez, S., Fonagy, P., & Strathearn, L. (2014) Unresolved trauma in mothers: intergenerational effects and
the role of reorganization. Frontiers in Psychology, 5, 966.
465 In practice, some of these cases may be treated by coders as if they were a frightening event, so that they can
be classified as unresolved, though this was ultimately not the approach adopted, on the advice of Main and Hesse,
in Goldwyn, R. & Hugh-Jones, S. (2011) Using the Adult Attachment Interview to understand reactive attachment
disorder: findings from a 10-case adolescent sample. Attachment & Human Development, 13(2), 169–91. It is inter-
esting that a study from the Leiden group found that reports of maltreatment in the AAI, but not U/d of maltreat-
ment (or loss), were associated with hippocampal volume: Riem, M.M., Alink, L.R., Out, D., van IJzendoorn, M.H.,
& Bakermans-Kranenburg, M.J. (2015) Beating the brain about abuse: empirical and meta-analytic studies of the
association between maltreatment and hippocampal volume across childhood and adolescence. Development &
Psychopathology, 27(2), 507–520. Such findings suggest that the U/d for trauma construct may be excluding some
relevant information. In support of this conclusion is the fact that the coding of maltreatment from the AAI used
by Reim and colleagues did include several items, such as chronic neglect, that are excluded by Main, Goldwyn,
and Hesse from the U/d classification. The question about the status of potentially traumatic experiences without
a locatable single event in the AAI in part reflects a wider discussion about the meaning of the concept of ‘trauma’
in psychiatric nosology. Van der Kolk, B.A. (2017) Developmental trauma disorder: toward a rational diagnosis for
children with complex trauma histories. Psychiatric Annals, 35(5), 401–408.
what status repeated hospitalisations has within the AAI coding system in terms of making
an unresolved classification, even though hospitalisation was the foundational experience of
trauma and loss in the emergence of attachment theory (Chapter 1). Pervasively frightening,
frightened, or dissociative caregiving may contribute to a disorganised attachment classifica-
tion in the Strange Situation and controlling behaviour on reunion at age six; but if the child
from this dyad grows up without a specific bereavement or identifiable subsequent trauma
that can be relayed discretely to the interviewer, it would be impossible for them to receive an
unresolved classification on the AAI in adolescence or adulthood.466
In the coding manual, Main and colleagues advised coders to scale unresolved traumatic
abuse and unresolved loss separately, and then to draw on both ratings in making a judge-
ment regarding whether a transcript should be placed in the U/d category. Seen in wider
context, this would appear a very surprising decision: abuse and bereavement are generally

not treated as psychological experiences of a kind and of equivalent significance. However,
from Main and Hesse’s perspective, what abuse and loss have in common is that an experi-
ence of the attachment figure may make thinking about the attachment figure alarming and
overwhelming, causing both the aversion and the intensification of attention. The scales
for unresolved traumatic abuse and unresolved traumatic loss are therefore situated in the
manual as means to the end of a categorical judgement regarding whether an unresolved/
disorganised/disoriented state of mind regarding attachment is present to a marked degree.
As such, following Main and Hesse’s lead, the relative distribution of unresolved loss and un-
resolved traumatic abuse in studies is rarely available in published studies; only the overall
U/d category is usually reported.467
However, a study by Fonagy and colleagues found that the association between unre-
solved traumatic abuse in a parent’s AAI and infant disorganised attachment classification
in the Strange Situation was higher than the usual association between U/d on the AAI and
disorganised attachment in the Strange Situation.468 Though acknowledging the need for
further replication, on the basis of these findings Fonagy and colleagues suggest that the
intrusion of unresolved traumatic experiences into everyday functioning may be a more po-
tent cause of frightening or dissociative behaviours by caregivers towards their child than
an unresolved loss.469 These conclusions are in line with a little-discussed finding reported
466 Lyons-Ruth, K., Yellin, C., Melnick, S., & Atwood, G. (2003) Childhood experiences of trauma and loss have
different relations to maternal unresolved and hostile-helpless states of mind on the AAI. Attachment & Human
Development, 5(4), 330–52. However, it is likely that such cases would, in practice, be placed by many coders as
‘Cannot Classify’, and included with the unresolved classification in analyses. See also Kisiel, C.L., Fehrenbach, T.,
Torgersen, E., et al. (2014) Constellations of interpersonal trauma and symptoms in child welfare: implications for
a developmental trauma framework. Journal of Family Violence, 29(1), 1–14.
467 One of the few studies to have done so is Weinfield, N.S., Whaley, G., & Egeland, B. (2004) Continuity, dis-
continuity, and coherence in attachment from infancy to late adolescence: sequelae of organization and disorgan-
ization. Attachment & Human Development, 6(1), 73–97. The researchers reported the important finding that,
when examined prospectively in the Minnesota study, ‘although maltreatment and disorganization share variance,
only disorganization contributes unique variance to the prediction of unresolved abuse’ (84).
468 They also worried that there may be unacknowledged construct variance, such that the associations of the U/
d classification (and their strength) may be different, depending on the relative proportion of trauma and loss in the
sample. Berthelot, N., Ensink, K., Bernazzani, O., Normandin, L., Luyten, P., & Fonagy, P. (2015) Intergenerational
transmission of attachment in abused and neglected mothers: the role of trauma-specific reflective functioning.
Infant Mental Health Journal, 36(2), 200–212. See also Ballen, N., Demers, I., & Bernier, A. (2007) A differential
analysis of the subtypes of unresolved states of mind in the adult attachment interview. Journal of Trauma Practice,
5(4), 69–93.
469 Lyons-Ruth and Jacobvitz have commented that unresolved trauma and unresolved loss have materially
different correlates in most of the studies that have reported them separately, even if they also share substantial
variance. Lyons-Ruth, K. & Jacobvitz, D. (2016) Attachment disorganization from infancy to adulthood: neuro-
biological correlates, parenting contexts, and pathways to disorder. In J. Cassidy & P.R. Shaver (eds) Handbook
by Main and Hesse from their Berkeley sample that lapses in reasoning and discourse in the
AAI occurred far more frequently in discussions of abuse experiences than in discussions
of loss.470 In 1995 Main also offered the provocative hypothesis that infant disorganised at-
tachment will usually resolve by adulthood unless the basis of the disorganisation lies in
traumatic abuse: ‘So long as direct maltreatment is not involved, many, perhaps most, are ex-
pected to have become ‘organised’ by adulthood, being either secure, dismissing or preoccu-
pied’.471 A decade later, a follow-up conducted by Main and Hesse with 44 of the Berkeley
sample at age 19 was consistent with this hypothesis. There was no association between in-
fant disorganised attachment and an unresolved/disorganised classification on the AAI in
this low-risk sample.472 However, a more adequate appraisal of Main’s hypothesis would re-
quire a cohort study including participants with abuse experiences.
Several criticisms of the operationalisation of unresolved traumatic abuse have been

raised. George and Solomon, among others, criticised the coding system for unresolved/
disorganised traumatic abuse as too limited. In their sample, most children who displayed
controlling-punitive and controlling-caregiving behaviours on reunion had caregivers who
were classified U/d on the AAI. However, examination of the cases where caregivers received
a different classification revealed transcripts showing lapses in reasoning or discourse—but
about kinds of events not included by the system, such as frightening occurrences in the im-
mediate life of the speaker rather than in childhood. This was also observed by Ainsworth
and Eichberg. George and Solomon concluded that ‘our findings suggested that trauma
should be defined as events that leave the individual feeling helpless and out of control,
including current/recent traumatic events in the caregiving relationship’.473
Levinson and Fonagy also criticised the boundaries placed by Main and Hesse around
the definition of traumatic abuse.474 In a study in the 1990s of individuals incarcerated for
violent crimes, they found that the prisoners reported histories of severe and appalling abuse
in childhood. However, in the AAI their participants systematically dismissed the import-
ance of these experiences, and so could not be coded as unresolved according to the coding
protocols. Levinson and Fonagy argued that these experiences are, in fact, best regarded as
indicating unresolved/disorganised states of mind regarding attachment, and that both un-
resolved and dismissed distress had contributed to the capacity of these speakers for callous
of Attachment: Theory, Research, and Clinical Applications, 3rd edn (pp.667–95). New York: Guilford; Byun, S.,
Brumariu, L.E., & Lyons-Ruth, K. (2016) Disorganized attachment in young adulthood as a partial mediator of
relations between severity of childhood abuse and dissociation. Journal of Trauma & Dissociation, 17(4), 460–79.
dei genitori. In M. Ammaniti & D. Stern (1992) Attaccamento e Psicoanalisi (pp.80–140). Rome: Gius, Laterza
& Figli.
Goldberg, R. Muir, & J. Kerr (eds) Attachment Theory: Social, Developmental and Clinical Perspectives (pp.407–
470). Hillsdale, NJ: Analytic Press, p.454.
472 Main, M., Hesse, E., & Kaplan, N. (2005) Predictability of attachment behavior and representational pro-
cesses at 1, 6, and 19 years of age: The Berkeley longitudinal study. In K.E. Grossmann, K. Grossmann, & E. Waters
(eds) Attachment from Infancy to Adulthood: The Major Longitudinal Studies (pp.245–304). New York: Guilford,
p.286. When Main and colleagues reworked their data so that participants with a primary unresolved or cannot
classify status were separated from those with a predominant organised pattern and a secondary unresolved classi-
fication, the authors reported that there was a statistically significant relationship with infant disorganised attach-
ment. However, they did not provide the strength of the association.
473 George, C. & Solomon, J. (1996) Representational models of relationships: links between caregiving and at-
tachment. Infant Mental Health Journal, 17(3), 198–216, p.213.

474 Levinson, A. & Fonagy, P. (2004) Offending and attachment: the relationship between interpersonal aware-
ness and offending in a prison population with psychiatric disorder. Canadian Journal of Psychoanalysis, 12(2),
225–51.
and violent behaviour.475 (In practice, contemporary coding norms would now likely place
the Levinson and Fonagy participants as ‘Cannot Classify’ rather than dismissing, since a
truly dismissing transcript would not report abuse in a way that would seem severe and ap-
palling. But this is an evolution in the culture of coding stemming from Hesse’s work on the
‘Cannot Classify’ category, rather than reflecting a change to the manual.) Like Levinson
and Fonagy, Lyons-Ruth and colleagues criticised the boundaries of the unresolved classifi-
cation. They proposed that it should be extended to encompass discourse suggesting unre-
solved/disorganised states of mind where no bereavement or specific trauma is identifiable.
They developed an additional ‘Hostile/Helpless’ coding system to identify unresolved/dis-
organised states apparent especially in the derogation of attachment figures or of speak-
ers themselves, or in strong identification with a hostile or a helpless caregiver.476 Fonagy,
Target, Steele, and Steele likewise expanded the boundaries of the unresolved classification

in their Reflective Functioning Scale, which includes assessment of ‘unintegrated or bizarre
statements, suggesting a lapse in reasoning, without this being around the topic of a bereave-
ment or trauma’.477
However, Main and Hesse have been reluctant to alter the AAI coding system much in
the past two decades. There are likely a few reasons for this. They perceived such changes
as risking the capacity to commensurate empirical studies over time using the AAI.
Furthermore, they clearly felt that it would be better to be overcautious in defining traumatic
experiences, given that the concept of ‘trauma’ has seen such widespread and diffuse use, and
their goal has ultimately been to pin down experiences that form the basis specifically for
unresolved/disorganised/disoriented states of mind.478 Bowlby’s emphasis on the need to
475 The preference of Fonagy and colleagues for Crittenden and Landini’s amended version of the AAI in recent
years may in part reflect the fact that one of these amendments was a more liberal definition of unresolved trauma,
which did encompass dismissed trauma. Strathearn, L., Fonagy, P., Amico, J., & Montague, P.R. (2009) Adult at-
tachment predicts maternal brain and oxytocin response to infant cues. Neuropsychopharmacology, 34(13), 2655;
Fonagy, P. (2015) An honest day’s work. DMM News, 18, p.2, September 2015. https://www.iasa-dmm.org/im-
ages/uploads/DMM%20News%20%2318-Sept%2015%20English.pdf. It should be noted, however, that ra-
ther than simply expanding the Main et al. system, available evidence suggests that the Crittenden and Landini
coding system for the AAI appears to have a different object: Baldoni, F., Minghetti, M., Craparo, G., Facondini,
E., Cena, L., & Schimmenti, A. (2018) Comparing Main, Goldwyn, and Hesse (Berkeley) and Crittenden (DMM)
coding systems for classifying Adult Attachment Interview transcripts: an empirical report. Attachment & Human
476 In a sample of 45 high-risk mothers, around half of whom had been known to social services, Lyons-Ruth
and colleagues reported that the ‘Hostile/Helpless’ system contributed additional prediction to disorganised at-
tachment assessed in the Strange Situation, over and above the unresolved classification as coded using the Main
et al. system. However, in interpreting these results it should be noted that, unusually, there was no association at
all in this sample between U/d on the Main et al. system and infant disorganised attachment classifications. Lyons-
Ruth, K., Yellin, C., Melnick, S., & Atwood, G. (2005) Expanding the concept of unresolved mental states: hostile/
helpless states of mind on the Adult Attachment Interview are associated with disrupted mother–infant commu-
nication and infant disorganization. Development & Psychopathology, 17(1), 1–23; Melnick, S., Finger, B., Hans,
S., Patrick, M., & Lyons Ruth, K. (2008) Hostile helpless states of mind in the AAI. A proposed additional AAI
category with implications for identifying disorganised infant attachment in high risk samples. In H. Steele & M.
Steele (eds) Clinical Application of the Adult Attachment Interview (pp.399–423). New York: Guilford. For further
empirical comparison of the Main et al. and Lyons-Ruth et al. coding systems see Frigerio, A., Costantino, E.,
Ceppi, E., & Barone, L. (2013) Adult Attachment Interviews of women from low-risk, poverty, and maltreatment
risk samples: comparisons between the hostile/helpless and traditional AAI coding systems. Attachment & Human
477 Fonagy, P., Target, M, Steele, H., & Steele, M. (1998) Reflective Functioning Manual, Version 5. London: UCL/
Anna Freud Centre.

478 One study presenting preliminary self-report associations between various forms of abuse, neglect, and ad-
versity with bearing on this question is Thomson, P. & Jaque, S.V. (2017) Adverse childhood experiences (ACE)
and Adult Attachment Interview (AAI) in a non-clinical population. Child Abuse & Neglect, 70, 255–63, Table 3.
attend to ‘actual historical events’ locatable in time and space (Chapter 1) also remained an
influence in the background for Main and Hesse despite their assertion that the AAI is not
a veridical representation of imputed historical experiences.479 In practice, however, many
but not all coders have circumvented the problem by coding cases with apparent unresolved
states of mind but no locatable traumatic experiences as ‘Cannot Classify’. This then allows
classification ‘by the back door’ since, by convention, ‘Cannot Classify’ cases are folded in
with Unresolved cases in statistical analyses.
Yet there has been some movement in the definition of the classification. For instance,
throughout the 1990s unresolved/disorganised speech regarding miscarriage and stillbirth
was classified with ‘loss of pets’ as not indicating a true trauma or bereavement. However,
studies by Bakermans-Kranenburg and colleagues and by Hughes and colleagues found
that infant disorganised attachment was predicted by lapses in reasoning or discourse re-

lating to experiences of miscarriage and stillbirth, when this was probed in the interview.480
Main and colleagues therefore acknowledged that these experiences ‘may be rated using the
same principles, although the nature of the loss should be marked on the coding form. At
present such cases should be analysed separately.’481 As a consequence, there has been some
variation between laboratories in how such cases are handled, with some groups probing
in interview and including unresolved trauma/loss regarding miscarriage and stillbirth as
sufficient basis for a U/d classification, and others following the letter of the manual and not
probing or including these experiences with the other cases in the U/d category.482 In most
samples, the number of cases is small enough that these variations are not critical. Rather,
the instability reflects structural tensions faced by the AAI as an instrument in attempting
to pin down the diffuse concept of trauma. It also reflects differences between Main/Hesse
and their critics in conceptualising the idea of ‘unresolved/disorganised/disoriented states
of mind’.
479 Mary Main, personal communication, August 2019.

480 Bakermans-Kranenburg, M.J., Schuengel, C., & van IJzendoorn, M.H. (1999) Unresolved loss due to miscar-
riage: an addition to the Adult Attachment Interview. Attachment & Human Development, 1(2), 157–70; Hughes,
P., Turton, P., Hopper, E., McGauley, G.A., & Fonagy, P. (2001) Disorganised attachment behaviour among infants
born subsequent to stillbirth. Journal of Child Psychology and Psychiatry, 42(6), 791–801.
manuscript, University of California at Berkeley, Department of Psychology. A recent self-report study found that
when college student participants had no losses other than miscarriages within two years of their birth, this was
not associated with higher scores on self-reported absorption. Self-reported absorption is, however, only a mod-
erate correlate of U/d. Granqvist, P., Fransson, M., & Hagekull, B. (2009) Disorganized attachment, absorption,
and new age spirituality: a mediational model. Attachment & Human Development, 11(4), 385–403; Bahm, N.I.G.,
Duschinsky, R., & Hesse, E. (2016) Parental loss of family members within two years of offspring birth predicts ele-
vated absorption scores in college. Attachment & Human Development, 18(5), 429–42.
482 Another ambiguous case may be unresolved states of mind regarding having a child with significant phys-
ical disabilities. The manual would not seem to include this as a possible instance of loss, since the parent has
not been bereaved. However, a meta-analysis revealed that the unresolved classification was overrepresented
among parents of physically disabled children. Bakermans-Kranenburg and van IJzendoorn offered their suspi-
cion that coders were making U/d classifications on the basis of parents’ ‘unresolved mourning about the loss of
their ideal child’. Bakermans-Kranenburg, M.J. & van IJzendoorn, M.H. (2009) The first 10,000 Adult Attachment
Interviews: distributions of adult attachment representations in clinical and non-clinical groups, Attachment &
Human Development, 11(3), 223–63, p.249. For theoretical discussion of unresolved mourning and disruption
of the caregiving system see Pianta, R.C., Marvin, R.S., & Morog, M.C. (1999) Resolving the past and present: re-
lations with attachment organization. In J. Solomon & C. George (eds) Attachment Disorganization (pp.379–98).
New York: Guilford; Oppenheim, D., Koren-Karie, N., Dolev, S., & Yirmiya, N. (2009) Maternal insightfulness and
resolution of the diagnosis are associated with secure attachment in preschoolers with autism spectrum disorders.
Child Development, 80, 519–27.
Dissociation
Some attachment researchers, for instance Mary Target and several Italian colleagues, have
accused Main and Hesse of self-contradiction and incoherence in theorising dissociation
and fear.483 They urge that there are significant outstanding questions for Main and Hesse in
this area. This latter point is undoubtedly true. However, the accusation of self-contradiction
and incoherence is overstated: fear, trauma, dissociation, and disorganisation have quite dis-
tinct and coherent places in Main and Hesse’s theory. In 1992, Main and Hesse published a

long and finely etched chapter discussing the mechanism that they saw as the basis for both
infant disorganised attachment and unresolved/disorganised/disoriented states of mind re-
garding attachment. Central to their concerns in the chapter was the idea of ‘dissociation’,
which had been gaining prominence in clinical and academic psychology through the pre-
vious decade, as well as public debates about ‘recovered memories’.484 Main and Hesse had
their attention drawn to the concept by Giovanni Liotti, who argued for the potential value of
the concept for interpreting seemingly contradictory, incomplete, or disrupted sequences of
behaviour or speech.485 The 1992 chapter by Main and Hesse offered an account of how dis-
sociation related to other key concepts in their theory: fear, trauma, and attention. However,
the work was only ever published in Italian.486 Without an English translation, the chapter
has not been widely known or discussed, contributing to a tendency for subsequent inter-
preters to treat fear, trauma, dissociation, and disorganisation as confused or as interchange-
able elements in Main and Hesse’s theory.487
In the model put forward in the 1992 chapter, the attachment system in childhood, the
caregiving system in adulthood, and the retrieval and communication of attachment-related
experiences in adulthood have something important in common. All three are underpinned
483 Seganti, A., Carnevale, G., Mucelli, R., Solano, L., & Target, M. (2000) From sixty-two interviews on ‘the
worst and the best episode of your life’: relationships between internal working models and a grammatical scale of
subject–object affective connections. International Journal of Psychoanalysis, 81(3), 529–51, p.532.
484 Kirshner, L.A. (1973) Dissociative reactions: an historical review and clinical study. Acta Psychiatrica
Scandinavica, 49(6), 698–711; Van der Hart, O. & Dorahy, M.J. (2009) Dissociation: history of a concept. In P.F. Dell
& J. O’Neill (eds) Dissociation and the Dissociative Disorders: DSM-V and Beyond (pp.3–26). London: Routledge.
See also Itzkowitz, S., Chefetz, R.A., Hainer, M., Hopenwasser, K., & Howell, E.F. (2015) Exploring dissociation and
dissociative identity disorder: a roundtable discussion. Psychoanalytic Perspectives, 12, 39–79.
485 Liotti, G. (1992) Disorganized/ disoriented attachment in the etiology of the dissociative disorders.
Dissociation, 4, 196–204. See also Hacking, I. (1992) Multiple personality disorder and its hosts. History of the
Human Sciences, 5(2), 3–31.
dei genitori. In M. Ammaniti & D. Stern (1992) Attaccamento e Psicoanalisi (pp.80–140). Rome: Gius, Laterza &
Figli. This is a translation of the chapter ‘Disorganized/disoriented attachment in infants as related to dissociative
states of mind in their parents’ from Hesse, E. (ed.) (1999) Unclassifiable and Disorganized Responses in the Adult
Attachment Interview and in the Infant Strange Situation Procedure: Theoretical Proposals and Empirical Findings.
Unpublished doctoral thesis, Leiden University. This English chapter is cited here rather than relying on a re-
translation of the text back from the Italian. Some elements are repeated in the discussion to Hesse, E. & Main, M.
(2006) Frightened, threatening, and dissociative parental behavior in low-risk samples: description, discussion,
and interpretations. Development & Psychopathology, 18(2), 309–343. However, they are exceptionally compressed
in the latter text, presumably given the challenges of the journal’s word limit, to the point that the claims are not
fully intelligible to a reader not already familiar with the 1992 chapter.
487 E.g. Schore, A.N. (2009) Attachment trauma and the developing right brain: origins of pathological dissoci-
ation. In P.F. Dell & J.A. O’Neill (eds) Dissociation and the Dissociative Disorders (pp.107–41). London: Routledge.
by the coordination of attention, and individual differences ‘follow upon alterations in the
focus of attention’ with respect to attachment-relevant information, including external
perceptions and memories.488 Attachment behaviour in infancy reflects these alterations
in the focus of attention most directly, since the threshold for activation of the attach-
ment system may be raised or lowered ‘by focusing attention either away from or toward
1) the attachment figure and 2) any cues to danger implicit in the situation’.489 Retrieval of
information and communication with the interviewer in the AAI also reflects individual
differences in the alteration of the focus of attention with respect to attachment-relevant
information. Dismissing states of mind are underpinned by a tendency to direct attention
away from attachment-relevant memories and perceptions in the past and in interaction
with the interviewer. Preoccupied states of mind are underpinned by an intense focus on
attachment-relevant memories and perceptions. Finally, the caregiving system is distinct

from the attachment system. But the assemblage of the caregiving system incorporates some
component elements from the attachment system, such as interpretations of the meanings
of physical touch (Chapter 1). As such, individual differences in the alteration of the focus of
attention with respect to attachment-relevant information may have some effect on the func-
tioning of the caregiving system.
In 1990, Main and Hesse argued that an approach/avoidance conflict is produced in the
Strange Situation for an infant with past experiences of a caregiver who displays alarming
behaviour, since the same figure will elicit a disposition to both withdraw and approach as
a safe haven, dispositions that are incompatible and mutually exacerbating. However, the
behavioural account of an approach/avoidance conflict was the external face of a hypothe-
sised mechanism occurring at the level of attention. In their 1992 chapter, the theory of this
attentional mechanism was extended as a model of adult caregiving behaviour and autobio-
graphical discourse. In all three cases, where attachment-relevant perceptions and memories
are also experienced as alarming, Main and Hesse argued that attention cannot simply be
turned away or turned to them since the other response intrudes. The result is a ‘looping’
of attention, and, if sustained or repeated over time, potential damage to the behavioural
system itself is expectable.490 This damage might be seen as weakened regulatory capacities,
localised holes or blockages in the functioning of the system, or potentially even the develop-
ment of relatively independent ‘nets’ of dispositional responses ‘potentially organised with
respect to one of the competing and incompatible goals’.491 In the chapter, Main and Hesse
appeared ambivalent as to whether dissociation represented a kind of defence or adaptation
more extreme in kind than the conditional strategies (as for Bowlby), or essentially a kind
of breakdown. In any case, in reducing environmental responsiveness and the integration
of information, dissociation was anticipated to have repercussions for functioning and for
mental health.
488 Main, M. & Hesse, E. (1992) Disorganized/disoriented attachment in infants as related to dissociative states
of mind in their parents. In E. Hesse (ed.) (1999) Unclassifiable and Disorganized Responses in the Adult Attachment
Interview and in the Infant Strange Situation Procedure: Theoretical Proposals and Empirical Findings. Unpublished
doctoral thesis, Leiden University.
489 Ibid.: ‘In contrast to Group B infants (whose attentional focus varies with circumstances) and Group A in-
fants (who utilise an organised shift in attention away from the attachment figure and her whereabouts), Group
C infants appear almost completely preoccupied with the attachment figure and her whereabouts throughout the
situation.’
490 Ibid.
491 Ibid.
In the 1992 chapter, Main and Hesse proposed a new ‘understanding of the qualitative
structure of trauma’ in the context of attachment.492 Where an attachment-relevant experi-
ence is itself alarming and the looping of attention occurs, there are consequences for the
encoding of the memory. The effective tagging and encoding of embodied memory, Main
and Hesse supposed, requires the attentional process lost to the loop. The looping of atten-
tion inhibits the integration and semantic extraction of experiences, so that these memories
may be accompanied by associations based on episodic rather than semantic resonances
and may lack important contextual markers about time and place. This accounts for the un-
housed, invasively intense quality of traumatic memories, and of the lapses in discourse and
reasoning seen in the AAI. The common mechanism is also proposed as accounting for the
fact that dissociation can be one consequence of such a wide variety of forms of trauma.
As van IJzendoorn and Schuengel among others have observed, ‘the construct of dissoci-

ation can easily be overstretched to include almost every defense mechanism’, and indeed
Main and Hesse were rather unclear in their use of the term, sometimes intending a nar-
rower ‘prototypical’ meaning and sometimes using the term loosely as a synonym for any
form of mental segregation.493 Nonetheless, it is quite possible to identify their key claims
about dissociation through a close reading. Main and Hesse argued that the looping of at-
tention can cause the most prototypically dissociated responses such as fugue states where
a child or an adult is unresponsive to the environment, perhaps accompanied by ‘blank,
unseeing eyes’ or ‘upward rolls of the eyes’ or a startle back to environmental alertness.494
However, problems in the encoding of experience caused by past looping of attention can
also produce segregation within behavioural systems, such that two incompatible responses
may be activated without coordination in response to the same cue from memory or from
the environment. This segregation was regarded by Main and Hesse as also supported by dis-
sociative processes, even if in itself it is not reducible to dissociation and less prototypically
dissociative than a fugue state.
In the chapter, Main and Hesse then applied their model one by one to frightening/fright-
ened caregiver behaviour, to the infant Strange Situation, and to lapses in reasoning or dis-
course in the AAI. In relation to frightening/frightened caregiver behaviour, one of the
three main categories is dissociative behaviours. Main and Hesse speculated that dissocia-
tive behaviours may occur when working memory—short-term perceptual and linguistic
processing—is overwhelmed by the looping of attention around loss or abusive experiences
relating to attachment figures. One potent cause of such looping is anticipated to be alarming
memories of or associations with these figures. It may be prompted by unlikely objects
492 Ibid.
493 Van IJzendoorn, M.H. & Schuengel, C. (1996) The measurement of dissociation in normal and clinical popu-
lations: meta-analytic validation of the Dissociative Experiences Scale (DES). Clinical Psychology Review, 16(5),
365–82, p.375. Ambiguities in the history of the concept between broader and narrower uses go back to the nine-
teenth century. Middleton, W., Dorahy, M.J., & Moskowitz, A. (2008) Historical conceptions of dissociation and
psychosis: nineteenth and early twentieth century perspectives on severe psychopathology. In A. Moskowitz, I.
Schäfer, & M.J. Dorahy (eds) Psychosis, Trauma and Dissociation. Emerging Perspectives on Severe Psychopathology
(pp.9–20). Oxford: Blackwell. An influential proposal was later made for detachment and mental segregation
as distinct phenomena under the label of ‘dissociation’. Holmes, E.A., Brown, R.J., Mansell, W., et al. (2005) Are
there two qualitatively distinct forms of dissociation? A review and some clinical implications. Clinical Psychology
Review, 25(1), 1–23.
because it has been poorly encoded, so it takes an overexpansive field of reference. This then
increases the circumstances that grant uncomfortable freedom to experiences of the past,
and place memory’s sharp edges up against the throat of the present. For instance, the touch
of an infant may evoke poorly encoded and frightening memories for a caregiver of abusive
touch by an attachment figure in childhood or adulthood.495 This may then elicit dissocia-
tive, frightening, or frightened responses by the caregiver towards the child. Or again, the
features of a child may recall those of a dead attachment figure, leading to loops of attention
or activation of the fear behavioural system if the deceased attachment figure and/or their
passing was in some way alarming.496
Even though they may co-occur, Main and Hesse were adamant, however, that not all
frightening/frightened caregiver behaviours should be reduced to the effects of dissociation.
There may be dissociative processes implicated in some or many of them, but not necessarily

(i) to the same degree or (ii) in the same way. Though Main and Hesse do not elaborate the
point, reasons for frightening behaviour by caregivers that do not require dissociation can
readily be identified. One process is acknowledged by Dozier and Bernard, who observed
in their work with at-risk dyads that ‘frightening behaviours can be rewarding to parents
because they are so powerful in eliciting reactions from children’.497 Another case could be
when a child’s safe haven is regularly under attack in the context of domestic violence. These
children may experience their caregiver’s fear as frightening, without the caregiver’s behav-
iour being dissociative. Main and Hesse’s claim that dissociation may not always operate in
the same way is also highlighted by their acknowledgement that other affects besides fear
may be implicated in looping attention and anomalous caregiving behaviours. For instance,
they later noted that if a loss has been profoundly confusing, the confusion may act in ways
analogous to alarm in relation to the caregiver in producing attentional loops and over-
whelming working memory.498 Fonagy and colleagues suggested that other difficult feelings
such as shame, guilt, anger, and disgust may be implicated in disrupting ordinary states of
mind following trauma.499 However, these are not matters considered by Main and Hesse.
495 A related point was made by Enlow and colleagues, who argued that some kinds of trauma may be more
likely to bring about frightening/frightened behaviour than others: Enlow, M.B., Egeland, B., Carlson, E., Blood,
E., & Wright, R.J. (2014) Mother–infant attachment and the intergenerational transmission of posttraumatic stress
disorder. Development & Psychopathology, 26(01), 41–65. ‘For example, normative displays of infant helplessness,
distress, and aggression may be especially threatening and triggering for mothers with PTSD resulting from in-
timate partner violence, particularly if the infant physically resembles the perpetrator’ (59).
496 Main, M. & Morgan, H. (1996) Disorganization and disorientation in infant Strange Situation be-
havior: phenotypic resemblance to dissociative states. In L. Michelson & W. Ray (eds) Handbook of
Dissociation: Theoretical, Empirical and Clinical Perspectives (pp.107–138). New York: Plenum Press, p.126.
497 Dozier, M. & Bernard, K. (2019) Coaching Parents of Vulnerable Infants: The Attachment and Biobehavioral
Catch-up Approach. New York: Guilford, p.87.

498 Bahm, N.I.G., Main, M., & Hesse, E. (2017) Unresolved/disorganized responses to the death of important
persons: relations to frightening parental behavior and infant disorganization. In S. Gojman de Millan, C.
Herreman, & L.A. Sroufe (eds) Attachment Across Clinical and Cultural Perspectives: A Relational Psychoanalytic
Approach (pp.53–74). New York: Routledge, p.56.
499 See e.g. Sharp, C., Fonagy, P., & Allen, J.G. (2012) Posttraumatic stress disorder: a social-cognitive perspec-
tive. Clinical Psychology: Science and Practice, 19(3), 229–40, pp.229–30. On guilt: in fact guilt appears alongside
fear as implicated in U/d in an early version of the lack of resolution scale: Main, M. & Hesse, E. (1987) Lack of
resolution of mourning, 15 November 1987. PP/Bow/B.3/36/1: ‘The individual may indicate excessive fear, guilt or
worry or regret regarding the previous relationship to the lost figure . . . guilt or fear may have become irrational.’
However, guilt was subsequently removed, in line with Main and Hesse’s increasing focus on fear from this period
onwards. On disgust: as we have seen, Main found that the mother in one of the dyads classified as disorganised
in her sample treated a child as too dirty to be allowed to touch her. And in the AAI there is already a classifica-
tion for speakers who show derogating disgust towards close others, even in brief passages of the transcript (Ds2),
though the classification system characterises Ds2 as dismissing rather than unresolved. The relationship between
disgust and unresolved states of mind regarding attachment is also discussed in Buchheim, A. & George, A. (2011)
Appraising disorganised/disoriented attachment behaviour in the Strange Situation in light

of their concern with dissociation, Main and Hesse observed that some of the behaviours used
for coding infant disorganised attachment appear dissociated ‘at a phenotypic level’. One cat-
egory is ‘freezing/stilling’. This, they suspected, directly reflects a lapse in serial processing in
the context of looping attention. Another category Main and Hesse identified as phenotypic-
ally dissociative was ‘direct indices of disorientation’. They regarded this kind of behaviour as
caused by ‘incompatible perceptions, experiences and impulses in which independent “nets”
have developed and momentarily control behaviour’.500 Some undirected behaviours could also
have this basis, for instance when children approach the stranger with arms raised directly on
reunion with their parent. Main and Hesse also argued that the most extreme and sharply de-
fined ‘sequential contradictory’ and ‘simultaneous contradictory’ behaviours shown towards
the caregiver on reunion may represent an expression of such independent ‘nets’ of responses.

For instance, ‘the infant may simultaneously scream for the parent and stretch as far out of the
parent’s arms as possible with eyes cast to the side’—this seems to entail fully developed behav-
ioural dispositions in contradiction.501 However, they identified two further kinds of behaviour
listed in the Main and Solomon indices where a ‘dissociative state need not be implied’: direct
apprehension of the caregiver, and forms of conflict about approach that remain environmen-
tally responsive. They argued that ‘not all disorganised-appearing behaviour listed by Main and
Solomon need imply more than momentary experiences of conflict, and expressions of conflict
between approach and avoidance behaviour towards the parent need not imply the intrusion of
a dissociated secondary plan or system’.502
Having considered caregiving behaviour and the Strange Situation, Main and Hesse
examined the potential relevance of dissociative processes in lapses of monitoring of rea-
soning or discourse in the AAI. In this case, they proposed that dissociation is the prox-
imal mechanism of most, and perhaps ‘virtually all’, lapses in reasoning and discourse in
the AAI.503 In some cases, the lapses are minor and suggest merely absorption of attention
Attachment disorganisation in borderline personality disorder and anxiety disorder. In J. Solomon & C. George
(eds) Disorganised Attachment and Caregiving (pp.343–82). New York: Guilford.
500 Main, M. & Hesse, E. (1992) Disorganized/ disoriented attachment in infants as related to dissociative
states of mind in their parents. In E. Hesse. (ed.) (1999) Unclassifiable and Disorganized Responses in the Adult
Attachment Interview and in the Infant Strange Situation Procedure: Theoretical Proposals and Empirical Findings.
Unpublished doctoral thesis, Leiden University.
501 Ibid. Main and Hesse report that a review of 300 Strange Situations from the Berkeley sample revealed
only three such sharply defined cases where a child seemed to have fully developed behavioural dispositions to
an avoidant conditional strategy and an ambivalent/resistant conditional strategy. In her doctoral project under
Ainsworth, Crittenden found simultaneous or sequential display of the two conditional strategies much more fre-
quently in maltreated children than in non-maltreated samples. Crittenden, P.M. (1988) Relationships at risk. In J.
Belsky & T. Nezworski (eds) Clinical Implications of Attachment (p.136–74). Hillsdale, NJ: Lawrence Erlbaum.
of mind in their parents. In E. Hesse (1999) Unclassifiable and Disorganized Responses in the Adult Attachment
doctoral thesis, Leiden University. See also Main, M. & Morgan, H. (1996) Disorganization and disorientation
in infant Strange Situation behavior: phenotypic resemblance to dissociative states. In L. Michelson & W. Ray
(eds) Handbook of Dissociation: Theoretical, Empirical and Clinical Perspectives (pp.107–138). New York: Plenum
Press: ‘Not all disorganised-disoriented behaviours have a clear relation to dissociative phenomena’ (108).
503 This would be restated again later: ‘Virtually all U/d lapses during the AAI appear to fit to a dissociative
model.’ Hesse, E. & Main, M. (2006) Frightened, threatening, and dissociative parental behavior in low-risk sam-
ples: description, discussion, and interpretations. Development & Psychopathology, 18(2), 309–343, p.311. This
strengthens an earlier, more qualified position: ‘Some lapses observed in the narratives of the parents of disorgan-
ised infants during discussions of traumatic events also appeared to fit to a dissociative model.’ Main, M. & Morgan,
H. (1996) Disorganization and disorientation in infant Strange Situation behavior: phenotypic resemblance to
during speech, conceptualised as a minor form of dissociation.504 In other cases, the lapses
are more major and suggest the operation of segregated processes, as when a speaker suffers
from an ‘intrusion of dissociated ideas, or holds two incompatible ideas regarding a loss or
abuse experience in parallel’.505 They gave out-of-context eulogistic speech about a lost at-
tachment figure in the AAI as an example of absorption. Such discourse suggests that the
question about losses evoked a memory which has been partially or wholly processed as an
immediate perception.506 The memory may have been encoded in such a way that it lacks
cues for context, and/or frightening aspects of the information about the attachment figure
are producing attentional loops that disturb working memory and the integration of remem-
bering with the interpersonal demands of the present interview. By contrast, a more intense
and potentially a qualitatively different form of dissociative processing may be seen in lapses
in reasoning, when incompatible ideas regarding a loss or abuse experience appear to be held

in parallel (e.g. dead/not dead). Usually such contradictions will be monitored and identified
before they can appear in speech, or will be corrected after they occur. However, if atten-
tional resources are tied up in loops, then monitoring can suffer as a result.507
In a 2006 article, Hesse and Main argued that future investigators should attempt to dis-
criminate between the degree of ‘dissociative components’ in lapses of discourse and rea-
soning, and examine their distinct antecedents and associations with anomalous caregiving
behaviours.508 More of the ‘transmission gap’ between caregiver states of mind regarding
dissociative states. In L. Michelson & W. Ray (eds) Handbook of Dissociation: Theoretical, Empirical and Clinical
Perspectives (pp.107–138). New York: Plenum Press, p.130.
504 Psychological absorption has been defined as ‘episodes of single (“total”) attention that fully engage one’s
representational (i.e. perceptual, enactive, imaginative and ideational) resources’. Tellegen, A. & Atkinson, G.
(1974) Openness to absorbing and self-altering experiences (‘absorption’), a trait related to hypnotic susceptibility.
Journal of Abnormal Psychology, 83, 268–77, p.268. This definition would later be the one cited in Hesse, E. & van
IJzendoorn, M.H. (1999) Propensities towards absorption are related to lapses in the monitoring of reasoning or
discourse during the Adult Attachment Interview. Attachment & Human Development, 1(1), 67–91.
506 See also Hesse, E. & Main, M. (2006) Frightened, threatening, and dissociative parental behavior in low-risk
samples: description, discussion, and interpretations. Development & Psychopathology, 18(2), 309–343: ‘As is clear
from the above, not all U/d lapses are indicative of extreme dissociation. For example, the use of funereal speech,
or unusual attention to detail, merely suggest elevated levels of the most “normative” component of dissociation,
absorption. In contrast, most of those cited under section c) above, suggest the presence of real dissociative phe-
nomena such as “segregated systems,” although in most cases we assume these are unlikely to involve multiple
executors capable of guiding action’ (333).
507 Later in the 1990s, Main and Hesse also proposed a dissociative basis when speakers in the AAI show ‘no
single attentional strategy’ with respect to attachment, and instead ‘the subject changes category in mid-interview
in a shocking manner, as though completely shifting state of mind with respect to attachment mid-interview’.
Transcripts showing this shift in states of mind regarding attachment would form one basis for the ‘Cannot Classify’
classification in the 1990s. Hesse, E. (1996) Discourse, memory, and the Adult Attachment Interview: a note with
emphasis on the emerging cannot classify category. Infant Mental Health Journal 17(1), 4–11, p.5. Another form
of Cannot Classify discourse is when low coherence scores make placement in the secure-autonomous category
impossible, but there are no elevated scores for dismissing or preoccupied speech. Main and Hesse did not spe-
cifically suggest a dissociative basis for this form of discourse, but stated they anticipate that frightening and/or
overwhelming historical experiences are implicated in derailing states of mind regarding attachment.
508 Hesse, E. & Main, M. (2006) Frightened, threatening, and dissociative parental behavior in low-risk sam-
ples: description, discussion, and interpretations. Development & Psychopathology, 18(2), 309–343, p.333. One
line of investigation pursued by Hesse has been examination of the role of absorption of attention. A self-report
measure of a tendency towards absorption of attention has been found to be associated with the U/d classifica-
tion in the AAI, which is in line with theory. However, associations have been moderate. See Hesse, E. & Van
IJzendoorn, M.H. (1999) Propensities towards absorption are related to lapses in the monitoring of reasoning
or discourse during the Adult Attachment Interview: a preliminary investigation. Attachment & Human
attachment and infant disorganised attachment might be closed, they argued, if the specific
contribution of dissociative processes was unpicked. Furthermore, the prediction of sequelae
such as later dissociative symptoms (Chapter 4) might be sharpened ‘through an examin-
ation of subtypes of disorganisation and disorientation. Among the likeliest candidates to
be predictive of the dissociative disorders are trance-like stilling and freezing, dissociated
actions, and simultaneous or rapid alternation of avoidance and resistance.’509 However, this
call has not been noted by researchers in part because, in the absence of the 1992 chapter, the
conceptual relationship between the U/d classification, dissociation, and fear has remained
blurry. Main and Hesse hoped that the 2006 article would make their position clear, and
regret that there appears to be little awareness of their account of how exactly U/d, dissoci-
ation, and fear interrelate.510 In particular, few attachment researchers seem to know that
Main and Hesse argued that lapses in reasoning or discourse in the AAI have varying degrees

of a dissociative basis, but that only some forms of frightening/frightened caregiver behav-
iour are regarded as especially dissociative. Likewise, few know that Main and Hesse signal
distinctions among more dissociative, apprehensive, and conflict behaviours (and a poten-
tial difference with stereotypic behaviours) within infant disorganised attachment. As a con-
sequence, too few studies of attachment have included measures of dissociation, in part due
to lack of recognition that the relationship between dissociation, trauma, and unresolved/
disorganised states rather urgently requires empirical disentanglement.
One major exception has been van IJzendoorn, Bakermans-Kranenburg, and colleagues.
Hesse’s doctorate was completed under van IJzendoorn’s supervision during the 1990s, and
there were regular reciprocal visits by the researchers between Berkeley and Leiden. A study
by Schuengel, van IJzendoorn, and Bakermans-Kranenburg provided data to examine Main
and Hesse’s proposal that the unresolved/disorganised/disoriented classification on the AAI
would have a more intimate and consistent relationship with adult dissociative processes than
frightening/frightened behaviour or the dyad’s classification in the Strange Situation procedure.
Following a review of measures of dissociation, the researchers selected a self-report measure—
the Dissociative Experiences Scale—with strong evidence of validity and reliability.511 The
measure captures dimensions of depersonalisation, derealisation, and selective amnesia, in add-
ition to absorption. In agreement with Main and Hesse’s proposal, they found that reports of
dissociative experiences are more common in speakers classified as unresolved/disorganised/
disoriented in the AAI.
Later research has in general terms supported this link between the Dissociative
Experiences Scale and unresolved/disorganised/disoriented speech in the AAI, but
found that it may hold only for unresolved traumatic abuse, not unresolved loss.512 On
Development, 1, 67–91; Granqvist, P., Fransson, M., & Hagekull, B. (2009) Disorganized attachment, absorption,
and new age spirituality: a mediational model. Attachment & Human Development, 11(4), 385–403.
509 Main, M. & Morgan, H. (1996) Disorganization and disorientation in infant Strange Situation be-
havior: phenotypic resemblance to dissociative states. In L. Michelson & W. Ray (eds) Handbook of
Dissociation: Theoretical, Empirical and Clinical Perspectives (pp.107–138). New York: Plenum Press, p.131.
510 Mary Main and Erik Hesse, personal communication, August 2019.
511 Van IJzendoorn, M.H. & Schuengel, C. (1996) The measurement of dissociation in normal and clinical popu-
lations: meta-analytic validation of the Dissociative Experiences Scale (DES). Clinical Psychology Review, 16(5),
365–82.
512 E.g. Zajac, K. & Kobak, R. (2009) Caregiver unresolved loss and abuse and child behavior problems: intergen-
erational effects in a high-risk sample. Development & Psychopathology, 21(1), 173–87; Madigan, S., Vaillancourt,
K., McKibbon, A., & Benoit, D. (2012) The reporting of maltreatment experiences during the Adult Attachment
Interview in a sample of pregnant adolescents. Attachment & Human Development, 14(2), 119–43. No associ-
ation was found by Stovall-McClough, K. & Cloitre, M. (2006) Unresolved attachment, PTSD, and dissociation
the other hand, Schuengel and colleagues found no significant association between the
self-report measure of dissociation and either frightening/frightened behaviours or with
infant disorganised attachment.513 Such findings suggest that more heterogeneous pro-
cesses are in play in these latter two assessments, or that self-report cannot capture the
forms of dissociation relevant to frightening/frightened behaviours or the kinds of care-
giving linked to infant disorganised attachment. With so few trained coders of the fright-
ening/frightened (FR) coding system, it is unsurprising there has been no later study to
have used both this coding system and a measure of dissociation. However, it is a mark of
the poor reception of Main and Hesse’s ideas about fear, trauma, dissociation, and disor-
ganisation that no later study has used a measure of caregiver dissociation alongside the
Strange Situation.514
Further interrogation of the relationship between unresolved/disorganised/disoriented

states of mind regarding attachment and the construct of trauma has been pursued by van
IJzendoorn and colleagues. They conducted the AAI and a clinical assessment for post-
traumatic stress disorder (PTSD) with 31 combat veterans in treatment for PTSD and 29
veterans not in treatment for PTSD. The researchers added a specific probe about combat
experiences to the AAI schedule, and coded both combat-related lapses in reasoning or
discourse and the non-combat-related lapses that are usually coded in other samples.
Rates of secure-autonomous attachment did not differ between the groups. However,
lapses in reasoning or discourse on the AAI in discussions of combat were found to be so
strongly associated with PTSD symptoms that it was almost as if they were the same con-
struct (r = .80):
The convergence between AAI unresolved state of mind and PTSD symptomatology is re-
markable as AAI unresolved state of mind and PTSD differ in severity of presentation, in
prevalence in general populations and in the theoretical perspective from which they were
constructed. These findings support the view that AAI unresolved state of mind and PTSD
symptomatology share lack of integration as a common core phenomenon. This core phe-
nomenon consists of the occurrence of discrete trauma-related disruptions of thought,
speech, and action.515
in women with childhood abuse histories. Journal of Consulting and Clinical Psychology, 74(2), 219–28; or by
Marcusson-Clavertz, D., Gušić, S., Bengtsson, H., Jacobsen, H., & Cardeña, E. (2017) The relation of dissoci-
ation and mind wandering to unresolved/disorganized attachment: an experience sampling study. Attachment
& Human Development, 19(2), 170–90. Thomson and Jaque found an association between U/d and pathological
forms of dissociation, but not absorption. Thomson, P. & Jaque, S.V. (2014) Unresolved mourning, supernatural
beliefs and dissociation: a mediation analysis, Attachment & Human Development, 16(5), 499–514. However, the
relationship with U/d is clouded by the diversity of measures of dissociation used across studies.
513 Schuengel, C., Bakermans-Kranenburg, M.J., & van IJzendoorn, M.H. (1999) Frightening maternal behavior
linking unresolved loss and disorganized infant attachment. Journal of Consulting and Clinical Psychology, 67(1),
54–63, p.59.
514 The only other relevant study was contemporaneous with the work of Schuengel and colleagues, and is now
over 20 years old. Lyons-Ruth and colleagues used both the Dissociative Experiences Scale and the Mississippi
Scale for Post Traumatic Stress Disorder in a study of 45 mother–infant dyads from low-income families. They
found that ‘most mothers of disorganized infants fell into the low symptom group (64%), while the remaining
third fell into the polysymptomatic group (36%)’. Such findings again suggest that dissociation in the context of
trauma may be only one process implicated in infant disorganised attachment. Lyons-Ruth, K. & Block, D. (1996)
The disturbed caregiving system: relations among childhood trauma, maternal caregiving, and infant affect and
attachment. Infant Mental Health Journal, 17(3), 257–75, p.268.
515 Harari, D., Bakermans-Kranenburg, M.J., De Kloet, C.S., et al. (2009) Attachment representations in Dutch
veterans with and without deployment-related PTSD. Attachment & Human Development, 11(6), 515–36, p.350.
Bakermans-Kranenburg and van IJzendoorn also reported meta-analytic findings that al-
most all adults with PTSD across different samples are classified as unresolved (U/d).516 They
aligned these findings with Main and Hesse’s proposal that some forms of unresolved/disor-
ganised discourse are caused by the intrusion of poorly processed perceptions which ‘may
disrupt attention . . . in the form of absorption and unmonitored intrusions of memories,
affects and sensory perceptions concerning the trauma’.517 However, they warned that ‘there
may be an asymmetric relation in the sense that not all AAI unresolved trauma involves
PTSD, while PTSD would almost always involve AAI unresolved trauma’, at least when the
trauma is probed for and coded.518
Van IJzendoorn and colleagues also examined lapses of reasoning or discourse appearing
in discussion of experiences unrelated to combat. Whereas in the control group 7% dis-
played lapses in reasoning or discourse relating to these other experiences, 42% of the group

in treatment for PTSD displayed U/d markers.519 This finding was interpreted as suggesting
that an unresolved state of mind regarding non-combat traumas may have predisposed vul-
nerability to lack of integration in mental processing of combat experiences. Traumas and
losses may layer on top of one another, with each contributing to greater vulnerability to
unresolved/disorganised states of mind. However, the researchers were led to reflect further
on the image of segregated processing or dissociation solely as a risk factor by their later
research. In a study of Holocaust survivors and a group of rigorously matched control par-
ticipants, they found, as expected, that dissociative symptoms were more common among
the Holocaust survivors. Yet there were no differences in the physical, psychological, or cog-
nitive functioning in the children of the survivors compared to the control. Van IJzendoorn
and colleagues proposed that some form of dissociation may here be serving, actually, as a
516 Bakermans- Kranenburg, M.J. & van IJzendoorn, M.H. (2009) The first 10,000 Adult Attachment
Interviews: distributions of adult attachment representations in clinical and non-clinical groups. Attachment
& Human Development, 11(3), 223–63, p.249. The ‘almost perfect’ classification of participants with PTSD as
showing U/d would likely have been yet higher if other researchers had, like the Leiden researchers in the Harari
study of combat veterans, included probes specific to relevant forms of trauma and loss, rather than relying on
the general questions in the interview protocol. Bailey and colleagues reported from their study of adolescent
mothers that ‘71% of women with a history of sexual abuse, as reported on either the AAI or the trauma inter-
view, were classified as Unresolved. This association may have been even stronger if a specific sexual abuse probe
were included on the AAI.’ Bailey, H.N., Moran, G., & Pederson, D.R. (2007) Childhood maltreatment, complex
trauma symptoms, and unresolved attachment in an at-risk sample of adolescent mothers. Attachment & Human
Development, 9(2), 139–61, p.153.
517 Out, D., Bakermans-Kranenburg, M.J., & van IJzendoorn, M.H. (2009) The role of disconnected and ex-
tremely insensitive parenting in the development of disorganized attachment: validation of a new measure.
518 Harari, D., Bakermans-Kranenburg, M.J., De Kloet, C.S., et al. (2009) Attachment representations in Dutch
veterans with and without deployment-related PTSD. Attachment & Human Development, 11(6), 515–36, p.351.
In fact, later research with involvement by van IJzendoorn qualified this picture, finding that only half of adoles-
cent patients who had experienced child sexual abuse and met clinical criteria for PTSD were classified U/d on the
AAI. See van Hoof, M.J., van Lang, N.D., Speekenbrink, S., van IJzendoorn, M.H., & Vermeiren, R.R. (2015) Adult
Attachment Interview differentiates adolescents with childhood sexual abuse from those with clinical depression
and non-clinical controls. Attachment & Human Development, 17(4), 354–75. In this sample, unresolved state of
mind had no association with dissociative symptoms.
519 These findings align with those of Nye and colleagues, who found that 50% of a group of Vietnam veterans
identified as disabled by PTSD received a U/d classification, compared to 16% in a control sample. The researchers
reported that U/d classification was associated with greater probability of a comorbid anxiety disorder. Nye, E.C.,
Katzman, J., Bell, J.B., Kilpatrick, J., Brainard, M., & Haaland, K.Y. (2008) Attachment organization in Vietnam
combat veterans with posttraumatic stress disorder. Attachment & Human Development, 10(1), 41–57. Parallel
findings are reported for other forms of trauma: 70% of those with U/d for trauma in a sample of women with
histories of childhood sexual and/or physical abuse were also identified by clinical interview as showing PTSD by
-Stovall McClough, K. & Cloitre, M. (2006) Unresolved attachment, PTSD, and dissociation in women with child-
hood abuse histories. Journal of Consulting and Clinical Psychology, 74(2), 219–28.
protective factor: ‘Removing traumatic memories from one’s mind may result in reduced
hyper vigilance, normal cortisol levels, and reduced fight or flight responses, all of which
might be adaptive’, and indeed may reduce the display of frightened/frightening behaviours
towards children.520 This conclusion is supported by recent work on the Dozier and col-
leagues Attachment and Biobehavioural Catch-up intervention, in which dissociative be-
haviour by caregivers towards their infants was strongly negatively associated with emotional
dysregulation in response to frustrating tasks when the children were aged three and four.521
In another later paper, van IJzendoorn and collaborators added a further qualification
to the image of early unresolved loss or trauma predisposing U/d for combat experiences.
A study of 184 twins revealed that genetic factors accounted for around half of variance
in dissociative symptoms, suggesting a role for genetic factors; however, the contribution
of genes associated with dissociation was intensified for individuals who had experienced

trauma.522 This suggests a complex gene × environment interplay in the activation of dis-
sociative processes. Genetic contributions to dissociation may then be implicated in the
higher rates of lapses of reasoning and discourse among both combat and non-combat ex-
periences in veterans with PTSD. However, an alternative/additional explanation could be
that the traumatic combat experiences activated a latent predisposition towards dissoci-
ation, undermining the retrieval of and communication about earlier non-combat-related
experiences in the AAI.
Van IJzendoorn and Bakermans-Kranenburg punned that the relationship between
trauma, dissociation, and the unresolved classification is an ‘unresolved issue’ for the field.523
Like an unresolved loss or trauma, the issue is barely recognised among researchers and the
audiences of attachment research, if at all, and reflects a lack of integration of different kinds
of information. Yet despite their punning, van IJzendoorn and Bakermans-Kranenburg
clearly regarded this as a serious problem. Unless matters are clarified by further research,
they even wondered whether lack of resolution for trauma on the AAI ‘shows sufficient in-
cremental validity beyond established measures for posttraumatic stress symptomatology’
to be worth continuing to use.524 The appearance of unresolved discourse in the AAI could
be an expression of PTSD, or—more likely—of some part of this somewhat heterogenous
phenomenon.525 Furthermore, van IJzendoorn and Bakermans-Kranenburg anticipate that
520 Fridman, A., Bakermans-Kranenburg, M.J., Sagi-Schwartz, A., & van IJzendoorn, M.H. (2011) Coping in old
age with extreme childhood trauma: aging Holocaust survivors and their offspring facing new challenges. Aging &
Mental Health, 15(2), 232–42, p.240.
521 Yarger, H.A. (2018) Investigating longitudinal pathways to dysregulation: the role of anomalous parenting
behaviour. Unpublished doctoral dissertation, University of Delaware.

522 Pieper, S., Out, D., Bakermans-Kranenburg, M.J., & van IJzendoorn, M.H. (2011) Behavioral and molecular
genetics of dissociation: the role of the serotonin transporter gene promoter polymorphism (5-HTTLPR). Journal
of Traumatic Stress, 24(4), 373–80.
523 van IJzendoorn, M.H. & Bakermans-Kranenburg, M.J. (2014) Confined quest for continuity: the categor-
ical versus continuous nature of attachment. Monographs of the Society for Research in Child Development, 79(3),
157–67, p.165.
524 Ibid. One approach to this question would be to see whether markers of U/d decline in AAI discourse fol-
lowing treatment for PTSD. Evidence from other fields to support the plausibility of this hypothesis was surveyed
by Stovall-McClough, K. & Cloitre, M. (2006) Unresolved attachment, PTSD, and dissociation in women with
childhood abuse histories. Journal of Consulting and Clinical Psychology, 74(2), 219–28.
525 Dutra, S.J. & Wolf, E.J. (2017) Perspectives on the conceptualization of the dissociative subtype of PTSD and
implications for treatment. Current Opinion in Psychology, 14, 35–9; Horwitz, A.V. (2018) PTSD: A Short History.
Baltimore: Johns Hopkins University Press. The exact relationship between U/d and PTSD, as an umbrella diag-
nosis, was queried already two decades ago by Cole-Detke, H. & Kobak, R. (1998) The effects of multiple abuse in
interpersonal relationships. Journal of Aggression, Maltreatment & Trauma, 2(1), 189–205. Several interpretations
of the AAI have assumed that the U/d classification was, in fact, intended as a measure of PTSD, e.g. Wilkins, D.,
Shemmings, D., & Shemmings, Y. (2015) A–Z of Attachment. London: Palgrave, pp.164–5.
clarification of the relationship between the AAI and PTSD will be hindered by the fact that
the AAI protocol does not adequately probe the topic of trauma. This is a point likewise
made by Riber, and by Bailey, Moran, and Pederson.526 Van IJzendoorn and Bakermans-
Kranenburg, Riber, and Bailey and colleagues have all responded to this predicament by
making adaptations to the questions asked in the AAI. Other researchers have not, or at least
have not reported doing so. This is likely to have contributed to variability in reports of the
association between the AAI and PTSD symptoms.
Another finding relevant here is that in a study of patients with a personality disorder
diagnosis, Fonagy and colleagues documented that whilst participants have a more difficult
time identifying their own unresolved losses, 87% of participants who regarded themselves
as traumatised were rated as unresolved/disorganised/disoriented on the AAI.527 If repli-
cated by other studies, this high level of agreement between self-report of post-traumatic

stress and unresolved trauma in the AAI would align with van IJzendoorn and Bakermans-
Kranenburg’s question about incremental validity.528 One forthcoming study offers relevant
findings. Whereas childhood PTSD is usually associated with lower hippocampal volume,
Cortes Hidalgo and colleagues found that infant disorganised attachment was associated
with larger hippocampal volume (d = .21) compared to other participants.529 However, this
is as yet only limited and preliminary evidence. The questions raised by van IJzendoorn,
Bakermans-Kranenburg, Pederson, and others regarding how the U/d classification for the
AAI relates or adds to existing assessments of PTSD remain an area for future research.
The status of depression
Another outstanding question left by the work of the Berkeley group has been the status
of depression in relation to the Berkeley assessments of attachment. Depression has mostly
been considered in relation to the AAI or in the mothers in dyads seen in the Strange
526 Bailey, H.N., Moran, G., & Pederson, D.R. (2007) Childhood maltreatment, complex trauma symptoms, and
unresolved attachment in an at-risk sample of adolescent mothers. Attachment & Human Development, 9(2), 139–
61; Riber, K. (2016) Attachment organization in Arabic-speaking refugees with post traumatic stress disorder.
527 Cirasola, A., Hillman, S., Fonagy, P., & Chiesa, M. (2017) Mapping the road from childhood adversity to
personality disorder: the role of unresolved states of mind. Personality and Mental Health, 11(2), 77–90: ‘61.1%
(n = 23) of those who had reported experiences of early loss were coded as U/d for loss, and 87.0% of participants
with a history of abuse were rated as U/d for abuse’ (82).
528 Though certainly not an exact replication, another study has bearing: Howard and Miriam Steele and col-
leagues studied the relationship between the unresolved classification on the AAI and self-report of the ‘Adverse
Childhood Experiences’ measure of early trauma. This differed from the Carasola study in that participants were
not asked whether they considered themselves traumatised, but nonetheless it has some conceptual similarities
since several of the ‘Adverse Childhood Experiences’ are explicitly various forms of experiences of abuse and neg-
lect. The researchers found a dose-response relationship, gradiated up to four or more discrete indices of child-
hood trauma, at which point 65% of participants received an unresolved AAI classification. Murphy, A., Steele, M.,
Dube, S.R., et al. (2014) Adverse Childhood Experiences (ACEs) questionnaire and Adult Attachment Interview
(AAI): implications for parent child relationships. Child Abuse & Neglect, 38(2), 224–33. See also Thomson, P. &
Jaque, S.V. (2017) Adverse Childhood Experiences (ACE) and Adult Attachment Interview (AAI) in a non-clinical
population. Child Abuse & Neglect, 70, 255–63.
529 Reported in Van IJzendoorn, M.H. (2019) Replication crisis lost in translation? Paper presented at
International Attachment Conference, Vancouver, 20 July 2019. Additionally, van Hoof and colleagues recently
showed that unresolved status on the AAI is correlated with atypical amygdala resting-state functional connectivity
even adjusting for mental health: van Hoof, M.J., Riem, M.M., Garrett, A.S., et al. (2019) Unresolved–disorganized
attachment adjusted for a general psychopathology factor associated with atypical amygdala resting-state func-
tional connectivity. European Journal of Psychotraumatology, 10(1), 1583525
Situation. Given that few infants or toddlers receive a diagnosis of depression, there has been
little pressure for Main and colleagues or the wider field of attachment research to consider
how depression in a young child might interact with or reflect disruption to the attachment
system. In general, apparent symptoms of depression in children under three have been re-
garded by the scientific community as reflecting parental mental health rather than a quality
or property of the child.530 The Berkeley six-year measures have not been in wide circulation,
so there has been almost no study of their relationship with depression in middle child-
hood.531 One of the few such studies was conducted by Gullone and colleagues, who used
the family drawing system with 326 children aged eight to ten. The researchers found a weak
but material association between attachment insecurity and the child’s report of depressive
symptoms (r = .25).532
Whereas the contribution of child depression to disruption of the child’s attachment

system has been understudied, there have, by contrast, been numerous studies examining
the association between maternal depression and infant attachment in the Strange Situation.
This was a topic of special and widespread interest, since Bowlby’s emphasis on caregiver
unavailability was readily extrapolated to maternal depression as a potential source of
major disruption to the infant–caregiver attachment relationship.533 Several attachment re-
searchers from diverse traditions—ranging from Hazan and Shaver to Cicchetti, Toth, and
Rogosch—claimed that disorganised attachment was an expectable result of caregiver de-
pression.534 Studies set out to examine the question empirically. By 1999, van IJzendoorn
and colleagues could report findings from a meta-analysis of 1053 dyads seen in the Strange
Situation where the mother had been identified, either by clinicians or by the researchers, as
showing symptoms of depression. To the surprise of the researchers, the percentage of dyads
receiving a disorganised attachment classification was only 21%, not significantly different
to incidence in the general population.535 Van IJzendoorn and colleagues tried separating
530 Joan Luby has been an important figure driving attention to depressive symptoms in children under three.
Luby demonstrated that depressive symptoms in early childhood predict a later diagnosis of depression. Luby, J.L.,
Si, X., Belden, A.C., Tandon, M., & Spitznagel, E. (2009) Preschool depression: homotypic continuity and course
over 24 months. Archives of General Psychiatry, 66(8), 897–905; Whalen, D.J., Sylvester, C.M., & Luby, J.L. (2017)
Depression and anxiety in preschoolers: a review of the past 7 years. Child and Adolescent Psychiatric Clinics, 26(3),
503–522.
531 In attachment research, the potential for child contributions to parental depression have generally been ig-
nored. For instance, attachment strategies—for instance controlling-punitive behaviour—may make a reciprocal
contribution to parental mental health once they have solidified. See Raposa, E.B., Hammen, C.L., & Brennan,
P.A. (2011) Effects of child psychopathology on maternal depression: the mediating role of child-related acute and
chronic stressors. Journal of Abnormal Child Psychology, 39(8), 1177–86.
532 Gullone, E., Ollendick, T.H., & King, N.J. (2006) The role of attachment representation in the relationship be-
tween depressive symptomatology and social withdrawal in middle childhood. Journal of Child and Family Studies,
15(3), 263–77.
533 Pound, A. (1982) Attachment and maternal depression. In C.M. Parkes & J. Stevenson-Hinde (eds) The Place
of Attachment in Human Behavior (pp.118–30). Oxford: Oxford University Press; Lyons-Ruth, K., Connell, D.B.,
Grunebaum, H.U., & Botein, S. (1990) Infants at social risk: maternal depression and family support services as
mediators of infant development and security of attachment. Child Development, 61(1), 85–98; DeMulder, E.K.
& Radke-Yarrow, M. (1991) Attachment with affectively ill and well mothers: concurrent behavioral correlates.
534 Hazan, C. & Shaver, P.R. (1994) Attachment as an organizational framework for research on close rela-
tionships. Psychological Inquiry, 5(1), 1–22, p.6; Cicchetti, D., Toth, S.L., & Rogosch, F.A. (1999) The efficacy of
toddler–parent psychotherapy to increase attachment security in offspring of depressed mothers. Attachment &
Human Development, 1(1), 34–66, p.36.
225–50.
the seven studies of caregivers recruited from community samples from the nine studies
recruited specifically on the basis of a diagnosis of clinical depression. In the former group
there was no association with disorganised attachment, even at trend level. In the latter
group there was a very weak (r = .13) association with disorganised attachment.536
One interpretation of these findings was that the severity of depressive symptoms may
play a role. However, severity of depressive symptoms has not generally been found to mod-
erate the relationship between maternal depression and infant disorganised attachment,537
though one study reported that intermittent symptoms have a weaker effect than continuous
symptoms of depression.538 Comorbidity with other mental health issues has likewise
generally not been found to strengthen the link between depression and disorganised at-
tachment.539 Tharner and colleagues hypothesised that the association between maternal
depression and child disorganised attachment seen in clinical samples may be due to an

interaction with environmental adversities such as poverty that might contribute to feelings
of despair or overloading of demands on cognitive resources.540 Equally, however, it is pos-
sible to use the same argument to suggest that clinical status is here serving as a weak index of
other factors—for instance Hofer’s ‘hidden regulators’ (Chapter 1)—with relevance to differ-
ences in infant–caregiver attachment, without depression itself making an independent con-
tribution. Bigelow, Beebe, and colleagues have provided evidence suggesting an additional
hypothesis: that the caregiver’s capacity to hold the child in mind may account for what asso-
ciation there is between parental depression and infant disorganised attachment.541
As mentioned earlier, a meta-analysis by Groh and colleagues found that avoidant attach-
ment in infancy predicted later depression and anxiety symptoms. But contrary to previous
narrative reviews and the expectations of attachment researchers in general, there was no
association between infant disorganised attachment and later symptoms of depression.542
536 Ibid. p.237. See also Atkinson, L., Paglia, A., Coolbear, J., Niccols, A., Parker, K.C., & Guger, S. (2000)
Attachment security: a meta-analysis of maternal mental health correlates. Clinical Psychology Review, 20(8),
1019–40.
537 McMahon, C.A., Barnett, B., Kowalenko, N.M., & Tennant, C.C. (2006) Maternal attachment state of mind
moderates the impact of postnatal depression on infant attachment. Journal of Child Psychology and Psychiatry,
47(7), 660–69; Tharner, A., Luijk, M.P., Van IJzendoorn, M.H., et al. (2012) Maternal lifetime history of depres-
sion and depressive symptoms in the prenatal and early postnatal period do not predict infant–mother attach-
ment quality in a large, population-based Dutch cohort study. Attachment & Human Development, 14(1), 63–81;
Flowers, A.G., McGillivray, J.A., Galbally, M., & Lewis, A.J. (2018) Perinatal maternal mental health and disorgan-
ised attachment: a critical systematic review. Clinical Psychologist, 22(3), 300–316.
538 Campbell, S.B., Brownell, C.A., Hungerford, A., Spieker, S.J., Mohan, R., & Blessing, J.S. (2004) The course
of maternal depressive symptoms and maternal sensitivity as predictors of attachment security at 36 months.
539 The relationship between maternal depression and insecure-organised attachment was, however, moderated
by comorbid personality disorder. Smith-Nielsen, J., Tharner, A., Steele, H., Cordes, K., Mehlhase, H., & Vaever,
M.S. (2016) Postpartum depression and infant–mother attachment security at one year: the impact of co-morbid
maternal personality disorders. Infant Behavior and Development, 44, 148–58.
540 Tharner, A., Luijk, M.P., Van IJzendoorn, M.H., et al. (2012) Maternal lifetime history of depression and de-
pressive symptoms in the prenatal and early postnatal period do not predict infant–mother attachment quality in a
large, population-based Dutch cohort study. Attachment & Human Development, 14(1), 63–81, p.75.
541 Bigelow, A.E., Beebe, B., Power, M., et al. (2018) Longitudinal relations among maternal depressive symp-
toms, maternal mind-mindedness, and infant attachment behavior. Infant Behavior and Development, 51, 33–44.
Child Development, 83(2), 591–610. Also of relevance here are findings from intervention research that suggest
that a video-feedback intervention with caregivers reduces disorganised attachment, but has no effect on children’s
symptoms of depression. Klein Velderman, M., Bakermans-Kranenburg, M.J., Juffer, F., Van Ijzendoorn, M.H.,
Mangelsdorf, S.C., & Zevalkink, J. (2006) Preventing preschool externalizing behavior problems through video-
feedback intervention in infancy. Infant Mental Health Journal, 27(5), 466–93.
Conclusion 323
Groh and colleagues identified a critical mismatch between theory and empirical findings.
Bowlby’s primary discussions of depression, at least in print, suggested that depression oc-
curs when a behavioural system cannot be terminated. From this account it would be expect-
able for unresolved loss and, probably, unresolved trauma to be associated with depression.
Since the 1990s, Main’s ideas have served as the field’s dominant source of theory; though her
expressive range is astonishing for a research scientist, she barely ever mentions depression.
The result has been a gap in explanatory resources for considering why disorganised attach-
ment has so little relationship with depressive symptoms, but has well-replicated associ-
ations with externalising symptoms.543 Groh and colleagues glumly conclude that ‘given the
current state of the literature on attachment and internalizing symptoms, relatively little can
currently be concluded with confidence’.544 Perhaps something about the attachment system
or attachment-relevant information can overcome many aspects of depression and nonethe-

less function sufficiently effectively. Attachment researchers have generally conceptualised
depression as a kind of parental unavailability, but maybe this assumption is too quick or
oversimple in some way. Or perhaps only some aspects of depression interact with the at-
tachment system and the processing of attachment-relevant information; this might be sug-
gested by Bowlby’s distinctions between levels of defensive exclusion (Chapter 1). It seems
likely that a requisite of further theory development in this area will be close attention to the
constructs of disorganised attachment and unresolved/disorganised/disoriented states of
mind, in order to better understand why they would have little association with depression.
Conclusion
See Table 3.1 in the Appendix to this c hapter for consideration of key concepts discussed here.
The theory, methodology, and even the basic conceptualisation of the research object of
attachment research entered a new era with the innovations introduced by Main and col-
leagues in the 1980s. Yet criticisms of attachment theory and research, for example by
anthropologists (Chapter 2), take Bowlby and Ainsworth as their targets and rarely dem-
onstrate direct knowledge of the work of Main and colleagues. Indeed, the ideas and ap-
proaches of the Berkeley group are surprisingly unknown in the wider academic community
and among the general public, and much of what is circulated has been cut-price, simplified
versions mistaken for the technical form (‘allodoxia’; Chapter 1).
543 Unresolved/disorganised states of mind are associated with depression. But it is not clear how much this is
driven by preoccupied states of mind. See Dagan, O., Facompré, C.R., & Bernard, K. (2018) Adult attachment rep-
resentations and depressive symptoms: a meta-analysis. Journal of Affective Disorders, 236, 274–90.
544 Groh, A.M., Roisman, G.I., van IJzendoorn, M.H., Bakermans-Kranenburg, M.J., & Fearon, R.P. (2012)
The significance of insecure and disorganized attachment for children’s internalizing symptoms: a meta-analytic
study. Child Development, 83(2), 591–610. More recently, Bakermans-Kranenburg and van IJzendoorn issued a
call to examine subtypes of depression, to support finer-grained hypothesis generation and testing regarding the
mechanisms that link or do not link unresolved/disorganised attachment and depression. Reiner, I., Bakermans-
Kranenburg, M.J., van IJzendoorn, M.H., Fremmer-Bombik, E., & Beutel, M. (2016) Adult attachment representa-
tion moderates psychotherapy treatment efficacy in clinically depressed inpatients. Journal of Affective Disorders,
195, 163–71, p.169.The call arises from a study conducted with 43 clinically depressed adults in an inpatient unit.
The researchers found that patients with higher scores for security-autonomy on the AAI at admission to the in-
patient unit saw greater improvements in their depression than other patients. No association was found with
U/d on the AAI. Bakermans-Kranenburg and colleagues also highlighted the heterogeneity of the category of
depression in Cao, C., Rijlaarsdam, J., van der Voort, A., Ji, L., Zhang, W., & Bakermans-Kranenburg, M.J. (2017)
Associations between dopamine D2 receptor (DRD2) gene, maternal positive parenting and trajectories of depres-
sive symptoms from early to mid-adolescence. Journal of Abnormal Child Psychology, 46(2), 365–79.
One contributing factor has been Main’s approach to disseminating her work. Main has
pursued little public engagement. Though she has sometimes delivered public lectures, she
has never given interviews to the popular media or written articles for popular venues.545 Key
ideas in print are often to be found as book chapters in now-obscure volumes (e.g. ‘Avoidance
in the service of proximity’), or were not published in English (e.g. ‘Attaccamento disor-
ganizzato/disorientato nell’infanzia e stati mentali dissociati dei genitori’, ‘Sicherheit und
Wissen’). Main’s two most comprehensive statements of her perspective and methodological
innovations were not published at all (Behaviour and the Development of Representational
Models of Attachment, 1986; Four Patterns of Attachment Seen in Behaviour, Discourse and
Narrative, 1995). Pivotal statements on the meaning of the ‘state of mind regarding attach-
ment’ construct also remained unpublished (e.g. the AAI coding manual; ‘Interview based
adult attachment classifications: Related to infant–mother and infant–father attachment’).

The Berkeley–Leiden Adult Attachment Questionnaire Unresolved States of Mind scale and
several empirical studies using it (Chapter 5) all remained unpublished. Main’s amendments
to the protocol manual for the Strange Situation are only available in manuscript to those
attending training institutes. The manuals for the FR coding system and the six-year sys-
tems remain unpublished. In accepting the criticisms raised in the present chapter, Main
and Hesse have pledged that they will attempt to get more of these texts into print, to re-
duce their contribution to misapprehensions about their work.546 This will include the AAI
coding manual, which is currently 304 pages (though no doubt will be longer by the time it
hits print).
Over the past 30 years or so, the restricted circulation of the work of Main and colleagues
has made it difficult for the developmental research community drawing on the theory and
measures originating in Berkeley to access, understand, and scrutinise their own constructs.
Over the decades, the most direct access to the ideas and methods of developmental attach-
ment research as oriented by Main has not been through published papers, but through
fortnight-long in-person trainings in the AAI and Strange Situation which serve as oppor-
tunities for enculturation in an oral culture and its interpretive framework.547 As we saw in
Chapter 2, van IJzendoorn had already identified the potential risks of inaccessibility and
insularity associated with an oral culture as one of the legacies of Ainsworth. One of the
legacies of Main and colleagues has been an intensified reliance on an oral culture among re-
searchers in the developmental tradition, such that full social citizenship of the community
has been partly based on graduation from methods training institutes. This worked against
integration between attachment researchers and the wider discipline of psychology: pub-
lications in journals such as Child Development or Psychological Bulletin by attachment re-
searchers time and again reported results from Main’s measures, but without the space in
methods sections and discussion to elaborate fully what these actually meant. Even the basic
explanation of measures like the AAI already requires an unusual amount of space for em-
pirical science journals. It has been difficult to justify yet more exposition of technical detail
in order to convey the concepts and theory held within the baroque coding systems.
545 Main’s acute concern with detail has likely made the compromises and simplifications of popularising dis-
course especially unappealing. A partial exception is an interview with Main and Hesse conducted by Dan Siegal,
available on YouTube: https://www.youtube.com/watch?v=YJTGbVc7EJY.
546 Mary Main and Erik Hesse, personal communication, August 2019.
547 For details of the availability of training in the coding systems of Main and colleagues for the AAI and Strange
Situation procedure see https://attachment-training.com/.

Conclusion 325
The limited circulation of detail about the Berkeley measures reduced—though cer-
tainly did not halt—their exposure to technical criticism, and to attempts at psychometric
appraisal or modification, by researchers outside of Main’s personal network.548 It also left
concepts such as ‘state of mind regarding attachment’ and ‘disorganised attachment’ circu-
lating widely through the discipline of psychology severed from the context that would have
clarified their sense.549 Neither concept, for instance, has been well understood by social
psychological attachment researchers (Chapter 5), since few—if any—social psychologists
have received socialisation into the oral culture of the developmentalists or gained reliability
in coding their measures. The restricted circulation of Main’s texts and full detail on her ideas
and measures also limited their effective reception beyond psychology. In particular, it has
made it difficult for attachment assessments in the developmental tradition to enter cognate
and applied areas, such as social work research and empirical inquiry in family law.550 The

measures were not built or adapted for travel, and so therefore generally did not, except in
simplified, summary form.
There was one significant exception to the restricted circulation of the ideas of Main and
colleagues. Main published several of her important publications in clinical journals, con-
tributing to an exceptionally warm reception for her work, especially among psychoanalytic
clinicians, and more recently among social workers. Her work was also advocated to the
clinical community by popularisers such as Dan Siegel and David Shemmings.551 The narra-
tive of infant disorganised attachment as caused by frightening caregiver behaviour, and as
leading to controlling-caregiving or controlling-punitive behaviour and to a wide range of
mental health problems, has circulated very widely.552 The AAI has also seen some clinical
application, especially among clinicians with a background in psychoanalytic practice.553
The contributions of Main and her colleagues have played a significant role, alongside the
548 An early example was the elaboration of Main’s incomplete metacognition scale into the reflective func-
tioning scale by Fonagy, Howard and Miriam Steele, and colleagues. Fonagy, P., Steele, M., Steele, H., Moran, G.S.,
& Higgitt, A.C. (1991) The capacity for understanding mental states: the reflective self in parent and child and
its significance for security of attachment. Infant Mental Health Journal, 12(3), 201–218. There have also been
modifications by researchers who attended AAI training with Main, but who are more distant from her personal
networks. These include Lyons-Ruth, K, Yellin, C., Melnick, S., & Atwood, G. (2005) Expanding the concept of un-
resolved mental states: hostile/helpless states of mind on the Adult Attachment Interview are associated with dis-
rupted mother–infant communication and infant disorganization. Development & Psychopathology, 17(1), 1–23;
Roisman, G., Fraley, R., & Belsky, J. (2007) A taxometric study of the Adult Attachment Interview. Developmental
Psychology, 43(3), 675–86; Crittenden, P.M. & Landini, A. (2011) Assessing Adult Attachment: A Dynamic-
Maturational Approach to Discourse Analysis. New York: Norton. For discussion of recent psychometric evaluation
of the AAI see Chapters 4 and 6.
550 Nina Koren-Karie is one of very few senior academics based in a social work department to regularly use
assessments of attachment from the developmental tradition. Koren-Karie was mentored by Sagi-Schwartz in the
psychology department at Haifa and has a doctorate in psychology, which helped this movement of knowledge
from developmental psychology into social work research.
551 Siegel, D.J. (1999) The Developing Mind. New York: Guilford; Shemmings, D. & Shemmings, Y.
(2011) Understanding Disorganized Attachment: Theory and Practice for Working with Children and Adults.
London: Jessica Kingsley; Shemmings, D. (2016) Making sense of disorganised attachment behaviour in preschool
children. International Journal of Birth and Parent Education, 4(1).
552 See e.g. Howe, D., Brandon, M., Hinings, D., & Schofield, G. (1999) Attachment Theory, Child Maltreatment
and Family Support: A Practice and Assessment Model. London: Macmillan; Golding, K.S. (2014) Nurturing
Attachments Training Resource: Running Parenting Groups for Adoptive Parents and Foster or Kinship Carers.
London: Jessica Kingsley.
553 Steele, H. & Steele, M. (eds) (2008) Clinical Applications of the Adult Attachment Interview.
New York: Guilford; Crowell, J.A. (2014) The Adult Attachment Interview. In S. Farnfield & P. Holmes (eds) The
Routledge Handbook of Attachment: Assessment (pp.144–55). London: Routledge.
contribution of others such as Fonagy and Holmes,554 in persuading the clinical community
to attend to attachment research, something Bowlby himself felt that he had failed to achieve.
Chefetz, for instance, described Main and Hesse’s work as an ‘explanatory godsend’ for clin-
icians, offering a ‘brilliant glow’ to difficult clinical phenomena.555
Among the second generation of developmental attachment research, the work of Main
and colleagues has had as much influence as that of Bowlby and Ainsworth. The Berkeley
group have been described as having ‘unprecedented resonance and influence’.556 In his final
years, Bowlby described Main’s contributions to attachment theory as ‘impressive’, ‘clinically
sophisticated’, and ‘striking’.557 When a critic spoke out at a conference against the complexity
of the descriptions of children and adults that Main was compacting within categories of in-
fant, child, and adult attachment, Bowlby replied ‘it is a language of its own, and one well-
worth learning’.558 Alongside such support from Bowlby, Main has widely been regarded as

receiving the baton of method-giver from Ainsworth, who used the Berkeley innovations in
theory and methodology in her final publications (Chapter 2). As such, positions assumed to
be held by Main have had a powerful authority for the second generation of attachment re-
searchers, delimiting the parameters of methodology, and therefore to an extent the param-
eters of theory.559 Unfortunately, the stances ascribed to Main frequently differ from, flatten,
or petrify into cliché the beliefs of an actual individual, Mary Main, whose ideas have an
underpinning architecture that is not widely understood.
Michael Rutter ascribed to Main and colleagues two of the five great advances in psych-
ology contributed by research in attachment on the basis of their introduction of the infant
disorganised attachment classification for the Strange Situation procedure, and the AAI.560
However, these two methodological innovations were not wholly distinct. They both re-
flected Main’s universal model of human emotion regulation, with special primacy given to
the role of attentional processes in the formation of individual differences. As discussed in
Chapter 2, the idea of behavioural systems was the basis for the Strange Situation procedure
but then became interred within it. Likewise, the theory of attentional processes, the ‘guess
and uncover’ approach, and the approach to evolutionary theory which formed the platform
for Main’s methodological innovations were all buried inside them. It might be thought that
the notion of minimising and maximising strategies has been enormously popular among
554 Holmes, J. (1996) Psychotherapy and memory—an attachment perspective. British Journal of Psychotherapy,
13(2), 204–218; Fonagy, P. (2000) Attachment Theory and Psychoanalysis. London: Karnac.
555 Chefetz, R.A. (2004) Re-associating psychoanalysis and dissociation: a review of Ira Brenner's ‘Dissociation
of Trauma: Theory, Phenomenology, and Technique’. Contemporary Psychoanalysis, 40(1), 123–33, p.130.
556 University of Haifa (2011) Honorary doctorate awarded to Prof. Mary Main. https://web.archive.org/web/
20160602061335/http://newmedia-eng.haifa.ac.il/?p=5107.
London: Routledge. The quotes are from, respectively, p.138, p.139, and p.147.
558 Bowlby, cited in Steele, H. & Steele, M. (1998) Response to Cassidy, Lyons-Ruth and Bretherton: a return to
exploration. Social Development, 7(1), 137–41, p.141.

559 Bourdieu, P. (1971, 1991) Genesis and structure of the religious field. Comparative Social Research,13, 1–45;
Bourdieu, P. (1971, 1993) The market of symbolic goods. In R. Johnson (ed.) The Field of Cultural Production.
New York: Columbia University Press, pp.112–42; Bourdieu, P. (1999, 2011) With Weber against Weber, trans.
S. Susen. In S, Susen & B.S. Turner (eds) The Legacy of Pierre Bourdieu: Critical Essays. London: Anthem Press,
pp.111–24; Bourdieu, P. (2004) The Science of Science and Reflexivity, trans. R. Nice. Chicago: University of
Chicago Press.
560 The other three were John Bowlby’s abandonment of the notion of ‘love’ in favour of finer-grained concepts,
the Ainsworth Strange Situation, and the introduction of the attachment disorder diagnosis within the DSM. See
Rutter, M., Kreppner, J., & Sonuga-Barke, E. (2009) Emanuel Miller Lecture: attachment insecurity, disinhibited
attachment, and attachment disorders: where do research findings leave the concepts? Journal of Child Psychology
and Psychiatry, 50, 529–43.
Conclusion 327
attachment researchers. But—seemingly without realising that there is a distinction—the

version that has come into use has been Cassidy’s notion as minimising or maximising of
attachment as individual strategies, rather than Main’s version as minimising or maximising
attention to attachment-relevant information as part of a repertoire of behaviours granted
by human evolutionary history. Main’s attentional theory has been little recognised, and has
seen almost no use, no critical scrutiny, and no operationalisation.
A long-time collaborator of Cassidy’s, Yair Ziv has considered the legacy of Ainsworth
and Main, and appraised the overarching authority given to the four-category Strange
Situation and AAI within the developmental tradition of attachment research. Ziv and
Hotam argued that in the wider context of academic psychology, overarching theories have
become rare, increasingly looked down on as departing or even distracting from the em-
pirical task of psychological research.561 They observed that, within the field of attachment

research, ‘scholars use the research tool as their guide, and, probably unintentionally, as the
theory’s ‘surrogate’, authorised by the predictive power of the methods.562 Ziv and Hotam
alleged that attachment research has become largely a hunt for the correlates of the Strange
Situation and AAI: further theoretical developments have been blocked by the identification
of methodological tools with theoretical constructs, combined with the conviction that the
methodology cannot be refined without undermining the basis for a cumulative research
paradigm.563 They express scepticism that methodological innovation—they give the ex-
ample of the use of Ainsworth’s dimensional scales advocated by Fraley and Spieker—would
really represent a threat to the possibility of cumulative attachment research, though they
know that others would disagree.564
It is plausible, as Ziv and Hotam argued, that attachment research has partially circum-
vented the decline of theory in academic psychology by embodying theory within meth-
odology, protecting theoretical propositions whilst supplanting the need to explicitly
acknowledge or discuss them.565 However, other factors may be mentioned as having ob-
structed understanding of the position of Main and colleagues. These include confusions
regarding the use of terminology, the complexity and gestalt-like quality to the coding of
observational measures of attachment, and restrictions on the accessibility of Main and col-
leagues’ texts containing details of methods and theory.
561 Some similar proposals have been made in Flis, I. (2018) Discipline through method: recent history and
philosophy of scientific psychology (1950–2018). Unpublished doctoral thesis, University of Utrecht. In a corpus
analysis of the 2,046 psychology articles published to date in Frontiers in Psychology, Beller and Bender found that
only 8.3% feature reference to a specific theory. A small number of theories were cited in more than three articles
among the 2,046, suggesting their ongoing relevance and durability. Among these were, unsurprisingly, ‘prob-
ability theory’ and ‘evolutionary theory’ as underpinning aspects of any contemporary life science. There were
also some domain theories such as the ‘theory of planned behaviour’. However, the most frequent of all theories
mentioned in at least three articles was ‘attachment theory’. It was also the only theory from developmental psych-
ology mentioned by at least three articles in the corpus. Beller, S. & Bender, A. (2017) Theory, the final frontier?
A corpus-based analysis of the role of theory in psychological articles. Frontiers in Psychology, 8, 951.
562 Ziv, Y. & Hotam, Y. (2015) Theory and measure in the psychological field: the case of attachment theory and
the strange situation procedure. Theory & Psychology, 25(3), 274–91, p.278.
563 Ibid. p.283.
564 Researchers will likely differ on the extent to which they believe that there can be innovations that would add
value, reward the resource investment, and still be plugged into the same meta-analyses. Kuhn, T.S. (1977) The
Essential Tension: Selected Studies in Scientific Tradition and Change. Chicago: University of Chicago Press.
565 The claims of Ziv and Hotam are restricted to the developmental tradition of attachment research, within
which they were trained. However, the argument that method has incorporated and supplanted theorising can
be applied to the social psychological tradition, where the two dimensions of anxiety and avoidance have formed
the central theoretical framework over the past two decades, stabilised and protected by the Experiences of Close
Relationships measure (Chapter 5).
The movement of Main’s ideas across disciplinary spaces has occurred at the price of a
mangled and simplified image of the methods and theory of the Berkeley group. Much like
the Strange Situation was misadministered in Japan by researchers without the social links or
deep theoretical knowledge to scaffold its appropriate use (Chapter 2), Main and colleagues
have seen misapplications of their work in child welfare contexts.566 This predicament can
be regarded at least in part as resulting from a combination of the compelling, magnetic,
and absorbative quality of the ideas such as attachment, fear, disorganisation, trauma, and
dissociation—combined with problems in the effective circulation of the technical ideas and
methods of Main and colleagues outside the communities of academic and psychoanalytic
psychology. The characterisation of disorganisation as a category may also have inadvertently
helped it meld with the diagnosis-focused infrastructures of clinical and welfare investiga-
tions of families. Third-generation attachment researchers have tended to avoid categories in

their adaptations of attachment measures for use by practitioners, not only because dimen-
sions often have better psychometric properties, but also to try to avoid the problems they
have seen with the reception of Main and Solomon’s disorganised classification.567
Main has had health difficulties at times which have reduced her capacity to write, to re-
spond to questions and clarify ambiguities, as well as—until recently—to offer trainings (or
train a trainer) in the six-year systems or the frightening/frightened coding system. This
has been compounded by her perfectionism and concern with detail, which has made it ex-
tremely difficult for Main to judge a coding system or a piece of writing finished enough to
go into circulation. As a result, over the decades vast swathes of her writing has remained
unpublished, misunderstandings have abounded, and there have been barely a handful of
accredited coders in any of the six-year systems or the frightening/frightened coding system.
The infant disorganised coding system was also left in an underdeveloped state that con-
tributes to it being difficult to learn; again, there are now remarkably few active, accredited
coders.568 Though dedicated researchers have found workarounds, such as the use of alter-
native assessments, the lack of training and reliable coders have obstructed lines of empirical
inquiry in the developmental tradition of attachment research. For example, though Main
and Hesse’s work is cited as an inspiration in many attachment-based interventions, none
has used frightened/frightening caregiving as an outcome measure.569 (Main and Hesse have
said that they will soon be certifying four trainers, contributing to the future availability of
the FR coding system.)
566 Main, M., Hesse, E., & Hesse, S. (2011) Attachment theory and research: overview with suggested appli-
cations to child custody. Family Court Review, 49(3), 426–63, p.441. See also Granqvist, P. (2016) Observations
of disorganized behaviour yield no magic wand: response to Shemmings. Attachment & Human Development,
18(6), 529–33. For examples of the garbled reception and understanding of Main’s work in policy discourse see
e.g. Moullin, S., Waldfogel, J., & Washbrook, E. (2014) Baby Bonds: Parenting, Attachment and a Secure Base for
Children. London: Sutton Trust; All Party Parliamentary Group for Conception to Age 2 (2015) Building Great
Britons. London: Wave Foundation.
567 Madigan, S. (2019) Beyond the academic silo: collaboration and community partnerships in attachment re-
search. Paper presented at International Attachment Conference, Vancouver, 20 July 2019.

569 Dozier’s group repeatedly described the central influence of Main and Hesse’s work on their Attachment and
Biobehavioural Catch-up intervention. However, in the absence of training in the FR system, they used an alter-
native assessment of caregiving behaviour in evaluating it: the Bronfman, Parsons, and Lyons-Ruth AMBIANCE
coding system. Yarger, H.A. (2018) Investigating longitudinal pathways to dysregulation: the role of anomalous
parenting behaviour. Unpublished doctoral dissertation, University of Delaware.
Appendix 329
Restricted circulation of knowledge of the Berkeley measures has fuelled misunder-

standings of Main’s theoretical position, which is closely tied to understanding the
meaning of the assessments she and colleagues developed. There are vast discrepancies
between the assumed positions generally ascribed to ‘Mary Main’ regarding the four-
category mode of individual differences, and the theoretical positions available on a close
reading of Main’s texts and coding manuals. These discrepancies are especially located
around the unrecognised technical meanings given to key elements of Main’s conceptual
vocabulary, including ‘disorganisation’, ‘fear’, ‘coherence’, ‘preoccupation’, ‘unresolved’, and
‘dissociation’. Yet there have been some researchers aware of Main’s position and trained
in use of her measures. One set of researchers with strong understanding and close per-
sonal ties to Main and Hesse has been the Minnesota group. The work of Sroufe, Egeland,
and colleagues with the Minnesota Longitudinal Study of Risk and Adaptation would

prove just as influential as the Berkeley study for shaping the second generation of attach-
ment research, providing independent validation of the measures developed in Baltimore
and Berkeley, and reporting findings to support and to qualify the claims by Bowlby,
Ainsworth, Main, and Hesse.
Appendix
Table 3.1 Some key concepts in Main and Hesse’s writings
Concept Mistaken for Explanation in longhand
Categories for Four boxes Ainsworth identified differences in the responses of

coding individual representing infants in the Strange Situation to separation from and
differences in different ‘kinds’ reunion with their familiar caregiver. These differences
attachment of attachment were characterised through various means, including a
variety of interactive rating scales, two latent dimensions
in a discriminant function analysis, and eight subtypes.
These eight subtypes were pragmatically grouped into
three categories. However, Ainsworth anticipated that
more categories would be identified on the basis of
further research.
Main, however, proposed that the Ainsworth
Strange Situation categories were a local form of
a wider phenomenon. This was that the output of
any behavioural system—including the attachment
system and the caregiving system—could be increased
or decreased based on how individuals direct their
attention. Where individuals direct their attention
away from the cues that might otherwise activate the
behavioural system, the system is difficult to initiate and
its output minimised. When individuals are vigilant
to cues that would activate the behavioural system,
the output of that system is maximised and the system
difficult to terminate. She termed these ‘conditional
strategies’ to highlight that these responses may not
directly express the behavioural system, but would
nonetheless likely have evolved because they helped
an individual survive under certain kinds of adverse
circumstances.
(continued )
Table 3.1 Continued

Main presumed there would be three forms of
expression of a behavioural system: a direct expression,
its minimisation, or its maximisation. Where the
expression of a behavioural system appeared to
be undermined by conflict or confusion, Main
characterised this with a fourth classification—
‘disorganisation’. However, no psychometric analysis was
conducted to assess whether disorganisation was best
considered as a category. The attachment categories are
therefore best regarded as pragmatic tools for describing
individual differences observed when using attachment
measures, and embodying Main’s thesis of minimising

and maximising strategies.
In unpublished written guidance provided to those
learning to code the Strange Situation at Minnesota,
Main advises that secure attachment in the Strange
Situation is coded on the basis of behaviours that
appear to flexibly direct attention to the caregiver or
the environment depending on the situation. Avoidant
attachment is coded on the basis of behaviours that
direct attention away from the caregiver and towards the
environment. Ambivalent/resistant attachment is coded
on the basis of behaviours that direct attention away
from the environment and vigilance is maintained about
the caregiver’s availability. Disorganised attachment is
coded on the basis of behaviours that suggest conflict or
confusion in the direction of attention either towards the
caregiver or to the environment.
Avoidance (for Ainsworth’s Ainsworth identified that some infants in the Strange
children) attachment Situation avoided approaching their caregiver on
Dismissing (for classification reunion; many turned away from the caregiver towards
adults) for infants who the toys, or drew the caregiver’s attention to a toy.
physically avoid Main developed a theory that these infants were
their mother directing their attention away from cues to the activation
on reunion. of the attachment system. In doing so, these infants
Extrapolated were not just physically avoiding the caregiver. They
metaphorically were also avoiding the conflict that would be evoked by
to characterise expectations about rebuff from the caregiver based on
speakers in the AAI past experiences, and the activation of the attachment
who ‘avoid’ the system which would prompt proximity-seeking.
topic of attachment
relationships Main drew on this account in developing the AAI
coding system. With colleagues, she developed a
‘dismissing’ classification for speakers who, in an
autobiographical interview, appeared to be directing
attention away from attachment-relevant information
about their past.
Both avoidant behaviour in the Strange Situation and
dismissing discourse in the AAI were understood
to represent use of a strategy that cuts off attention
to environmental cues that would otherwise feed
information to a behavioural system—thereby avoiding
disruption of this system and the ensuing reduced
responsiveness to the environment.
Appendix 331
Table 3.1 Continued

Ambivalence/ Ainsworth’s Ainsworth identified that some infants in the Strange
resistance (for attachment Situation showed distress and frustration with their
children) classification for caregivers on reunion, and were not readily able to be
Preoccupation infants who show comforted by their caregiver.
(for adults) anger towards their Main developed a theory that these infants were
mother on reunion. directing their attention vigilantly towards potential
Extrapolated cues to the activation of the attachment system, and
metaphorically away from cues to the termination of the system. The
to characterise continual activation of the system was regarded as
speakers in the prompting frustration. In turn, frustration was seen as
AAI who remain helping to maintain vigilance, and also to maintain the
angry with

attention of the caregiver.
their childhood
attachment figures Main drew on this account in developing the AAI
coding system. With colleagues, she developed a
‘preoccupied’ classification for speakers who, in an
autobiographical interview, appeared to be directing
attention and frustration towards attachment-
relevant information about their past at the expense of
cooperation with the interviewer in answering their
exact questions.
Both ambivalence/resistance in the Strange Situation
and preoccupied discourse in the AAI were understood
to represent a strategy that intensifies attention to
environmental cues that would feed information to
initiate a behavioural system, and the cutting off of
attention from cues that would terminate the system.
The unsatisfied system prompts frustration; at the
same time, frustration may help maintain attention on
relevant cues and attracting the attention of attachment
figures.
Disorganisation Random, chaotic The term ‘disorganisation’ has regularly been used in five
behaviour and different ways by Main. She did so to draw links between
mental states behaviour and mental processes, and to identify
characteristic of potential similarities in mental processes across the life
certain infants course.
in the Strange A first use of the term has been as an umbrella term for
Situation. The three different kinds of behaviour shown towards the
effect of child object of a behavioural system: conflict, confusion, and/
maltreatment or apprehension. These behaviours may, for instance, be
shown by infants towards their caregiver in the Strange
Situation, where the attachment system is presumed to
be activated by the separations and reunions.
A second use of the term has been to characterise a
significant disruption of a behavioural system. It is this
(invisible) disruption at the level of motivation that is
presumed to cause the visible conflicted, confused, or
apprehensive behaviour seen, for instance, in the Strange
Situation.
A third use of the term has been as a category label for
infant–caregiver dyads seen in the Strange Situation,
where conflicted, confused, and/or apprehensive
behaviour is seen to a significant degree.
(continued )
Table 3.1 Continued

A fourth use of the term was as the category label
for controlling-punitive and controlling-caregiving
behaviour in the Main and Cassidy six-year reunion
system. The behaviour was generally smoothly
sequenced and goal-oriented, and often resulted in
some form of caregiver availability—so it was not
technically disorganised at a behavioural level. However,
Main and Cassidy used the term ‘disorganised’ to
signal developmental continuities from infancy, and
to highlight that controlling-punitive and controlling-
caregiving behaviour likely arises in the context of
disruption to the child–caregiver relationship and its

usual hierarchies.
A fifth use of the term has been to characterise the
psychological process indicated by unresolved loss and
trauma in the AAI. In their 1992 paper, Main and Hesse
conceptualised this state as representing an overloading
of working memory, as attention cannot be directed
coherently either away from or towards attachment-
relevant information.
Disorganised Random, chaotic Main and Solomon defined seven indices of
attachment behaviour and disorganised attachment for coding the Strange
behaviour mental states Situation in their 1990 chapter. The first five were
characteristic of drawn from Hinde’s 1966 discussion of ‘conflict
certain infants behaviours’: behaviours shown when an organism
in the Strange experiences strong, conflicting motivations. The sixth
Situation was apprehension of the caregiver. The seventh was
behaviours suggestive of confusion or disorientation. It
was not assumed that all of these would mean the same
thing. In Main and Hesse’s later work, they suggested
that many of the behaviours would to varying degrees
reflect a history of alarming behaviours by caregivers.
They also suggested that some, though certainly not all,
may entail dissociative processes.
Attachment- A child scared of The term ‘fear’ has a variety of connotations. Main and
related fear a frightening or Hesse used the term in a technical sense, to describe the
abusive parent situation in which an attachment figure, their behaviour,
or their absence in a major separation has been a source
of alarm for a child. When they use the phrase ‘fright
without solution’, this was intended to highlight that if a
child is alarmed, the attachment system would prompt
them to approach their caregiver. However, if the
caregiver is themselves associated with alarm, then the
child is placed in a conflict situation.
In characterising ‘fear without solution’ as one pathway
to disorganised attachment, Main and Hesse were not
intending to imply that the caregiver has necessarily
frightened the child directly. A situation in which a child
observed violence perpetuated by their father towards
their mother might also lead them to associate their
mother with feelings of alarm.
Appendix 333
Table 3.1 Continued

Attachment Stable images held Main regarded the AAI as assessing the speaker’s
representations by people about capacity for flexibility in organising information relevant
their attachment to attachment and for obtaining or limiting access to
figures that information. Secure-autonomous speakers could
answer questions about their childhood fluently whilst
also cooperating with their interviewer, paying attention
to what is relevant for them, what level of detail is helpful
and not overwhelming or confusing, etc. Dismissing
speakers directed attention away from attachment-
relevant experiences. Preoccupied speakers directed
attention towards their dissatisfaction with attachment-

relevant experiences at the expense of cooperation with
the interviewer.
Over time, Main characterised these individual
differences in various ways: sometimes as ‘internal
working models’, sometimes as ‘attachment
representations’, and sometimes as ‘states of mind
regarding attachment’. None of these terms is an
especially effective characterisation of what Main
thought she had identified, and there remains a good
deal of confusion about their meaning today. Generally,
attachment researchers today use the terms ‘attachment
representations’ and ‘state of mind regarding
attachment’ to mean individual differences as measured
by the AAI.
Coherence of Ordered and Coherence represents an individual’s presentation
discourse meaningful speech of a plausible, unified, yet free-flowing picture of
experiences, feelings, and viewpoints—as well as their
potential causes—within the AAI.
Main anticipated that individuals whose speech
displayed these characteristics would also meet Grice’s
maxims for cooperative discourse with the interviewer:
Quality—‘be truthful and have evidence for what you say’
Quantity—‘be succinct, and yet complete’
Relation—‘be relevant to the topic as presented’
Manner—‘be clear and orderly’
Lack of resolution A trauma or loss Lack of resolution is a technical term used by Main,
that still preys Goldwyn, and Hesse to characterise a particular
on the mind of phenomenon in the AAI. There are many ways in
a speaker, for which a past loss or trauma might influence the present.
instance causing For instance, individuals may block all thoughts of it.
denial or distress Or they may become distressed just thinking about
when it is raised it. Neither of these is lack of resolution in Main and
colleagues’ technical sense. In both cases, attention has
been directed in an effective way away from, or towards,
the loss or trauma. The loss or trauma is not resolved.
But it is not technically ‘unresolved’ in the meaning
intended by Main and colleagues.
(continued )
Table 3.1 Continued

Lack of resolution occurs when the speaker’s response
when discussing a bereavement or a determinate
traumatic event suggests disorganisation or
disorientation in the capacity to attend to it. It is coded
primarily on the basis of lapses in reasoning or discourse
in discussions of childhood attachment relationships,
from which a coder is able to infer this disorganisation
or disorientation in mental processing around the loss
or trauma.
There are further technical qualifications on what is
meant by lack of resolution. The most important is

that the loss or trauma has to be a historically locatable
event. Hence there may be experiences that, despite
appearances, are not technically considered unresolved
even if relevant disturbances of reasoning or discourse
are present. These include: anticipated bereavement in
the case of a terminally ill attachment figure; experiences
of chronic neglect in childhood; exposure to an
atmosphere of emotional abuse between parents during
childhood but without there being a stand-out ‘event’;
and fear of an intimidating current partner or ex-partner.
Earned security Individuals’ ‘earned’ Early in their work on the AAI Main and colleagues
achievement of the found that participants who could speak coherently
benefits of secure about their early attachment relationships tended
early relationships towards two types. A first group had experienced
through their positive early relationships. A second group had
own efforts in experienced difficult childhoods; nonetheless, they
later childhood, spoke with flexible attention to the good and the bad,
adolescence, or and could turn fluently from considering memories of
adulthood the past to communication with the interviewer in the
present, taking stock of their own perspective and that of
others. Main and colleagues termed these latter speakers
‘earned secure’. The term is misleading, since it may seem
to imply that individuals were able to rise above their
early adversities through force of will. It also assumes
that the earned-secure speakers did, indeed, have a more
difficult childhood than non-earned-secure speakers; in
fact, the AAI cannot assure this. Roisman and colleagues
have been especially active in exploring whether an
earned-secure classification in the AAI is actually
associated with greater childhood adversity than other
speakers classified as ‘autonomous-secure’.
As a consequence, Main and Hesse have renamed the
classification ‘evolved secure’, to avoid the implication
that it is the achievement of individual willpower. To
date, attachment researchers have, however, largely
ignored this rechristening, perhaps due to concerns
about the term ‘evolved’. Main and Hesse have also
placed stringent criteria on the classification (a score of
lower than 2.5 on the scale for inferred loving parental
behaviour). So only speakers with credible evidence
of extremely adverse caregiving who nonetheless
show coherent speech can be considered as ‘evolved
secure’. However, these criteria are so stringent that few
individuals can be identified in most samples.
Appendix 335
Table 3.1 Continued

Beyond its exact operationalisation, the functional
meaning of the earned/evolved secure status in the
AAI for attachment researchers has been to signal
acknowledgement that states of mind regarding
attachment can change over time, and that these changes
themselves are potentially important.
Dissociation Blocked The term ‘dissociation’ is one with various meanings in
communication discussions of mental health. Main has tended to use the
between parts of term to mean one of two things. Either it has been used
the self to mean much the same as ‘segregation’ (see Bowlby’s
use of the term). Or it has been used to mean specific

behaviours suggestive of the segregation of perception,
such as fugue states.
Note: The table has been confirmed by Mary Main and Erik Hesse.
Illustrative statement: ‘Whereas avoidant attachment is associated with caregiver insensitivity and rebuff, Main’s
disorganised attachment classification is associated with frightening experiences of the caregiver. The caregivers of
infants with this classification often have attachment representations characterised by lack of resolution of loss or
trauma. An important mechanism underpinning this lack of resolution is proposed to be dissociation.’
Mistaken for: Main provided an exhaustive four-category system. Some infants, classified as avoidant, experi-
enced less warm care. There is another category, comprising of ambivalent infants who also resist their caregiver’s
attempts to comfort them. A further category, disorganised infants, experienced abusive care, breaking their at-
tachment system and causing random, chaotic behaviour in the Strange Situation. The behaviour of these care-
givers can be explained by their own adverse experiences, which still disturb them, and which cause them to
behave irrationally and dangerously towards their child.
Technical meaning: Some caregivers struggle to perceive and to interpret accurately the signals and commu-
nications implicit in an infant’s behaviour, and given this understanding, to respond to them appropriately and
promptly (i.e. sensitive care). For instance, caregivers may rebuff their infants’ attachment behaviour. Main sup-
posed that such behaviour by caregivers will expectably prompt a conditional strategy from the infant, in which at-
tention is directed away from the caregiver in the Strange Situation so as to maintain regulation and environmental
responsiveness. This is in contrast to another conditional strategy (ambivalent/resistant attachment) in which at-
tention is directed vigilantly towards potential cues to the activation of the attachment system and away from cues
to the termination of the system. However, both the primary (secure) strategy and the conditional strategies can
be disrupted—for instance by conflicting affects. There may be a variety of causes of such disruption and a variety
of forms of conflict, as Hinde had already shown. Main and Hesse identified that an especially important case is
conflict between the disposition to approach the caregiver when alarmed and experiences of the caregiver as them-
selves in some way alarming. One potential cause of such experiences is when caregivers are abusive. However,
other forms of caregiver behaviour can be alarming for infants, such as sudden and inexplicable lapses in a caregiv-
er’s attentional availability. Main and Hesse found that forms of alarming caregiving behaviour are more common
among caregivers who show marked lapses in discourse or reasoning in the AAI, as defined by a technical set of
criteria that identifies in these lapses indications of disrupted attentional processing of attachment-relevant infor-
mation when discussing specific alarming events in their history.
4
Alan Sroufe, Byron Egeland, and the Minnesota
Longitudinal Study of Risk and Adaptation
Biographical sketch
The Minnesota Longitudinal Study of Risk and Adaptation began in the early 1970s, initiated

by Byron Egeland as a study of a large cohort of mothers living in poverty. In the 1970s and
early 1980s, empirical attachment research had been primarily pursued by Ainsworth and
her students. The research group led by Alan Sroufe and Egeland was important in providing
a second pillar to hold up the paradigm from the 1970s to the present. The Strange Situation
was conducted with the sample by Sroufe and his graduate students Everett Waters and Brian
Vaughn. Early work by the research group documented the role of caregiving in shaping
patterns of attachment in the Strange Situation, and also the capacity of infant attachment
patterns to predict later social competence and mental health. Sroufe and Egeland created
an ‘electric atmosphere’ in their research group, as they provided the first longitudinal evi-
dence of the implications of attachment relationships.1 Students described their ‘imperturb-
able optimism’, ‘wisdom about human nature’, and ‘compassion’ as important qualities in the
creation of the atmosphere, along with the sense of contributing to meaningful and cutting-
edge developmental science.2 They were a great stabilising and integrative presence for the
field of attachment research. Though Egeland and Sroufe have now retired, research with the
Minnesota Longitudinal Study of Risk and Adaptation has continued.
Introduction
The Minnesota research group has been of critical importance to the establishment and de-
velopment of attachment research. Whilst to the general public Bowlby remains the most
visible and prominent representative of attachment research, within the American de-
velopmental science community this role has arguably been held by Sroufe and Egeland.
Mary Ainsworth wrote to Sroufe in 1982, expressing her deep gratitude: ‘One of the very
best things that ever happened to attachment research was your decision to participate in
it!’3 In the introduction to their six-volume edited work on attachment theory, Slade and
Holmes described the contribution of the Minnesota group as no less than a ‘revolution,’ one
that ‘brought attachment theory into mainstream academic psychology’.4 In part as a conse-
quence, attachment theory has at times been referred to as the Bowlby–Ainsworth–Sroufe
1 Dante Cicchetti interviewed by Robert Karen, cited in Karen, R. (1998) Becoming Attached: First Relationships
and How They Shape Our Capacity to Love. Oxford: Oxford University Press, p.167.
2 Marvinney, D. (1988) Sibling relationships in middle childhood: implications for social-emotional develop-
ment. Unpublished doctoral dissertation, University of Minnesota, p.i.

3 Ainsworth, M. (1982) Letter to Alan Sroufe, 27 March 1982. Mary Ainsworth papers, Box M3173, Folder 3.
4 Slade, A. & Holmes, J. (2014) Introduction. In Attachment Theory, Vol. 1 (pp.5–37). London: Sage, p.13.
338 Alan Sroufe and Byron Egeland
tradition.5 The primary scientific contribution of the Minnesota group under the leadership
of Sroufe and Egeland has been to show the developmental implications of early caregiving
and attachment. This was of essential importance for consolidating Ainsworth’s findings and
for showing why they mattered for development. Furthermore, on the basis of their em-
pirical research, Sroufe, Egeland, and colleagues made important theoretical contributions
to the conceptualisation of the role of attachment within development as a whole. Fonagy
would describe Sroufe and the Minnesota group as having played a ‘seminal role in ex-
tending the scope of attachment theory from an account of the developmental emergence
of a set of social expectations to a far broader conception of attachment as an organizer of
physiological and brain regulation’.6
In the 1940s, Bowlby felt unable to report on his clinical observations of child abuse
(Chapter 1). By the 1970s, however, the clinical community had come to acknowledge the

potential for parents to abuse and neglect their children.7 There was growing policy interest
in the role of abuse and neglect in contributing to later mental illness, aggressive behaviour,
and social difficulties. There was also growing interest in the lives of single-parent families
living in poverty, in the context of both changing family demography in America and the
political problematisation of these families.8 The Minnesota Longitudinal Study of Risk and
Adaptation began in the 1970s as the world’s first attempt to conduct a prospective study of
child abuse. Until that point, all research on child abuse had been pursued retrospectively.
This raised questions of validity, especially given the significant potential stigma for adults in
revealing experiences of childhood abuse. Egeland, Amos Deinard, and Ellen Elkin wrote a
proposal to the Centre of Maternal and Child Health to initiate a longitudinal cohort study.
Funding was provided in 1975 for three years in the first instance. The researchers’ belief
in the value of prospective research exploring abuse and neglect proved well founded: dec-
ades later, only half of the adults in their sample with prospectively documented histories of
childhood abuse self-reported that they had experienced maltreatment.9
Shortly before the study began, Arnold Sameroff visited the department at Minnesota and
shared a draft of his classic paper proposing an ‘ecological’ model in which human develop-
ment is shaped by the interaction between risk factors and environmental supports.10 This
model had a profound influence on the methodology of the study, which sought to identify
and capture various risk and protective factors in line with Sameroff ’s idea of development
5 E.g. Belsky, J. (2002) Developmental origins of attachment styles. Attachment & Human Development, 4(2),
166–70.
6 Fonagy, P. (2015) Mutual regulation, mentalization, and therapeutic action. Psychoanalytic Inquiry, 35(4),
355–69, p.358.
7 Hacking, I. (1991) The making and molding of child abuse. Critical Inquiry, 17(2), 253–88; Ferguson, H. (2004)
Protecting Children in Time: Child Abuse, Child Protection and the Consequences of Modernity. London: Palgrave.
8 Jones, J.P. & Kodras, J.E. (1990) Restructured regions and families: the feminization of poverty in the US.
Annals of the Association of American Geographers, 80(2), 163–83; Hancock, A.M. (2004) The Politics of Disgust: The
Public Identity of the Welfare Queen. New York: NYU Press.
9 Shaffer, A., Huston, L., & Egeland, B. (2008) Identification of child maltreatment using prospective and self-
report methodologies: a comparison of maltreatment incidence and relation to later psychopathology. Child Abuse
& Neglect, 32(7), 682–92. These results agree with later meta-analytic findings: Baldwin J.R., Reuben, A., Newbury,
J.B., & Danese, A. (2019) Agreement between prospective and retrospective measures of childhood maltreat-
ment: a systematic review and meta-analysis. JAMA Psychiatry, 76(6), 584–93.
10 The term ‘risk’ is multivalent and potentially obfuscating, as noted by Lupton, D. (1999) Risk.
London: Routledge. However, Sroufe’s usage has been clear and consistent throughout his career: Sroufe, L.A.,
Egeland, B., Carlson, E.A., & Collins, W.A. (2005) The Development of the Person. New York: Guilford: ‘A risk factor
is anything that increments the probability of some negative outcome’ for individuals or families within a given
population (28).
Introduction 339
as multiply determined.11 Particular attention was given to risk and protective factors in the
life of the child’s caregiver or caregivers. Egeland and colleagues recruited 267 women in
the third trimester of pregnancy and living in poverty.12 However, in line with Sameroff ’s
ideas, Egeland and colleagues were intent on assessing the multiple adversities faced by the
families, documenting the widespread prevalence of domestic violence, mental health prob-
lems, and drug and alcohol use. Forty-one percent of the mothers had not finished high-
school education;13 86% of the pregnancies were unplanned, with a third of families having
obtained no equipment necessary for the baby or made arrangements for somewhere for the
baby to sleep;14 the mothers were often socially isolated, with very low average levels of social
support;15 59% were single at the time of their baby’s birth, compared to a 13% national
average in 1975;16 and only 13% of the fathers were living in the same home as their children
by the time they had reached 18 months of age. Soon after the study began, Egeland con-

tacted Sroufe, a colleague at Minnesota, and together they conducted the Ainsworth Strange
Situation with the sample, with input from two graduate students—Everett Waters and Brian
Vaughn.17 The Strange Situation was attractive as a procedure offering a window into the
experiences of care and emotional development of the children in the sample. And the clas-
sifications for the procedure had the initial basis for construct validity through Ainsworth’s
Baltimore study (Chapter 2). For his dissertation Waters also conducted a parallel study of
50 mother–infant dyads from Minnesota who were not living in poverty. This allowed the
Minnesota group to draw some comparisons to explore the role of socioeconomic adversity
in predicting outcomes of interest, such as distribution of classifications in the Ainsworth
Strange Situation.18
Egeland, Sroufe, and colleagues in the Minnesota group were sensitive to the cultural
and moral context of developmental research with a sample facing such adversities.19 One
such factor was acrimonious debates at the time over whether abuse was in fact harmful to
children’s development, or whether associations between abuse and with negative outcomes
could better be attributed to poverty.20 One measure the Minnesota group took to respond to
these controversies was to expend every effort to carefully measure both abuse and adversity
11 Sameroff, A.J. & Chandler, M.J. (1975) Reproductive risk and the continuum of caretaking casualty.
Review of Child Development Research, 4, 187–244; Mangelsdorf, S.C. (2011) The early history and legacy of the
Minnesota Parent–Child Longitudinal Study. In D. Cicchetti & G.I. Roisman (eds) The Minnesota Symposia on
Child Psychology, Volume 36: The Origins and Organization of Adaptation and Maladaptation (pp.1–12). Hoboken,
NJ: Wiley.
12 Egeland, B. & Sroufe, L.A. (1981) Attachment and early maltreatment. Child Development, 52(1), 44–52.
13 Pianta, R.C., Sroufe, L.A., & Egeland, B. (1989) Continuity and discontinuity in maternal sensitivity at 6, 24,
and 42 months in a high-risk sample. Child Development, 60, 481–7.

14 Egeland, B., Jacobvitz, D., & Papatola, K. (1987) Intergenerational continuity of abuse. In R.J. Gelles & J.B.
Lancaster (eds) Child Abuse and Neglect: Biosocial Dimensions. New York: Aldine De Gruyter, 255–76, p.259.
15 Appleyard, K., Egeland, B., & Sroufe, L.A. (2007) Direct social support for young high risk children: relations
with behavioral and emotional outcomes across time. Journal of Abnormal Child Psychology, 35(3), 443–57: ‘27%
of the sample had only one or two individuals in their network’ (447).
16 US Bureau of the Census (1975) Marital Status and Living Arrangements: March 1995 Update. Washington,
DC: US Government Printing Office. https://www.census.gov/library/publications/1975/demo/p20-287.html.

17 The coders of the Strange Situations conducted at 12 months were Waters and Vaughn; the coders of the 18-
month procedures were Sroufe and Lyle S. Joffe.

18 Vaughn, B., Egeland, B., Sroufe, L.A., & Waters, E. (1979) Individual differences in infant–mother attachment
at twelve and eighteen months: stability and change in families under stress. Child Development, 50(4), 971–5.
19 Sroufe, L.A. (1970) A methodological and philosophical critique of intervention- oriented research.
20 Elmer, E. (1977) Fragile Families, Troubled Children: The Aftermath of Infant Trauma. Pittsburgh: University of
Pittsburgh Press, discussed in Sroufe, L.A., Egeland, B., Carlson, E.A., & Collins, W.A. (2005) The Development of
the Person. New York: Guilford, p.51.
in a variety of ways, building convergent validity piece by piece out of multiple sources of
information. These included researcher observation, parent report, and professional assess-
ments.21 Such an approach allowed the Minnesota group to distinguish the shared and non-
shared contribution of abuse and other adversities to later outcomes. Shaffer, Huston, and
Egeland later reported that 40% of the non-maltreated child participants in their sample met
diagnostic criteria for at least one mental disorder in late adolescence. Reports from abuse
from one source made an additional contribution to later mental health, with an additional
20% of participants diagnosed with at least one mental disorder. There was, however, an add-
itional 32% where abuse was validated by more than one source.22
A sense of the serious ethical commitment of Egeland, Sroufe, and colleagues in com-
mencing a longitudinal study with a large sample recruited on the basis of low income and
adversity, and the value of making sense of parents’ experiences, comes through with spe-

cial clarity in their early publications. In examining the antecedents of abusive caregiving,
this did not imply that they ascribed blame to the caregivers.23 Egeland and colleagues had
specifically chosen a sample in poverty because deprivation of resources was so predictable
as an antecedent of abuse. Responsibility for the care of children did not reside solely with
their parent or parents: ‘As a member of society one shares a responsibility with respect to
the quality of care available to all children’, wrote Sroufe in 1985, and ‘if responsibility for the
child’s well-being does not reside in his or her inborn variation, then it is ours’.24
At the time the longitudinal study was beginning, there was a great deal of scepticism
regarding claims, such as those of Bowlby, that early experience has a lasting role in human
development. In the 1960s, researchers such as Jerome Kagan and Howard Moss had found
little continuity in children’s behaviour or cognitive responses over time.25 However, these
researchers assumed that continuity meant that the same behaviour or cognitive responses
would be seen. The Minnesota group did not assume mimetic (homotypic) continuity in the
forms of cognition, behaviour, or emotion shown by children in the context of maturation.26
In this, they differed from Bowlby who tended to expect early adversities to be mirrored in
later behaviour (Chapter 1). All development builds on itself, they believed, but straight-
forward continuity should not be expected. Even if the present is heavy with the past, this
legacy is always altered by the need to respond to present circumstances.27 In this perspec-
tive, the likelihood and extent of behaviours like abusive parenting are influenced by the
interaction of risk and protective factors over time. Continuity would lie in the prediction of
22 Shaffer, A., Huston, L., & Egeland, B. (2008) Identification of child maltreatment using prospective and self-
report methodologies: a comparison of maltreatment incidence and relation to later psychopathology. Child Abuse
& Neglect, 32(7), 682–92, Table 3. As well as careful work to ensure the convergent and predictive validity of their
quantitative measures, the Minnesota group made particular efforts to understand the perspectives of their parti-
cipants. They therefore conducted extensive interviews with their participants. The interviews were used to create
rating scales, though it is a shame that qualitative analyses of these interviews were never reported.
23 Alan Sroufe interviewed in Karen, R. (1998) Becoming Attached: First Relationships and How They Shape Our
Capacity to Love. Oxford: Oxford University Press: ‘The poor single mothers in our study all want the best for their
kids. They maybe can’t do it. They may be so beaten down by their histories and their circumstances that they’re
doing a terrible job, but I’ve never seen one that didn’t want to do it right’ (378).
fant temperament. Child Development, 56(1), 1–14, p.12.

25 Kagan, J. & Moss, H.A. (1962) Birth to Maturity: A Study in Psychological Development. New York: Wiley.
26 Described in the historical section of Sroufe, L.A., Coffino, B., & Carlson, E.A. (2010) Conceptualizing the role
of early experience: lessons from the Minnesota Longitudinal Study. Developmental Review, 30(1), 36–51, p.36.
27 Sroufe, L.A. (1979) The coherence of individual development: early care, attachment, and subsequent devel-
opmental issues. American Psychologist, 34(10), 834–41.

Introduction 341
later maladaptation by earlier maladaptation, even if these forms of maladaptation were con-
ditioned by quite different developmental challenges. Such difficulties would be held in place
not only by individual-level factors, but also by the insistent compounding of environmental
risks and the paucity of forms of support over time, which provides poor soil for anything
but problems to grow. The Minnesota group therefore focused their attention on factors that
could contribute to change and stability, and the role of family life as a contributory to these
processes.
A multigeneration longitudinal study was the necessary testing-ground for many of the
ideas that Bowlby and Ainsworth had sketched about attachment. For Ainsworth, ‘longi-
tudinal research deals with the very stuff of life and human development, with all its conse-
quent complexities and difficulties’.28 The participant is sought and attended to in the round.
In her time it was a general consensus among developmental researchers that ‘long-term lon-

gitudinal studies have proved unmanageable’, which was a good part of why she had limited
herself to a study of the first year of infant–caregiver interaction.29 Yet the importance as-
signed to early attachment relationship could only be accepted if it predicted later outcomes
anticipated by the theory, and the most valid way to do this was through a prospective study
rather than retrospective recall, with all its ensuing biases.30 Sroufe, Egeland, and colleagues
regarded the construct validation of infant attachment as demanding two tasks. One was
already sufficiently available from the work of Bowlby and Ainsworth: ‘a rich, interwoven
network of interconnected theoretical proposition’. However, secondly, ‘this structure must
be secured through observation’.31 A set of observations linking infant attachment not only
to caregiving behaviour, but also to relevant developmental sequelae provided a protective
belt for the emergent paradigm.32
Ultimately, ‘the validity of the attachment concept, as with any construct, hinges on the
network of empirical relations built up around it and its power in yielding a coherent picture
of individual adaptation’.33
In line with Ainsworth’s approach, the Minnesota group adopted two priorities in judging
their methodology, oriented by a holism that aimed to train attention on the ‘whole person’,
28 Ainsworth, M. (1962) The effects of maternal deprivation: a review of findings and controversy in the context of
research strategy. In Deprivation of Maternal Care: A Reassessment of its Effects (pp.87–195). Geneva: WHO, p.106.
29 Ainsworth, M. (1962) Letter to John Bowlby, 11 December 1962. Mary Ainsworth papers, Box M3168,
Folder 1.
30 Waters, E., Wippman, J., & Sroufe, L.A. (1979) Attachment, positive affect, and competence in the peer
group: two studies in construct validation. Child Development, 50(3), 821–29: ‘As a developmental construct, se-
curity of attachment can be validated only by confirming predicted external correlates’ (822).
31 Carlson, E.A., Sroufe, L.A., & Egeland, B. (2004) The construction of experience: a longitudinal study of repre-
sentation and behavior. Child Development, 75(1), 66–83, p.77.

32 On the concept of the protective belt for a scientific paradigm see Lakatos, I. (1978) The Methodology of
Scientific Research Programmes. J. Worrall & G. Currie (eds). Cambridge: Cambridge University Press.
33 Sroufe, L.A., Fox, N.E., & Pancake, V.R. (1983) Attachment and dependency in developmental perspective.
Child Development, 54(6), 1615–27, p.1616. See also Main M. (1999) Mary D. Salter Ainsworth: tribute and por-
trait. Psychoanalytic Inquiry, 19, 682–776: ‘if the links found in Ainsworth’s small Baltimore sample were to endure
the test of time, a number of vital empirical questions would have to be addressed. Perhaps the most central and
enduring leader in this initial quest was Alan Sroufe of the Institute of Child Development at the University of
Minnesota . . . Sroufe’s pioneering work made an immeasurable difference to Mary Ainsworth’s acceptance in the
empirically oriented psychological circles of the 1970’s, since the Institute at the University of Minnesota was con-
sidered to be “hardheaded”, and psychometrically sophisticated. As soon as Sroufe (shortly to be joined by Byron
Egeland) began his investigations of infant–mother interaction, infant strange situation behavior, and the child’s
later development (the study with Egeland is presently continuing into young adulthood), Mary Ainsworth felt
secured on two sides: in England, by her mentor, John Bowlby, and within the United States, by the much younger
Alan Sroufe and his new student Everett Waters’ (731).
not just cut-up aspects of the person’s behaviour.34 One was to seek opportunities for nat-
uralistic observation, where possible, especially in the early years of the study. Whereas
video technology was just becoming available to researchers in the late 1960s, for use by
Ainsworth’s doctoral students but not in her home observations, the Minnesota group made
extensive use of video observations as well as observer report in order to capture the behav-
iour of their families in detail, and in a way that permitted repeated coding and inter-rater
reliability.35 Critics alleged that Ainsworth’s observational data by no means amounted to
science. This criticism has been often repeated over the decades.36 By contrast, use of video
technology and multiple observers helped the Minnesota group combine the naturalistic
observation valued by Ainsworth with a strengthened apparatus of scientific reliability and
replication.
A second priority of the Minnesota group was to focus on broad-band competencies,

rather than discrete measures.37 Egeland, Sroufe, and colleagues tended to create ‘com-
posite’ variables out of their measures. For instance, attachment classifications and maternal
sensitivity and intrusiveness were often, though not always, folded together into an ‘early
caregiving composite’. The fact that sometimes they reported attachment classifications in-
dependently and sometimes as part of a composite, and sometimes the 12-and 18-month
assessments independently and sometimes within the composite, led critics to accuse the
Minnesota group of a lack of transparency. Lamb and colleagues, for instance, accused the
Minnesota team of privately running the analysis for all the permutations and reporting
those that had the best prediction.38 Certainly it is true that the choice of the use of the com-
posite or attachment measures is generally left unjustified, which has at times proven an obs-
tacle for attempts at replication or meta-analysis. However, on review of the publications of
the Minnesota group, recourse to composite measures does not seem arbitrary. It occurs
more often when the researchers were interested in broad-band or diffuse psychological pro-
cesses. Additionally, for many of the investigations they wished to conduct, for instance in
exploring the relative contribution of different adversities and of early care and attachment
to later adaption, the Minnesota group had no precedents to follow. To an extent, they had
to feel their way, trying out different approaches, hoping to hear the general trends in their
data through composite measures, as well as to capture developmental processes anticipated
34 Sroufe, L.A., Egeland, B., Carlson, E.A., & Collins, W.A. (2005) The Development of the Person.

35 Duschinsky, R. & Reijman, S. (2016) Filming disorganized attachment. Screen, 57(4), 397–413.
36 Most recently, on review of the notes in the Ainsworth archive, Vicedo expressed scorn that ‘I do not believe
any scholar would consider the notes taken by the diverse observers during home visits reliable scientific data’.
Vicedo, M. (2018) On the history, present, and future of attachment theory. European Journal of Developmental
Psychology, 17(1).
37 Elicker, J., Englund, M., & Sroufe, L.A. (1992) Predicting peer competence and peer relationships in child-
hood from early parent–child relationships. In R. Parke & G. Ladd (eds) Family–Peer Relationships: Modes of
Linkage (pp.77–106). Hillsdale, NJ: Lawrence Erlbaum: ‘There can be multiple pathways to adaptive success (see
Sroufe & Jacobvitz 1989). Discrete measures, such as social participation or sharing, are used to provide concur-
rent validity for the broad band competence measures, or in follow up analyses to examine specific aspects of
overall competence. Additional assessment strategies compatible with the organizational perspective include: em-
phasis on naturalistic observation, rather than on highly structured laboratory tasks, especially in the early stages
of research; emphasis on situations in which there is a clear need for the individual to coordinate affect, condition
and behaviour; and special attention to situations that tax the adaptive capacity of the individual’ (83).
38 A detailed evaluation and critique of all the Minnesota papers published by the early 1980s is presented as
Chapter 9 of Lamb, M., Thompson, R.A., Gardner, W., & Charnov, E.L. (1985) Infant–Mother Attachment: The
Origins and Developmental Significance of Individual Differences in the Strange Situation. Hillsdale, NJ: Lawrence
Erlbaum. The issue of lack of transparency and the potential for cherry-picking strong results through compos-
iting variables is, however, perhaps the primary criticism.
Introduction 343
to be diffuse and pervasive rather than localised. Many of the assessment measures originally
developed in Minnesota were used in subsequent cohort studies.
The concern with observation and with assessing broad-band competencies led the
Minnesota group to utilise the laboratory as a site for ‘critical situations’ (Chapter 2), where
aspects of ordinary experience could be dramatised. They designed and set tasks that re-
quired the child or dyad to coordinate affect, cognition, and behaviour—the components
of a behavioural system—to solve the problem. The laboratory-based challenge might differ
in degree from those encountered within the rich, obscuring colour and camouflage of
everyday life. Nonetheless it was anticipated to present a broad-band demand upon the re-
sources, expectations, and capacities of the individual or dyad, and in this way offer a window
into their history and present-day functioning. The paradigmatic example of such a chal-
lenge for the Minnesota group was the Ainsworth Strange Situation, which taxed children’s

regulatory capacities and their trust in the availability of their caregiver, thereby offering
some window into the dyad’s history of secure base and safe haven provision (Chapter 2).39
A later example was the ‘friendship interview’, a measure designed by the Minnesota group,
which asked participants in sixth grade to describe their relationship with a close friend.
Again, this was a broad-band demand relevant to the tasks characteristic of children’s de-
velopmental stage. Not all such measures were equally successful, however. The friendship
interview had excellent associations with social competence in girls, and was predicted well
by early infant–caregiver interactions. However, the assessment did not work as well with
boys in the sample: ‘Minnesota males, and perhaps even more among our subpopulation of
youth, are “men of few words”.’40
The Minnesota Longitudinal Study of Risk and Adaptation has now been running for over
40 years, and is rightly regarded as a masterpiece of developmental scientific research and the
‘gold standard in prospective studies of neglect and abuse’.41 Even decades later, the sample
has retained 163 out of the original 267 participants, an astonishing achievement and a tes-
tament to the commitment of the participants and researchers. The Minnesota Longitudinal
Study of Risk and Adaptation has provided a special opportunity to explore the lives of the
grandchildren of the parents originally recruited to the sample, and to trace developmental
processes across generations. This multigenerational research has revealed both substantial
change and marked continuities in parenting practices between the generations, with sta-
bility calculated at r = .43, controlling for a variety of covariates. In a regression analysis with
all the correlates of second-generation caregiving, the care that these adults had themselves
received accounted for 12% of variance.42 This is far from the deterministic rhetoric some-
times deployed by Bowlby, especially in his populist writings (Chapter 1). However, a 12%
difference in day-in-day-out parenting and care represents a very substantial difference in
the family life of a young person. This finding exemplifies the contribution of the Minnesota
study: considering a large range of risk and protective factors, following participants over
Linkage (pp.77–106). Hillsdale, NJ: Lawrence Erlbaum, p.83.

41 Brown, D. (2009) Assessment of attachment and abuse history, and adult attachment style. In C.A.
Courtois & J.D. Ford (eds) Treating Complex Traumatic Stress Disorders: An Evidence-Based Guide (pp.124–44).
42 Kovan, N.M., Chung, A.L., & Sroufe, L.A. (2009) The intergenerational continuity of observed early parent-
ing: a prospective, longitudinal study. Developmental Psychology, 45(5), 1205–13.

long periods, pursuing multivariate analyses, and making powerful empirical contributions
that in turn have implications for theory and child welfare practice.
In this, one area of particular contribution to the field of attachment research was in the con-
ceptualisation and study of emotions. In order to situate the findings regarding attachment from
the Minnesota study, this chapter will begin by outlining Sroufe’s general model of emotion. The
selection of measures and the interpretation of findings by the Minnesota group were deter-
mined by a variety of factors which could vary from study to study. Nonetheless nearly all stud-
ies by the research group draw on Sroufe’s theory in the justification for their research design or
in making sense of their results. It is therefore important to consider this contribution before
describing the major empirical reports from the cohort study. And whilst Sroufe’s headline ideas
are familiar and well cited, the development of his ideas, and their full depth, are less well known
as they are often scattered across a variety of texts. As a result, many important aspects of Sroufe’s

theory of human emotional life have been surprisingly untapped by later researchers.
Theory of affect
Emotional development
Bowlby identified three components to behavioural systems: behavioural components, cog-

nitive components, and affective components (Chapter 1). In the 1960s and 1970s he gave
special attention to behaviour. This left its mark in many ways on the theory, not least in the
name ‘behavioural’ systems. Though already present from the 1960s in his concept of ‘in-
ternal working models’, from the 1980s Bowlby focused attention on the cognitive aspects
of behavioural systems. By contrast, the affective component of behavioural systems was al-
ways acknowledged but never elaborated by Bowlby. In his private notes, he observed this
tendency in his writings, and attributed it to a desire to backlight the importance of phe-
nomena downplayed by psychoanalytic theory. However, the result was a hole in attachment
theory where an account of affect, motivation, and development should have been. For stud-
ies on other topics, this hole might have been of less consequence. For conceptualising the
results of a longitudinal study with a central focus on child maltreatment, it was unsustain-
able. Sroufe felt that ‘emotions and motivation are downplayed despite the fact that Bowlby’s
observations led him to describe attachment as an “affective” bond’.43
The neglect of emotion in Bowlby’s work, in part an attempt to distance his approach from
psychoanalysis, mirrored a trend in developmental psychology more generally. Though at-
tention to emotion was on the rise, the main protagonist in the story of psychology in the late
1970s was cognition. Psychoanalytic theory retained a strong interest in emotion, but came
into disrepute in American academic psychology departments, and there were few effective
bridges between psychoanalytic theory and the empirical psychology of the period.44 Yet
44 For discussion of the standing of emotion within the psychological science of this period see Cicchetti, D.
& Schneider-Rosen, K. (1984) Theoretical and empirical considerations in the investigation of the relationship
between affect and cognition in atypical populations of infants: contributions to the formulation of an integrative
theory of development. In C. Izard, J. Kagan, & R. Zajonc (eds) Emotions, Cognition, and Behavior (pp.366–406).
Cambridge: Cambridge University Press. There were some exceptions and attempts at bridge-building, e.g. James,
A.E. (1976) Freud, Piaget, and human knowledge. Annual Review of Psychoanalysis, 4, 253–77; Schlesinger, H.J.
Theory of affect 345
through the 1970s, across the social sciences, attention to emotional life, emotion regula-
tion, and the social situatedness of emotions was growing, reflecting rising attention to these
topics in wider society and politics.45 In a landmark book chapter from 1979, Sroufe argued
that ‘the child grows not as a perceptive being, not as a cognitive being, but as a human being
who experiences anxiety, joy, and anger, and who is connected to its world in an emotional
way’.46 Together with Minnesota colleagues, Sroufe highlighted the flexibility of human be-
havioural systems, in contrast to reflexes, and the role of psychosocial challenges such as
trust, autonomy, and control as providing the context of their progressive elaboration. Many
behavioural systems include reflex or relatively inbuilt components, but become elaborated
as they are fed by social and cultural experiences. For instance, the laugh response is predis-
posed for human infants by particular kinds of situation, such as incongruity in the context
of mid–high arousal, but also by the availability of a caregiver who gives a positive appraisal

to the incongruity.47 Sroufe narrated infant development in terms of the gradual differenti-
ation of arousal into various, initially quite general affects, and from there into more refined
emotions that are differentiated by their specificity and directedness.48
These refinements are based partly on the social support the child receives in identifying
and modulating affects when they arise in different contexts. Sroufe ascribed an important
role to the provision of a safe haven when a child is distressed for the modulation of af-
fect. However, he also elaborated on an expanded role for exploration. For Bowlby, the ex-
ploratory system has the function of learning about new and/or complex phenomena. Sroufe
added to this that play, especially in the context of a secure base and stable routines, helps the
infant learn, over time, to remain regulated in the context of higher levels of arousal: ‘As care-
giver and infant play, tension is escalated and de-escalated, to the edge of overstimulation
and back again, commonly ending in the bursts of positive affect so rewarding for caregivers.
Episode by episode, day by day, the infant’s own capacity to modulate (and tolerate) tension
is developed, and a reservoir of shared positive affect is created.’49
This reservoir, which Sroufe largely identified with the attachment system, is nestled ini-
tially within the relationship. However, over time it becomes in part the property of the child,
as a sense of confidence and fluency in accessing, modulating, and expressing feelings, as
well as recouping or finding relief afterwards. This becomes a skill, one that can help make
for successful family and peer interactions. Frustration and conflicts can be experienced
(1971) Clinical-cognitive psychology: models and integrations. Psychoanalytic Quarterly, 40, 366–68. However,
these had little traction.
45 For psychological science see e.g. Izard, C. (1977) Human Emotions. New York: Plenum; Lewis, M. &
Rosenblum, L. (1978) Introduction: issues in affect development. In M. Lewis & L. Rosenblum (eds) The
Development of Affect: The Genesis of Behavior, Vol. 1 (pp.1–10). New York: Plenum. As a point of comparison, see
the growing attention to emotions in anthropology. Rosaldo, M.Z. (1980) Knowledge and Passion: Ilongot Notions
of Self and Social Life. Cambridge: Cambridge University Press. On the wider context of emotional citizenship in
the late 1970s and early 1980s see Berlant, L. (2008) The Female Complaint. Durham, NC: Duke University Press.
46 Sroufe, L.A. (1979) Socioemotional development. In J. Osofsky (ed.) Handbook of Infant Development
(pp.462–515). New York: John Wiley, p.462.

47 Ibid. p.502; Sroufe, L.A. & Wunsch, J.P. (1972) The development of laughter in the first year of life. Child
Development, 43(4), 1326–44; Sroufe, L.A. & Waters, E. (1976) The ontogenesis of smiling and laughter: a perspec-
tive on the organization of development in infancy. Psychological Review, 83(3), 173–89.
48 In general, though without full consistency, Sroufe used the term ‘affect’ to refer to the embodied reaction to a
salient event, and ‘emotion’ to refer to the subjective experience of this reaction, including awareness of its source.
An affect seems to have more of the early, embodied prototype in play, whereas an emotion seems to be more in-
flected by contemporaneous cognitive resources. However, this contrast is only elaborated implicitly by Sroufe.
without feeling unacceptable, dangerous, or spiralling into persistent activation or diffusion

into other nearby emotions such as distress, fear, or despair:
Resolving conflicts is an important building block of the child’s emerging sense of com-
petence at problem solving, and as Erikson argues, it also deepens the child’s trust in the
caregiving relationship. Such prototypical conflict experiences within the security of the
caregiving relationship can also represent a model for later close relationships, providing
an abiding confidence that relationships may be sustained despite strife, which allows a
person to risk conflict in relationships and, ultimately, to even see its value.50
However, as well as social scaffolding by caregivers, the refinements of emotional develop-

ment of young childhood depend on cognitive maturation. Unlike Bowlby for whom em-

phasis on emotional processes was felt to risk underplaying the cognitive components of
behavioural systems, Sroufe emphasised that the two are intertwined. Cognitive achieve-
ments are spurred and textured by affects. In turn, the registering and differentiation of emo-
tions is scaffolded by cognitive development:
Cognitive factors underlie the unfolding of emotions: only with recognition is there
pleasure and disappointment; only with some development of causality, object perman-
ence, intentionality and meaning are there joy, anger, and fear; only with self-awareness
is there shame. Also, distinctions among affects and their precursors call upon cognitive
achievements—for example, fear, as reflected in more immediate distress upon a second
exposure to a stranger, has been referred to as a categorical reaction, dependent upon as-
similation to a negative scheme. Finally, the effects of sequence, setting, and other aspects
of context on emotion are obviously mediated by cognition.51
Certain emotions, such as shame, pride, and guilt, are only possible once sufficient cognitive
infrastructure is available, since, together with high arousal, they require identification of
discrepancies between behaviour and the representation of our value from the perspective
of others. Think, for instance, of the development over childhood of shame in what we have
drawn with pencil and paper, or in our bodies, to the point that it is to an extent typical in
adults. Sroufe observed that since toddlers are only just piecing together the representation
of their value in the eyes of others, this impression may still be especially fragile. A conse-
quence is that ‘the toddler is vulnerable to a global feeling of dissolution when being pun-
ished for a specific behaviour (especially if done harshly or in a degrading way)’.52 Whereas
in the first year of life, anxiety in relationships is associated particularly with separation,
Sroufe argued that over the second year, cognitive advances allow psychological separations,
such as scolding, to be pierced by as much anxiety as physical separations.53
In his account of emotional development, Sroufe took particular interest in the basic
smile reflex, highlighting the critical role of social interaction. He wrote beautifully of the
social smile as the ‘crowning achievement’ of the first three months of life, initiating the use
50 Ibid. p.206.
53 Ibid. p.126.
of positive affects within communication: ‘the infant smiles and coos at its own feet, at its
toy giraffe, and especially at the caregiver. For the first time it laughs in response to vigorous
stimulation. “Pleasure has become an excitatory phenomenon”, associated with high states
of excitation (Escalona, 1968, p.159). With eyes sparking, caregiver and infant set out upon
the task of establishing reciprocal exchanges.’54 The infant does not yet have a ‘representa-
tion’ of the relationship—that will only come later, along with language—but they do have
‘action schemes’ which allow for procedure-level memories and expectation of how the care-
giver will respond, coordinated recognition, and sharing of differentiated emotions, and
some basic forms of independent emotion regulation.55 Action schemes are encoded in a
preverbal way that is qualitatively different to semantic memory, and that makes them more
difficult to verbally analyse and re-evaluate.
With cognitive development, more complex and differentiated emotions are possible. Yet

‘The more mature emotions do not displace precursor emotions. The smooth execution of
behaviour can still produce pleasure, unfathomability and vague threat can produce wari-
ness, and diffuse frustration can lead to intense distress at any age.’56 Within the mature anger
behavioural system of a one-or two-year-old remains the earlier more global, intense, and
diffuse form, which becomes vocally evident in children’s tantrums. The same, Sroufe ar-
gued, was true of other behavioural systems like fear: abject fear loses specificity and direct-
edness, as well as its differentiation from nearby emotions such as distress.57 The process of
emotional maturation exemplified the broader commitment of Sroufe and the Minnesota
group to the principle that all development is built on itself, with earlier forms remaining
operative even if reforged by new opportunities, capabilities, and demands: ‘Development
is a dynamic process wherein what evolves derives from what was there before in a logical
way, while, at the same time, prior experience or prior adaptation has a fundamentally new
meaning in the now more complex system.’58 Even if early experiences of support and care
are supplanted by later adversity, they remain a resource that may be reactivated by a facilita-
tive later context. Likewise, early vulnerability in the context of adversity can be reactivated,
even many years later, when someone is faced with depriving or traumatic conditions.59
Within emotional maturation, Sroufe gave an important place to attachment pro-
cesses. His stance was that ‘attachment is critical and has a central place in the hierarchy
(pp.462–515). New York: John Wiley, p.478; Escalona, S.K. (1969) The Roots of Individuality: Normal Patterns of
Development In Infancy. Oxford: Aldine.
55 Sroufe, L.A. (1989) Relationships, self, and individual adaptation. In A.J. Sameroff & R.N. Emde (eds)
Relationship Disturbances in Early Childhood: A Developmental Approach (pp.70–94). New York: Basic Books, p.76.
The term ‘emotion regulation’ remains undefined in the writings of Sroufe and Egeland. However, in practice their
usage resembles Thompson, R.A., Flood, M.F., & Lundquist, L. (1995) Emotional regulation and developmental
psychopathology. In D. Cicchetti & S. Toth (eds) Rochester Symposium on Developmental Psychopathology, Volume
6: Emotion, Cognition, and Representation (pp.261–99). Rochester, NY: University of Rochester Press: ‘emotional
regulation consists of the extrinsic and intrinsic processes responsible for monitoring, evaluating and modifying
emotional reactions, especially their intensive and temporal features, to accomplish ones goals . . . this definition of
emotion regulation includes maintaining and enhancing emotional arousal as well as inhibiting it’ (265).
56 Sroufe, L.A. (1996) Emotional Development. Cambridge: Cambridge University Press, p.64–5.
57 Sroufe, L.A. (1979) Socioemotional development. In J. Osofsky (ed.) Handbook of Infant Development, (pp.462–
515). New York: John Wiley, p.488; see also Sroufe, L.A. (1996) Emotional Development. Cambridge: Cambridge
University Press, p.55.
58 Sroufe, L.A., Egeland, B., Carlson, E.A., & Collins, W.A. (2005) The Development of the Person. New York:
Guilford, p.11.
59 See also Sroufe, L.A., Egeland, B., & Kreutzer, T. (1990) The fate of early experience following developmental
change: longitudinal approaches to individual adaptation in childhood. Child Development, 61(5), 1363–73.
of development because of its primacy. The infant–caregiver attachment relationship is

the core, around which all other experience is structured, whatever impact it may have.’60
However, in occupying this central place, Sroufe held that attachment processes are none-
theless dependent on an infrastructure of various factors that stretch beyond them. One is
the dependence of the attachment system on cognitive maturation. He assigned a particular
role to cognitive anticipation and expectation, which is reflected in the partly innate capacity
for joyous greeting of attachment figures. He proposed that ‘bouncing, smiling, arms-raised
gestures are differential to attachment figures, reflecting the positive value of the special
scheme to which the attachment figures are immediately assimilated’.61 Cognitive anticipa-
tion is also reflected in the capacity to hold an angry mood, even with a caregiver, or to in-
hibit the activation of a behavioural system that will result in disapproval or punishment.
The infant begins to include his or her own potential feelings within the field of awareness,

influencing responses. For instance, the knowledge that punishment and sadness will ensue
may help hold the inhibition of a behavioural system in place.
Sroufe proposed that the infant’s emerging sense of self is organised especially around
‘patterns of behavioural and affective regulation, which grant continuity to experience des-
pite development and changes in context’.62 This is a good part of why attachment relation-
ships are central to socioemotional development, and why the recognition, communication
and modulation of emotion is central to the quality of attachment relationships. On the basis
of this account, Sroufe offered rebuttal to Kagan’s strongly worded accusation that tempera-
ment was a sufficient explanation for apparent differences in the quality of attachment re-
lationship.63 Sroufe argued that findings, such as those of Main and Weston (Chapter 3),
showing barely any overlap between attachment classifications with different caregivers
are important evidence here. They indicate that differences in attachment lie in the history
of infant–caregiver interaction, and have no basis in differences in temperament.64 Over
time, Behrens observed that ‘the attachment–temperament debate has subsided because
most attachment researchers today incorporate some temperamental or biological assess-
ment of children, recognizing the potential biological contribution in forming relation-
ships’.65 Van IJzendoorn and Bakermans-Kranenburg later criticised Sroufe and Kagan as
partially responsible for the ‘deadly war over dominance and territory’ between attachment
and temperament researchers, sustaining a polemic that obscured the way that parenting,
60 Sroufe, L.A. (2005) Attachment and development: a prospective, longitudinal study from birth to adulthood.

61 Sroufe, L.A. (1979) Socioemotional development. In J. Osofsky (ed.), Handbook of Infant Development

63 Kagan, J. (1982) Psychological Research on the Human Infant: An Evaluative Summary. New York: W.T. Grant
Foundation. See also Kagan, J. (1995) On attachment. Harvard Review of Psychiatry, 3(2), 104–106; Kagan, J. (1996)
Three pleasing ideas. American Psychologist, 51(9), 901–908. On Kagan’s role as a ‘spokesperson’ for critics of at-
tachment research see Karen, R. (1998) Becoming Attached: First Relationships and How They Shape Our Capacity
to Love. Oxford: Oxford University Press, p.253.
fant temperament. Child Development, 56(1), 1–14.

Online Readings in Psychology and Culture, 6(1), 7, p.25. See Vaughn, B.E. & Bost, K.K. (2016) Attachment and
temperament as intersecting developmental products and interacting developmental contexts throughout in-
fancy and childhood. In J. Cassidy & P.R. Shaver (eds) Handbook of Attachment: Theory, Research, and Clinical
Applications, 3rd edn (pp.202–22). New York: Guilford.
attachment, and temperament interrelate (Chapter 7).66 If, as Karen has suggested, Kagan in-
tentionally helped to block access to key potential funders for attachment research through
the 1980s, the war becomes quite intelligible.67 Yet with Kagan’s relevance for attachment re-
search receding into the past, the opposition between attachment and temperament has lost
some of its necessity. In an article from 2000, Sroufe and colleagues accepted that the majority
of variance in internalising and anxiety symptoms in middle childhood is likely accounted
for by temperamental and genetic factors.68 And in a 2005 article, Sroufe highlighted the
interaction between infant temperament and caregiving sensitivity in predicting infant am-
bivalent/resistant attachment in the Strange Situation in the Minnesota Longitudinal Study
of Risk and Adaptation.69 Nonetheless, with the exception of the ambivalent/resistant clas-
sification, a meta-analysis by Groh found little evidence for main effects of temperamental
variation in the development of differences in infant attachment.70 Exciting examinations of

potential interactions between temperament and caregiving experiences are ongoing, but on
the basis of a scientific consensus that ascribes an irreducibly important place to caregiving.71
Felt security
Attention to the behavioural, cognitive, and emotional components of behavioural systems

led Sroufe to a different position to Bowlby regarding the set-goal of the attachment system.
In his writings for a psychoanalytic audience, Bowlby identified important emotional com-
ponents related to the set-goal of the attachment system. He specified that the desire to
achieve the set-goal would be felt as ‘anxiety’; when the caregiver is accessible if the child
feels frightened, he or she will feel ‘secure’; when the caregiver is accessible after the child has
felt frightened, he or she will feel ‘comforted’.72 Yet, in Attachment, Volume 1, Bowlby gave
66 Van IJzendoorn, M.H. & Bakermans-Kranenburg, M.J. (2012) Integrating temperament and attachment. The
differential susceptibility paradigm. In M. Zentner & R.L. Shiner (eds) Handbook of Temperament (pp.403–24).

68 Warren, S.L., Emde, R.N., & Sroufe, L.A. (2000) Internal representations: predicting anxiety from children’s
play narratives. Journal of the American Academy of Child & Adolescent Psychiatry, 39(1), 100–107: ‘The variance
accounted for by our variables in the prediction from 5-year-old narratives to 6-year-old internalizing and anxiety
symptoms was substantial for this kind of research (20%), [and] there is a strong likelihood that genetic predispos-
ition and temperament account for much of the remaining variance’ (106).
Attachment & Human Development, 7(4), 349–67. These results were earlier reported in Susman-Stillman, A.,
Kalkoske, M., Egeland, B., & Waldman, I. (1996) Infant temperament and maternal sensitivity as predictors of at-
tachment security. Infant Behavior and Development, 19(1), 33–47.

71 Spangler, G. (2013) Individual dispositions as precursors of differences in attachment quality: why maternal
sensitivity is nevertheless important. Attachment & Human Development, 15(5–6), 657–72.

72 Bowlby, J. (1960) Separation anxiety. International Journal of Psycho-Analysis, 41, 89–113, pp.93–4. In fact, use
of the term ‘comfort’ to describe the set-goal of the attachment system in infancy led a subterranean existence over
decades. For example, the concept of comfort can be seen dodging in and out of view in Separation: ‘accessibility
in itself is not enough. Not only must an attachment figure be accessible but he, or she, must be willing to respond
in an appropriate way; in regard to someone who is afraid this means willingness to act as comforter and protector.
Only when an attachment figure is both accessible and potentially responsive can he, or she, be said to be truly
available. In what follows, therefore, the word “available” is to be understood as implying that an attachment figure
is both accessible and responsive.’ Bowlby, J. (1973) Separation. New York: Basic Books, p.234. Following Bowlby,
Ainsworth tended to use ‘availability’ in descriptions of the set-goal in infancy in written presentations. However,
in oral presentations she tended to use the term ‘comfort’.
these subjective states primarily the role of concomitants to the set-goal of the attachment
system, which he positioned as ‘proximity’ (Chapter 1). In the 1970s, however, there were
growing concerns that the attachment paradigm was being hindered by a lack of precision in
situating the subjective, felt aspects of the phenomena under discussion. The German eth-
ologist and systems theorist Norbert Bischof, for example, observed that ‘authors dealing
with attachment quite commonly talk about the mother as a source of infantile “security”. It
should be noted, however, that this term does not clearly distinguish between an emotional
state (feeling more or less secure) and an environmental fact.’73 Sroufe and Waters likewise
argued that Bowlby’s neglect of emotion obscured, and even distorted, his account of the at-
tachment behavioural system.
In ‘Attachment as an organizational construct’, published in 1977, Sroufe and Waters
raised several objections to the sidekick role allocated to emotion.74 They noted that some

infants, such as those in dyads classified B1 by Ainsworth, appeared to have had their at-
tachment system terminated by merely the return of the caregiver and distance interaction,
without physical proximity-seeking. The ambivalent/resistant infant also achieved prox-
imity in the Strange Situation, and yet the attachment system remains activated. Sroufe and
Waters argued that emotions not only are important parts of the motivational system of at-
tachment, but also condition its terminating conditions. In contrast to Bowlby, who offered
a behavioural set-goal for the attachment system, Sroufe and Waters broadened this out.
They proposed that the attachment system would be terminated when infants ‘feel secure’.
Proximity is one facilitating and sometimes a necessary contributor to this affective state.
But it is the affective state that is the set-goal.75
Elsewhere, Sroufe proposed that ‘the infant secure in his or her attachment has experi-
enced the caregiver as a reliable source of comforting, as responsive to his or her signals,
and as available and sensitive. The infant has learned that stimulation in the context of the
caregiver will generally not be overwhelming and that when arousal threatens to exceed the
infant’s organizational capacity, the caregiver will intervene.’76 Indeed, Sroufe was one of the
very few researchers who, following Ainsworth, highlighted that the capacity to feel secure
in attachment relationships would be influenced by feelings of security or insecurity in other
domains (Chapter 2). Furthermore, ‘the availability of cues associated with security (familiar
procedures or objects, predictability, control) and a generalised expectancy concerning
likely outcomes when tension is high’ were assumed to feed into felt security within attach-
ment relationships,77 perhaps contributing to the small overlap in variance in attachment
relationships with different caregivers.
Ainsworth did not appear to regard the broader emphasis on felt security as discrepant
with the orthodox position represented by Bowlby. She initially reported reading the Sroufe
and Waters paper with approval. However, over the years, she began to feel uncomfortable.
Many second-generation attachment researchers began to rally to ‘felt security’, against what
was increasingly regarded in the 1980s as the somewhat wooden quality to Bowlby’s account
73 Bischof, N. (1975) A systems approach toward the functional connections of attachment and fear. Child
Development, 46(4), 801–17, p.802.

74 Sroufe, L.A. & Waters, E. (1977) Attachment as an organizational construct. Child Development, 48(4),
1184–99.
75 Ibid. p.1191.
76 Sroufe, L.A. (1979) The coherence of individual development: early care, attachment, and subsequent devel-
opmental issues. American Psychologist, 34(10), 834–41, p.837.

of attachment relationships in terms of activating and terminating conditions.78 Ainsworth’s

former student Mark Cummings was even developing a new coding system for the Strange
Situation focused on felt security, as an alternative to coding proximity-seeking and contact
maintaining.79 Ainsworth brought the matter to Bowlby’s attention in 1987, and the collab-
orators reflected together.80 They were worried, justifiably as it turned out (Chapter 5), that
the term ‘felt security’ left ambiguity regarding whether an individual would be consciously
aware of whether or not they had a secure attachment. They were sure that individuals would
not. However, they also registered that the set-goal of the attachment system, even in in-
fancy, was not merely proximity. That early proposal of Bowlby’s became rather battered as
time passed since the 1960s.
Ultimately, Bowlby and Ainsworth came to the conclusion that the set-goal of the care-
giver is the feeling of the caregiver’s availability when called upon, which includes the possi-

bility of proximity as needed:
I am in agreement with you . . . I have for my part referred to ‘availability’ as the set-goal in
the case of children and adults. It requires belief in the effectiveness of communication &
belief that the attach figure will respond if called upon & that physical proximity is readily
attainable when desired.81
The set-goal of the attachment system in infancy was revised by Ainsworth to be caregiver
‘availability’, assumed to incorporate behavioural, cognitive, and emotional components.
In this, access to proximity was maintained as one of the proximal goals of the attachment
system at high levels of activation, but not the ultimate set-goal regardless of level of ac-
tivation. Unfortunately there was only one book chapter in which Ainsworth explicitly
flagged the shift in position, and it was framed as a rejection rather than partial acceptance
of Sroufe and Waters’ position.82 Furthermore, most readers assumed that when Bowlby and
Ainsworth were referring to availability, they meant exclusively older children.83 Attachment
researchers still today have generally not integrated the different positions, and the set-goal
of the attachment system remains variously defined as ‘proximity’, ‘availability’, ‘accessibility’,
and ‘felt security’.84 Ainsworth’s student Roger Kobak, recognising this problem, roundly
78 E.g. Cicchetti, D. & Pogge-Hesse, P. (1981) The relation between emotion and cognition in infant devel-
opment. In M. Lamb & L. Sherrod (eds) Infant Social Cognition (pp.205–72). Hillsdale, NJ: Lawrence Erlbaum;
Erickson, M.F., Korfmacher, J., & Egeland, B. (1992) Attachments past and present: implications for therapeutic
intervention with mother–infant dyads. Development & Psychopathology, 4(4), 495–507, p.496. In fact, Erickson
and colleagues appear to have regarded Bowlby’s behavioural system model as superseded already by his own no-
tion of internal working models.
79 Cummings, M.E. & El-Sheikh, M. (1986) An Organizational Scheme for the Classification of Attachments on
a Continuum of Felt-Security. Morgantown: West Virginia University. https://files.eric.ed.gov/fulltext/ED288653.

pdf.
80 Ainsworth, M. (1987) Letter to John Bowlby, 28 November 1987. PP/BOW/B.3/8.
81 Ainsworth, M. (1987) Letter to John Bowlby, 30 December 1987. PP/BOW/B.3/8.
82 Ainsworth, M.D.S. (1990) Some considerations regarding theory and assessment relevant to attachments be-
yond infancy. In M.T. Greenberg, D. Cicchetti, & E.M. Cummings (eds) Attachment in the Preschool Years: Theory,
Research, and Intervention (pp.463–88). Chicago: University of Chicago Press.
83 See e.g. Kerns, K.A. & Brumariu, L.E. (2016) Attachment in middle childhood. In J. Cassidy & P. Shaver (eds)
Handbook of Attachment, 3rd edn (pp.349–65). New York: Guilford.

84 In the first and second editions of Cassidy and Shaver’s Handbook of Attachment, a chapter by Kobak at-
tempted to adjudicate the issue in favour of ‘availability’ as the set-goal of the attachment behavioural system. This
section was removed from the chapter for the third edition. Besides Kobak’s chapter, there has been little attempt to
adjudicate the issue, except where it has caused problems on the borders between the developmental and the social
psychology attachment traditions (Chapter 5). An early exception is Greenberg, M.T., Siegel, J.M., & Leitch, C.J.
(1983) The nature and importance of attachment relationships to parents and peers during adolescence. Journal of
criticised everyone. Bowlby and Ainsworth were criticised for recognising so late in the
day that caregiver availability, even in infancy, was broader than proximity. And Sroufe and
Waters were criticised for underspecifying the role of physical proximity, and for defining
felt security too loosely. Not least, Kobak was concerned that a whole host of factors could
contribute to feeling secure, not all of which would be based in attachment processes.85
Kobak reserved particular criticism, however, for the way that Sroufe and Waters’ pro-
posal contributed to the split in attachment theory between the developmental tradition and
the social psychology tradition (Chapter 5). Kobak regarded the split as based in part on a
lack of clarity about how the attachment system actually works in adulthood, caused in part
by Sroufe and Waters’s overencompassing and subjective notion of felt security.86 More re-
cently, Kobak has moderated his criticisms, likely in the context of increased rapprochement
between the developmental and social psychological traditions of attachment research. He

has allowed that the shift from a focus on proximity to availability helped support a shift in
‘the focus of attachment research from the study of young children’s separations from care-
givers to investigations of the quality of emotional communication in maintaining attach-
ment bonds’, a shift that he regards as overall a positive development for the field.87
Social affects
In the 1980s, Sroufe’s theoretical reflections on attachment and emotional development were
elaborated by the Minnesota group’s empirical research. An important early study was ‘The
role of affect in social competence’, published in 1984 by Sroufe, Fox, and Pancake, reporting
a mixed methods study of the experiences of children in the sample in preschool.88 Two pre-
school classes (15 children and 25 children) were constituted out of children in the sample
who had received the same attachment classification at 12 and 18 months, stratified to in-
clude children from each of the Ainsworth classifications. The preschool lasted for 20 weeks.
Youth and Aolescence, 12(5), 373–86. Greenberg and colleagues retain proximity and felt security as two different
set-goals of the attachment system, one or both of which may be met. For instance, the ambivalent/resistant infant
was conceptualised as achieving proximity but not felt security.
85 This criticism of Sroufe and Waters has recently been made again in a book chapter collectively written by
a group of anthropologists and developmental psychologists: Gaskins, S., Beeghly, M., Bard, K.A., et al. (2017)
Meaning and methods in the study and assessment of attachment. In H. Keller & K.A. Bard (eds) Contextualizing
Attachment: The Cultural Nature of Attachment (pp.321–33). Cambridge, MA: MIT Press.
86 Kobak, R. (1994) Adult attachment: a personality or relationship construct? Psychological Inquiry, 5(1), 42–4;
Kobak, R. (1999) The emotional dynamics of disruptions in attachment relationships: implications for theory, re-
search, and clinical intervention In J. Cassidy & P.R. Shaver (eds) Handbook of Attachment: Theory, Research, and
Application (pp.21 – 43). New York: Guilford, p.31.
87 Kobak, R., Zajac, K., & Madsen, S.D. (2016) Attachment disruptions, reparative processes, and psychopath-
ology: theoretical and clinical implications. In P.R. Shaver & J. Cassidy (eds) Handbook of Attachment, 3rd edn
88 Sroufe, L.A., ???Schork, E., Motti, F., Lawroski, N., & LaFreniere, P. (1984) The role of affect in social com-
petence. In C. E. Izard, J. Kagan, & R. B. Zajonc (Eds.), Emotion, cognition, and behavior (pp. 289-319).
Cambridge: Cambridge University Press. Mary Jo Ward later recalled this period: ‘My fondest memories of Alan
are from the year of the preschool project. He was in his element: choosing just the right subjects to compose a
class, working daily with some very gifted teachers, supervising grad students, planning assessments and coming
up with methods to capture the richness of what we were observing each day. But he was most happy about getting
to know each of the precious children. His insights into each child’s coping, his joy in their accomplishments, his
concern for their trials gave us an amazing opportunity to experience the human side of developmental science. It’s
an example that we’ll never forget.’ Ward, M.J. (2009) Tribute to L. Alan Sroufe on the Occasion of his Retirement, 17
October 2009, Minneapolis. Unpublished manuscript shared by the author.
Based on detailed ethnographic observations, Sroufe and colleagues were struck by the role
of positive emotion, which seemed to serve as the major currency of social interaction.
The children in the sample who found it difficult to socialise seemed often those who
were unable to resource positive emotions to serve this role of currency.89 From childhood
onwards, Sroufe and colleagues theorised, something has to be shared in order to make or
sustain a relationship. Emotion offers potent forms of sharing, and indeed of not sharing.90
Relative access to shared positive affect therefore contributes to the creation and stabilisa-
tion of social hierarchies, which in turn channel the expression and acceptance of emotion
between the preschool students. In the two different classes, Sroufe and colleagues asked the
teachers to rank the students in order of social competence. They found a strong association
between competence ranking and the frequency of affectively positive social interaction
(ρ = .70, p < .002). Notably, however, the frequency of social interaction with a negative af-

fective tone made no additional contribution, once shared positive emotions were taken into
account. This suggests that the emotional tone of peer interactions was not simply a product
of social competence, but that positive social interaction was linked, likely bidirectionally, to
social success.91 The researchers also documented that positive affect in itself was not critical
for success with peers at the school, but shared positive social interaction specifically: ‘the
sharing of good feelings and the rewarding of other’s overtures with positive affect’.92
Sroufe worked with his partner June Fleeson (later June Sroufe) in reflecting further on
the social quality of affects. They made two key claims. A first was that if the relationship
depends on sharing, then all partners in a relationship have to learn one another’s roles in
order to understand their own. As such, children learn all the ‘parts’ of a relationship with
an attachment figure, not just their own role. There is no sense of determinism in Sroufe
and Fleeson’s reflection. Individuals may develop thoughts, feelings, or behaviours in rela-
tionships that develop out of earlier roles only in the most non-linear way.93 However, early
roles may press themselves forward at times, particularly when an adult is exhausted, bored,
depleted, and/or poorly resourced. Across our lives, Sroufe and Fleeson argued, when we
respond in ways that are intuitive and unreflective, this may be especially coloured by our
history, and especially our experiences of attachment and of trauma. Bowlby had argued
that children who have experienced abuse may find it hard to find and sustain the role of
adequate caregiver because their representations in adulthood mirror those of their child-
hood. By contrast, Sroufe and Fleeson proposed that children who have experienced abuse
not only have the abusive role available, but also may have relative difficulties accessing pro-
cedural components of sensitive caregiving—for instance, gentle, non-intrusive touch and
the accurate recognition and modulation of affects. Children who have experienced abuse
may also have comparatively less access, even into adulthood, to social relationships that
help them sustain the caregiving role. In fact, in some cases, they may only have access to
89 Sroufe, L.A., Schork, E., Motti, E., Lawroski, N., & LaFreniere, P. (1984) The role of affect in social competence.
In C. Izard, J. Kagan, & R. Zajonc (eds) Emotion, Cognition and Behavior (pp.289–319). New York: Plenum, p.290.
90 Sroufe, L.A. (1996) Emotional Development. Cambridge: Cambridge University Press: ‘Emotion is part of all
critical transactions with the environment. It guides, directs, and sometimes disrupts action. And it is the currency
of personal relationships’ (12).
91 Sroufe, L.A., Schork, E., Motti, E., Lawroski, N., & LaFreniere, P. (1984) The role of affect in social compe-
tence. In C. Izard, J. Kagan, & R. Zajonc (eds) Emotion, Cognition and Behavior (pp.289–319). New York: Plenum,
p.298–9.
92 Ibid. p.305.
93 Sroufe, L.A. & Fleeson, J. (1986) Attachment and the construction of relationships. In W. Hartup & Z. Rubin
(eds) Relationships and Development (pp.51–71). Hillsdale, NJ: Lawrence Erlbaum.

relationships with individuals who have some investment in maintaining earlier patterns
of power and identity. So efforts these children undertake as adults to avoid the abusive role
may still leave them disadvantaged in attempting to parent well, especially when contextual
demands elicit intuitive and unreflective responses.
A second key claim made by Sroufe and Fleeson was that the organisation of life within
the family and the psychological coherence and emotional life of its members influence one
another. If the integration of behaviour, thought, and emotion within the family over time
is fragile or punctuated by crises, the members of the family will find emotion regulation
difficult. Likewise, if a member of the family has difficulty organising behaviour, thinking,
and emotions in coherent ways, this makes integration at the level of the family a greater
challenge.94 At both the individual and the family level, the integration of emotion with
thoughts and feelings permits the modulation and direction of emotional expression. In turn

this allows emotional experiences to inform and guide action but without threatening to
overwhelm or disrupt its effectiveness.95 Families regularly face external stresses and devel-
opmental change—some especially. This means that the organisation of family life is never a
fully stable achievement, and for some families is under constant bombardment and active
attempts at repair. Nonetheless, when integrative processes are linking behaviour, thinking,
and emotions well, experiences shared by families, even difficult ones, can help inform the
actions of its members in ways that help them respond to the present; likewise, experiences
by individual members contribute to action at a family level.
By contrast, when family members find integration behaviour, thinking, and affect con-
fusing or difficult, there will often be patterns to the fallout and response. Two broad possible
‘strategies for adaptation’ were highlighted by Sroufe and Fleeson, corresponding to the ‘two
primary patterns of anxious attachment’. A first is for the family to treat a wide array of situ-
ations as potentially combustable or distressing, and to seek continual emotional contact
with one another around these troubles. A second is for the family to cut themselves off from
interaction or communication when this might raise potentially troubling or distressing ex-
periences for individuals.96 Families adopting the latter strategy, who experience one another
as unreachable or threatening under certain conditions, may outlaw those conditions rather
than feel confident in their capacity to modulate and be informed by them. For instance, any
expression of anger may be completely forbidden, with tense situations evoking ‘ritualist (re-
petitive) patterns of avoidance’.97 Sroufe and Fleeson observed that it is quite usual for one or
even both of the strategies to be deployed by families handling the fallout from disruption
or challenges, before a new organisation has been re-established. What signals a pathway to-
wards problems for the family and its members, and may predispose or contribute to mental
health problems, is if such strategies become rigid to the point that they block opportunities
for repair, learning, or development.
From her doctoral research, a study of family relationships, Fleeson reported supportive
findings.98 Children were likely to display a tight and rigid control over their emotions when
94 Sroufe, L.A. & Fleeson, J. (1988) The coherence of family relationships. In R.A. Hinde & J. Stevenson-Hinde
(eds) Relationships within Families: Mutual Influences (pp.27–47). Oxford: Oxford University Press.
95 See also Sroufe, L.A., Cooper, R.G., DeHart, G.B., & Marshall, M.E. (1996) Child Development: Its Nature and
Course, 3rd edn. New York: McGraw-Hill, p.381.

97 Sroufe, L.A. (1989) Relationships and relationship disturbances. In A.J. Sameroff & R.N. Emde (eds)
Relationship Disturbances in Early Childhood (pp.97–124). New York: Basic Books, p.107.
98 Sroufe, J.W. (1991) Assessment of parent–adolescent relationships: implications for adolescent development.
Journal of Family Psychology, 5(1), 21–45.

raised in families where anger was disguised or distorted and in which members of the
family did not feel able to communicate safely. Conversely, children were likely to display
dysregulation of emotion under two conditions, which could occur separately or together.
One was when integration was hindered by intrusive sharing of emotions by the parents. The
other was when the roles in the family were distorted, with the opposite-sex parent seeking a
closer relationship with the child than with his or her partner. In either case, intimacy occurs
primarily to the parent’s pacing, without modulation for the child’s needs. Both situations
were theorised by Fleeson to contribute to a feeling of lack of control for the child. The child’s
attachment relationship with the parent means that insulation against intrusive parental in-
timacy is a complex task at best. It may, in fact, be ‘essentially impossible’ for many younger
children and early adolescents.99 At that age, especially, our parents’ concerns can be all but
inescapable.

Intrusive intimacy
Pursuing the implications of intrusive intimacy with the parent as a distortion of the attach-
ment relationship, Sroufe and Ward filmed mothers’ behaviour during a toy cleanup task in
the laboratory with the Minnesota sample when the children were toddlers. The researchers
measured the extent to which the mothers showed sensual physical contact, sexual teasing,
or requests for affection in the course of trying to persuade their toddler to put away the toys.
Of the 173 participants who attended the toddler call-back, 16 showed one or more of these
behaviours.
Sroufe and Ward distinguished forms of boundary dissolution from warmth or affection
on a number of grounds: the behaviours were directed almost exclusively by the mothers
towards their male toddlers; they were associated with more, not less, physical punishment
and threats of punishment; they were not associated with measures of cooperative, encour-
aging, supportive, or affectionate behaviour; and they tended to interrupt the child’s behav-
iour putting away the toys rather than facilitate this goal.100 In fact, half of the punishing
behaviours shown in the entire sample were displayed in these 16 dyads; this suggests a lack
of monitoring of the environment by the mothers, as they knew they were being filmed.101
Mothers who showed seductive behaviour towards their son did not display the same be-
haviour towards their daughter; instead they showed derision.102 Eight of the 16 mothers
were included in a subsample of 36 participants given a full family history interview, and
assessed independently for indicators of abuse, including incest. Seven of the eight had his-
tories suggesting that they had been required to supply emotional or sexual intimacy to their
father or stepfather during childhood in ways that were ‘suggestive of an incestuous-type
99 Ibid. p.34. Reflecting on Sroufe and Fleeson’s work in his private notes, Bowlby considered it a brilliant con-
tribution, showing that ‘it looks as though “self ” is one pole of a relationship and so cannot be conceptualised ex-
cept in terms of a relationship’. Bowlby, J. (1983) Concept of self . PP/Bow/H.8.
100 Sroufe, L.A. & Ward, M.J. (1980) Seductive behavior of mothers of toddlers: occurrence, correlates, and
family origins. Child Development, 51(4), 1222–9.

101 Ibid. p.225.
102 At age 13, the relationship between mother and daughter was also more peer-like than the rest of the sample,
and the mothers displayed more child-like and needy behaviours towards their daughters: Sroufe, L.A. (1989)
Relationships and relationship disturbances. In A.J. Sameroff & R.N. Emde (eds) Relationship Disturbances in
Early Childhood (pp.97–124). New York: Basic Books, pp.77–8.
relationship’.103 In a follow-up at 42 months, nearly half of the women reported experiences

of intrafamilial sexual abuse (compared to 8% of the total sample).
The behaviours shown by adults displaying intimately intrusive behaviours should not be
regarded as imitations of their own caregivers, as anticipated by a simple learning approach
to intergenerational transmission of abuse—as for instance held by Bowlby (Chapter 1).
Instead, Sroufe and colleagues argued that children learn both roles within attachment rela-
tionships, and that aspects of both can appear within adult behaviour:
These mothers do not simply show deferred imitation of their parents’ previous behav-
iour. For example, those who were exploited by their fathers do not literally do what their
fathers did to them. Rather, they engage in the culturally specified adult female form of
cross-gender child exploitation. They have internalised a relationship and not simply a set

of behaviours.104
For this reason, the fact that some of the mothers were doing well with their children in
other regards was not a surprise to the Minnesota group. The problem in some instances
was quite localised: the attachment system may inform the caregiving system, and they will
share some components, but they are distinct. The adult caregiving system is influenced in
powerful ways, for instance, by contemporaneous social support. Therefore a parent’s be-
haviour towards his or her child would not necessarily be expected to mirror the parent’s
past. For instance, if other relationships are available to provide a secure base and safe haven,
children would be less likely to be treated as objects of the attachment system and elicit the
sexuality and derision that had been woven into it by these mothers’ childhood experiences.
In support of the protective role of contemporaneous support, the Minnesota group found
that only 9% of the mothers who engaged in boundary dissolution behaviours had a stable
relationship with a partner, compared to 25% of the rest of the sample.105
The construct of boundary-dissolving behaviours was later found to have substantial pre-
dictive value. There was also substantial stability over time. In observations made at age 13
of dyads who had earlier shown seductive behaviours, boundary-dissolving behaviour was
displayed by both adolescents and their parent.106 It had become an organisational quality of
the relationship involving both members of the dyad. A decade later, the Minnesota group
reported a strong association between early boundary-dissolving behaviours by moth-
ers and their child’s number of sexual partners in adolescence.107 Boundary dissolution in
the age 13 assessment was also found to predict conduct and attention problems at age 16,
103 Sroufe, L.A., Jacobvitz, D., Mangelsdorf, S., DeAngelo, E., & Ward, M.J. (1985) Generational boundary dis-
solution between mothers and their preschool children: a relationship systems approach. Child Development,
56(2), 317–25, p.322.
104 Sroufe, L.A. (1989) Relationships and relationship disturbances. In A.J. Sameroff & R.N. Emde (eds)
Relationship Disturbances in Early Childhood (pp.97–124). New York: Basic Books, p.103.
105 Sroufe, L.A., Jacobvitz, D., Mangelsdorf, S., DeAngelo, E., & Ward, M.J. (1985) Generational boundary dis-
solution between mothers and their preschool children: a relationship systems approach. Child Development,
56(2), 317–25, p.324.
106 Additionally, mothers with substance abusing partners were more likely to show peer-like or spouse-like
behaviour with their 13-year-old sons; they were also more likely to have daughters who showed caregiving behav-
iour towards them. Hiester, M. (1993) Generational boundary dissolution between mothers and children in early
childhood and early adolescence. Unpublished doctoral dissertation, University of Minnesota.

Correlates of attachment 357
over and above behavioural problems shown at age 13 and other measures of family func-
tioning.108 Such predictive validity offered support for the construct of boundary dissol-
ution, though the measure has not subsequently seen use in later research by developmental
scientists. One reason may have been that the idea of mothers behaving seductively towards
their toddlers was simply too controversial and unthinkable an idea to be taken up by re-
searchers. A broader trend may have been the decline of reference to family systems theory
within developmental science over the subsequent decades. Furthermore, ideas of ‘child
abuse and exploitation’ and the ‘parentification of children’ were increasingly salient in clin-
ical discourses during the 1980s and 1990s: these covered some of the same ground as Sroufe
and colleagues’ work on boundary dissolution, but in both cases trained attention more on
older children and on dyadic interaction.109 Nonetheless, the fact that Sroufe’s approach to
thinking about affects and relationships contributed to the novel finding of the importance

of boundary dissolution offered demonstration of its value. And the finding also highlighted
the importance of caregiving experienced within attachment relationships for children’s de-
velopment.110 Exploration of the antecedents and sequelae of early care was central to the
concerns of the Minnesota group, with particular interest in measures of individual differ-
ences in infant attachment.
Correlates of attachment
Antecedents of attachment
Sroufe, Waters, and Vaughn conducted the Strange Situation with the sample of 267 mothers
when their infants were 12 and 18 months. Sixty percent of the sample had stable attachment
classifications over the six-month period. Of the 267 mothers, there were 31 cases of infant
maltreatment, confirmed scrupulously through multiple methods.111 Of these 31 mother–
infant dyads, at 12 months, 62% were classified as insecurely attached, compared to 45%
in the total Minnesota sample and 34% of the combined middle-class samples reported in
Ainsworth and colleagues’ Patterns of Attachment.112
108 Nelson, N.N. (1994) Predicting adolescent behavior problems in late adolescence from parent–child inter-
actions in early adolescence. Unpublished doctoral dissertation, University of Minnesota; Shaffer, A. & Sroufe,
L.A. (2005) The developmental and adaptational implications of generational boundary dissolution: findings from
a prospective, longitudinal study. Journal of Emotional Abuse, 5(2–3), 67–84.
109 If intimately intrusive behaviours by parents towards their child had been framed as child abuse, then re-
searchers would have been obliged to report the families to child protective services. This may have been a disin-
centive to include measure of such behaviours within a research study, given the ethical and administrative issues it
would have raised.
110 The research on boundary-dissolving behaviour at Minnesota also likely influenced the development of a
scale for role-reversing behaviour in Main and Hesse’s work on the Adult Attachment Interview (AAI).
According to a later report, using slightly different criteria to Egeland and Sroufe’s 1981 paper, 47 children were
abused and/or neglected by their caregivers in infancy (of 211 with adequate data; 22%), 66 in early childhood and
in middle childhood (of 185 in early childhood and 190 in middle childhood; 35–36%), and 21 in adolescence
(of 179; 12%). Johnson, W.F., Huelsnitz, C.O., Carlson, E.A., et al. (2017) Childhood abuse and neglect and phys-
ical health at midlife: prospective, longitudinal evidence. Development & Psychopathology, 29(5), 1935–46. The
reasons for the change in definition of infant maltreatment from those used by Egeland and Sroufe are not fully
provided by these later researchers.
112 Egeland, B. & Sroufe, L.A. (1981) Attachment and early maltreatment. Child Development, 52(1), 44–52. Of
studies using the Ainsworth classifications, there has been a very substantial association between maltreatment
In the total Minnesota sample, 21% of the dyads were classified as ambivalent/resistant.
This was substantially higher than the 13% seen by Ainsworth.113 The Minnesota group
found that there was a powerful overlap between ambivalent/resistant attachment in the
Strange Situation and the nurse’s report on the regulatory capacities of the child as a newborn
(r = .46), including their startle response (r = .20).114 However, these findings, suggestive
of a role for temperament, were reported in an undertone by the Minnesota group in their
publications of the 1980s and 1990s, and were rarely mentioned in their review or theoret-
ical works. (The field would have to wait for a recent meta-analysis by Groh and colleagues
for the resistance–temperament link to resurface for discussion.)115 Instead, the Minnesota
group focused on the portion of the variance in ambivalent/resistant attachment that ap-
peared to be environmental. Ainsworth had situated the broad notion of ‘inconsistent care’
as the origin of ambivalent/resistant attachment, though her more specific theory seemed

to be that these children had caregivers who struggled to hold them in mind (Chapter 2).
Egeland and Sroufe argued that inconsistent care may have been made more likely by the
poverty and other adversities faced by their sample.116 This interpretation was buttressed by
their finding that half the cases of neglect were in dyads receiving a classification of ambiva-
lent/resistant attachment. And a qualitative examination of the cases indicated a high level of
drug use, which contributed to leaving the infant unattended for long stretches.117
A puzzle in the findings from this study was that 12 of the 31 abused infants were clas-
sified as securely attached, using Ainsworth’s coding protocols. They displayed protest on
separation, proximity-seeking, and some contact-maintenance on reunion, but could be
comforted and were able to return to play. Yet they also displayed other behaviours that
seemed to interrupt or contradict the secure pattern described by Ainsworth. Informally,
the Minnesota group called these 12 cases ‘unhealthy Bs’: though they met the Ainsworth
criteria for Group B, their behaviour in the Strange Situation did not seem secure.118 In
watching the tapes closely, Egeland and Sroufe identified behaviour that was ‘neither avoid-
ant nor resistant’, but which nonetheless suggested insecurity, for instance responses to the
caregiver that were ‘apathetic or disorganized’.119 Rather than disregarding these cases as
and insecure attachment. Baer, J.C. & Martinez, C.D. (2006) Child maltreatment and insecure attachment: a meta-
analysis, Journal of Reproductive and Infant Psychology, 24(3), 187–97.
113 This is only somewhat higher than the 17% reported in a recent meta-analysis: Verhage, M.L., Schuengel, C.,
Madigan, S., et al. (2016) Narrowing the transmission gap: a synthesis of three decades of research on intergenera-
tional transmission of attachment. Psychological Bulletin, 142(4), 337–66, Table 4.
114 Warren, S.L., Huston, L., Egeland, B., & Sroufe, L.A. (1997) Child and adolescent anxiety disorders and early
attachment. Journal of the American Academy of Child & Adolescent Psychiatry, 36(5), 637–44, p.640. See also
Waters, E., Vaughn, B.E., & Egeland, B.R. (1980) Individual differences in infant–mother attachment relationships
at age one: antecedents in neonatal behavior in an urban, economically disadvantaged sample. Child Development,
51(1), 208–16.

116 Egeland, B. & Sroufe, L.A. (1981) Attachment and early maltreatment. Child Development, 52(1), 44–52,
p.50. The same factors would also, however, have reduced the caregivers’ attentional availability to their infant,
increasing the relevance of a strategy that increased attachment signals in order to have their caregiver heed and
respond to their desire for felt security. See Chapter 2 for criticism of the attribution of ‘inconsistent care’ as the
ultimate cause of ambivalent/resistant attachment behaviour.
117 A relationship between parental drug use and ambivalent/resistant attachment was also later documented
by Seifer, R., LaGasse, L.L., Lester, B., et al. (2004) Attachment status in children prenatally exposed to cocaine and
other substances. Child Development, 75(3), 850–68.
118 Alan Sroufe and Byron Egeland, personal communication, July 2012.
119 Egeland, B. & Sroufe, L.A. (1981) Developmental sequelae of maltreatment in infancy. New Directions for
Child and Adolescent Development, 1981(11), 77–92, p.84.

‘noise’ produced by imprecision in the Strange Situation procedure, they suspected that the
cases represented an attachment-relevant process of some significance. However, because
Sroufe was coding the 18-month Strange Situations, he needed to be ‘blind’ as to which re-
cordings were those of maltreated children. He therefore went rigorously out of his way to
avoid speculating about the aetiology of this behaviour until data analysis was complete.
Instead, Sroufe gave copies of these recordings to Main to examine (Chapter 3).
The Minnesota group were interested in the antecedents of the Ainsworth patterns
of attachment, to see whether their findings aligned with those of the Baltimore study.
Confirming Ainsworth’s conclusions, Egeland and Farber reported that caregivers in secure
dyads were more likely to display sensitivity during feeding episodes with their infants at
three and six months and free play at six months. Though the association between sensitivity
and security in the Minnesota study was much less strong than in Ainsworth’s data, it was

in the expected direction. The Minnesota group identified other variables that might sug-
gest a more favourable and settled caregiving environment for infants in securely attached
dyads. Mothers who lived with their partner for the first 12 months of the child’s life were
more likely to be part of a secure dyad.120 It was also noted that mothers who reported fewer
current adversities or stressors were more likely to have a child with a secure attachment re-
lationship, a finding that ‘shows the foolishness of blaming parents for developmental prob-
lems of the child. As life stress is reduced, quality of care improves, as does the quality of the
child’s attachment relationship.’121
Egeland and Farber examined nurses’ ratings of the mothers’ interest in their newborn
baby. They found that infants in dyads later classified as avoidant had mothers who were re-
ported to show less interest. Three months after the birth, caregivers in ambivalent/resistant
dyads also scored lower on the Maternal Attitude Scale. This was a self-report measure with
three scales: appropriate vs inappropriate control of child’s aggression, encouragement vs
discouragement of reciprocity, and acceptance vs denial of emotional complexity in child
care. Furthermore, they found that infants in ambivalent/resistant dyads scored lower on
scales of motor and mental development at nine months. The researchers suspected that if
the infants had been easier to care for, or had caregivers who accepted the complexity of
child care and could adapt to their child’s needs, many of these dyads would have developed
secure attachment relationships.122 They found no difference between groups in the amount
of warmth demonstrated to the infants by their caregivers. However, the mothers of infants
who changed from secure to anxious over the six months were rated by observers as showing
less delight in their baby than other caregivers at six months. Self-report measures of ma-
ternal personality, such as aggression and impulsivity, had little relationship with caregiving
behaviours or Strange Situation classifications.123
One potential form of adverse caregiving that especially interested the Minnesota group
was emotional neglect. This was a relatively new concept. An early advocate of the category
of emotional neglect had been John Bowlby. Though he tended to ignore the role of emotion
120 Egeland, B. & Farber, E.A. (1984) Infant–mother attachment: factors related to its development and changes
over time. Child Development, 55(3), 753–71.

52(4), 743–6, p.744.

122 Egeland, B. & Farber, E.A. (1984) Infant–mother attachment: factors related to its development and changes
over time. Child Development, 55(3), 753–71, p.767.

123 Egeland, B. & Brunnquell, D. (1979) An at-risk approach to the study of child abuse: some preliminary find-
ings. Journal of the American Academy of Child Psychiatry, 18(2), 219–35.

as a cause or process in psychological life in his theorising, he was highly attentive to child
emotion as an index of problems with the availability of adequate care. Already in 1971,
Bowlby was arguing to the Home Office select committee on the adoption of children that
there is an ‘urgent need’ for ‘a really adequate research study of emotional neglect’.124 The use
of emotional life in the identification of abuse and neglect was a matter of contention in the
early 1980s, as it seemed to involve an even stronger value judgement by social workers than
physical or sexual abuse or physical neglect.125 Yet emotional abuse and neglect swiftly grew
into the primary basis for referrals to child protection services.126 In 1985, Sroufe reported
on the antecedents of emotional neglect/psychological unavailability in the Minnesota
sample.127 Though the infants who received this form of care received typical scores for
mental and motor development at ten days and three months, by six months their develop-
ment was impeded, and they departed more and more from the rest of the cohort with each

further assessment of their development. At 12 months, 42% of the infant–caregiver dyads
were classified as avoidant, and at 18 months 86% received this classification.128 Much later,
at age 17, Egeland reported that the large majority met diagnostic criteria for a psychiatric
disorder: 53% met criteria for an anxiety disorder compared to 34% of the rest of the sample;
53% met criteria for a conduct disorder compared to 19% in the wider sample; and 25% met
criteria for PTSD compared to 12% in the cohort as a whole.129
In the 1990s, after the introduction of the disorganised/disoriented attachment classifica-
tion by Main and colleagues (Chapter 3), Elizabeth Carlson recoded the Minnesota Strange
Situation recordings for disorganised attachment and examined caregiving antecedents.
In her analyses she used the highest rating at either 12 or 18 months to represent a score
for disorganisation. Examining antecedents of disorganised attachment, she found that the
best predictor was a composite variable comprised of intrusive and insensitive caregiving
(r = .38), and that various forms of neglect and maltreatment made a modest additional con-
tribution (overall abuse/neglect r = .29; physical abuse r = .20; emotional neglect/psycho-
logical unavailability r = .23). She found no association between disorganised attachment
and maternal mental health or drug and/or alcohol abuse.130 This last finding, however, dif-
fers from later studies. A meta-analysis by Cyr and colleagues found that parental drug and/
124 Bowlby, J. (1971) Evidence presented on 6/4/1971 to the Committee on the Adoption of Children, Home
Office and Department of Health and Social Security, The National Archives, Kew, BN 29/2340.
125 Rohner, R.P. & Rohner, E.C. (1980) Antecedents and consequences of parental rejection: a theory of emo-
tional abuse. Child Abuse & Neglect, 4(3), 189–98; Trowell, J. (1983) Emotional abuse of children. Health Visitor,
56(7), 252–5.
126 Cicchetti, D. & Manly, J.T. (1990) A personal perspective on conducting research with maltreating fam-
ilies: problems and solutions. In G. Brody & I. Sigel (eds) Methods of Family Research: Families at Risk, Vol. 2
(pp.87–133). Hillsdale, NJ: Lawrence Erlbaum. More recently, see Bilson, A., Featherstone, B., & Martin, K. (2017)
How child protection’s ‘investigative turn’ impacts on poor and deprived communities. Family Law, 47, 316–19.
fant temperament. Child Development 56(1), 1–14. An early report on findings from this analysis was covered in
the New York Times:Brody, J.E. (1983) Emotional deprivation seen as devastating form of child abuse. New York
Times, 20 December 1983. This was unusual: over the decades there has been relatively little press coverage of the
Minnesota Longitudinal Study of Risk and Adaptation, as compared with other large longitudinal studies. Egeland
and Sroufe did not much seek out media appearances.
fant temperament. Child Development 56(1), 1–14, p.9.

129 Egeland, B. (1997) Mediators of the effects of child maltreatment on developmental adaptation in adoles-
cence. In D. Cicchetti & S. Toth (eds) Rochester Symposium on Developmental Psychopathology, Volume VIII: The
Effects of Trauma on the Developmental Process (pp.403–34). Rochester, NY: University of Rochester Press, Table 5.
Development, 69(4), 1107–28, Table 4.

or alcohol abuse did predict disorganised infant attachment, especially when compounded
by other risk factors.131 Likewise, some aspects of maternal mental health has been found to
be relevant to the prediction of infant insecure and insecure-disorganised attachment, such
as PTSD.132 (On the relationship between caregiver depression and infant attachment see
Chapter 3.)
Early social competence
Sroufe, Egeland, and colleagues did not regard secure attachment as any guarantee of posi-
tive later outcomes, but as a promotive or protective factor. Nonetheless, the Minnesota
Longitudinal Study of Risk and Adaptation demonstrated a number of important and valu-

able findings regarding the later positive correlates of infant secure attachment. In a follow-
up when the children were two years old, the dyads were invited to the laboratory; 190
families attended, of the 267 who started the study. Mother–child dyads were observed as
they completed a series of tool-based problem-solving tasks together, several of which were
more demanding than most two year olds would find comfortable. The reason for this was
that assessments of behaviour or cognition alone would risk being confounded with other
variables like IQ. Completion of a problem-solving task with a caregiver was assumed to
represent a developmentally appropriate challenge to a toddler, requiring the integration of
affect, cognition, and behaviour in order to stay on task and coordinate together effectively.
This meant that the task could serve as an assessment of the ability of the children to mobilise
and coordinate personal and social resources to realise the opportunities and potentials of
the environment.133
This capacity was termed by Waters and Sroufe ‘social competence’. The term was drawn
from scholars such as Diana Baumrind, for whom social competence signified the ‘social
responsibility, independence, achievement orientation, and vigor’ required of citizens of
American capitalism.134 Waters and Sroufe appreciated the metaphor of individual possibil-
ities as shaped by the resources available to them. However, they departed from Baumrind
and other individualising competency discourses in emphasising that qualities such as social
responsibility and independence are not best thought of as individual properties, and cer-
tainly not in children. Rather, they should be regarded as transactional effects of individual–
environment interactions, shaped by experiences of early care. The importance of early care
for Waters and Sroufe lay in the fact that it shapes both the inner and external resources
available to the child, and also influences how these resources are interpreted and used. In
line with this emphasis on early experiences of care, the Minnesota group found that, ac-
cording to the parent-report, children from dyads who had previously been classified as inse-
cure attached showed more behaviour problems at home. Those caregivers from previously
131 Cyr, C., Euser, E.M., Bakermans-Kranenburg, M.J., & Van Ijzendoorn, M.H. (2010) Attachment security and
disorganization in maltreating and high-risk families: a series of meta-analyses. Development & Psychopathology,
22(1), 87–108, Table 4.
132 E.g. Enlow, M.B., Egeland, B., Carlson, E., Blood, E., & Wright, R.J. (2014) Mother–infant attachment and the
intergenerational transmission of posttraumatic stress disorder. Development & Psychopathology, 26(1), 41–65.
133 Waters, E. & Sroufe, L.A. (1983) Social competence as a developmental construct. Developmental Review, 3,
79–97, p.81.
134 Baumrind, D. (1978) Parental disciplinary patterns and social competence in children. Youth & Society,
9(3), 239–67, p.249. See also White, R. (1959) Motivation reconsidered: the concept of competence. Psychological
Review, 66, 297–333.
insecure dyads who reported that their children showed fewer behavioural problems were
observed to be respectful of their autonomy, offered more support, provided clearer struc-
ture, and were less hostile during the problem-solving tasks. These were also mothers who
reported greater access to family support and friendship networks.135 This suggested that the
capacities of the caregiver were likewise shaped integrally by the availability of supportive
provisions, and should not be regarded merely as an individual quality.
In later assessments at preschool age, the Minnesota group found that an interaction be-
tween supportive maternal care and secure attachment predicted stronger capacity for man-
aging and delaying behaviour and wishes in order to flexibly achieve an overarching goal
(‘executive function’).136 In the two preschool classes constituted from children from the
sample, there were also distinct sequelae of infant attachment. Children who so infuriated
the teacher that they were sent to the corner were, on every occasion, those who had been

classified as having an avoidant attachment relationship with their caregiver in infancy.137
Sroufe and colleagues found that the children who were identified by observers—unaware of
the attachment classification—as unable to engage effectively in independent activities were
those who had earlier been in dyads classified as insecurely attached. The children from am-
bivalent/resistant dyads were observed to exhibit chronic, low-level contact seeking and at-
tention seeking. By contrast, the children who had earlier been classified as part of avoidant
dyads did not seek contact from the teacher, except when the intimacy of the contact would
be reduced, such as in large-group time.138 The preschoolers from insecure dyads were in
general less socially competent than their peers, and found making and retaining friends
more difficult.139
In the preschool classroom, the children were assigned into pairs and observed for a series
of free-play interactions. The researchers found that only children with histories of avoidant
attachment with their caregiver were seen to attempt to bully or victimise the other children.
When children ended up in a sustained role of bullied or victimised, these were always chil-
dren who had earlier received an insecure classification with their caregiver in the Strange
Situation.140 Sroufe and colleagues were interested in the fact that children with experience
of an avoidant attachment relationship could assume either role, victimiser or victim. They
interpreted this as evidence that, in receiving rejection or rebuff from their caregiver in re-
sponse to their attachment behaviour, the infant in an avoidantly attached dyad is learning
both how to reject and how to be rejected by others. Later researchers confirmed the link
between infant avoidant attachment relationships and later aggressive behaviour, with meta-
analytic research reporting an association of d = .58 for observation-based studies.141 Sroufe
135 Erickson, M., Sroufe, L.A., & Egeland, B. (1985) The relationship of quality of attachment and behavior prob-
lems in preschool in a high risk sample. Monographs of the Society for Research in Child Development, 50(1–2),
147–86, p.157, 164.
136 Meuwissen, A.S. & Englund, M.M. (2016) Executive function in at-risk children: importance of father-figure
support and mother parenting. Journal of Applied Developmental Psychology, 44, 72–80.
137 Sroufe, L.A. (1983) Infant–caregiver attachment and patterns of adaptation in preschool: the roots of mal-
adaptation and competence. In M. Perlmutter (ed.) Minnesota Symposium in Child Psychology, Vol. 16 (pp.41–83).
Hillsdale, NJ: Lawrence Erlbaum, p.76.

Relationship Disturbances in Early Childhood: A Developmental Approach (pp.70–94). New York: Basic Books.
140 Troy, M. & Sroufe, L.A. (1987) Victimization among preschoolers: the role of attachment relationship his-
tory. Journal of the American Academy of Child and Adolescent Psychiatry, 26(2), 166–72.
141 Fearon, R.P., Bakermans-Kranenburg, M.J., Van IJzendoorn, M.H., Lapsley, A.M., & Roisman, G.I. (2010)
The significance of insecure attachment and disorganization in the development of children’s externalizing
and colleagues put forward a hypothesis to account for the association between avoidant
attachment relationships and later aggressive behaviour.142 They argued that the attachment
relationship serves as a non-linear prototype for later relationships, with avoidant attach-
ment based on a history of unavailability, disappointment, and frustration. Viewing others
as hostile and remote, children with avoidant attachment relationships may respond with
aggression fed by their frustration, leading in turn to further rejection,143 and selective asso-
ciation with aggressive peers.144
Curiously, three decades later, this remains one of the only hypotheses proposed for
the association between avoidant attachment and aggressive behaviour, and no test of the
hypothesis has ever been specifically conducted.145 Writing with Michael Rutter, Sroufe
later acknowledged that the exact mechanism leading to conduct problems remains un-
clear: ‘Avoidant attachment in infancy is associated with later conduct problems, but avoid-

ant attachment is associated with low self-esteem, negative attributional biases, and rejection
by peers and teachers, all of which may predispose to conduct problems. We are only begin-
ning to understand how these chain effects operate.’146 One factor that may have inhibited
the generation and testing of hypotheses in this area has been the growth of findings that
suggest that disorganised infant attachment is a somewhat better predictor of later external-
ising behaviours than avoidant attachment.147
In their study of preschool behaviour, the Minnesota team additionally found differences
in the play of the children, in pairs, and also in other contexts. Preschoolers who had been
classified as secure had narratives to the stories and games they made that routinely assumed
a positive resolution to dangers, sickness, or other serious problems. If someone was hurt,
they were taken to hospital and made better. This follows what Waters and Waters later called
the ‘secure base script’ (Chapter 2). Children from avoidantly attached dyads were less likely
to create such secure base stories. In fact, ‘what is more noteworthy is the almost complete
behavior: a meta-analytic study. Child Development, 81(2), 435–56. Questionnaires of parents revealed an associ-
ation of d = .22. Questionnaires of teachers revealed an association of d = .30.
142 The hypothesis put forward later by Sroufe regarding the link between disorganised attachment and con-
duct problems could equally be applied as an alternative hypothesis in the case of avoidant attachment, given that
both showed substantial prospective links to dissociative tendencies in Carlson’s 1998 report. Sroufe, L.A. (2005)
Attachment and development: a prospective, longitudinal study from birth to adulthood. Attachment & Human
Development, 7(4), 349–67: ‘Disorganized attachment also predicts conduct disorder, we believe, because of the
dissociative tendencies and attendant problems with impulse control’ (361).
143 Renken, B., Egeland, B., Marvinney, D., Mangelsdorf, S., & Sroufe, L.A. (1989) Early childhood antecedents
of aggression and passive-withdrawal in early elementary school. Journal of Personality, 57(2), 257–81.
Multilevel Dynamics in Developmental Psychopathology: Pathways to the future. The Minnesota Symposia on Child
145 The matter has been examined in a meta-analytic treatment by Fearon, R.P., Bakermans-Kranenburg, M.J.,
Van IJzendoorn, M.H., Lapsley, A.M., & Roisman, G.I. (2010) The significance of insecure attachment and dis-
organization in the development of children’s externalizing behavior: a meta-analytic study. Child Development,
81(2), 435–56. An alternative/additional hypotheses for the association between avoidant attachment and exter-
nalising behaviours is the idea that when an avoidant strategy fails or is disrupted, anger may be evoked. This
may be as the intrusion of a segregated system, as proposed for example by Crittenden, P.M. (2016) Raising
Parents: Attachment, Representation, and Treatment, 2nd edn. London: Routledge. Mikulincer and Shaver also
suggested the idea of a dominance behavioural system, which may be engaged to supplement or replace a condi-
tional strategy (Chapter 5).
146 Rutter, M. & Sroufe, L.A. (2000) Developmental psychopathology: concepts and challenges. Development &
Psychopathology, 12(3), 265–96, p.271.

havior: a meta-analytic study. Child Development, 81(2), 435–56.
absence of fantasy play concerning people. Such fantasies dominate the play of almost all
preschool children and were well represented in the play of those with secure histories in
our sample. These data reveal sharp contrasts in the working models of the two groups—one
world is richly peopled, the other is not.’148 This finding was partially supported by later re-
search with the sample. As discussed in Chapter 3, when participants were asked to complete
a family drawing, the drawings of children from avoidantly attached dyads were not found
to be less populated than those of other children. However, adult figures were assessed by
coders blind to the infant attachment classifications as showing more emotional distance be-
tween figures, more tension or anger, and figures were more likely to have their arms rigidly
held at their sides rather than offering contact (e.g. holding hands).149
When the children in the sample were aged either five or eight, thirty families who also had
another child no more than three years younger were called back to the lab. The parent was

asked to leave and the siblings were given a set of games to complete together. For instance, in
one game the siblings had to build a house out of blocks; in another the siblings had to work
together to write using a stylus. The Minnesota group found that in sibling dyads where the
elder sibling took on the role of guide in an effective way, this elder sibling was more likely to
have had a history of secure attachment as assessed in earlier Strange Situations (r = .37).150
Furthermore, the extent to which the elder sibling took the role of guide, and the extent
to which they treated their sibling with sensitivity, independently predicted observer and
teacher assessments of their social competence at school, over and above attachment rela-
tionships. There was no association between attachment classifications and measures of how
much fun the siblings seemed to be having together or their cooperativeness. As such, the
Strange Situation seemed to predict, not a positive relationship in general, but specifically
secure base provision by older siblings to their younger brother or sister. This aligned with
Sroufe’s idea that children learn both roles in their relationships with attachment figures.
This quality of the sibling relationship, even accounting for parent–child attachment, was
found in turn to predict the elder sibling’s peer competence at school. Sroufe and colleagues
observed that such findings offer evidence for the potential importance of the attachment
components of the relationships with siblings in the development of peer competence.151
Overall, the Minnesota Longitudinal Study of Risk and Adaptation was able to offer ample
evidence that early attachment experiences are associated with early social competence. This
provided suggestive evidence towards a transactional model of individual–environment
interactions, in which experiences in relationships with attachment components con-
tribute to forms of social competence, perhaps especially through the availability of secure
base scripts. Slade and Holmes observed that the preschool findings at Minnesota were of
particular importance for establishing attachment research as a paradigm within develop-
mental psychology.152 Sroufe, Egeland, and colleagues situated their findings as demon-
strating the predictive validity of the Ainsworth Strange Situation, for instance in the finding
149 Fury, G., Carlson, E.A., & Sroufe, L.A. (1997) Children’s representations of attachment relationships in family
drawings. Child Development, 68(6), 1154–64.

150 Marvinney, D. (1988) Sibling relationships in middle childhood: implications for social-emotional develop-
ment. Unpublished doctoral dissertation, University of Minnesota. This relationship did not, however, hold when
the elder sibling was themselves secondborn.

152 Slade, A. & Holmes, J. (eds) (2014) Attachment Theory. London: Sage.
of an association between avoidant attachment and later conduct problems. Indeed, this
association has gone on to be well replicated. In retrospect, however, it is interesting that
in their analyses the Minnesota group did not quite treat the Strange Situation as the priv-
iledged measure of early attachment that it would become subsequently.153 For instance,
the Minnesota researchers would at times composite the Strange Situation with, or study
its interactions with, other measures of early care. Sroufe and Egeland appear to have felt
that ‘attachment’ itself cannot be directly measured, even by the Strange Situation, but must
be inferred from the network of correlations that link early care to a child’s later behaviour
in theoretically expectable ways (Chapter 2). Having established correlations in early child-
hood, a next step in appraising the influence of early attachment relationships would be to
see whether associations could still be found as the cohort matured into late childhood and
adolescence.

Adolescence
When the sample reached 11 years old, in 1986, 48 children were recruited to attend a
summer camp lasting four weeks. At this point, research on adolescence within academic
psychology was on the cusp on becoming established: empirical study of adolescence barely
existed in the late 1970s, and yet would be a central concern of developmental psychology
by the early 1990s. The Society for Research on Adolescence was founded in the winter of
1984, and its first meeting was held in in Madison, Wisconsin, in 1986.154 Adolescence was
an object of increasing public and policy concern. However, more proximally, the attention
given to the topic in developmental psychology can also be situated in the context of the ma-
turing of a number of longitudinal cohort studies established in the 1970s. The Minnesota
Longitudinal Study of Risk and Adaptation was a few years ahead of this general curve.
In the mid-to late 1980s, a central concern in the research literature was with how to
achieve the psychological measurement of aspects of adolescent social development in an
ecologically valid way.155 The use of three summer camps with members of the cohort study
was an ingenious solution. Most of the children were drawn from the subsample who had
been studied in the preschool and who were already therefore stratified by attachment clas-
sification. Each of the three camps had 16 children. The programme of activities included
group games, singing, swimming, art and craft projects, and sports. Weekly day trips were
taken to recreation parks, and there was also one overnight camp-out.156 During the camp,
the children were observed, and in the last week each child was interviewed twice about his
or her experiences of the camp and relationships with other campers.
Sroufe and colleagues examined the maintenance of appropriate gender boundaries
at the summer camp as one particular index of social competence at 11 years of age. The
construct of ‘gender boundary violations’ encompassed two rather different kinds of
153 Ziv, Y. & Hotam, Y. (2015) Theory and measure in the psychological field: the case of attachment theory and
the strange situation procedure. Theory & Psychology, 25(3), 274–91.

154 See Steinberg, L. & Lerner, R.M. (2004) The scientific study of adolescence: a brief history. Journal of Early
Adolescence, 24(1), 45–54.

155 See e.g. Petersen, A.C. (1988) Adolescent development. Annual Review of Psychology, 39, 583–607.
behaviour. One was hovering near opposite-sex groups or joining in their activities. The
other was forms of sexual harassment or cross-gender aggression. No report was made
on whether these behaviours were correlated. These two kinds of behaviours were aggre-
gated by the researchers, though the way that the scale was constructed gave more weight
to sexual harassment or cross-gender aggression than cross-gender interaction in gen-
eral.157 Elicker, Englund, and Sroufe found that there were marked negative associations
between gender boundary violations and peers’ ratings of campers social skill (r = –
.33).158 Children with a history of ambivalent/resistant relationships engaged in gender
boundary violations much more frequently. They also sat next to opposite-gender chil-
dren during large-group circle times more often. Gender boundary violations were pre-
dicted by earlier intergenerational boundary dissolution in early childhood.159 Gender
boundary violation would turn out to be an important variable in the Minnesota study,

with strong predictive value for later outcomes. For instance, gender boundary viola-
tions at age 11 had a negative association with contemporary (r = .58) and later (r = .31)
competence with peers in social situations,160 lower satisfaction in adolescent dating re-
lationships (r = .47),161 and a dramatic association with age of commencement of sexual
activity for girls (r = .75).162
During the summer camps, the 11-year-olds with histories of avoidant attachment rela-
tionships were indistinguishable from those with secure relationships when the tasks and
challenges of the camp related to physical tasks or the inanimate environment. They also
appeared confident in the first day of the camp, with less of the ordinary clumsy vulner-
ability and reserve of the other early adolescents. However, difficulties in social situations
became more apparent over time.163 For example, they were less effective when given tasks
that required inferences about the thoughts and feelings of others, and they displayed a
higher level of negative bias in their evaluations of their own group when asked to engage
157 Sroufe, L.A., Bennett, C., Englund, M., Urban, J., & Shulman, S. (1993) The significance of gender bound-
aries in preadolescence: contemporary correlates and antecedents of boundary violation and maintenance. Child
159 Sroufe, L.A., Bennett, C., Englund, M., Urban, J., & Shulman, S. (1993) The significance of gender bound-
aries in preadolescence: contemporary correlates and antecedents of boundary violation and maintenance. Child
160 Sroufe, L.A., Egeland, B., & Carlson, E.A. (1999) One social world: the integrated development of parent–
child and peer relationships. In W. . Collins & B. Laursen (eds) Relationships as Developmental Contexts: The
Minnesota Symposia on Child Psychology, Vol. 30 (pp.241–61). Mahwah, NJ: Lawrence Erlbaum, Table 11.4.
161 Collins, W.A., Hennighausen, K.H., Schmit, D.T., & Sroufe, L.A. (1997) Developmental precursors of ro-
mantic relationships: a longitudinal analysis. In S. Shulman (ed.) New Directions for Child Development. San
Francisco: Jossey-Bass (pp.69–84), p.78; Sroufe, L.A., Egeland, B., Carlson, E.A., & Collins, W.A. (2005) The
Development of the Person. New York: Guilford, p.186.
162 Ibid. p.194. Despite these intriguing results, in the two decades since the study took place, no further research
has been done using the gender violations scale. It might be suspected that the amalgamation of cross-gender soci-
ality and sexual harassment folded together phenomena that were of interest to two rather different communities
of researchers; both have subsequently been considered in the literature on gender and education, but generally
not together. An exception is research in qualitative feminist sociology, e.g. Renold, E. (2002) Presumed inno-
cence: (hetero) sexual, heterosexist and homophobic harassment among primary school girls and boys. Childhood,
9(4), 415–34.
163 Sroufe, L.A. & Egeland, B. (1991) Illustrations of person and environment interaction from a longitudinal
study. In T. Wachs & R. Plomin (eds) Conceptualization and Measurement of Organism–Environment Interaction
in groupwork.164 Kagan argued that the unruffled behaviour of infants in avoidant dyads in
the Strange Situation did not represent any kind of problem. He suggested that it may just
reflect infants with a bold temperament, or infants who had been socialised by their parents
to healthy independence.165 Elicker, Englund, and Sroufe argued that the sequelae of infant
avoidant attachment showed that neither interpretation was especially plausible. Instead, the
behaviours shown at the summer camp by participants who had been in avoidant attach-
ment relationships as infants could better be regarded as an effect of their experience of less
caregiver availability as a safe haven to help develop successful emotion regulation and social
competence.166
However, even if attachment does not begin as a trait, the Minnesota group concluded
that individual differences related to attachment may become stable with development.
When the children were recruited back for a summer camp reunion weekend at age 15,

assessments of peer interaction were conducted. Whereas there was little stability in in-
dividual behaviours, global assessments of peer competence were very stable between the
summer camp and the reunion.167 A curious finding reported by Sroufe, Egeland, and
Carlson was that peer competence at the reunion was predicted well by Strange Situation
classifications and preschool behaviour a decade earlier, with little additional prediction
accounted for by observations of social skills in the intervening years. The researchers were
led to conclude that ‘by age 5, much of the variance in adolescent social competence can be
accounted for. Even though numerous capacities are only nascent or not at all apparent in
the preschool period (e.g. coordinated group activities, intimacy).’168 Soufe and colleagues
highlighted that since preschool social competence was itself predicted by infant Strange
Situation classifications, there seemed to be both a direct and an indirect effect of attach-
ment on peer competence through childhood and adolescence. The researchers attributed
this especially to the role that secure base and safe haven availability have for the modula-
tion of emotion. Such capacity for modulation can contribute to skills in handling closeness
in the face of potential peer conflicts and to tolerating the stimulation of group activities.169
The emphasis on early attachment was also supported by another finding from the reunion
weekend, reported by Englund, Levy, Hyson, and Sroufe in 2000. Within the group task
at the reunion, the groups were asked to elect a spokesperson to report back to the wider
group. The long-term implications of early experiences within attachment relationships
were suggested by the fact that ‘out of a total of 16 adolescents named spokesperson, 3 were
165 Kagan, J. (1984) The Nature of the Child. New York: Basic Books; Jerome Kagan interviewed by Robert Karen,
1 January 1989, cited in Karen, R. (1998) Becoming Attached: First Relationships and How They Shape Our Capacity
to Love. Oxford: Oxford University Press: ‘My view is, if you’re attached, you are motivated to adopt the values of
your parents. If your parent values autonomy, you’ll be autonomous; if your parent values dependency, you’ll be
dependent’ (151).
167 Englund, M.M., Levy, A.K., Hyson, D.M., & Sroufe, L.A. (2000) Adolescent social competence: effectiveness
in a group setting. Child Development, 71(4), 1049–60.

168 Sroufe, L.A., Egeland, B., & Carlson, E. (1999) One social world: the integrated development of parent–
child and peer relationships. In W.A. Collins & B. Laursen (eds) Relationships as Developmental Context: The 30th
Minnesota Symposium on Child Psychology (pp.241–62). Hillsdale, NJ: Lawrence Erlbaum, 241–62, p.256.
169 Ibid. p.258.
classified as insecurely attached in infancy and 13 were classified as securely attached in

infancy’.170
W. Andrew (Andy) Collins was the Director of the Minnesota Institute of Child
Development from 1982. One of his particular research interests was the psychology of teen-
agers, and when the sample of the Minnesota Longitudinal Study of Risk and Adaptation
reached this age he became heavily involved with the cohort study. At age 16 years, partici-
pants were invited back to the laboratory and participated in interviews about their dating
and social lives, considered by the researchers to be key developmental challenges of mid-
adolescence. Participants who had been in dyads classified as ambivalent/resistant in the
Strange Situation were less likely than other participants to have dated, a finding Collins and
Sroufe attributed to the relative emotional and social immaturity of this group. Participants
who had been in dyads classified as securely attached in infancy were more likely to have had

a romantic relationship that lasted at least three months.171
When the sample were 20–21 years old, 78 participants who were in a romantic relation-
ship were invited to the laboratory along with their partner. Each member filled out a bat-
tery of self-report assessments of their relationship, and were also observed completing a
collaborative task and discussing a touchy subject for their relationship. First publications
from analysis of these data began to appear in 2001.172 Study of the observations of the call-
back at 20–21 years was supported by the appointment of Jeff Simpson as a new professor at
Minnesota in 2004. Simpson’s background as a social psychologist in the self-report trad-
ition of attachment research meant that he was well equipped for the study of adult romantic
relationships.173 The Minnesota group reported that infant attachment as assessed by the
Strange Situation predicted several aspects of adult relationship functioning, including a
robust association between infant disorganised attachment and the couple’s hostility with
one another (r = .42).174 They found that secure infant attachment assessed by the Strange
Situation was also associated with twice the rate of effective resolution of conflict among the
young adults regarding the touchy subject.175
Other variables were also found to be very important. One concurrent factor was gender.
Associations between Strange Situation classifications in infancy and the behaviour of the
couple were generally stronger for men than for women. The researchers observed that male
participants tended to set the tone in the collaborative task, even when female participants
170 Englund, M.M., Levy, A.K., Hyson, D.M., & Sroufe, L.A. (2000) Adolescent social competence: effectiveness
in a group setting. Child Development, 71(4), 1049–60, p.1056.

171 Collins, A. & Sroufe, L.A. (1999) Capacity for intimate relationships: a developmental construction. In W.
Furman, B.B. Brown, & C. Feiring (eds) (1999) The Development of Romantic Relationships in Adolescence (pp.125–
47). New York: Cambridge University Press, p.135.
172 Roisman, G.I., Madsen, S.D., Hennighausen, K.H., Sroufe, L.A., & Andrew Collins, W. (2001) The coherence
of dyadic behavior across parent–child and romantic relationships as mediated by the internalized representation
of experience. Attachment & Human Development, 3(2), 156–72.
173 In historical perspective, Simpson’s appointment can be regarded as an important moment for the social
psychological tradition (Chapter 5) which has held substantial sway at Minnesota since the retirement of Egeland
and Sroufe. Subsequently, the leaders of the Minnesota Longitudinal Study of Risk and Adaptation have been Jeff
Simpson—a social psychologist—and Glenn Roisman, whose time with Chris Fraley at Illinois was influential for
his attitude towards attachment methodology and theory (Chapter 3).

175 Salvatore, J.E., Kuo, S.I.C., Steele, R.D., Simpson, J.A., & Collins, W.A. (2011) Recovering from conflict in ro-
mantic relationships: a developmental perspective. Psychological Science, 22(3), 376–83.

had been confident and forthright in individual interviews.176 Other antecedent variables
were also highlighted.177 Of particular importance was the finding by Simpson, Collins,
Trans, and Haydon that the link between Strange Situation classifications and interaction
with the romantic partner was partially mediated by the quality of peer relationships at
school age.178 Though their findings qualified the emphasis on continuity of Bowlby’s early
and populist writings, the researchers regarded their findings as confirmation of the proposal
of developmental pathways from Bowlby’s late theoretical writings (Chapter 1).179 Aspects of
early care remained relevant to some aspects of adult functioning, even after decades.180 The
factors proposed by the Minnesota group were of particular importance for these develop-
mental pathways, such as emotion regulation and effective social relationships, and were also
those that would suggest links to mental health. Following the centrality of mental health to
Bowlby’s conceptualisation of attachment and its implications, the Minnesota group gave

this topic special attention.
Mental and physical health
When child participants in the cohort study reached 17 years of age, they were brought back
for a detailed psychiatric interview. Other assessments were conducted at the time, including
a holistic interview with the young people about their school, work, living situation, and
dating life—and a parallel interview about participants with their mother.181 If, as Slade and
Holmes argued, the preschool study at Minnesota provided important support for attach-
ment research as a paradigm in the 1990s, a parallel role was played by the findings from the
psychiatric interview at age 17 in the 2000s, reported by Carlson.182 In general, Kobak and
Bosmans observed that infant attachment has ultimately been disappointing as a predictor
176 Hennighausen, K.H. (1999) Developmental antecedents of young adult romantic relationships. Unpublished
doctoral dissertation, University of Minnesota.

177 See e.g. Madsen, S.D. & Collins, W.A. (2011) The salience of adolescent romantic experiences for romantic
relationship qualities in young adulthood. Journal of Research on Adolescence, 21(4), 789–801; Labella, M.H.,
Johnson, W.F., Martin, J., et al. (2018) Multiple dimensions of childhood abuse and neglect prospectively predict
poorer adult romantic functioning. Personality and Social Psychology Bulletin, 44(2), 238–51.
178 Simpson, J.A., Collins, W.A., Tran, S., & Haydon, K.C. (2007) Attachment and the experience and expression
of emotions in romantic relationships: a developmental perspective. Journal of Personality and Social Psychology,
92(2), 355–67. The mediational account has, however, later been criticised by Gillath and colleagues. Reviewing
findings from other cohort studies, they argued that the model presented by Simpson and colleagues underesti-
mates the extent to which early attachment experiences provide scaffolding for the quality of peer relationships.
The quality of peer relationships is therefore not an independent intervening variable. Gillath, O., Karantzas,
G.C., & Fraley, R.C. (2016) Adult Attachment: A Concise Introduction to Theory and Research. London: Academic
Press, p.69.
179 Van Ryzin, M.J., Carlson, E.A., & Sroufe, L.A (2011) Attachment discontinuity in a high- risk sample.
180 In a later meta-analysis based on 80 independent samples (N = 4441), Groh and colleagues found a robust
association between attachment security with mother and peer competence (d = .39). The association was of the
same strength no matter how long after the attachment assessment peer competence was measured: there was no
moderation by age, suggesting effects were enduring. Groh, A.M., Fearon, R.P., Bakermans-Kranenburg, M.J., van
IJzendoorn, M.H., Steele, R.D., & Roisman, G.I. (2014) The significance of attachment security for children’s social
competence with peers: a meta-analytic study. Attachment & Human Development, 16(2), 103–136, p.128.
181 https://innovation.umn.edu/parent-child/wp-content/uploads/sites/35/2018/08/M001-List-of-Measures.pdf.
Development, 69(4), 1107–28.

of adult mental health, with ‘lack of specificity and relatively small effect sizes’.183 However,
Carlson’s report helped to support the idea that early attachment as measured by the Strange
Situation does play a distinct and important role, even if this role is of less magnitude than
Bowlby tended to suggest. In line with the general preference at Minnesota for composite
measures, Carlson treated the psychiatric interview as a measure of overall mental health,
coded on a seven-point scale. She did not differentiate by kind of mental health problem. She
found that infant avoidant attachment, infant disorganised attachment, middle childhood
behaviour problems, and poor observed child–parent boundaries at age 13 each made an
independent contribution to total mental health problems at age 17, together accounting for
a third of the variance.
In the past few years, meta-analytic research has led several senior figures in the devel-
opmental tradition of attachment research to query the conclusion drawn by Carlson, and

taken by her readers, that Strange Situation classifications are associated with general mental
health problems. Meta-analytic work indicated that infant avoidant and disorganised attach-
ment in the Strange Situation tended to be much more associated with externalising prob-
lems, and less with internalising problems such as anxiety or depression.184 In fact, Carlson
was not actually intending to make the specific claim that avoidant and disorganised attach-
ment would be generally associated with all mental ill health185—though technically this
is what she reported and what readers naturally understood. Instead, her report was based
on the Minnesota creed of using composite measures, especially for pioneering work, with
the expectation that later researchers would pursue differentiation.186 Today, the question of
whether the Strange Situation predicts primarily externalising problems remains an area of
ongoing discussion among researchers, not least since several phenomena of clinical con-
cern (e.g. PTSD, ADHD) are poorly captured by the division between externalising and
internalising mental health issues.187
As well as studying prospective links between the Strange Situation and general mental
health at age 17, building on the ideas of Main and colleagues (Chapter 3) Carlson conducted
a hypothesis-driven analysis of associations with dissociative symptoms. She found that in-
fant avoidant attachment, infant disorganised attachment, and middle childhood behaviour
problems made independent contributions, and together accounted for 17% of the symp-
toms. Infant disorganised attachment and child abuse also made independent contributions
to later self-injurious behaviour and other symptoms of borderline personality disorder. The
association between the Strange Situation classifications and later borderline symptoms was
183 Kobak, R. & Bosmans, G. (2018) Attachment and psychopathology: a dynamic model of the insecure cycle,
Current Opinion in Psychology, 25, 76–80, p.76.

184 Groh, A.M., Fearon, R.P., van IJzendoorn, M.H., Bakermans-Kranenburg, M.J., & Roisman, G.I. (2017)
Development Perspectives, 11(1), 70–76.
185 Betty Carlson, personal communication, May 2019.
186 Though not discussed in the paper, in Table 5 Carlson presented findings from teacher report of participants
suggesting that disorganised attachment was associated weakly with internalising problems in elementary and
high school (r = .19, .18), and had no association with externalising problems. Carlson, E.A. (1998) A prospective
longitudinal study of attachment disorganization/disorientation. Child Development, 69(4), 1107–28, Table 5. The
findings may be the product of chance. However, it is conceivable that the distinctive predominance of child neg-
lect in the Minnesota sample compared to other studies may have played a role in these results running contrary to
later meta-analyses of samples.
187 See e.g. Jacobvitz, D., Hazen, N., Zaccagnino, M., Messina, S., & Beverung, L. (2011) Frightening maternal
behavior, infant disorganization, and risks for psychopathology. In D. Cicchetti & G.I. Roisman (eds) The Origins
and Organization of Adaptation and Maladaptation: The Minnesota Symposia on Child Psychology, Vol. 36 (pp.283–
322). Hoboken, NJ: Wiley.
mediated by responses to the family drawing task at age eight and interview measures of rep-
resentations of the self and key relationships (including the ‘friendship interview’) in early
adolescence.188
Another paper by the Minnesota group reported findings that infant ambivalent/resistant
attachment was distinctively associated with anxiety problems in adolescence (r = .26), even
controlling for maternal anxiety and child temperament, but had no other links to mental
illness.189 This was important in providing a source of prospective validity for the category.
Following Main (Chapter 3), Sroufe proposed that the infant ambivalent/resistant attach-
ment pattern entails a strategy of hypervigilance and hyperattentiveness regarding the avail-
ability of the caregiver: ‘Such a stance may be adaptive in insuring contact with the caregiver
when there is a genuine threat, but a price is paid for such chronic wariness and vigilance.’190
However, attempts to replicate the finding of a link between infant ambivalent/resistant at-

tachment and later anxiety problems have failed, at least when observational and interview
measures have been used.191
Another finding that ran contrary to expectations was a report by the Minnesota group
using a composite of Strange Situation classifications and measures of early caregiving. The
researchers examined the association of this composite with depression at age 17, antici-
pating based on available theory that insecure attachment and early care would both be ro-
bust predictors of later depression.192 Yet Duggal, Carlson, Sroufe, and Egeland reported in
2001 that in the Minnesota study the composite accounted for only 3% of variance, control-
ling for maternal depression. Furthermore, it lost significance entirely in a regression once
account was taken of the amount of support from family and friends the mother received in
the child’s first five years. By contrast, prospectively documented child abuse remained sig-
nificant in predicting depression: together, maternal social support and documented child
abuse accounted for 19% of variance in adolescent depression.193 However, intriguingly,
188 Carlson, E.A., Egeland, B., & Sroufe, L.A. (2009) A prospective investigation of the development of border-
line personality symptoms. Development & Psychopathology, 21(4), 1311–34. A later study by Lyons-Ruth and
colleagues found only a marginal association between infant disorganised attachment and later self-injurious be-
haviours, but there was a marked relationship between maternal withdrawing behaviours in response to child dis-
tress, forms of disorganised attachment that entailed approaching the caregiver, and suicidality in early adulthood.
Lyons-Ruth, K., Bureau, J.F., Holmes, B., Easterbrooks, A., & Brooks, N.H. (2013) Borderline symptoms and sui-
cidality/self-injury in late adolescence: prospectively observed relationship correlates in infancy and childhood.
Psychiatry Research, 206(2–3), 273–81.
189 Warren, S.L., Huston, L., Egeland, B., & Sroufe, L.A. (1997) Child and adolescent anxiety disorders and early
attachment. Journal of the American Academy of Child & Adolescent Psychiatry, 36(5), 637–44.

Child Development, 83(2), 591–610; Madigan, S., Atkinson, L., Laurin, K., & Benoit, D. (2013) Attachment and in-
ternalizing behavior in early childhood: a meta-analysis. Developmental Psychology, 49(4), 672–89. In contrast, a
positive association (r = .37) between resistant attachment and anxiety has been reported when self-report measures
of anxiety are included in the meta-analysis: Colonnesi, C., Draijer, E.M., Jan, J.M., et al. (2011) The relation between
insecure attachment and child anxiety: a meta-analytic review. Journal of Clinical Child and Adolescent Psychology,
40, 630–45. This would suggest that one of the distinct sequelae of resistant attachment is participant report of anx-
iety symptoms on self-report measures, perhaps as a consequence of a maximising strategy (Chapter 5).
192 E.g. Cicchetti, D. & Schneider-Rosen, K. (1986) An organizational approach to childhood depression. In M.
Rutter, C. Izard, & P. Read (eds) Depression in Young People: Clinical and Developmental Perspectives (pp.71–134).
New York: Guilford.
193 Duggal, S., Carlson, E.A., Sroufe, L.A., & Egeland, B. (2001) Depressive symptomatology in childhood and
adolescence. Development & Psychopathology, 13(1), 143–64. The finding that insecure attachment does not pre-
dict later depression after taking into account maternal social support ran too far contrary to expectations, and has
generally been ignored by subsequent researchers who have continued to treat insecure childhood attachment as a
more recently when the sample were followed up to adulthood, symptoms of depression
were more likely to remit among those participants who had been in a secure attachment re-
lationship as infants.194 Early attachment relationships may play less of a role in predisposing
or triggering depression than in stabilising symptoms.
Somewhat better specificity and greater effect sizes were seen with respect to responses
to trauma. Enlow, Egeland, and colleagues reported an analysis of PTSD symptoms in data
from the adolescent psychiatric interview. Rates of PTSD were doubled for dyads where
mothers were assessed as showing emotional neglect/psychological unavailability.195 There
was no association between Strange Situation classifications and later traumatic experiences
per se.196 However, for those participants with insecure attachment relationships who did
experience later trauma, there was a dose-response relationship with PTSD symptoms: 12%
of infants in dyads classified as secure at both 12 and 18 months in the Strange Situation

developed PTSD following trauma exposure; 28% received the diagnosis when assessed as
belonging to an insecure dyad in either the 12-or 18-month Strange Situation; and 46% had
PTSD in adolescence following trauma exposure if they had an insecure attachment rela-
tionship at both 12 and 18 months.197 Disorganised infant attachment appeared to make no
additional contribution beyond insecurity.
Sroufe, Carlson, Levy, and Egeland also reported findings regarding the respective contri-
bution of Strange Situation classifications and middle childhood behaviour problems to later
mental health. Children with a history of secure attachment relationships who showed few
conduct problems at school had the lowest scores for mental illness. Children with a history
of insecure attachment relationships who showed significant conduct problems at school
had the very highest scores for mental illness, and the lowest scores for their education and
social relationships. The researchers reported, however, that children with a mixed history
(secure attachment and later behaviour problems; or insecure attachment and later positive
functioning) had equivalent scores. They were likewise functioning at an equivalent level
in their education and social relationships. Sroufe and colleagues concluded that children
who were struggling in middle childhood but who had a secure attachment relationship in
infancy ‘drew upon a more positive foundation’, which accounted for their better outcomes
than their peers with insecure infant attachment relationships.198
In educational research in the 1990s, adolescent conduct problems and poor achieve-
ment were identified as cardinal predictors of young people dropping out of school. The
Minnesota Longitudinal Study of Risk and Adaptation was unusual in being able to examine
cause of depression in adult caregiers. See e.g. Toth, S.L., Rogosch, F.A., & Cicchetti, D. (2008) Attachment-theory
informed intervention and reflective functioning in depressed mothers. In H. Steele & M. Steele (eds) The Adult
Attachment Interview in Clinical Context (pp.154–72). New York: Guilford: ‘Retrospective studies have found that
depressed adults report histories replete with inadequate or abusive care. Thus insecure childhood attachment
relationships may contribute to mothers’ depression’ (155). It is possible, however, that there are two different diag-
nostic entities in play. Duggal and colleagues were reporting on depressive symptoms; Toth and colleagues were
discussing major depressive disorder.
194 Sroufe, L.A., Coffino, B., & Carlson, E.A. (2010) Conceptualizing the role of early experience: lessons from
the Minnesota longitudinal study. Developmental Review, 30(1), 36–51, p.45.

195 Sroufe, L.A., Egeland, B., Carlson, E.A., & Collins, W. A. (2005) The Development of the Person.

196 Enlow, M.B., Egeland, B., Carlson, E., Blood, E., & Wright, R.J. (2014) Mother–infant attachment and the
intergenerational transmission of posttraumatic stress disorder. Development & Psychopathology, 26(01), 41–65.
197 Ibid. p.54.
198 Sroufe, L.A., Carlson, E.A., Levy, A.K., & Egeland, B. (1999) Implications of attachment theory for develop-
mental psychopathology. Development & Psychopathology, 11(1), 1–13, p.8.

this question prospectively, including variables from early childhood. A regression analysis
indicated that the child’s gender, the degree of chaos in the early home environment, and a
composite of infant attachment and caregiver sensitivity could predict 77% of children who
had dropped out of school by age 16. When these variables were included, more proximal
variables such as truancy, disciplinary problems, and failing grades in high school were no
longer significant. The same was true for peer rejection, conduct problems, and low school
achievement. As such, Jimerson, Egeland, Sroufe, and Carlson concluded that ‘many estab-
lished “predictors” may be better conceptualized as markers of presence on the pathway’,
leading from early caregiving to dropping out of school. The reason for this, they argued,
is that ‘success in school calls upon numerous capacities for self-regulation that begin to
be formed in the early years’.199 Raby and colleagues, in fact, later documented that early
maternal sensitivity predicted academic achievement by adulthood at least as well as it pre-

dicted social competence, controlling for covariates such as maternal education and socio-
economic status.200 The picture is complicated, unsurprisingly, by the fact that experiences
of education have a bidirectional relationship with development. Englund, Egeland, and
Collins elsewhere documented that developmental outcomes could be altered, for some
children, by their experience of the school itself.201 As we saw in Chapter 2, Ainsworth
anticipated that feelings of security in different domains would influence each other. And
Bowlby specifically identified that a school may serve attachment functions for a young
person.202
Though social psychologists had pursued research on the topic for over a decade,203 in
the 2010s the relationship between attachment and physical health outcomes was heralded
as a ‘new frontier’ for developmental attachment research.204 Puig, Englund, Simpson, and
Collins examined the links between early attachment and later medically unexplained or
‘non-specific’ symptoms at age 32, controlling for a variety of covariates including adult
socioeconomic status, life stress, and social support. Individuals who had been part of am-
bivalent/resistant dyads in infancy as assessed in the Strange Situation were three times more
likely to report symptoms such as dizzy spells, headaches, skin trouble, back ache, and in-
digestion than those adults who had been part of secure dyads. They were also six times
more likely to report physical health conditions. Individuals who had been part of avoidantly
attached dyads in infancy were three times more likely than those from secure dyads to re-
port inflammation-related illnesses such as heart disease, high blood pressure, and asthma.
Participants who were part of dyads classified as secure at both 12 months and 18 months re-
ported the fewest physical illnesses in adulthood compared to those who were part of dyads
classified as insecure at one or both assessments.
199 Jimerson, S., Egeland, B., Sroufe, L.A., & Carlson, B. (2000) A prospective longitudinal study of high school
dropouts examining multiple predictors across development. Journal of School Psychology, 38(6), 525–49, p.544.
200 Raby, K.L., Roisman, G.I., Fraley, R.C., & Simpson, J.A. (2015) The enduring predictive significance of early
maternal sensitivity: social and academic competence through age 32 years. Child Development, 86(3), 695–708.
201 Englund, M.M., Egeland, B., & Collins, W.A. (2008) Exceptions to high school dropout predictions in a low-
income sample: do adults make a difference? Journal of Social Issues, 64(1), 77–94.
202 Bowlby, J. (1969, 1982) Attachment. London: Penguin, p.207. The potential for school to serve as a secure
base has been elaborated by Geddes, H. (2006) Attachment in the Classroom: A Practical Guide for Schools.
London: Worth Publishing.
203 For a review see Pietromonaco, P.R. & Beck, L.A. (2019) Adult attachment and physical health. Current
Opinion in Psychology, 25, 115–20.

204 Cassidy, J. (2015) Early relationships, later functioning: why and how a secure base matters. Keynote address
presented at 7th International Attachment Conference, New York; Ehrlich, K.B. & Cassidy, J. (2019) Attachment
and physical health: introduction to the special issue. Attachment & Human Development, 21(1), 1–4.
The researchers proposed that caregiver provision of a secure base and safe haven in the
context of stable routines offers an external source of biological, as well as psychological,
regulation. Insensitive care may heighten stress reactivity across the lifespan, inscribed on
the body in the form of ‘non-specific’ symptoms like headaches, skin trouble, and back ache,
where physical and psychological stress together may contribute to symptom development
and maintenance over time. Conversely, sensitive care may reduce stress reactivity across the
lifespan, leading to fewer non-specific symptoms and fewer symptoms relating to the human
inflammatory response, which is activated by alarming or threatening experiences.205 This
hypothesis has been confirmed by other findings from the Minnesota group showing that
assessments of maternal sensitivity blocked the association between stressful life events and
physical ill health, and also predicted a lower body-mass index and fewer physical symp-
toms for participants with fewer stressful life events.206 Also relevant to this hypothesis were

findings that child neglect contributed to adult physical health according to both self-report
measures of physical symptoms and biomarkers of cardiometabolic risk.207
The study of attachment and health has been an exciting area of research in the social
psychological tradition over the past decade.208 It seems likely that investigation of the im-
plications of attachment for physical health is likely to be a growth area over the coming
years in the developmental tradition, pursuing the pioneering and striking findings from
Minnesota.209 This will be facilitated by the collection of genetic data from the cohort. It
will also be facilitated by the fact that a number of new large cohort studies were initiated
in the 2000s in Europe with a rich array of measures of physical health. For example, the
Generation R cohort study in the Netherlands includes Strange Situation procedures with
900 infants, a number of observational measures of parent and child behaviour, and a vast
array of assessments of physical health over the subsequent decade, as well as genetic in-
formation on the sample.210 The prospect of new integrations between attachment research
and biological and medical research using the Minnesota Longitudinal Study of Risk and
Adaptation and other cohort studies seems not only possible, but likely.
Another area where future reports are to be expected is on three-generational processes.
The Minnesota group recently followed up the grandchildren of the original women recruited
to the sample. Strange Situation procedures were conducted with these grandchildren. Raby,
Steele, Carlson, and Sroufe found little evidence of intergenerational continuities in attach-
ment security. One pattern in the data was that mothers who had themselves had insecure
attachment relationships in the 1970s but who formed secure attachment relationships with
their own infant had higher-quality concurrent support from family and friends. Though
205 Puig, J., Englund, M.M., Simpson, J.A., & Collins, W.A. (2013) Predicting adult physical illness from infant
attachment: a prospective longitudinal study. Health Psychology, 32(4), 409–417.

206 Farrell, A.K., Simpson, J.A., Carlson, E.A., Englund, M.M., & Sung, S. (2017) The impact of stress at different
life stages on physical health and the buffering effects of maternal sensitivity. Health Psychology, 36(1), 35–44; see
also Farrell, A.K., Waters, T.E.A., Young, E.S., et al. (2019) Early maternal sensitivity, attachment security in young
adulthood, and cardiometabolic risk at midlife. Attachment & Human Development, 21(1), 70–86.
207 Johnson, W.F., Huelsnitz, C.O., Carlson, E.A., et al. (2017) Childhood abuse and neglect and physical health
at midlife: prospective, longitudinal evidence. Development & Psychopathology, 29(5), 1935–46.

208 Pietromonaco, P.R. & Beck, L.A. (2019) Adult attachment and physical health. Current Opinion in Psychology,
25, 115–20.
209 Ehrlich, K.B. & Cassidy, J. (2019) Attachment and physical health: introduction to the special issue.

210 E.g. Windhorst, D.A., Mileva- Seitz, V.R., Linting, M., et al. (2015) Differential susceptibility in a de-
velopmental perspective: DRD4 and maternal sensitivity predicting externalizing behavior. Developmental
Psychobiology, 57(1), 35–49.
there was little continuity between generations in security or insecurity in general, there was
specifically substantial continuity in disorganised infant attachment: half of the mothers who
had been part of dyads that received a classification of disorganised attachment in the 1970s
had children who received the same classification. This was in contrast to only 20% of dyads
where the mother had no history of an infant disorganised attachment classification.211
Dissociation
For Sroufe, an individual’s experience of ‘self ’ was best conceptualised as an emergent or-
ganisation, constituted through the routine integration of cognition, emotion, and behav-
iour. Within this organisation, the whirring clockwork that sustains ordinary adaptation

is assembled out of both historical and contemporary machinery. Earlier components are
incorporated into later behavioural systems and may remain accessible. This produces the
strange and uneven inner space of thoughts and feelings characteristic of human experience,
as well as contributing to the way that our own behaviour has the capacity to surprise or dis-
concert us. Sroufe’s emphasis, however, was above all on the coherence of the assemblage of
components, an expression of his methodological principle of holism.212 He regarded this
constituted organisation as experientially stabilised by memory processes, which register
continuities in an individual’s qualities and capacities across contexts and relationships.213
Nonetheless, memory, expectations, and images of the value of the self in the eyes of others
can become a facet of self-monitoring and a guide for behaviour, filtering the meaning of
later social experiences. Experiences in relationships, in turn, contribute to the further sta-
bilisation or reorganisation of the inner organisation: ‘the self was forged within vital rela-
tionships; within such relationships, it continues to evolve’.214
The identification of selfhood with organisation over the course of development contrib-
uted to the special attention given by Sroufe and colleagues to dissociative symptoms. In
Defences that Follow Loss, Bowlby had situated dissociation as a form of defensive exclusion,
a psychological process that blocked episodic information about the perception of events
in time and space from reaching and activating a behavioural system (Chapter 1). It was a
‘particular sort of segregating process’, and in this had elements in common with attachment
avoidance, even if Bowlby believed that the two were distinguishable.215 However, Bowlby
did not publish his speculations on dissociation. Though he had Robertson’s qualitative ob-
servations, he lacked the data to press further. Writing in 1996, Egeland and Susman-Stillman
211 Raby, K.L., Steele, R.D., Carlson, E.A., & Sroufe, L.A. (2015) Continuities and changes in infant attachment
patterns across two generations. Attachment & Human Development, 17(4), 414–28.
212 Sroufe, L.A. & Rutter, M. (1984) The domain of developmental psychopathology. Child Development, 17–29, p.20.
Holism is listed top of the list of principles governing any true developmental perspective; Sroufe, L.A. (1996) Emotional
Development. Cambridge: Cambridge University Press, p.8. On the history of holism within psychology see Shelley, C.
(2008) Jan Smuts and personality theory: the problem of holism in psychology. In R. Diriwächter & J. Valsiner (eds)
Striving for the Whole: Creating Theoretical Syntheses (pp.89–109). New Brunswick: Transaction Publishers.
213 Sroufe, L.A. (1990) An organizational perspective on the self. In D. Cicchetti & M. Beeghly (eds) The Self in
Transition: Infancy to Childhood (pp.281–307). Chicago: University of Chicago Press: ‘The self should be conceived
as an inner organisation of attitudes, feelings, expectations and meanings, which arises itself from an organised
caregiving matrix (a dyadic organisation that exists prior to the emergence of the self) and which has organisational
significance for ongoing adaptation and experience. The self is organisation. It arises from organisation. It influ-
ences ongoing organisation of experience’ (281).
214 Ibid. p.303.
215 Bowlby, J. (1962) Defences that follow loss: causation and function. PP/Bow/D.3/78.
acknowledged that, in the wake of Bowlby’s threadbare published comments, ‘there is some
confusion over the relation of attachment and dissociation’.216
The Minnesota group included questions relevant to child dissociative symptoms in
twelve different assessments made by mothers and teachers, the first in kindergarten and the
last at 16 years. At age 19, the 168 participants remaining in the study provided self-report
of their own experiences of dissociative symptoms on the Carlson and Putnam Dissociative
Experiences Scale.217 Mothers were asked about potential traumatic life events experienced
by their child during the 12, 18, 30, 42, 48, 54, and 64 month assessments and when their
child was 19 years old. At 19 years, participants themselves were also asked to complete this
assessment. Ogawa, Sroufe, and colleagues found that in preschool, dissociative symptoms
were predicted well by neglect in infancy.218 In elementary school, 27% of variance in chil-
dren’s dissociative symptoms was predicted by their mother’s experience of abuse as a child,

physical abuse experienced by the child, and avoidant attachment in the Strange Situation. In
both preschool and elementary school, the child’s IQ made a small negative contribution to
the score. In adolescence, the best predictor of dissociative symptoms was avoidant attach-
ment assessed in the Strange Situation. Additional contributions were made by exposure to
domestic violence in infancy, disorganised attachment as assessed in the Strange Situation,
and experiences of physical abuse in adolescence. Together these variables accounted for
19% of variance in dissociative symptoms. At age 19, 30% of variance in dissociation could
be accounted for by only three variables, all from infancy: caregiver psychological unavail-
ability, infant disorganised attachment, and maternal-report measures of infant distracti-
bility and attention-span. Psychological unavailability in infancy alone accounted for 19%
of variance. It was curious that, besides infant attachment, the 19-year assessment was pre-
dicted so powerfully by two variables that had previously not proven relevant, both from
infancy. However, it should be noted that the 19-year assessment was a self-report of dis-
sociative symptoms by the young persons themselves, whereas previous waves had been ma-
ternal or teacher report.
Ogawa, Sroufe, and colleagues also examined factors that predicted dissociative symp-
toms in the clinical range, participants with high but not clinical levels of symptoms, and
participants with few symptoms. In preschool, they found that infant neglect predicted
symptoms in the clinical range, whereas witnessing domestic violence and parental psy-
chological unavailability predicted high but not clinical levels of symptoms. The two groups
were quite distinct in terms of their antecedents. Witnessing domestic violence was also as-
sociated with high but not clinical levels of symptoms in elementary school. Symptoms in
the clinical range were differentially predicted by neglect in infancy, infant disorganised at-
tachment in the Strange Situation, and concurrent physical abuse experienced by the school-
aged child. In adolescence, the clinical group was distinguished strongly by infant avoidant
and infant disorganised attachment in infancy. High but not clinical levels of symptoms were
predicted by concurrent physical abuse. In early adulthood, the clinical group was distin-
guished from the others on the basis of maternal psychological unavailability in infancy and
216 Egeland, B. & Susman-Stillman, A. (1996) Dissociation as a mediator of child abuse across generations. Child
Abuse & Neglect, 20(11), 1123–32, p.1129. The work by Main and Hesse offering the most sustained attention to
these issues remained published only in Italian (Chapter 3).
217 Carlson, E.B. & Putnam, F.W. (1993) An update on the Dissociative Experiences Scale. Dissociation, 6, 16–27.
218 Ogawa, J.R., Sroufe, L.A., Weinfield, N.S., Carlson, E.A., & Egeland, B. (1997) Development and the frag-
mented self: longitudinal study of dissociative symptomatology in a nonclinical sample. Development &
disorganised attachment in the Strange Situation. The authors interpreted their data as sug-
gesting two pathways. A first pathway was to clinical levels of dissociation, based to a sur-
prising degree on experiences in infancy. These include experiences of neglect in infancy,
and of avoidant and disorganised attachment. In fact, ‘an examination of the zero-order cor-
relations between disorganized attachment and dissociation at each time period reveals a
linear trend with a positive slope (Time 2 r = .14, n.s.; Time 3 r = .20 p < .05; Time 4 r = .25 p
< .01; Time 5 r = .35 p < .001)’.219 A separate analysis was run which found that disorganised
infant attachment alone marginally elevated dissociative symptoms in early adulthood, but
that disorganised attachment and the experience of at least one form of trauma had a strong
link to dissociative symptoms (F[2, 125] = 12.0, p < .001).220
The second pathway was to subclinical but substantial levels of dissociative symptoms.
This pathway was predicted by concurrent physical abuse, and tended to stop once the par-

ticipants reached the greater independence of young adulthood. Ogawa, Sroufe, and col-
leagues concluded that ‘psychopathological dissociation should not be viewed as the top end
of a continuum of dissociative symptomatology, but as a separate taxon’.221 However, they
highlighted that avoidant attachment appeared to make an independent and robust con-
tribution to dissociative symptoms, except to the participant self-report measure. Whereas
dissociation reduces environmental responsiveness by blocking a form of perception, avoid-
ance retains awareness of the environment whilst bracing against the attachment system,
directing attention away from cues that might activate it (Chapter 1). Nonetheless, Ogawa,
Sroufe, and colleagues reflected that the contribution of infant attachment to later observed
dissociative symptoms perhaps lies in the fact that avoidance may at times shade into dis-
sociative processes, or may predispose a vulnerability to dissociation. Nonetheless, the lack
of correlation between avoidance and self-report of dissociation required explanation. The
Minnesota group drew on Bowlby’s published reflections to speculate that avoidant attach-
ment as a strategy of defensive exclusion may be more mature in young adulthood, in the
context of cognitive maturation and social independence, allowing the strategy to work
more effectively, with less ‘leakage’ in the form of dissociation. It might also be supposed
that avoidant attachment strategies could suppress awareness and/or reporting of dissocia-
tive symptoms.
Giovanni Liotti had proposed that infant neglect and avoidant and disorganised at-
tachment may all have in common that they disrupt the early integrative processes that
pull together behavioural systems (Chapter 3).222 Building on these ideas from Liotti,
Carlson, Yates, and Sroufe suggested that dissociation may be regarded as a strategy of
self-regulation in the face of threat, where such integration has been poor, and where the
caregiver cannot be approached directly to offer a safe haven and an external source of inte-
gration.223 Experiences of domestic violence and physical abuse that a child cannot escape
are liable to cause some dissociative symptoms, as a strategy developed for the purposes of
219 Ibid.
220 Ibid. p.868.
221 Ibid. p.855.
222 Liotti, G. (1992) Disorganized/ disoriented attachment in the etiology of the dissociative disorders.
Dissociation, 5, 196–204.
223 Carlson, E.A., Yates, T.M., & Sroufe, L.A. (2009) Development of dissociation and development of the
self. In P.F. Dell & J.A. O’Neil (eds) Dissociation and the Dissociative Disorders: DSM-V and Beyond (pp.39–52).
New York: Routledge: ‘The process of dissociation, which begins as a protective mechanism to promote the in-
tegrity of the self in the face of trauma, may directly threaten optimal functioning when employed routinely or
pervasively’ (47).
self-regulation. However, clinical levels of symptoms are more likely when the need for this
strategy occurs in infancy, since it may be used pervasively in the absence of other strategies
and because infancy may be a period of special importance for the formation of behavioural
systems and psychological coherence. This might account for the interaction between
trauma and disorganised attachment. Traumatic events are those that can be anticipated to
overwhelm established coping strategies. They may therefore be a catalyst for dissociative
symptoms for those children who were unable to lay a solid foundation for integration and
emotion regulation in infancy.224
More recently, Haltigan and Roisman reported findings from the NICHD Study of Early
Child Care and Youth Development of 1,149 families.225 Dissociative symptoms were as-
sessed by mothers at six time-points and by teachers at three time-points between age 4
and age 15. Participants themselves reported symptoms at age 15. The researchers found

no evidence that infants from dyads classified as disorganised in the Strange Situation
later showed dissociative symptoms. Infants from dyads classified as avoidant did have
elevated dissociative symptoms as rated by their teachers and mothers, though not in their
self-report. Haltigan and Roisman commented that ‘the claim that infant attachment dis-
organization represents a unique risk factor for the development of dissociative symp-
tomatology has been given much attention and cachet in the literature, perhaps because
of a theoretical account that provides an especially compelling rationale for this associ-
ation. That said, the bedrock of science is the replicability of its findings.’226 Given that the
NICHD study is a community-risk sample, and the Minnesota sample was selected specif-
ically on the basis of adversities, Haltigan and Roisman raised the prospect that infant dis-
organised attachment may contribute to later dissociative symptoms in interaction with
severe or chronic trauma. A link might be drawn here with Main’s prediction (Chapter 3)
that infant disorganised attachment will usually resolve by adulthood unless the basis of
the disorganisation lies directly in child maltreatment.227 Most researchers have, how-
ever, responded by ignoring the Haltigan and Roisman study, which remains relatively
comparatively cited four years after its publication. There has been informal discussion
of whether the NICHD sample was coded effectively for disorganisation, given that the
classification does not predict well in this cohort study. However, a further possibility, one
with quite serious implications for attachment research, is that the forms of disorganised
attachment may differ between high-risk and low-risk samples (Chapter 3). These ques-
tions will likely be resolved over the coming years as more recent cohort studies enter ado-
lescence and adulthood, if funds can be secured for later follow-up and measures are used
to assess dissociative symptoms.
224 Ibid. p.43. For aligned reflections by other attachment researchers see also Schuder, M. & Lyons-Ruth, K.
(2004) ‘Hidden trauma’ in infancy: attachment, fearful arousal, and early dysfunction of the stress response system.
In J. Osofsky (ed.) Young Children and Trauma (pp.69–104). New York: Guilford; Liotti, G. (2004) Trauma, dis-
sociation, and disorganized attachment: three strands of a single braid. Psychotherapy: Theory, Research, Practice,
Training, 41(4), 472.
225 Haltigan, J.D. & Roisman, G.I. (2015) Infant attachment insecurity and dissociative symptomatology: find-
ings from the NICHD Study of Early Child Care and Youth Development. Infant Mental Health Journal,
36(1), 30–41.
226 Ibid. p.8.
Goldberg, R. Muir, & J. Kerr (eds) Attachment Theory: Social, Developmental and Clinical Perspectives (pp.407–
470). Hillsdale, NJ: Analytic Press: ‘So long as direct maltreatment is not involved, many, perhaps most, are ex-
pected to have become “organised” by adulthood, being either secure, dismissing or preoccupied’ (454).
Developmental psychopathology 379
Developmental psychopathology
The domain of developmental psychopathology
The work of Sroufe and Egeland was fundamental to the emergence of developmental
psychopathology, as a movement within developmental psychology.228 In turn, develop-
mental psychopathology has supported the continuation of attachment theory as a signifi-
cant reference-point within contemporary developmental psychology. An important part
of the legacy of the Minnesota Longitudinal Study and of Sroufe and Egeland as attach-
ment researchers is therefore revealed through consideration of the rise of developmental

psychopathology.
Though based on earlier ideas,229 the movement was initiated primarily by a confer-
ence hosted by Norman Garmezy and Michael Rutter at the Center for Advanced Study
in Behavioral Sciences at Stanford from 1979 to 1980.230 The conference was sponsored
by the William T. Grant Foundation, as part of a growing public and policy concern about
child abuse and maltreatment, contributing to an increased availability of research funds in
this area.231 The meeting was prompted especially by findings from high-risk longitudinal
studies set up around specific diagnostic entities such as schizophrenia that nonetheless
highlighted the role of poverty, contextual risks and conflictual family relationships in the
prediction of a variety of mental health outcomes for individuals and across generations.232
The compounding of adversities seemed to be much more predictive than any ‘main effect’
from imputed single causes. The researchers’ concerns were also galvanised by interest in
the potential role of individual and social protective factors in buffering risks faced by indi-
viduals and families.233 Garmezy and Rutter drew on the term ‘resilience’ in conceptualising
228 Masten, A.S. (2006) Developmental psychopathology: pathways to the future. International Journal of
Behavioral Development, 30(1), 47–54; Cicchetti, D. (2013) An overview of developmental psychopathology. In P.

Zelazo (ed.) Oxford Handbook of Developmental Psychology (pp.455–80). Oxford: Oxford University Press.
229 The movement took its name from Achenbach’s 1974 book Developmental Psychopathology
(New York: Ronald Press). A former trainee of Norman Garmezy at Minnesota, in his book Achenbach argued
that classifications of adult mental disorder should not be imposed backward on childhood. Achenbach called for
a new paradigm thinking about mental health in a way commensurate with developmental processes. Cicchetti
reviewed other precursors in Cicchetti, D. (1990) A historical perspective on the discipline of developmental psy-
chopathology. In J. Rolf, A. Masten, D. Cicchetti, K. Nuechterlein, & S. Weintraub (eds) Risk and Protective Factors
in the Development of Psychopathology (pp.2–28). New York: Cambridge University Press.
230 There were a number of other relevant meetings at the time, such as the High Risk Consortium, San Juan,
Puerto Rico, 11–13 March 1980; and the 15th Annual Minnesota Symposium on Child Development, Minneapolis,
30 October–1 November 1980. See W.A. Collins (ed.) The Concept of Development: The Minnesota Symposia
on Child Psychology, Vol. 15. Hillsdale, NJ: Lawrence Erlbaum. In-person meetings of the movement would be
supported by subsequent annual Minnesota Symposia, and also the Rochester Symposia on Developmental
Psychopathology which were initiated in 1987 by Cicchetti.
231 Other key sources of funding that supported the emergence of developmental psychopathology included
the National Center on Child Abuse and Neglect, which was founded in 1974, and the National Institute of Mental
Health in the 1980s.
232 Garmezy, N. (1971) Vulnerability research and the issue of primary prevention. American Journal of
Orthopsychiatry, 41, 101–116; Garmezy, N. (1974) Children at risk: the search for the antecedents of schizophrenia.
Part I. Conceptual models and research methods. Schizophrenia Bulletin, 1(8), 14–90; Rutter, M. & Quinton, D.
(1984) Parental psychiatric disorder: effects on children. Psychological Medicine, 14(4), 853–80.
233 Rutter, M. (1979) Protective factors in children’s responses to stress and disadvantage. In M.W. Kent & J.E.
Roif (eds) Primary Prevention of Psychopathology, Vol. 3: Social Competence in Children. Hanover: University Press
of New England.
this process.234 The proceedings of the Stanford conference were published by Garmezy and
Rutter as Stress, Coping, and Development in Children in 1983.235
Also in 1983, Sameroff proposed the dual-risk model of psychopathology, in which the
particular role of multicausal models of cumulative risk factors was highlighted in the role of
mental illness.236 Emergent longitudinal findings appeared to show that any one risk factor
had little association with later disturbance. What seemed to be critical was the combin-
ation of liabilities and supports. Additionally, in his 1983 chapter, Sameroff began to sketch
the ideas that would later form his distinction between ‘protective’ supports and ‘promo-
tive’ support for development, the former contributing to beneficial stability and the latter
contributing to thriving.237 Overall, this account firmly emphasised the contextual embed-
dedness of mental health, and the transactional relationship between different obstacles and
difficulties individuals face in their lives. As mentioned above, a presentation by Sameroff of

his emerging ideas prior to this publication proven influential in the choice of initial meas-
ures in the Minnesota Longitudinal Study of Risk and Adaptation.
Garmezy and Sroufe were colleagues and friends, and taught classes together at
Minnesota, where there had been growing momentum around developmental approaches
to mental health since the 1970s.238 Sroufe and Rutter’s 1984 article ‘The domain of devel-
opmental psychopathology’ was a turning-point for developmental psychopathology, as the
movement resolved into an explicit and self-identified alternative paradigm for the study
of mental health. Whereas the dominant medical model started from adult diagnostic
234 This term was introduced to psychology from ecology by Jack Block in the 1950s to refer to a quality of
individual children that allows some to flexibly return to adaptive functioning after a disturbance; ‘resilience’
was then transferred to the wider literature on adversity and development by Emmy Werner in the early 1970s.
Block J. & Thomas, H. (1955) Is satisfaction with self a measure of adjustment? Journal of Abnormal and Social
Psychology, 51(2), 254–9; Werner, E.E., Bierman, J.M., & French, F.E. (1971) The Children of Kauai Honolulu.
Hawaii: University of Hawaii Press; Werner, E. & Smith, R. (1988) Vulnerable but Invincible: A Longitudinal Study
of Resilient Children and Youth. New York, Adams, Bannister & Cox; Masten, A.S. & Garmezy, N. (1985) Risk,
vulnerability, and protective factors in the developmental psychopathology. In B.B. Lahey & A.E. Kazdin (eds)
Advances in Clinical Child Psychology, Vol. 8 (pp.1–51). New York: Plenum.
235 Garmezy, N. & Rutter, M. (eds) (1983) Stress, Coping, and Development in Children. Baltimore: Johns Hopkins
University Press. The influence of Garmezy on the emergence of developmental psychopathology in the 1980s is
addressed in a festschrift from 1990: Rolf, J., Masten, A.S., Cicchetti, D., Nuechterlein, K., & Weintraub, S. (eds)
(1990) Risk and Protective Factors in the Development of Psychopathology. Cambridge: Cambridge University Press.
Meanwhile, another important development was the Social Science Research Council’s Committee on Social and
Affective Development during Childhood, which held a symposium in 1982. Members included Michael Rutter,
Carroll Izard, and Carol Dweck and the committee was chaired by Martin Hoffman. A further development in
1983 was the first meeting of the MacArthur Working Group on the study of Attachment in the Transition Period,
which was concerned with the application of attachment theory and methods beyond infancy (Chapter 3). The
Network met twice yearly between 1983 and 1987. Members of the network included Mary Ainsworth and several
of her former students including Inge Bretherton, Jude Cassidy, Mark Cummings, Mark Greenberg, Mary Main,
and Robert Marvin. The group also included other researchers who would play an important role in the subse-
quent development of attachment theory including Jay Belsky, Dante Cicchetti, and Susan Spieker.
236 Sameroff, A.J. (1983) Developmental systems: contexts and evolution. In P. Mussen (ed.) Handbook of Child
Psychology, Vol. 1 (pp.237–94). New York: Wiley.

237 The concept of ‘promotive factors’ to describe processes that enhance the likelihood of adaptation (in contrast
to protective factors, which buffer the likelihood that adversities will have an effect) was introduced by Sameroff,
A.J. (1999) Ecological perspectives on developmental risk. In J.D. Osofsky & H.E. Fitzgerald (eds) WAIMH
Handbook of Infant Mental Health: Vol. 4. Infant Mental Health Groups at Risk (pp.223–48). New York: Wiley. The
terms ‘protective-stabilising’ and ‘protective-enhancing’ were also introduced by Luthar in making this distinc-
tion: Luthar, S.S. (1993) Annotation: methodological and conceptual issues in the study of resilience. Journal of
Child Psychology and Psychiatry, 34, 441–53.
238 Garmezy later attributed the influence of both Sroufe and Rutter to his reorientation away from research on
adult schizophrenia and towards the study of child and adolescent development. Masten, A.S. (1995) Interview
with Norman Garmezy. SRCD oral history interview. http://srcd.org/sites/default/files/documents/garmezy_
norman_interviewweb.pdf.
categories and explored precursors, developmental psychopathology was defined by a con-

cern to study ‘the origins and time course of a given disorder, its varying manifestation with
development, its precursors and sequelae, and its relation to nondisordered patterns of be-
haviour. Thus, developmental psychopathologists may be just as interested in a group of chil-
dren showing precursors of a disordered behavior pattern, but not developing the disorder
proper, as the group that in time manifested the complete pathology.’239
Sroufe and Rutter highlighted that the distinction between adaptation and pathology
was context-dependent. They give the example of an avoidant response to an abusive at-
tachment figure, which might include not only physical avoidance but also ‘blunting or con-
trolling emotional experiences’. This response is ‘adaptive’ in the sense of being intelligible
and helpful for a young child, for whom proximity or displays of distress or desire for com-
fort may not be safe. However, the very same adaptation ‘may later compromise the child’s

ability to maximally draw upon the environment’.240 Even stereotypic behaviour (Chapter 3)
could be seen in this light: these odd, seemingly purposeless behaviours can best be re-
garded as ‘meaningful’ responses that at some point helped modulate arousal, and which
have therefore been retained as a response to stress.241 Sroufe and Rutter argued that adult
mental health should be understood as, in part, the residue of the history of ways in which
an individual responded to earlier challenges, and in part a new, inventive attempt to turn
existing expectations, skills, and strategies towards contemporary problems and opportun-
ities.242 Where continuities are seen across periods of development these will rarely be in the
display of identical forms of behaviour, but rather in the meaning of the behaviours as a way
of engaging with the environment. For instance, there are vast differences between weeping
and clinging in infancy, on the one hand, and impulsivity, jealousy, and continual dramas
in adulthood, on the other. But both profiles may serve similar interpersonal functions and
have their basis in a low threshold of activation of the attachment system.
As such, Sroufe and Rutter made a plea for broad-band indices of adaptation, whether
in peer-relations or mental health. In line with Sroufe and Waters’ earlier ‘Attachment as an
organizational construct’ (Chapter 2), Sroufe and Rutter argued that little stability of behav-
iour would be expected in the context of an individual’s changing developmental challenges
and environment. By contrast, broader indices would predict well to later adaptation in gen-
eral, since they would capture the extent to which adaptation had required significant com-
promise or even led to failures in meeting challenges in earlier life. In this, their argument lay
the ground for proposals by Rutter’s former collaborator Avshalom Caspi and colleagues in
2014 for a general ‘p-factor’ representing general mental ill-health, rather than the division
of symptoms into specific diagnoses.243
239 Sroufe, L.A. & Rutter, M. (1984) The domain of developmental psychopathology. Child Development, 55(1),
17–29, p.18.
240 Ibid. p.23.
241 Sroufe, L.A. (1990) Considering normal and abnormal together: the essence of developmental psychopath-
ology. Development & Psychopathology, 2(4), 335–47: ‘Stereotypy increased in frequency and intensity during
demanding tasks, was preceded by heart rate acceleration, and was followed by a pronounced heart rate deceler-
ation . . . We suggested that both stereotypies and behavioural negativism were serving the purpose of modulating
arousal’ (337).
17–29, p.21.
243 Caspi, A., Houts, R.M., Belsky, D.W., et al. (2014) The p factor: one general psychopathology factor in the
structure of psychiatric disorders? Clinical Psychological Science, 2(2), 119–37. The debt to developmental psycho-
pathology is clear in this paper: Rutter’s work is cited as an ‘initial motivation’, and ‘developmental psychopath-
ology’ is one of the article’s three keywords.
Though Sroufe tended to regard positive internal working models of self and relation-
ships as characteristic of mental health,244 in his writings on developmental psychopath-
ology he was keen to argue that there is no content of internal working models that is itself
adaptive. Positive models are primarily beneficial when combined with environments in
which expressions of positive models such as self-esteem or a capacity to trust can generate
opportunities and rewards, which is not always the case. Against essentialism about either
adaptation or pathology, in their 1984 article Sroufe and Rutter argued that it is the inter-
action with context that is critical for the meaning of even patterns of attachment.245 For
example, though generally a protective factor over time in the wider context of development,
secure attachment may in theory function variously as a vulnerability, protective or a risk
factor, depending on local circumstances.246 Indeed, attachment was a central concern for
developmental psychopathology, since it provided both theory and empirical results sup-

porting the role of transdiagnostic, developmental processes for mental health, with any
factor compounded by the network of other risk and protective factors experienced by a
group or population.247
Though an outspoken critic of Bowlby’s notion of ‘maternal deprivation’, Rutter none-
theless praised Bowlby’s work as having ‘ushered in the era’ in which the developmental
psychopathology perspective was possible, especially through his adoption of the concept
of ‘developmental pathways’ from embryology (Chapter 1).248 Rutter also observed that
‘Bowlby’s (1951) World Health Organization monograph had raised expectations of the
power of prediction from experiences in infancy and on the preventive value of psycho-
social interventions, but also it had run into a storm of academic criticism. Both proved
only half-right. The over-riding potency of infantile experiences has not been supported by
subsequent research, but the focus on children’s attachments to their parents has paid off
richly.’249 Sroufe, perhaps, if forced to give proportions, would likely have said that Bowlby
was three-quarters right.250 Even if the direct effects of early attachment have been smaller
hood from early parent–child relationships. In R. Parke & G. Ladd (eds) Family-Peer Relationships: Modes of
245 Sroufe, L. & Rutter, M. (1984) The domain of developmental psychopathology. Child Development, 55(1),
17–29, p.23.
246 Egeland, B., Carlson, E., & Sroufe, L.A. (1993) Resilience as process. Development & Psychopathology, 5(4),
517–28: ‘Any constitutional or environmental factors may serve as vulnerability, protective, or risk variables’ (517).
247 Rutter, M. (1983) Stress, coping and development: some issues and some questions. In N. Garmezy &
M. Rutter (eds) Stress, Coping, and Development in Children (pp.1–42). Baltimore: Johns Hopkins University
Press, p.23.
248 Rutter, M. (1985) Resilience in the face of adversity: protective factors and resistance to psychiatric disorder.
British Journal of Psychiatry, 147(6), 598–611, p.598.

249 Rutter, M. (1986) Child psychiatry: looking 30 years ahead. Journal of Child Psychology and Psychiatry, 27(6),
803–840, p.803. See also Rutter, M.L. (1999) Psychosocial adversity and child psychopathology. British Journal of
Psychiatry, 174(6), 480–93: ‘Bowlby’s 1951 monograph, although overstating the case in some respects, led to a
major change in the approach to experiences during the preschool years’ (482).
250 Sroufe, L.A. (1986) Appraisal: Bowlby’s contribution to psychoanalytic theory and developmental psych-
ology; attachment: separation: loss. Journal of Child Psychology and Psychiatry, 27(6), 841–9. Regarding the debate
between Bowlby and Rutter over what priority to assign to early experience, the Minnesota group argued that no
global position was acceptable, or in line with evidence from embryology. Sroufe, L.A., Coffino, B., & Carlson, E.A.
(2010) Conceptualizing the role of early experience: lessons from the Minnesota longitudinal study. Developmental
Review, 30(1), 36–51: ‘It is clearly not always the case that early experience is most critical . . . Maternal nutrition
during pregnancy is a dramatic contrary example. While certain nutrients like folic acid are critically important
in the early embryonic period, poor general maternal nutrition has almost no demonstrable effects in the first tri-
mester, because the tiny developing organism can simply take many of the nutrients it needs from maternal stores.
However, in the third trimester, when rapid fetal size and weight gain is occurring, adequate maternal nutrition is
crucial’ (37–8).
than anticipated by Bowlby, the Minnesota Longitudinal Study of Risk and Adaptation pro-
vided ample evidence of the continuing importance of early experiences in some areas of
life, such as enduring associations between early experiences of abuse and later relation-
ships with peers and romantic partners.251 Looking back on the results of the study, Sroufe,
Coffino, and Carlson concluded that ‘early experience can be conceptualized in terms of
creating vulnerabilities or strengths with regard to later experience, including what experi-
ences are sought and how they are interpreted, rather than as directly producing particular
outcomes’.252
This emphasis on the power of prediction, for which Bowlby was a forerunner, was central
to developmental psychopathology as a domain of inquiry. It was assumed that reasonable
certainties were possible regarding the probabilities entailed by developmental pathways.
In particular, hypotheses could be developed for testing against prospective data, to explore

the ‘characteristics or histories that buffer individuals against stress or that provide them
with attitudes, orientations, and skills that promote successful coping with stress, and . . . fac-
tors that produce vulnerability to stress or coping failures’.253 In wider social and political
context, the emergence of developmental psychopathology can be regarded as aligned with
a rise of public health and primary prevention approaches to mental health.254 It was also
supported by a broader trend in social and policy science towards a concern with the cap-
acities of individuals to handle the potential fallout of societal and economic uncertainties
and problems.255 In turn, the prominence and credibility given to ‘resilience’ by the emergent
field of developmental psychopathology helped contribute to its resonance and deployment
across contexts in the 1990s—albeit generally completely shorn of developmental concerns,
and formulated instead as a quality of individual ‘invulnerability’.256 In their contribution to
Anthony and Cohler’s influential book on The Invulnerable Child, Farer and Egeland casti-
gated the editors and other authors for what they regarded as a false and unethical framing of
the question of adaptation in the context of adversity.257
Though considered one of the ‘founders’ of developmental psychopathology, Sroufe was
also regarded as a distinct voice. He was known for his characteristic criticisms of medical
models of diagnosis; his emphasis on lifespan development and intergenerational processes;
New York: Guilford: ‘We may think of early experience as the foundation for the building. The foundation cannot
be more important than solid supporting beams or a sturdy roof; without these, the house will not last. But at the
same time, a house cannot be stronger than its foundations’ (11). For a recent and comprehensive report see Raby,
K. L., Roisman, G.I., Labella, M.H., Martin, J., Fraley, R.C., & Simpson, J.A. (2019) The legacy of early abuse and
neglect for social and academic competence from childhood to adulthood. Child Development, 90(5), 1684–701.

17–29, p.19.
254 Spaulding, J. & Balch, P. (1983) A brief history of primary prevention in the twentieth century: 1908
to 1980. American Journal of Community Psychology, 11, 59–80; Cowen, E. L. (1983) Primary prevention in
mental health: past present and future. In R.D. Felner, L.A. Jason, J.N. Montsugu, & S.S. Farber (eds) Prevention
Psychology: Theory, Research and Practice (pp.290–97). New York: Pergammon Press.
255 E.g. Rose, N. & Lentzos, F. (2017) Making us resilient: responsible citizens for uncertain times. In S. Trnka
& C. Trundle (eds) Competing Responsibilities: The Ethics and Politics of Contemporary Life (pp.27–48). Durham,
NC: Duke University Press.
256 Bourbeau, P. (2018) A genealogy of resilience. International Political Sociology, 12(1), 19–35. See e.g. Howard,
D.E. (1996) Searching for resilience among African-American youth exposed to community violence. Journal of
Adolescent Health, 18(4), 254–62.
257 Farber, E. & Egeland, B. (1987) Abused children: can they be invulnerable? In J. Anthony & B. Cohler (eds)
The Invulnerable Child. New York: Guilford, 253–88, p.286.

his concern with the interdependence of different family relationships; his hostility to tem-
perament researchers where they neglected the role of caregiving factors; and his dismay at
images of ‘resilience’ that constructed it as an individual quality rather than a social effect.258
He and Egeland were also among the strongest advocates for the value of considering typical
and atypical forms of development within the same frame, based on the assumption that the
processes involved would be mutually revealing. Sroufe and Egeland argued that the role of
many social factors in ordinary life is best revealed when they are absent or intensified to
an unusual degree. They pointed, for instance, to the importance of the family and friend-
ship networks available to the mothers in their sample: when there were improvements of
social support to mothers who had otherwise been especially isolated, their children’s mental
health and conduct in school saw dramatic improvements. However, in the sample in gen-
eral, social support had only a modest overall relationship with child behaviour.259

Furthermore, the Minnesota Longitudinal Study of Risk and Adaptation led Sroufe,
Egeland, and colleagues to a distinctive position on the role of risk factors across the lifespan.
They agreed with the consensus in developmental psychopathology that risk factors were
cumulative. However, they were unusual in offering some qualifications to this position.
Kaufman and Zigler had argued that various risks were interchangeable, and what mattered
was their number: for instance, various forms of child abuse were argued to have the same
implications.260 The Minnesota group were definitely interested in the use of summary vari-
ables and the study of cumulative effects: they generally found stronger effects when mul-
tiple forms of abuse compounded. However, they held that risk factors would often interact
in specific ways. Egeland reported that two-thirds of participants who had been physically
abused showed oppositional defiant disorder. This was in line with other research groups
who had found that physical abuse predicted clinically significant levels of conduct prob-
lems. However, the multi-assessment and longitudinal nature of the Minnesota data allowed
Egeland to show that this association was, at least in his sample, a secondary effect of the links
between physical abuse and other forms of maltreatment and inadequate care.261 Conduct
problems were in fact best predicted by the co-occurrence of caregiver psychological un-
availability, neglect, and sexual abuse. The Minnesota group also reported findings showing
the implications of specific forms of abuse. Berzenski, Yates, and Egeland reported that emo-
tional abuse, whether alone or in combination with other forms of abuse, made a particu-
larly powerful contribution to anxiety and depression, while the combination of physical
and emotional abuse was especially strongly associated with conduct problems.262 They also
258 These positions can already be seen in comparing Sroufe’s chapter to others in Ciccheti, D. (1989) (ed.)
The Emergence of a Discipline: Rochester Symposium on Developmental Psychopathology, Vol. 1. Hillsdale,

NJ: Lawrence Erlbaum. Some of these differences would also later be remarked upon by Michael Rutter,
looking back on the period: Rutter, M. (2008) Developing concepts in developmental psychopathology. In J.J.
Hudziak (ed.) Developmental Psychopathology and Wellness: Genetic and Environmental Influences (pp.3–22).
New York: American Psychiatric Publications.
study. In T. Wachs & R. Plomin (eds) Conceptualization and Measurement of Organism-Environment Interaction
260 Kaufman, J. & Zigler, E. (1989) The intergenerational transmission of child abuse. In D. Cicchetti & V.
Carlson (eds) Child Maltreatment: Theory and Research on the Causes and Consequences of Child Abuse and Neglect
cence. In D. Cicchetti & S. Toth (eds) Rochester Symposium on Developmental Psychopathology: Vol. VIII. The
Effects of Trauma on the Developmental Process (pp.403–434). Rochester, NY: University of Rochester Press.
262 Berzenski, S.R., Yates, T.M., & Egeland, B. (2014) A multidimensional view of continuity in intergenerational
transmission of child maltreatment. In J.E. Korbin & R.D. Krugman (eds) Handbook of Child Maltreatment, Vol.
highlighted research findings from other groups that indicated that parental depression re-
duces the likelihood of physical abuse, but increases the likelihood of sexual abuse.263
The Minnesota group were distinctive among developmental psychopathologists for
their interest in the capacity of risk factors at one stage to go dormant until the develop-
mental challenges of a later stage made them salient again: ‘Even following change, early
patterns of attachment retain a potential for reactivation . . . Certain issues and certain
arenas of functioning—those tapping anxiety about the availability of others or apprehen-
sion regarding emotional closeness—will be especially likely to reveal the legacy of early
attachment, even during periods of generally adequate functioning.’264 For instance, Yates,
Dodds, Sroufe, and Egeland found that exposure to interpartner violence in the preschool
years reappeared as a powerful predictive factor in contributing to behavioural problems
at age 16, at a time when the key developmental challenges are forming intimate relation-

ships, handling commitment and its conflicts, establishing boundaries with peers, and co-
ordinating same-and cross-gender relationships.265 The Minnesota group later reported
that exposure to interpartner violence in the preschool years predicted later partner vio-
lence perpetration and victimisation for both male and female participants. Participants
who initiated partner violence perpetration and/or victimisation between age 26 and age
32 had greater exposure to partner violence in the preschool years than participants who
showed no or declining domestic violence over these years.266 Collins, LaFreniere, and
Simpson argued that such findings confirm that ‘one must understand the trajectory of
an individual’s relationship history to fully appreciate, situate, and comprehend his or her
2 (pp.115–29). New York: Springer. Another example from the Minnesota group was work by Jeff Simpson and
colleagues examining the implications of early environmental unpredictability—operationalised as changes in
residence, cohabitatio, and parental occupation that occur before a child is five years old. These early experiences
have been found to predict later aggressive behaviour, more substance use at age 16, and more sexually risky be-
haviour at age 23. This is in contrast to changes in residence, cohabitation, and parental occupation after the age
of five, which have barely any effect. Furthermore, the combination of these factors and harsh maternal caregiving
additionally predicted adolescent risk-taking behaviours. See Simpson, J.A., Griskevicius, V., & Kuo, S.I. (2012)
Evolution, stress, and sensitive periods: the influence of unpredictability in early versus late childhood on sex and
risky behavior. Developmental Psychology, 48(3), 674–86; Doom, J.R., Vanzomeren-Dohm, A.A., & Simpson, J.A.
(2016) Early unpredictability predicts increased adolescent externalizing behaviors and substance use: a life his-
tory perspective. Development & Psychopathology, 28(4), 1505–16.
transmission of child maltreatment. In J.E. Korbin & R.D. Krugman (eds) Handbook of Child Maltreatment, Vol.
2 (pp.115–29). Dordrecht: Springer. For parental depression reducing the likelihood of physical abuse see Pears,
K.C. & Capaldi, D.M. (2001) Intergenerational transmission of abuse: a two-generational prospective study of
an at-risk sample. Child Abuse & Neglect, 25(11), 1439–61. For parental depression increasing the likelihood of
sexual abuse see Leifer, M., Kilbane, T., & Kalick, S. (2004) Vulnerability or resilience to intergenerational sexual
abuse: the role of maternal factors. Child Maltreatment, 9(1), 78–91.
mental psychopathology. Development & Psychopathology, 11(1), 1–13, p.6; see also Sroufe, L. & Rutter, M. (1984)
The domain of developmental psychopathology. Child Development, 55(1), 17–29, p.20.
265 Yates, T.M., Dodds, M.F., Sroufe, L.A., & Egeland, B. (2003) Exposure to partner violence and child behavior
problems: a prospective study controlling for child physical abuse and neglect, child cognitive ability, socio-
economic status, and life stress. Development & Psychopathology, 15(1), 199–218, p.209; Sroufe, L.A., Egeland, B.,
Carlson, E.A., & Collins, W.A. (2005) The Development of the Person. New York: Guilford, Table 4.2.
266 Narayan, A.J., Labella, M.H., Englund, M.M., Carlson, E.A., & Egeland, B. (2017) The legacy of early child-
hood violence exposure to adulthood intimate partner violence: variable-and person-oriented evidence. Journal
of Family Psychology, 31(7), 833–43. On the direct and indirect pathways between exposure to domestic violence
in childhood and perpetration or victimisation in adulthood see also Narayan, A.J., Englund, M.M., Carlson,
E.A., & Egeland, B. (2014) Adolescent conflict as a developmental process in the prospective pathway from ex-
posure to interparental violence to dating violence. Journal of Abnormal Child Psychology, 42(2), 239–50; Zamir,
O., Szepsenwol, O., Englund, M.M., & Simpson, J.A. (2018) The role of dissociation in revictimization across the
lifespan: a 32-year prospective study. Child Abuse & Neglect, 79, 144–53.
adult relationships. This view of how development and relationships continually intersect
across the lifespan is one of the major and lasting legacies of Sroufe and Egeland’s work on
the entire field of psychology.’267
Another finding illustrates the ongoing role of early experience as a ‘resource’ to later
adaptation.268 A clinical interview was conducted when participants in the Minnesota study
were aged 28, assessing ‘global adaptive functioning’. This scale was based on the clinician’s
impression of the individual’s overall psychological, social, and occupational functioning,
and scored on a continuous scale. This measure is not just an assessment of mental illness
or difficulties, then, but also positive adaptation and thriving. Infant Strange Situation at-
tachment classifications had a very substantial association with this measure of adaptation
and thriving several decades later (r = .41). Path analysis revealed that this association was
partly direct (r =.26), even taking into account a large variety of mediators. Infant attach-

ment appeared to remain, even after so many years, directly entangled in the coping and
thriving of the Minnesota study participants at the most general level. The remainder of the
association between Strange Situation classification and global adaptive functioning was
mediated through a complex path including peer relationships at age 16, effectiveness in
handling issues in romantic relationships in early adulthood, and total experience of life
stress. Englund and colleagues within the Minnesota group characterised infant attachment
as a species of ‘social capital’, and cumulative with other forms of social capital such as sup-
portive friendships and romantic relationships in contributing to psychological, social, and
occupational functioning.
Developmental pathways
The journal Development and Psychopathology was established in 1989 by Dante Cicchetti.
The initial idea for the journal had come to Cicchetti during his years teaching seminars on
developmental psychopathology for Sroufe’s classes, and then coteaching with Sroufe in his
final year of graduate school.269 Sroufe’s articles for Development and Psychopathology in the
1990s are magisterial and programmatic. In particular, they set out his particular problems
with medical models of diagnosis. The ‘essence of developmental psychopathology’, Sroufe
argued, was attention to developmental processes that span both mental health and illness.270
He agreed with Egeland that developmental psychopathology needed to have an operational
definition of targets of inquiry and intervention, like child abuse or conduct disorders.271
A collectively agreed definition would permit convergent validity in measurement, stability
in theory-building, and clarity of message in communicating with policy-makers. However,
267 Collins, W.A., LaFreniere, P., & Simpson, J.A. (2011) Relationships across the lifespan: the benefits of a the-
oretically based longitudinal-developmental perspective. In D. Cicchetti & G.I. Roisman (eds) The Origins and
Organization of Adaptation and Maladaption: The Minnesota Symposium on Child Psychology, Vol. 36 (pp.155–83).
New York: Wiley, p.177.
268 Englund, M.M., Kuo, S.I.C., Puig, J., & Collins, W.A. (2011) Early roots of adult competence: the significance
of close relationships from infancy to early adulthood. International Journal of Behavioral Development, 35(6),
490–96.
269 Cicchetti, D. (2013) The legacy of development and psychopathology. Development & Psychopathology,
25(4.2), 1199–200, p.1200.

270 Sroufe, L.A. (1990) Considering normal and abnormal together: the essence of developmental psychopath-
ology. Development & Psychopathology, 2(4), 335–47, p.335.

271 Egeland, B. (1991) From data to definition. Development & Psychopathology, 3(1), 37–43.
Sroufe argued that too much research to date had assumed that medical diagnoses could
serve to define these targets of interest:
The disease model requires syndromic integrity. If the disease model is apt for children’s
behavioral and emotional problems, children generally should manifest tight clusters of
symptoms, with unique indicators of other syndromes being absent. But in reality chil-
dren commonly manifest problems that cut across established categories. To be sure, one
disorder may potentiate another in medicine as well, but not nearly to the extent implied
by the prevalence of co-morbidity of childhood disturbances . . . Comorbidity is the rule,
not the exception. Moreover, broad classes of problems such as externalizing behaviors are
predictive of a myriad of later conditions, including depression.272

Sroufe, Carlson, Levy, and Egeland argued that these concerns should be approached
through the lens of Bowlby’s metaphor of developmental pathways, which implies a series of
specific hypotheses:
(a) At any age, current quality of care will add to early attachment history in predicting
pathology, given that adaptation is always the joint product of current circumstances and
early history. (b) Likewise, broader aspects of current contexts, including relationships
outside of the family and stresses and challenges of the period, also will increase prediction
beyond early attachment; (c) a cumulative history of maladaptation will be more patho-
genic than a single early period of poor functioning with pathology ever more likely the
longer a maladaptive pathway has been followed, and (d) change itself will be predictable
in light of changes in stress and/or support.273
This approach led Sroufe and colleagues to highlighted three ways that a developmental psy-
chopathology perspective was a necessary counterweight to the disease model that domin-
ates medicine.
1. Sroufe and colleagues argued that a diagnosis-focused science would end up miss-
ing broad causal factors that cross-cut and precede mental health and illness.
Though ignored by a focus on diagnostic categories, attention to subclinical symp-
toms is critically important, both for understanding how clinical conditions de-
velop and for understanding how mental health can be maintained. In this regard,
Sroufe and colleagues claimed that family relationships are not a risk factor like
any other, but play a fundamental role in shaping patterns of emotion regulation.
This means that family relationships are especially important for the initiation and
prolongation of most mental health symptoms.274 For instance, Carlson, Jacobvitz,
and Sroufe found that relationship stability and social support for the mothers in
251–68, p.257. See also Rutter, M. & Sroufe, L.A. (2000) Developmental psychopathology: concepts and chal-
lenges. Development & Psychopathology, 12(3), 265–96.
mental psychopathology. Development & Psychopathology, 11(1), 1–13, p.3.

274 Sroufe, L.A., Duggal, S., Weinfield, N.S., & Carlson, E. (2000) Relationships, development, and psycho-
pathology. In M. Lewis & A. Sameroff (eds) Handbook of Developmental Psychopathology, 2nd edn (pp.75–92).
New York: Kluwer Academic/Plenum Press, p.84.
their sample emerged as a powerful predictor of ADHD symptoms in their child

participants.275
2. Second, they highlighted that the same diagnosis could reflect various diverse devel-
opmental pathways, with varying implications for treatment. ADHD was a clear ex-
ample: Sroufe and colleagues felt that problems around attention and the integration
of affect, cognition, and behaviour were best regarded as common signals of a devel-
opmental pathway rocked by adversities, rather than primarily a diagnostic entity in
themselves.276 Supporting the idea that the same diagnosis could reflect different path-
ways, Aguilar, Sroufe, Egeland, and Carlson found two quite distinct routes to con-
duct problems and aggression in adolescence: one based on factors beginning in early
childhood and leading to early-onset and persistent symptoms, another beginning in
adolescence and associated with concurrent life stress and internalising symptoms.277

Sroufe and colleagues expressed concern that a diagnosis called ‘attachment disorder’,
applicable only to a tiny proportion of children who had not formed a stable selective
attachment, would magnetise attention away from the broader relevance of patterns of
attachment to developmental pathways.278
3. Third, Sroufe and colleagues suggested that developmental pathways could result in
various diagnoses. A pathway beginning from an avoidant attachment relationship in
infancy could help predispose various forms of mental illness, depending on what then
ensued in that child’s life and what other coping tactics were elaborated.279 Avoidant
attachment suggests that key matters of concern cannot be brought to the attention
of the caregiver, that the child must suppress or redirect distress, and that they may
receive less support from caregivers in handling difficult feelings when they do un-
avoidably arise. Whilst not a clinical pathology in themselves, these aspects of avoid-
ant attachment might contribute to the initiation of mental health symptoms and/or
their prolongation. Clinical interventions and support focused on particular diagnosis
could well, Sroufe and colleagues argued, miss the underlying developmental trajec-
tory and hence the causal processes in play. This would reduce the effectiveness of
the clinical work.280 In an important paper from 2005, Appleyard, Egeland, Dulmen,
and Sroufe examined the impact of six risk factors in childhood (child maltreatment,
interparental violence, family disruption, low socioeconomic status, and high parental
stress) on symptoms relating to conduct problems and aggression in adolescence.
The study showed that the number of risks in early childhood predicted the extent
275 Carlson, E.A., Jacobvitz, D., & Sroufe, L.A. (1995) A developmental investigation of inattentiveness and
hyperactivity. Child Development, 66(1), 37–54, p.52.

276 E.g. Sroufe, L.A. (2007) The place of development in developmental psychopathology. In A. Masten (ed.)
277 Aguilar, B., Sroufe, L.A., Egeland, B., & Carlson, E. (2000) Distinguishing the early-onset/persistent and
adolescence-onset antisocial behavior types: from birth to 16 years. Development & Psychopathology, 12(2),
109–132.
278 Rutter, M. & Sroufe, L.A. (2000) Developmental psychopathology: concepts and challenges. Development
& Psychopathology, 12(3), 265–96, pp.273–4; Sroufe, L.A., Duggal, S., Weinfield, N., & Carlson, E. (2000)
Relationships, development, and psychopathology. In A.J. Sameroff, M. Lewis, & S.M. Miller (eds) Handbook of
Developmental Psychopathology (pp.75–91). New York: Kluwer Academic/Plenum Publishers, p.83.
279 Cicchetti, D. & Sroufe, L.A. (2000) Editorial: the past as prologue to the future: the times, they’ve been
a’changin’. Development & Psychopathology, 12, 255–64.

280 See also Sroufe, L.A. & Jacobvitz, D. (1989) Diverging pathways, developmental transformations, multiple
etiologies and the problem of continuity in development. Human Development, 32(3–4), 196–203.
of symptoms in adolescence in a linear fashion. The authors therefore concluded that

‘the experience of an additional risk factor does not increase the odds of problems
in a multiplicative way. For interventionists, such information might imply that there
does not appear to be a “point of no return” beyond which services for children are
hopeless, and that every risk factor we can reduce matters.’ The implication was that
‘interventions should be designed as comprehensive programs that enhance as many
aspects of family life as possible’.281
The approach to developmental psychopathology of Sroufe, Egeland, and colleagues is well il-
lustrated by their study of dissociative symptoms, discussed earlier. It was characteristic that the
starting point of the Minnesota group was inquiry into dissociative processes in a population-
based perspective. Rather than contrasting a group of patients with a diagnosis of a dissociative

condition with healthy controls, Sroufe and colleagues explored the factors that can contribute
to clinically significant and subclinical levels of dissociative symptoms. They leveraged the col-
lection of data at multiple time-points, from multiple informants, to provide a rich picture of de-
velopmental pathways. The factors explored ranged from what may be individual-level factors
such as infant distractibility, to relational factors such as infant attachment or present-day phys-
ical abuse, to the mother’s own childhood experiences—which her child might not even know.
The researchers found that relational factors, more than individual factors, were important for
predicting dissociative symptoms, which supported the researchers’ broader conceptualisation
of risk and resilience.
The report on the antecedents of dissociative symptoms, however, also illustrates a limi-
tation of the approach adopted by Sroufe and colleagues. Their analysis was solely within-
sample. Without justifying the exclusion, they did not include structural factors such as
socioeconomic disadvantage as a possible predictor within their analyses of dissociation.
Socioeconomic disadvantage was characteristic of their whole sample as a consequence of
the original recruitment strategy, which reduced relevant variance. Nonetheless, Sroufe and
colleagues included socioeconomic disadvantage as an independent variable in other ana-
lyses, such as the sequelae of abuse and antecedents of ADHD, and found that it contributed
to prediction.282 It was not as if the Minnesota group were uninterested in socioeconomic
disadvantage, even if they regarded it as a distal risk factor and always mediated by prox-
imal factors such as family behaviours and social support.283 As Egeland observed, ‘even in
281 Appleyard, K., Egeland, B., Dulmen, M.H., & Sroufe, L.A. (2005) When more is not better: the role of cu-
mulative risk in child behavior outcomes. Journal of Child Psychology and Psychiatry, 46(3), 235–45, p.242–3. This
stance would be contested by researchers such as Bakermans-Kranenburg, as we shall see.
282 Carlson, E.A., Jacobvitz, D., & Sroufe, L.A. (1995) A developmental investigation of inattentiveness and
hyperactivity. Child Development, 66, 37–54; Egeland, B. (1997) Mediators of the effects of child maltreatment on
developmental adaptation in adolescence. In D. Cicchetti & S. Toth (eds) Rochester Symposium on Developmental
Psychopathology: Vol. 8. The Effects of Trauma on the Developmental Process (pp.403– 434). Rochester,
NY: University of Rochester Press. Similarly socioeconomic status has been used by other researchers conducting
analyses of the dataset, e.g. Raby, K.L., Roisman, G.I., Fraley, R.C., & Simpson, J.A. (2015) The enduring pre-
dictive significance of early maternal sensitivity: social and academic competence through age 32 years. Child
Development, 86(3), 695–708.
283 Yates, T.M., Egeland, B., & Sroufe, A. (2003) Rethinking resilience. In S. Luthar (ed.) Resilience and
Vulnerability: Adaptation in the Context of Childhood Adversities (pp.243–66). Cambridge: Cambridge University
Press: ‘Poverty is a distal risk factor whose effects are mediated by proximal risk factors such as parenting behav-
iors, family structure, community variables, and the broader social networks within which the child and her or his
family are embedded’ (245).
our poverty sample, socioeconomic status was a confound’.284 Many of the findings from
the the Minnesota Longitudinal Study of Risk and Adaptation have replicated with low-
risk samples.285 However, Haltigan and Roisman found that infant avoidant attachment,
but not disorganisation, was associated with later dissociative symptoms in the NICHD
study. In cases such as this, where findings differ, the exclusion of socioeconomic factors as
a potential confound by the Minnesota group has contributed to difficulty in interpreting
results.286
Resilience
‘Resilience’ was one of the headline concepts from the emergence of developmental psycho-

pathology as a movement within developmental science. Colleagues of Sroufe and Egeland
at Minnesota, such as Ann Masten, Norm Garmezy, and Auke Tellegan, played especially
important roles in fostering this discourse.287 Some members of the developmental psycho-
pathology movement were interested in examining individual-level factors that could lead
research participants to do better than expected in response to adversity. 288 This is the con-
cept of resilience that would generally get absorbed by cognate disciplines, and by public and
policy discourses.289 Sroufe and Egeland regretted this interpretation of the concept. In what
became the dominant interpretation within developmental psychopathology over time, they
took an opposing stance, emphasising that resilience is a relational process.290 They demon-
strated that many of the studies that have shown links between temperamental robustness
and positive outcomes in the face of adversity have had significant methodological flaws. For
instance, a key work in the characterisation of resilience as a form of individual ‘invulner-
ability’ was Werner and Smith’s cohort study of 505 men and women who were born in 1955
cence. In D. Cicchetti & S. Toth (eds) Rochester Symposium on Developmental Psychopathology: Vol. 8. The Effects of
Trauma on the Developmental Process (pp.403–434). Rochester, NY: University of Rochester Press, p.428.
285 Fraley, R.C. & Roisman, G.I. (2015) Do early caregiving experiences leave an enduring or transient mark on
developmental adaptation? Current Opinion in Psychology, 1, 101–106.

286 Lyons-Ruth and colleagues reported findings from a high-risk sample of 76 families, comprising 41 referred
to child protective services and 35 matched comparisons. Data were available on 56 families who remained in the
sample 20 years later. Lyons-Ruth and colleagues reported that observed lack of responsiveness of parents at home
and in the laboratory to their infants accounted for half of the variance in participants’ self-report of dissociative
symptoms at age 19 on the Carlson and Putnam Dissociative Experiences Scale. The researchers did not find that
either disorganised attachment or trauma predicted later dissociative symptoms; no analysis appears to have been
conducted for avoidant attachment. However, verbal abuse by parents, which was not assessed by the Minnesota
study, made an additional contribution to dissociative symptoms in adolescence, over and above caregiver emo-
tional neglect/psychological unavailability. Dutra, L., Bureau, J.F., Holmes, B., Lyubchik, A., & Lyons-Ruth, K.
(2009) Quality of early care and childhood trauma: a prospective study of developmental pathways to dissociation.
Journal of Nervous and Mental Disease, 197(6), 383–90.
287 Garmezy, N., Masten, A.S., & Tellegen, A. (1984) The study of stress and competence in children: a building
block for developmental psychopathology. Child Development, 55(1), 97–111.

288 E.g. Anthony, E.J. (1974) Introduction: the syndrome of the psychologically vulnerable child. In E.J. Anthony
& C. Koupernik (eds) The Child in his Family: Children at Psychiatric Risk, Vol. 3 (pp.3–10). New York: Wiley.
289 Harper, D. & Speed, E. (2012) Uncovering recovery: the resistible rise of recovery and resilience. Studies
in Social Justice, 6, 9–25; Joseph, J. (2013) Resilience as embedded neoliberalism: a governmentality approach.
Resilience, 1(1), 38–52.
290 On later regret among developmental psychopathologists regarding the widespread popular interpretation
of the concept of resilience, and the role of some of their members in contributing to this interpretation see Luthar,
S.S., Cicchetti, D., & Becker, B. (2000) The construct of resilience: a critical evaluation and guidelines for future
work. Child Development, 71(3), 543–62.
on the Hawaiian island of Kauai.291 Sroufe and Egeland noted, however, that only one tem-
peramental variable in childhood was linked to positive outcomes, and this was the mother’s
description of her child as ‘loveable’ at two years old. They expressed doubt that this can con-
fidently be regarded as reflecting inherent child variation.292
The term ‘resiliency’ implies an individual trait, and was therefore regarded by the
Minnesota group as unacceptably misleading. But the term ‘resilience’ may more readily
imply a process, even if often it has been used to refer to a trait. In 2003, Yates, Egeland, and
Sroufe set out their definition of ‘resilience’ as a description of the capacity of children to
garner resources that allow them to negotiate the present effectively and set up internal and
external resources for future challenges.293 The emphasis on both internal and external re-
sources provided a place for individual-level factors. However, the placement of these factors
in developmental perspective suggested that it is ‘as supports accumulate’ that an individual

develops the ‘acquired attitudes, expectations, and capacities to marshal resources [to] en-
able them to cope better with the additional challenge they may face’.294
The characteristics usually ascribed to ‘resilient individuals’, such as a calm temperament
or optimism, were firmly regarded by Yates, Egeland, and Sroufe as developmental and social
outcomes, mediated through the inner calibration of behavioural systems and by ongoing
family and peer relationships. This made their position a profoundly destigmatising one,
since it emphasised that mental health symptoms are an ordinary and expectable response to
the accumulation of adversities, especially through childhood and adolescence, in the con-
text of inadequate support. They were also early critics of a unidimensional concept of re-
silience, and urged recognition that positive outcomes despite adversities in one domain of
life may in fact corrode opportunities in another domain that requires quite a different set of
resources and capacities to the first. The paradigmatic case for Yates, Egeland, and Sroufe was
avoidant attachment, which was conceptualised as an intelligible and likely helpful response
to a less-sensitive caregiving context, but as making later emotion regulation and peer rela-
tionships more difficult in certain regards.295 Yates and colleagues suggested that keeping
our worries and desires to ourselves reduces our exposure to hurt and disappointment, but if
we lack the capacity to do otherwise then this saps something from us that might have other-
wise been used elsewhere in building a richer and more stable life, as we lose heart in the pro-
spects of attaining or sharing anything more than what we can muster alone. This then may
fuse the avoidance into place, as we miss the opportunities that may arise for recognition and
response to our worries and desires by others.296
291 Werner, E.E. & Smith, R.S. (1982) Vulnerable but Invincible: A Study of Resilient Children. New York: McGraw-
Hill; Werner, E.E. & Smith, R.S. (1992) Overcoming the Odds: High Risk Children from Birth to Adulthood. Ithaca,
NY: Cornell University Press.
Press, pp.249–50.
294 Supkoff, L., Puig, J., & Sroufe, L.A. (2013) Situating resilience in developmental context. In M. Ungar (ed.)
The Social Ecology of Resilience (pp.127–42). New York: Springer, p.128.

517–28. See also Luthar, S.S., Doernberger, C.H., & Zigler, E. (1993) Resilience is not a unidimensional con-
struct: insights from a prospective study of inner-city adolescents. Development & Psychopathology, 5(4), 703–717.

Yates, Egeland, and Sroufe illustrated their argument by pointing to the developmental tra-
jectories of children who displayed substantial behavioural problems at school. Causadias,
Salvatore, and Sroufe found that, of children showing conduct problems, those who had experi-
enced the most favourable early care showed the fewest mental health problems and had more
supportive friendships in adolescence.297 Yates, Egeland, and Sroufe argued that a positive foun-
dation had served as a resource for meeting the challenges of adolescence. The early foundation
may have proven a ‘promotive factor’, enhancing the probability of later adaptation, not merely
buffering the effects of contemporary adversity. Such findings were taken as showing that re-
silience is a description of the availability and utilisation of resources, rather than an inherent
quality of individuals: ‘Consider the alternative interpretation that an early history of positive
adaptation reflects an underlying individual trait called resilience. In this view, some of these
children were resilient, then were not, and then were again.’298 Had the longitudinal study begun

in preschool, the fact that some of the children had better outcomes than others would have
seemed to be an effect of their individual characteristics; viewed in developmental context, these
differences in fact were predictable from their experiences of early care.299
Perhaps the single most famous finding regarding resilience of the Minnesota
Longitudinal Study of Risk and Adaptation was the identification of factors that could ‘break
the cycle of abuse’, the transmission of abusive caregiving from one generation to the next.
When parents who were abused as children were asked whether they would raise their child
differently, all but four said yes, vowing that they would not repeat their own maltreatment.
Yet based on interviews with the mothers in their sample, and multimethod assessments of
child wellbeing and safety, the Minnesota group found that some mothers were not able to
fulfil these vows. They found that about a third of caregivers who reported that they had been
abused went on to abuse their own child: a much higher proportion than matched controls
but overall a minority of cases. There was a substantial gap in intergenerational transmis-
sion, a finding that has subsequently been well replicated.300 Egeland, Jacobvitz, and Sroufe
reported that depression was exceptionally prevalent among the abusing mothers, and they
‘tended to view their children in an entirely negative or positive light and failed to recognize
the ambivalence, which, of necessity, accompanies child care’.301
297 See also Causadias, J.M., Salvatore, J.E., & Sroufe, L.A. (2012) Early patterns of self-regulation as risk and
promotive factors in development: a longitudinal study from childhood to adulthood in a high-risk sample.
International Journal of Behavioral Development, 36(4), 293–302.
Press, p.251.
299 Another illustration is Supkoff, L., Puig, J., & Sroufe, L.A. (2013) Situating resilience in developmental con-
text. In M. Ungar (ed.) The Social Ecology of Resilience (pp.127–42). New York: Springer: ‘We found that we could
distinguish between those who recovered from depression in adulthood following formation of a romantic part-
nership from those who did not recover after partnering on the basis of different histories of infant attachment
(Sroufe, Corffino & Carlson 2010) ’ (138). In fact, however, these findings are not reported in Sroufe, Corffino, and
Carlson (2010). Some analogous results are reported in Salvatore, J.E. (2011) Moderating processes in the link be-
tween early caregiving and adult individual and romantic functioning: the distinctive contributions of early adult
romantic relationships. Unpublished doctoral dissertation, University of Minnesota. However the interaction is
between the early childhood composite variable, not attachment, and adult relationships, in predicting depressive
symptoms at age 23.
300 E.g. Widom, C.S., Czaja, C.J., & DuMont, K.A. (2015) Intergenerational transmission of child abuse and neg-
lect Science, 347(6229), 1480–85; Ben-David, V., Jonson-Reid, M., Drake, B., & Kohl, P.O. (2015) The association
between childhood maltreatment experiences and the onset of maltreatment perpetration in young adulthood
controlling for proximal and distal risk factors. Child Abuse & Neglect, 46, 132–41.
301 Egeland, B., Jacobvitz, D., & Sroufe, L.A. (1988) Breaking the cycle of abuse. Child Development, 59(4), 1080–
1088, p.1087.
Egeland, Jacobvitz, and Sroufe reported that the abused mothers who did not go on to abuse
their child were distinguished by one or more of three experiences. First, some had emotional
support from a non-abusive adult during their childhood. Second, some had participated in
therapy. Third, some had a stable and emotionally supportive relationship with their adult
partner. Egeland and colleagues observed that what these three experiences have in common
is that they are experiences in which a secure base and safe haven has been provided, a ‘basis for
developing alternative or transformed models of relationship’.302 One of the case studies they
provide is of a mother ‘VF’.303 VF grew up in a family of ten children, all of whom ended up in
foster care. Her mother would regularly beat the children, and as an adult VF still had scars from
the physical abuse. Scared of being taken away from her mother like her older siblings, VF lied
repeatedly at school to account for her bruises and cuts. However, when she was 11 years old, her
mother asked child protection services to take VF. At the point that she left her mother’s care she

had never been to a doctor or dentist—and did not know what these were. She spent two years
in a first foster home, then two more foster homes. She then spent two years in a centre for treat-
ment of substance dependency. During this time, she had psychotherapy. At age 18 she moved
into an apartment with another girl, and very shortly after that became pregnant. When her
son was a year old, she entered into a long-term relationship and eventually married. She and
her husband both found work and had three more children together. Egeland and colleagues
reported that ‘VF seems particularly aware of her past’, and was able to use this awareness to
consider how it affects her caregiving. VF regarded her period of therapy and the support of her
husband as important for helping her care for her children.
The researchers were powerfully struck that certain social experiences helped some moth-
ers interrupt the continuities in abusive caregiving between generations.304 However, a later
report by Enlow, Englund, and Egeland found that maternal experiences of child maltreat-
ment predicted behavioural problems in their children at age seven, even if the mothers were
not maltreating. This pathway was mediated by social isolation and negative life events such
as financial difficulties, bereavements, and illness. The researchers concluded that ‘mothers
able to “break the intergenerational cycle of maltreatment” may still be at risk for creating a
negative caregiving context through increased stress exposures and reduced social support,
part of the developmental sequelae of their maltreatment histories’.305 In a review chapter,
Egeland and colleagues drew attention to research from other laboratories showing that,
when abused parents do abuse their children, the kinds of abuse can be quite dissimilar.306
302 Ibid. p.1082.

303 Egeland, B., Jacobvitz, D., & Papatola, K. (1987) Intergenerational continuity of abuse. In R.J. Gelles & J.B.
Lancaster (eds) Child Abuse and Neglect: Biosocial Dimensions (pp.255–76). New York: Aldine De Gruyter, p.269.
304 Subsequent research has suggested that it is concrete ecological and financial supports that especially predict
the interruption of the transmission of abusive care, more than individual psychological factors. St-Laurent, D.,
Dubois-Comtois, K., Milot, T., & Cantinotti, M. (2019) Intergenerational continuity/discontinuity of child mal-
treatment among low-income mother–child dyads: the roles of childhood maltreatment characteristics, maternal
psychological functioning, and family ecology. Development & Psychopathology, 31(1), 189–202.
305 Bosquet Enlow, M., Englund, M.M., & Egeland, B. (2018) Maternal childhood maltreatment history and
child mental health: mechanisms in intergenerational effects. Journal of Clinical Child & Adolescent Psychology,
47(1), 47–62. Later research with the sample also indicated competence in romantic relationships as a potent
mediator of the association between experiences of abuse and neglect in childhood and parenting identified as
maltreating. Labella, M.H., Raby, K.L., Martin, J., & Roisman, G.I. (2019) Romantic functioning mediates pro-
spective associations between childhood abuse and neglect and parenting outcomes in adulthood. Development &
transmission of child maltreatment. In J.E. Korbin & R.D. Krugman (eds) Handbook of Child Maltreatment, Vol. 2
(pp.115–29). New York: Springer.
For instance, Thompson and colleagues found that mothers who had experienced sexual
or physical violence as children were substantially more likely to have a child subject to in-
quiry from child protective services. Yet 75% of these referrals were for neglect.307 The move-
ment of abuse between the generations appears not to be mimetic, as Bowlby had assumed
(Chapter 1).
The Minnesota group emphasised that resilience is not a singular entity. Individuals can
do well in one area of their life, but less well in others, as a result of the internal and ex-
ternal resources available to them—as well as the cut and colour of the challenges they face.
The availability of a secure base as a form of social support helped promote non-abusive
caregiving behaviours in mothers in their sample who had themselves experienced abuse.
However, the early life stress these mothers experienced in receiving abuse nonetheless
had implications for the mental health of their children. Sroufe and Egeland emphasised

the importance of both life stress and social support in developmental perspective. When
individuals responded to adversity with more positive outcomes, Sroufe and colleagues
claimed that in regression analyses—not reported in print—the combination of life stress at
different points and social support at different points could account for 80% of variance in
outcomes: ‘Thus, little is left to mystery with adequate developmental data. Our research has
found, time and time again, that resilience is not simply a function of good outcomes despite
bad experience, but also an example of prior good experience facilitating the mobilisation of
resources.’308 However, the analyses that lead to the 80% figure are not described in the text
or in other published work.
Gender
In ‘The domain of developmental psychopathology’, Sroufe and Rutter proposed that gender
is an important category to investigate.309 Both researchers found in their longitudinal stud-
ies that after puberty males facing risk factors were more likely to display aggression, con-
duct problems, and other ‘externalising’ forms of mental health problems, whereas girls
were more likely to display depression, anxiety, and other ‘internalising’ symptoms. Sroufe
and Rutter acknowledged that biological factors play a role, but overall their emphasis fell
on social and cultural causes of this difference. This conclusion was later reinforced by
Bakermans-Kranenburg and colleagues, who found no such gender differences in the dis-
tribution or trajectory of depressive symptoms in a Chinese sample, in contrast to European
and American samples.310 Sroufe and Rutter pointed to the role of powerful cultural norms
307 Thompson, R. (2006) Exploring the link between maternal history of childhood victimization and child
risk of maltreatment. Journal of Trauma Practice, 5(2), 57–72. A recent meta-anlysis of 142 studies showed
modest intergenerational transmission of abuse (k = 80; d = .45), and of kind of abuse (neglect: d = .24; physical
abuse: d = .41; emotional abuse: d = .57; sexual abuse: d = .39). See Madigan, S., Cyr, C., Eirich, R., et al. (2019)
Testing the cycle of maltreatment hypothesis: meta-analytic evidence of the intergenerational transmission of
child maltreatment. Development & Psychopathology, 31(1), 23–51.
308 Supkoff, L., Puig, J., & Sroufe, L.A. (2013) Situating resilience in developmental context. In M. Ungar (ed.)
The Social Ecology of Resilience (pp.127–42). New York: Springer, p.134. It is not clear, however, what outcome
measures were used.
309 Sroufe, L. & Rutter, M. (1984) The domain of developmental psychopathology. Child Development,
55(1), 17–29.
310 Cao, C., Rijlaarsdam, J., van der Voort, A., Ji, L., Zhang, W., & Bakermans- Kranenburg, M.J. (2017)
Associations between dopamine D2 receptor (DRD2) gene, maternal positive parenting and trajectories of depres-
sive symptoms from early to mid-adolescence. Journal of Abnormal Child Psychology, 46(2), 365–79.
around gender that shape the expression of emotions, the activation of the caregiving system,
the modulation of closeness and intimacy, and acceptable bases for self-esteem. They give
particular emphasis to gender socialisation, since ‘girls in our culture are socialized toward
compliance, inhibition, passivity, and reliance on others’. Depression and anxiety are con-
gruent with these forms of socialisation in European and American societies. Sroufe and
Rutter also refer to the experience of girls of sexual vulnerability, harassment, and the threat
of rape.311
In the Minnesota Longitudinal Study of Risk and Adaptation, Pianta, Sroufe, Egeland,
and colleagues observed aspects of gender socialisation. For instance, boys who were rated
by observers and by their mothers as less socially skilled received more sensitive care at later
follow-up, whereas this compensation by mothers was not provided to girls. In fact, girls
who were more self-controlled received more sensitive care, which the researchers suspect

may have been a form of reward and reinforcement.312 Gender also played a role in shaping
the context of mental health symptoms. For instance, Pianta, Sroufe, and Egeland found pro-
found effects of the kind of role model provided by mothers to their daughters, both as a po-
tential risk and a potential protective factor for internalising symptoms. There was no such
effect for boys.313
Sroufe and colleagues argued that ‘gender is a key aspect of the preschooler’s emerging
self concept. Being a boy or a girl is central to the definition of the self. Development of
a gender-based self-concept involves three steps. First, children gradually adopt sex-typed
behaviour—actions that conform to cultural expectations regarding what is appropriate for
boys or for girls. Second, children simultaneously acquire a gender-role concept—a begin-
ning knowledge of the cultural stereotypes regarding males and females. Finally, children
develop an emotional commitment to their particular gender.’314 Pursuing this reasoning,
the Minnesota group conducted research to examine antecedents of the extent to which
children regarded gender as a permanent characteristic.315 They asked preschool partici-
pants whether someone’s gender could change in different situations, such as if they put on
different clothes. They found that maternal tendencies towards angry or depressive mood
reduced girls’ perception of gender as a permanent characteristic. By contrast, mothers’ sup-
port and guidance for problem-solving and their responsiveness at home predicted greater
gender concept stability for boys. There was no effect for attachment classification in the
Strange Situation.
Indeed, it appeared to Sroufe and Egeland that gender and the attachment system
were mostly unrelated phenomena. Given the importance of gender for so many
aspects of life, the relatively modest role of gender differences in attachment research is
311 Sroufe, L. & Rutter, M. (1984) The domain of developmental psychopathology. Child Development, 55(1),
17–29, p.26.
312 Pianta, R.C., Sroufe, L.A., & Egeland, B. (1989) Continuity and discontinuity in maternal sensitivity at 6, 24,
and 42 months in a high-risk sample. Child Development, 60(2), 481–7, p.486.

313 Pianta, R.C., Egeland, B., & Sroufe, L.A. (1990) Maternal stress in childrens’ development: predictions of
school outcomes and identification of protective factors. In J.E. Rolf, A. Masten, D. Cicchetti, K. Neuchterlen, &
S. Weintraub (eds) Risk and Protective Factors in the Development of Psychopathology (pp.215–35). Cambridge,
MA: Cambridge University Press.
314 Sroufe, L.A., Cooper, R.G., DeHart, G.B., & Marshall, M.E. (1996) Child Development: Its Nature and Course,
3rd edn. New York: McGraw-Hill, p.373.

study. In T. Wachs & R. Plomin (eds) Conceptualization and Measurement of Organism–Environment Interaction
(pp.68–84). Washington, DC: American Psychological Association.
curious.316 This was already remarked upon by Ainsworth. In correspondence with Bowlby,
Ainsworth wrote:
I believe that the main reason why the issue of sex differences has been so neglected in
regard to attachment is that such differences seem to be absent in infancy. Certainly, the
three main attachment patterns (strategies) do not differ in incidence among males and fe-
males. That makes sense, because in infancy protection is just as important for males as for
females, and the kind of protection that is needed is essentially the same for both sexes.317
Yet gender would similarly prove of relatively little relevance to the distribution of attachment
classifications in middle childhood or adolescence, at least when observational measures
were used.318 Whilst developmental psychopathology has generally been concerned with the

interaction of risk and protective factors, there is no implication that all aspects of life will
interact to the same extent. Gender and attachment were regarded by Sroufe and Egeland as
both important aspects of a child’s experience and relevant to their developmental outcomes.
In particular, both gender socialisation and attachment were regarded as contributing to the
elaboration and calibration of the caregiving and sexual behavioural systems. However, they
considered these contributions as largely independent, rather than in interaction—though
this was not formally tested. After Sroufe and Egeland’s retirement, some small associations
were found by Raby, Roisman, and colleagues on the Adult Attachment Interview (AAI)
when this was coded dimensionally rather than categorically: dismissing is slightly more
common in male participants (r = .17), and preoccupation in female participants (r = .22).319
However, such associations are not large enough to imply a need to substantially renegotiate
Sroufe and Egeland’s position on gender and attachment.
Two case studies
Case descriptions are woven in elegantly to the reports of the Minnesota group, particularly
in Sroufe’s early book chapters.320 However, it was rare for the group to describe in detail the
result of the assessments that had taken place over the years with particular dyads. Two case
studies, however, were reported in substantial depth at the end of the 1990s. One was the case
316 Sroufe, L.A., Cooper, R.G., DeHart, G.B., & Marshall, M.E. (1996) Child Development: Its Nature and Course,
3rd edn. New York: McGraw-Hill: ‘Gender is a central organising theme in development. It plays a key role in the
way people define and experience their worlds. In all societies, parents and others treat boys and girls differently
and expect different things from them. Because of this, children learn cultural stereotypes regarding male and
female behaviours and characteristics. This learning begins early and is pervasive. It manifests itself in children’s
activities, preferences, and social styles. Even among preschoolers, gender is so salient that a child’s most advanced
thinking is often applied to it. Preschoolers label and categorise different activities in terms of gender (Fagot et al.
1992); they remember modelled behaviours better when they are “gender appropriate” (Bauer 1993); and in gen-
eral they use gender as a basis for organising information (Serbin et al. 1993)’ (372–3).
317 Ainsworth, M. (1989) Letter to John Bowlby, 16 November 1989. PP/BOW/B.3/8.
318 Bakermans-Kranenburg, M.J. & van IJzendoorn, M.H. (2009) No reliable gender differences in attachment
across the lifespan. Behavioral and Brain Sciences, 32(1), 22–3.

319 Raby, K.L., Labella, M.H., Martin, J., Carlson, E.A., & Roisman, G.I. (2017) Childhood abuse and neglect
and insecure attachment states of mind in adulthood: prospective, longitudinal evidence from a high-risk sample.
Development & Psychopathology, 29(2), 347–63, Table 4.
320 E.g. Sroufe, L.A. (1983) Infant–caregiver attachment and patterns of adaptation in preschool: the roots of
maladaptation and competence. In M. Perlmutter (ed.) Development and Policy Concerning Children with Special
Needs: The Minnesota Symposium in Child Psychology, Vol. 16 (pp.41–83). Hillsdale, NJ: Lawrence Erlbaum.
of Beth. At 12 months, Beth and her mother received a classification of ambivalent/resistant

attachment, and at 18 months a classification of secure attachment. Beth’s father was killed
in a car crash, but at 24 months Beth and her mother both seemed to be doing well under
the circumstances. Beth was able to cooperate effectively with her mother in the 24-month
tool-use situation, and at 42 months was persistent and enthusiastic in attempting the diffi-
cult Barrier Box task.321 At the end of first grade, Beth’s mother had remarried. Her teacher
described Beth as a bit of a loner, but doing quite well academically. In an interview during
second grade, the interviewer noticed that Beth seemed afraid of her stepfather, who was
regularly drunk. A year later, Beth and her mother moved into a women’s shelter. In the sixth
grade follow-up, Beth and her mother were living with the stepfather again.
At the follow-up when participants were 17 years old, it took some time for the re-
searchers to locate Beth. In interview, she described that her stepfather had been sub-

jecting her to multiple forms of maltreatment since she was 12 years old. Child protection
services inquiries had been deflected by her mother and father, who presented Beth herself
as the problem and disengaged when this narrative was questioned. She had left home at
16, and her parents left the state in order to avoid a court order that the stepfather leave
the home to protect the younger siblings. Beth first lived with a foster family, then a boy-
friend, then her grandparents. She had remained in school throughout this period. The
clinical interview at age 17 revealed no current psychiatric diagnosis. The interview docu-
mented depression in early adolescence, and a subclinical phobia of small spaces. She also
reported one past suicide attempt. At age 19, Beth had been awarded a scholarship to at-
tend college and had completed one semester. She had recently withdrawn due to prob-
lems with her boyfriend, including abuse. However, she planned to return to school, had
a part-time job, and seemed to be doing very well in other areas of her life. She was in the
process of seeking counselling.
A second detailed case study offered by the Minnesota group was that of Laura Miller (a
pseudonym). In the 1990s, a furious controversy raged regarding whether adult ‘recovered
memories’ of child sexual abuse were fabulations or truthful recollections of previously re-
pressed experience.322 In 1998, the journal Development and Psychopathology dedicated a
special issue to this concern, which highlighted the need for thinking about how memory
processes following trauma might be drawn upon for coping and for other adaptive chal-
lenges.323 In line with this interest, in an article from the same year, Duggal and Sroufe
reported the first ever prospective report in which sexual abuse was documented by mul-
tiple sources, and the young person subsequently reported no recall of the abuse for a time,
and then ‘recovered’ the memory. Data for the case study were drawn from the longitu-
dinal study, from a follow-up interview with Laura and her mother, from the records of a
cence. In D. Cicchetti & S. Toth (eds) Rochester Symposium on Developmental Psychopathology: Vol. VIII. The
Effects of Trauma on the Developmental Process (pp.403–434). Rochester, NY: University of Rochester Press, p.417.
322 E.g. Herman, J.L. (1992) Trauma and Recovery. New York: Basic Books; Ceci, S.J. & Loftus, E.F. (1994)
‘Memory work’: a royal road to false memories? Applied Cognitive Psychology, 8, 351–64. For professionals’ views
on recovered memories see Ost, J., Wright, D.B., Easton, S., Hope, L., & French, C.C. (2013) Recovered memories,
satanic abuse, dissociative identity disorder and false memories in the UK: a survey of clinical psychologists and
hypnotherapists. Psychology, Crime & Law, 19(1), 1–19. For discussions of this debate in historical context see
e.g. Hult, J. (2005) The re-emergence of memory recovery: return of seduction theory and birth of survivorship.
History of the Human Sciences, 18(1), 127–42.
323 Toth, S. L. & Cicchetti, D. (1998) Remembering, forgetting, and the effects of trauma on memory: a develop-
mental psychopathology perspective. Development & Psychopathology, 10(4), 589–605.

therapist who worked with the family after child protection services became involved when
Laura was four years old, and from interviews with Laura’s child protection worker and her
therapist.
A first sign of sexual trauma was documented in preschool. Laura was one of the children
who was part of the preschool class made up of members of the sample. The head teacher of
the preschool reported sexual behaviours shown by Laura in the preschool classroom. In the
interview with mothers when their children were 54 months, Ms Miller (Laura’s mother)
stated that Laura had described a dream in which her father, who had some limited custody,
had come into her bed while she was sleeping and touched her, while she pleaded with him to
stop. When confronted, Laura’s father denied sexually abusing Laura, but agreed to see Laura
for only three hours per week and only with a third party present. Ms Miller asked child pro-
tection services if visitation could be prevented entirely, but physical examination for pene-

tration was negative and by this time Laura would not talk any further about the incident, so
it would have been difficult to evidence by the legal standards of the time. However, Laura
was referred for therapy. The therapist notes from the period agree with Ms Miller’s account.
They also state that Laura spontaneously announced to the therapist, a month after intake,
that ‘Daddy knows he did it, he’s just scared’. The notes included record of Laura’s anxiety and
anger towards her father.324 The father was interviewed by the therapist and said that he had
no memory of sexually abusing his daughter, and that if it occurred it would have been while
he was on drugs. At the end of first grade, Laura’s teacher related that Laura had told her that
the reason her parents had divorced was that her father had abused her, but that he was not
doing it any more. Duggal and Sroufe concluded that ‘Multiple sources of evidence discussed
in this section consistently support the conclusion that Laura experienced trauma of a sexual
nature during early childhood’.325 The last record in which Laura evidenced knowledge of
abuse was at age eight, in the therapist notes.
In her engagements with mental health services in seventh and ninth grade, she did not
mention childhood sexual abuse as a presenting issue. At age 16, participants in the study
were asked to complete a broad-ranging written health interview. One of the items was the
question: ‘Have you ever been sexually abused? Sexual abuse is when someone in your family
or someone else touches you in a place you did not want to be touched, or does something to
you sexually which they shouldn’t have done.’ Laura indicated that she had not been sexually
abused. The interviewer noted that Laura appeared unguarded in her replies, and disclosed
various other personal information with candour. At age 17, in a clinical interview to assess
symptoms of post-traumatic stress disorder, Laura was asked whether she had ever had any
terrible and unusual experiences. She reported, without hesitation, that she had experienced
nothing of this kind. In response to a query regarding experiences of abuse, including sexual
abuse, Laura replied that this had not happened to her. She was again candid in discussing
other difficult past and present experiences, including contemporary mental health prob-
lems. Ms Miller recalled that during adolescence, Laura reported being happy to spend time
with her father and wanted to see him more. Ms Miller did not bring up the issue of abuse or
ask direct questions because she felt that she had been warned by child protection services to
avoid ‘putting words in Laura’s mouth’.326
324 Duggal, S. & Sroufe, L.A. (1998) Recovered memory of childhood sexual trauma: a documented case from a
longitudinal study. Journal of Traumatic Stress, 11(2), 301–321, p.306.

325 Ibid. p.311.
326 Ibid. p.313.
When Laura was 18, she took part in an AAI (Chapter 3), which includes a question about
experiences of abuse. In response to this question, Laura described a recent conversation
with her boyfriend about her first memories of her parents. This triggered a ‘really over-
whelming feeling’, and led Laura to seek out a trusted teacher. During the conversation with
the teacher, she reportedly recovered memories of her father kissing her while she was in
bed, with a sense that there was a sexual component to the interaction:
I told her this first memory and then it just came to me. Like, it just was so clear in my
mind, um, you know he like did this to me, whatever. And um, and I like went into com-
plete shock . . . I just felt like I was really dizzy and, and I couldn’t really see very well and,
and there was like all these noises going on, but I didn’t know what they were, and and it
wa-I was really confused and, like I couldn’t cry but I felt like I should be.327

She added: ‘You know he bent, he was like walking over to me and he bent down and kissed
me. And there’s something after that and I don’t know what it is.’328 Following the conversa-
tion with the teacher, Laura’s relationship with her father changed strikingly. She felt scared
of him most of the time, interspersed with missing him terribly. Duggal and Sroufe argued
that exclusion of memory of the event made sense if Laura was forced to see her father dur-
ing her childhood. There was a logic to the memory resurfacing at age 18, since this was
the threshold of Laura’s independence from her parents and the need for visitation. Duggal
and Sroufe situated Laura’s case as illustrating the role of defensive processes as adaptive re-
sponses to situations that cannot be controlled or escaped. They conjectured that it seemed
a good sign for Laura’s later development that she could release the exclusion of the painful
memory at the point that she was no longer required to see her father.
There is no sense that the two published case studies are representative of the sample as
a whole. They nonetheless offer a valuable window into the way that the multiple assess-
ments of the Minnesota Longitudinal Study of Risk and Adaptation together offered the re-
search group an encompassing picture of human lives, each hard-won and remarkable. The
cases illustrate the approach to child development that framed Egeland and Sroufe’s work on
the Minnesota study. The assessments conducted by the Minnesota group allowed lines of
continuity and discontinuity in development to be examined, alongside exploration of the
interdependence and lurching along of different social relationships. Comparison of the two
cases illustrates the important role of changing adversities and social supports for the kinds
of resilience they could muster. For instance, the cases of Beth and Laura illustrate disinhib-
iting parental drug and alcohol use as a risk factor for the children, which compounded with
other adversities and forms of unforgiving isolation they faced.
Both case studies were reported in the late 1990s. There have subsequently been several
further waves of data collection. The field has also seen a ‘changing of the guards’ with the
retirement of Sroufe and Egeland. Such a significant developmental transition has come with
new challenges, and, as could be anticipated, has required the elaboration and change in the
field’s inner organisation. One angle on the lines of developmental continuity between past
and present is to consider one of Sroufe and Egeland’s former students, Dante Cicchetti, who
now holds Sroufe’s William Harris Professorship in Child Development at the University of
327 Ibid. p.315–16.

328 Ibid.
Minnesota. Cicchetti did not primarily regard himself as an attachment researcher. However,
the approach of Sroufe and Egeland and the findings of the Minnesota Longitudinal Study
of Risk and Adaptation have been a central component of his influential approach to de-
velopmental psychology. Consideration of Cicchetti’s work offers an illustration of one way
that the legacy of the study of attachment at Minnesota has been inherited by subsequent
researchers.
Two former students
Dante Cicchetti

Part of Mary Ainsworth’s achievement was to have translated a new and exciting methodo-
logical innovation into a new generation of attachment researchers in the 1970s and 1980s,
including many important future leaders (Chapter 2). The other primary hub of attachment
research during the period was the Minnesota group, which was also critical for mentoring a
second generation of attachment researchers. Sroufe commented that his proudest achieve-
ment has been the contribution made by his former students to the field.329 One of Sroufe’s
last doctoral students was Glenn Roisman, whose work was discussed in Chapter 3. At the
other end of his career, one of his first doctoral students was Dante Cicchetti. Cicchetti had
a drive to examine experiences of risk and improve outcomes for families rooted in his own
childhood. As he later recalled in interview, ‘much of my research has been influenced by my
own experiences. Early encounters with poverty and harsh conditions played a major role
in fuelling my research interest in child maltreatment.’330 He later dedicated his Award from
the American Psychological Association for Distinguished Senior Career Contribution to
Psychology in the Public Interest to his younger sister ‘Candace, who died in the fall of 2000.
She represents one of all too many individuals who unfortunately lose battles due to ad-
verse circumstances and mental illness.’331 Cicchetti has also written about his own struggles
around mental health.332
Cicchetti’s doctoral thesis examined the affective and cognitive development of in-
fants with Down syndrome.333 Cicchetti badged his work as an application of the organ-
isational perspective, announced in the Sroufe and Waters 1977 paper ‘Attachment as an
organizational construct’ (Chapter 2).334 Over four decades there is an explanation of this
329 Alan Sroufe, personal communication to Dante Cicchetti, cited in Cicchetti, D. (2011) Champions of psych-
ology: Dante Cicchetti—an interview with APS Student Caucus. Association for Psychological Science Observer,
24(9). http://www.psychologicalscience.org/index.php/publications/observer/champions-ofpsychology-dante-
cicchetti.html.
330 Ibid.
331 Cicchetti, D. (2004) Biography. American Psychologist, 59(8), 1–4 p.728–31.
332 Cicchetti, D. (2015) Reflections on Carroll Izard’s contributions: influences on diverse scientific disciplines
and personal recollections. Emotion Review, 7(2), 104–109.

333 Cicchetti, D. & Sroufe, L.A. (1976) The relationship between affective and cognitive development in Down’s
syndrome infants. Child Development, 47, 920–29.

334 Several other later researchers in developmental psychopathology situated themselves as under the banner of
Sroufe’s ‘organisational perspective’. See for instance Wyman, P.A., Cowen, E.L., Work, W.C., & Kerley, J.H. (1993)
The role of children’s future expectations in self-system functioning and adjustment to life stress: a prospective
study of urban at-risk children. Development & Psychopathology, 5(4), 649–61. However, Cicchetti has been the
most consistent advocate for this as an explicit approach to and model of developmental psychopathology.
Two former students 401
perspective in nearly every first-author article and book chapter he has written. Following
Sroufe, Cicchetti emphasised that individual behaviours should not be the unit of analysis
in the study of child development. Instead, researchers should focus on the integration of
resources in responding to the most important challenges from the environment relevant to
a person at that time in their lives.335 Bowlby’s concept of probabilistic developmental path-
ways had just been introduced during Cicchetti’s time as a graduate student, and this proved
an important influence on his exploration of the interaction of factors in both risk and re-
silience, and the reciprocal interpretation of typical and atypical development in light of one
another.336 As a graduate student at Minnesota, Cicchetti co-authored a review paper with
Egeland on antecedents of child maltreatment, with particular attention paid to the role of
poverty, family conflict, and alcohol and substance use.337 He documented that parents who
maltreat their children show no single set of personality traits or fit into specific psychiatric

diagnostic categories. Instead there are a wide variety of factors that together contribute to
the likelihood of maltreatment, as well as other potential difficulties for children and fam-
ilies. The fact that a diversity of pathways could lead to the same outcome would be dubbed
by Cicchetti as ‘equifinality’; the fact that a single pathway could contribute to diverse out-
comes would be dubbed ‘multifinality’.338
After leaving Minnesota in 1977, his central focus in the 1980s was basic research on child
maltreatment. Cicchetti took up a faculty position at Harvard, where he won a grant for
over $2 million from the National Center on Child Abuse and Neglect for a study of the eti-
ology, intergenerational transmission, and developmental sequelae of children. Funding for
follow-up research with the sample was provided by the National Institute of Mental Health
and the William T. Grant Foundation. He became Director of the Mt. Hope Family Center
at the University of Rochester in 1985. During the 1980s, Cicchetti played a pivotal role in
the initiation of developmental psychopathology as a movement within developmental sci-
ence. He sustained a punishing schedule of productive work during the decade as the move-
ment was established. One metric is his work as an editor: as well as initiating the journal
Development and Psychopathology and the annual Rochester Symposium on Developmental
Psychopathology, Cicchetti co-edited 10 books through the 1980s. He also published 26 art-
icles as first author.
By the end of the decade, the National Academy of Sciences had embraced develop-
mental psychopathology as the overarching framework for its report Research on Children
335 Cicchetti, D. & Sroufe, L.A. (1978) An organizational view of affect: illustration from the study of Down’s
syndrome infants. In M. Lewis & L. Rosenblum (eds) The Development of Affect (pp.309–350). New York: Plenum
Press; Cicchetti, D. & Schneider- Rosen, K. (1986) An organisational approach to childhood depression.
In M. Rutter, C. Izard, & P. Read (eds) Depression in Young People: Clinical and Developmental Perspectives.
New York: Guilford.
336 Cicchetti, D. & Greenberg, M.T. (1991) The legacy of John Bowlby. Development & Psychopathology, 3(4),
347–50.
337 Cicchetti, D., Taraldson, B., & Egeland, B. (1978) Perspectives in the treatment and understanding
of child abuse. In A. Goldstein (ed.) Prescriptions for Child Mental Health and Education (pp.301–378).
New York: Pergamon.
338 Cicchetti, D., Cummings, E.M., Greenberg, M.T., & Marvin, R.S. (1990) An organisational perspective on
attachment beyond infancy: implications for theory, measurement and research. In M.T. Greenberg, D. Cicchetti,
& E.M. Cummings (eds) Attachment in the Preschool Years (pp.1–50). Chicago: University of Chicago Press,
p.35; Cicchetti, D. & Rogosch, F.A. (1996) Equifinality and multifinality in developmental psychopathology.
Development & Psychopathology, 8, 597–600. These terms were drawn from general systems theory and embry-
ology: see von Bertalanffy, L. (1968) General System Theory. New York: Braziller; Weiss, P. (1969) Principles of
Development. New York: Hafner.
and Adolescents with Mental, Behavioural, and Developmental Disorders,339 and the National
Institute of Mental Health had situated developmental psychopathology as the organising
framework for future research funding priorities for child and adolescent mental health.340
In 1990, the eminent developmental researcher Robert Emde described the work of
Cicchetti, Mary Main, and others as commencing the ‘third phase of attachment research’.341
If Bowlby’s initiation of attachment theory had represented the first phase, and the valid-
ation of the Strange Situation by Ainsworth, Sroufe, Egeland, and others in the late 1970s and
1980s had represented the second phrase, the third phase was characterised by innovations
in methodology to examine attachment beyond infancy and new insights into risk and pro-
tective factors for mental illness.
Yet the continuities between the work of Cicchetti and Sroufe were strong and fun-
damental. Cicchetti picked up from Sroufe and colleagues an abiding concern for the

implications of early attachment relationships, even if this concern would be integrated
within a broader frame. For instance, he argued for greater attention to the implications
of cultural differences in constituting and shaping risk and promotive factors, and indi-
vidual and group responses to challenges.342 However, the prime extension of Sroufe’s
model was the inclusion of biological factors. Cicchetti was led to a concern with bio-
logical differences from the start of his career in making sense of the experiences of fam-
ilies with an infant with Down syndrome. However, he was also part of a generation of
researchers with much more ready access to a diversifying array of biomedical measures.
Cicchetti argued that the account of resilience provided by Sroufe and colleagues could
be expanded and enhanced through attention to biological factors, such as neurological,
endocrinal, genetic, and immunological processes, as sources of stability and change in
developmental pathways.343
Behavioural systems such as attachment were reconceived as having affective, cognitive,
behavioural, cultural, and neuroendocrinal components. Cicchetti therefore argued that as-
sessments of maltreated children by child protective services as well as researchers should
include both biological and psychological measures. However, for Cicchetti, this was only as
a more fiercely realised version, not different in kind, from Sroufe’s stance that the develop-
ment of the ‘whole person’ must be kept in view.344 It was also an extension of Sroufe’s focus
on the emotional components of behavioural systems. Cicchetti regarded emotions as having
four components: non-verbal components, verbal components, experiential components,
339 National Academy of Sciences (1989) Research on Children and Adolescents with Mental, Behavioral, and
Developmental Disorders. Washington, DC: National Academy Press.

340 US Department of Health and Human Services (1990) National Plan for Research on Child and Adolescent
Mental Disorders. Rockville, MD: US Department of Health and Human Services, National Institute on Mental
Health.
341 Emde, R. (1990) Preface. In M.T. Greenberg, D. Cicchetti, & E.M. Cummings (eds) Attachment in the
Preschool Years (pp.ix–xii). Chicago: University of Chicago Press, p.x.

342 Coll, C.G., Akerman, A., & Cicchetti, D. (2000) Cultural influences on developmental processes and out-
comes: implications for the study of development and psychopathology. Development & Psychopathology, 12(3),
333–56. On the role of cultural components within the organisation of the attachment behavioural system see also
Morelli, G., Chaudhary, N., Gottlieb, A., et al. (2017) Taking culture seriously: a pluralistic approach to attachment.
In H. Keller & K.A. Bard (eds) The Cultural Nature of Attachment (pp.139–69). Cambridge, MA: MIT Press.
343 Cicchetti, D. (2002) How a child builds a brain: insights from normality and psychopathology. In W. Hartup
& R. Weinberg (eds) Child Psychology in Retrospect and Prospect: The Minnesota Symposia on Child Psychology, Vol.
32 (pp.23–71). Mahwah, NJ: Lawrence Erlbaum.
344 Curtis, W.J. & Cicchetti, D. (2003) Moving research on resilience into the 21st century: theoretical and meth-
odological considerations in examining the biological contributors to resilience. Development & Psychopathology,
15(3), 773–810.
and biological components.345 For instance Cicchetti and colleagues utilised the Egeland
and Sroufe methodology of inviting their child participants to take part in summer camps
in order to see how individuals form and respond to the developmental challenge of peer
relationships. However, unlike Egeland and Sroufe, Cicchetti and colleagues interpreted the
behaviour of the individuals and the interactions of the groups in light of neuroendocrinal
data, as well as self-report and observation.346
Egeland and Sroufe did not have available such neurological and hormonal measures
and did not give biological processes a specific place in their thinking about behavioural
systems.347 However, to some extent both researchers were attentive to biological factors.
For instance, Sroufe and Waters pioneered the use of physiological measures of cardiac
response to show the distress of the apparently unruffled infants in avoidant dyads. And
there is every reason to assume that if other physiological measures of sympathetic nervous

system response had been available, for instance saliva cortisol, Sroufe and Waters would
have measured these too.348 Nonetheless, of the two principal investigators, Egeland appears
to have been the more sympathetic to biological assessments. He drew on assessments of pu-
berty development in elaborating structural equation models of the antecedents of teenage
sexual behaviour and alcohol use. Attentive to the multicausal processes involved in pu-
berty timing, Egeland and colleagues regarded physical maturity as a ‘combined effect of
hormones and social interactions’.349 He also offered speculations on the neurophysiological
processes that provide the infrastructure for avoidant and resistant attachment.350 Cicchetti
and Egeland have subsequently worked together on a study of contribution of genotype to
insecure attachment in the Minnesota sample. The researchers found that variance in the
serotonin-transporter-linked polymorphic region (5-HTTLPR) had no association with
whether a child was classified as secure or insecure. However, for those classified as insecure,
it contributed to the prediction of an ambivalent/resistant or avoidant attachment classifi-
cation, through children’s emotional reactivity.351 5-HTTLPR also predicted the subtype of
secure attachment, contributing to the more distressed (B3 and B4) or less distressed (B1
and B2) form of security. The effect seemed to be independent of the contribution made to
Development: Today and Tomorrow (pp.177–98). San Francisco: Jossey-Bass.

346 Cicchetti, D. & Rogosch, F.A. (2001) The impact of child maltreatment and psychopathology upon neuro-
endocrine functioning. Development & Psychopathology, 13, 783–804.

347 Egeland, B. (1983) Book review: the second year: the emergence of self awareness by J. Kagan. Journal
of Orthopsychiatry, 53, 365–7: ‘Most of us tend to ignore the issue of inheritance and the biological basis of
behavior’ (366).
348 Sroufe had a career-long concern with medications given to children, and this concern was generally a sense
of dismay at how social and family relationships got sidelined by the hunt for neuroendocrinal solutions to devel-
opmental problems. See e.g. Sroufe, L.A. & Stewart, M.A. (1973) Treating problem children with stimulant drugs.
New England Journal of Medicine, 209, 407–413; Sroufe, L.A. (2012) The ‘other’ drug dilemma. USA Today, May
2012, p.21.
349 Siebenbruner, J., Zimmer-Gembeck, M.J., & Egeland, B. (2007) Sexual partners and contraceptive use: a
16-year prospective study predicting abstinence and risk behavior. Journal of Research on Adolescence, 17(1),
179–206, p.184.
Health, Vol. 4 (pp.35–89). New York: Wiley: ‘Suboptimal social environments tend to promote physiological sys-
tems that are dominated by the parasympathetic branch of the autonomic nervous system (avoidant histories) or
the sympathetic branch of the autonomic nervous system (resistant histories)’ (42).
351 Raby, K.L., Cicchetti, D., Carlson, E.A., Cutuli, J.J., Englund, M.M., & Egeland, B. (2012) Genetic and
caregiving-based contributions to infant attachment: unique associations with distress reactivity and attachment
security. Psychological Science, 23(9), 1016–23.
attachment classification by caregiver sensitivity: there was no interaction effect between the
variables. On the basis of such findings, Raby, Cicchetti, Egeland, and colleagues have called
for ‘reconciliation’ between Sroufe’s focus on attachment and Kagan’s focus on temperament,
arguing that both can make a contribution.352 However, it should be noted that Raby’s later
work on the NICHD sample has not offered support for the role of molecular genetic factors
in predicting attachment classifications.353
In agreement with Sroufe and Egeland, Cicchetti viewed it as an important inter-
vention against the stigma attached to mental illness to view symptoms as ordinary
and statistically expectable responses to a dynamic accumulation of adversities, and
relative lack of supports, in the course of development.354 Whilst less adverse to use
of diagnostic categories in assessments than Sroufe, Cicchetti has been mindful of the
sociological processes that led to the dominant role of diagnoses within psychology,

and in his own work on children and adolescents has focused much more on cross-
diagnostic developmental and relational processes.355 Cicchetti and Toth have been
extensively involved, for example, in studies that measure traumatic experiences, but
have expressed doubts about the criteria used to diagnose trauma in clinical practice,
for instance the need for a precipitating single event.356 They have also expressed con-
cern regarding the varied connotations of the term ‘trauma’, which can lead to mis-
communication between researchers, for instance in discussion of the implications of
trauma across developmental periods.
The decision to not intervene with the families in the Minnesota study, even the maltreat-
ing families not known to child protective services, in their study, but to watch and try to
understand was one that weighed heavily on Sroufe and colleagues.357 They felt that their
participants had signed up for a cohort study, not an intervention study, and that it was also
important to understand the naturalistic antecedents and implications of child maltreat-
ment in order to inform future intervention efforts. By contrast, Egeland was actively in-
volved in the creation of the STEEP intervention (which is discussed in the section ‘Holism
in interventions’). Like Egeland, Cicchetti was concerned with both epidemiology and clin-
ical intervention, a position he and Toth described as ‘Bowlby’s dream come full circle’ in
contributing to clinical practice.358 However, Cicchetti and Toth argued that Bowlby’s atten-
tion to psychological unavailability must be elaborated into a more multicausal model, cap-
able of acknowledging factors such as economic, medical and housing difficulties, without
352 Ibid. p.1017.

353 Raby, K.L., Roisman, G.I., & Booth-LaForce, C. (2015) Genetic moderation of stability in attachment se-
curity from early childhood to age 18 years: a replication study. Developmental Psychology, 51(11), 1645–9.
354 Hinshaw, S.P. & Cicchetti, D. (2000) Stigma and mental disorder: conceptions of illness, public attitudes, per-
sonal disclosure, and social policy. Development & Psychopathology, 12(4), 555–98.
355 E.g. Richers, J.E. & Cicchetti, D. (1993) Mark Twain meets DSM-III-R: conduct disorder, development, and
the concept of harmful dysfunction. Development & Psychopathology, 5(1–2), 5–29.

356 Cicchetti, D. & Toth, S.L. (1997) Preface. In D. Cicchetti & S.L. Toth (eds) Rochester Symposium on
Developmental Psychopathology: Trauma: Perspectives on Theory, Research, and Intervention, Vol. 8 (pp.xi–xvi).
Rochester, NY: University of Rochester Press.
357 Sroufe, L.A., Egeland, B., Carlson, E.A., & Collins, W.A. (2005) The Development of the Person. New York:
Guilford, p.84.
358 Cicchetti, D., Toth, S.L., & Lynch, M. (1995) Bowlby’s dream comes full circle. In Advances in Clinical Child
Psychology (pp.1–75). Boston: Springer. Cf. Bowlby, J. (1988) A Secure Base. London: Routledge: It was ‘unex-
pected that, whereas attachment theory was formulated by a clinician for use in the diagnosis and treatment of
emotionally disturbed patients and families, its usage . . . has been mainly to promote research in developmental
psychology’ (ix).
collapsing levels of analysis.359 Cicchetti and Toth also supplemented Bowlby’s approach
with Sroufe’s more fine-grained attention to different developmental challenges across the
lifecourse. For instance, Cicchetti and Toth argued that it follows from an organisational per-
spective that even when a child has been supported to achieve good outcomes following mal-
treatment, developmental reorganisations—such as the transition to parenthood—could
result in vulnerabilities. The reorganisation may well draw upon earlier affective, cognitive,
behavioural, and neuroendocrinal components that had previously not been much in evi-
dence.360 On the other hand, Cicchetti and Toth emphasised that developmental transitions
such as adolescence, and their ensuing shifts in challenges and relevant resources, can also
represent opportunities for reorganisation away from pathology, especially if there are re-
sources available on the basis of past experience or present support.361
Additionally, intervention research offered the prospect of experimental rather than nat-

uralistic evidence for theoretical hypotheses. Intervention studies allow mechanisms of
change and stability for families to be specified, and causation more readily distinguished
from correlation.362 Cicchetti, Toth, and colleagues were early adopters of the use of ran-
domised control trials to evaluate interventions with families and the use of attachment
measures as intervention outcomes.363 Even as attachment theory has seen a decline in
authority and acceptance within the developmental science community, Cicchetti has re-
mained supportive of research in this area. In his role as a prominent and influential editor,
he has allocated space to attachment theory as a key paradigm within developmental psycho-
pathology. From his perspective, the preferential attachment relationships with a caregiver
or caregivers established in infancy ‘provide a context for children’s emerging bio-behavioral
organization’, influencing the integration of ‘neurobiological, cognitive, affective, and behav-
ioral capacities that influence ongoing and future relationships, as well as the understanding
of self ’.364 Stating their agreement with Sroufe, Egeland and colleagues, Cicchetti, and Boyle
have argued that ‘variations in early attachment are not considered to be pathology, or as
directly causing pathology. However, they do lay the foundation for disturbances in develop-
mental processes which can eventuate in psychopathology’.365 For instance, the researchers
359 Cicchetti, D. & Toth, S.L. (1995) Child maltreatment and attachment organization: implications for interven-
tion. In S. Goldberg, R. Muir, & J. Kerr (eds) Attachment Theory: Social, Developmental, and Clinical Perspectives
(pp.279–308). Hillsdale, NJ: The Analytic Press, p.280.
360 Cicchetti and colleagues termed this ‘hierarchic motility’: the way that early components may be incorp-
orated into later behavioural systems and remain accessible to be activated, especially during periods of extreme
stress. This can be an asset, if earlier resources can be made available and repurposed for new tasks. However, it can
also be a liability if these earlier components are called upon but not well suited to contemporary challenges. See
Cicchetti, D. (1994) Integrating developmental risk factors: perspectives from developmental psychopathology. In
C. Nelson (ed.) Threats to Optimal Development: Integrating Biological, Psychological, and Social Risk Factors: The
Minnesota Symposia on Child Psychology, Vol. 27 (pp.229–72). Hillsdale, NJ: Lawrence Erlbaum.
361 Cicchetti, D. & Toth, S.L. (eds) (1996) Rochester Symposium on Developmental Psychopathology: Adolescence
: Opportunities and Challenges, Vol. 7. Rochester, NY: University of Rochester Press.

362 Cicchetti, D. & Hinshaw, S.P. (2002) Prevention and intervention science: contributions to developmental
theory. Development & Psychopathology, 14(4), 667–71; Toth, S.L., Sturge-Apple, M.L., Rogosch, F.A., & Cicchetti,
D. (2015) Mechanisms of change: testing how preventative interventions impact psychological and physiological
stress functioning in mothers in neglectful families. Development & Psychopathology, 27(4), 1661–74.
363 Cicchetti, D., Toth, S.L., & Rogosch, F.A. (1999) The efficacy of toddler–parent psychotherapy to increase
attachment security in offspring of depressed mothers. Attachment & Human Development, 1(1), 34–66; Toth,
S.L., Rogosch, F.A., & Cicchetti, D. (2008) Attachment-theory informed intervention and reflective functioning in
depressed mothers. In H. Steele & M. Steele (eds) The Adult Attachment Interview in Clinical Context (pp.154–72).
New York: Guilford.
364 Cicchetti, D. & Doyle, C. (2016) Child maltreatment, attachment and psychopathology: mediating relations.
World Psychiatry, 15(2), 89–90, p.89.

365 Ibid. p.90.
found that secure attachment at 26 months fully mediated the association between a psycho-
therapy intervention for parents suffering from depression and their child’s display of con-
duct problems at school eight years later.366
Cicchetti and colleagues were the first researchers to use the Main and Solomon dis-
organised attachment classification in a sample of maltreated children.367 The Harvard
Child Maltreatment Project was initiated in 1979 by Cicchetti and Aber, as a large longi-
tudinal study, modelled on the Minnesota Longitudinal Study of Risk and Adaptation.368
A small subsample were seen in the Strange Situation and classified using the Ainsworth
categories.369 Twenty-two dyads were known to child protection services for abuse and/
or neglect of the child or an older sibling; 21 were a matched comparison group. However,
a third of the sample were classified as securely attached, which was an unexpected and
concerning finding. Like Egeland and Sroufe, the infants in these dyads displayed dis-

rupted behaviour that seemed to interrupt or contradict their use of the caregiver as a
secure base. There were a few different proposals available for additions to Ainsworth’s
system. One was that of Ainsworth’s student Patricia Crittenden, who observed that
many abused infants show both avoidant and resistant attachment (A–C).370 Another
was the Main and Solomon disorganised attachment classification, which appealed to
Cicchetti and colleagues as more encompassing; it was assumed that finer-grained dis-
tinctions could be worked out later. The disorganised classification also had the advan-
tage of evidence from Main’s laboratory of relevant correlations with interviews with the
caregivers and with the child’s later behaviour, offering a source of predictive validity.371
The analysis used a coding of the data conducted by Cicchetti and Barnett; Barnett was
blind to the children’s maltreatment status: 82% of the infant–caregiver dyads known to
child protection services were classified as showing disorganised attachment and 14%
classified as secure; in the comparison group 19% were classified as disorganised and
52% classified as secure. At 18 months, 61% of dyads known to child protection services
were classified as showing disorganised attachment compared to 29% in the comparison
group.372
366 Guild, D.J., Toth, S.L., Handley, E.D., Rogosch, F.A., & Cicchetti, D. (2017) Attachment security mediates the
longitudinal association between child–parent psychotherapy and peer relations for toddlers of depressed moth-
ers. Development & Psychopathology, 29(2), 587–600.
Cambridge: Cambridge University Press.
368 Another early, important study of a sample with substantial rates of maltreatment was conducted by Mary
Jo Ward, also a former student of Sroufe: Ward, M.J. & Carlson, E.A. (1995) Associations among adult attachment
representations, maternal sensitivity, and infant–mother attachment in a sample of adolescent mothers. Child
369 Schneider-Rosen, K., Braunwald, K.G., Carlson, V., & Cicchetti, D. (1985) Current perspectives in attach-
ment theory: illustration from the study of maltreated infants. Monographs of the Society for Research in Child
Development, 50(1–2), 194–210.
370 Crittenden, P.M. (1988) Relationships at risk. In J. Belsky & T. Nezworski (eds) Clinical Implications of
Attachment Theory (pp.136–74). Hillsdale, NJ: Lawrence Erlbaum.

372 Barnett, D., Ganiban, J., & Cicchetti, D. (1999) Maltreatment, negative expressivity, and the development
of type D attachments from 12 to 24 months of age. Monographs of the Society for Research in Child Development,
64(3), 97–118.
In a later study, Cicchetti, Rogosch, and Toth used the Strange Situation as part of the
evaluation of two different forms of intervention with 137 families known to child protec-
tion services for infant abuse and/or neglect.373 They also included 52 matched families as a
non-maltreating comparison sample. The maltreating families were allocated to one of three
conditions. In the first condition, they received usual social work support. In the second
condition, they received an intervention that focused on teaching parents about infants’
needs. In the third condition, the mother–child dyad received once-a-week psychotherapy,
using the model of child–parent psychotherapy developed by Selma Fraiberg374 and Alicia
Lieberman (a former student of Ainsworth).375 In child–parent psychotherapy, the inter-
vention focuses on exploring the caregiver’s personal history of adversities and difficult
feelings, especially as evoked in the context of the present-day relationship with their child.
Therapeutic interactions are responsive to spontaneous child–parent interactions and play,

and aim to provide a safe base for the caregiver in providing a safe base for the child. As in
their earlier study, Cicchetti and colleagues found high rates of disorganised attachment in
the maltreatment group (89%, compared to 42% in the non-maltreating comparison dyads).
At 26 months, the Strange Situation was repeated. For the maltreating families who had seen
social work intervention as usual, 78% received a disorganised attachment classification and
the rest of the dyads were classified as avoidantly attached. In the parent education condition,
45% received a disorganised attachment classification and 54% were securely attached. In
the child–parent psychotherapy condition, 32% of dyads received a disorganised attachment
classification and 61% a secure classification. To the researchers’ surprise, the change was not
mediated by measures of adult representations of their child, caregiver sensitivity, parenting
attitudes, child-rearing stress, or social support. Cicchetti and colleagues described that the
results of the trial were ‘gratifying and sobering’:
The fact that plasticity is possible during infancy and that even the most disorganized form
of attachment is modifiable in extremely dysfunctional mother–child dyads offers signifi-
cant hope for thousands of young children and for their families. By fostering secure at-
tachment, costlier interventions such as foster care placement, special education services,
residential treatment, and incarceration can be averted. Unfortunately, our results also
shed light on the harsh reality of the ineffectiveness of services currently being provided.376
More recently, Cicchetti, Rogosch, and Toth reported results from a study of genotypic vari-
ation in their sample, examining the 5-HTTLPR (serotonin-related) and DRD4 (dopamine-
related) polymorphisms. 377 This followed up a hypothesis Cicchetti had put forward in the
373 Cicchetti, D., Rogosch, F.A., & Toth, S.L. (2006) Fostering secure attachment in infants in maltreating fam-
ilies through preventive interventions. Development & Psychopathology, 18(3), 623–49, p.646. The research was
funded by the Administration of Children, Youth, and Families, the National Institute of Mental Health, and the
Spunk Fund, Inc.
374 Fraiberg, S., Adelson, E., & Shapiro, V. (1975) Ghosts in the nursery: a psychoanalytic approach to impaired
infant–mother relationships. Journal of the American Academy of Child Psychiatry, 14, 387–421.
375 Lieberman, A.F. (1992) Infant– parent psychotherapy with toddlers. Development & Psychopathology,
4, 559–74. See also Lieberman, A.F. & Van Horn, P. (2008) Psychotherapy with Infants and Young Children.
New York: Guilford.
376 Cicchetti, D., Rogosch, F.A., & Toth, S.L. (2006) Fostering secure attachment in infants in maltreating fam-
ilies through preventive interventions. Development & Psychopathology, 18(3), 623–49, p.646.
377 Cicchetti, D., Rogosch, F.A., & Toth, S.L. (2011) The effects of child maltreatment and polymorphisms of
the serotonin transporter and dopamine D4 receptor genes on infant attachment and intervention efficacy.
1980s that dyads would be most likely to receive a disorganised attachment classification if
the infant was both abused and had a genetic predisposition towards affective reactivity.378
The hypothesis was not precisely supported. The researchers found that among the non-mal-
treated infants, 5-HTTLPR and DRD4 polymorphisms influenced attachment security and
disorganisation at 26 months and the stability of attachment disorganisation between 12 and
26 months. However, among the maltreated infants, they found no such associations be-
tween genotype and attachment classifications. The researchers concluded that the high rate
of disorganisation ‘overpowered the genetic contribution’ to attachment patterns among the
maltreated infants.379 Genetic variation between infants was not found to have any implica-
tions for the effectiveness of the interventions.
In a later follow-up of the sample 12 months after the end of treatment, dyads who had
received psychotherapy had rates of secure (56%) and disorganised (26%) attachment that

resembled the distribution in non-maltreating samples facing financial adversities. By con-
trast, the parent education intervention had high rates of disorganised attachment (59% dis-
organised; 23% secure), as did those who had received social work intervention as usual
(49% disorganised; 12% secure). Cicchetti and colleagues proposed that the parent edu-
cation intervention had supported the parents only in dealing with the needs of infants. It
had not supported them in responding to past adversities and traumas that might disrupt
their ongoing and unfolding relationship with their growing child.380 Aligned findings come
from another study by Toth, Cicchetti, and colleagues who studied a group of maltreated
preschoolers who received parent–child psychotherapy. The researchers found that the pre-
schoolers’ beliefs about themselves and about their caregivers became more positive and less
negative over the course of the intervention, in contrast to the parent education condition or
the standard child protective services intervention.381
Glenn Roisman
As discussed in Chapter 3, another student of Sroufe’s later in his career was Glenn Roisman.
Roisman is now one of the leaders of the Minnesota Longitudinal Study of Risk and
Adaptation. Roisman has been fascinated by the discoveries of Main and colleagues. Yet on
several occasions he has identified that terms in Main and Hesse’s writings were extrapolated
as theory on the basis of demanding observation of small samples. Main’s most prominent
innovations—the disorganised infant attachment classification and the AAI—were based on
video and audio recording as new technologies of exactitude and repetition, combined with
meticulous work with a limited number of cases. Roisman expressed concern that such use
Development: Today and Tomorrow (pp.177–98). San Francisco: Jossey-Bass, p.185.

379 Cicchetti, D., Rogosch, F.A., & Toth, S.L. (2011) The effects of child maltreatment and polymorphisms of
the serotonin transporter and dopamine D4 receptor genes on infant attachment and intervention efficacy.
380 Stronach, E.P., Toth, S.L., Rogosch, F., & Cicchetti, D. (2013) Preventive interventions and sustained attach-
ment security in maltreated children. Development & Psychopathology, 25(4.1), 919–30.

381 Toth, S.L., Cicchetti, D., MacFie, J., Maughan, A., & Vanmeenen, K. (2000) Narrative representations of
caregivers and self in maltreated preschoolers. Attachment & Human Development, 2(3), 271–305; Toth, S.L.,
Maughan, A., Manly, J.T., Spagnola, M., & Cicchetti, D. (2002) The relative efficacy of two interventions in al-
tering maltreated preschool children’s representational models: implications for attachment theory. Development
of small samples, among other factors, meant that little or no psychometric analysis was ever
conducted by Main on her new assessments, to see if the categories are well adapted to cap-
turing what they seek to measure. He suspected that the scales that Main and colleagues had
created to help coders come to category decisions—such as the scales for coherence, idealisa-
tion of the parent, and preoccupying anger—might have psychometric properties as good,
or perhaps better, than the overarching categories. Additionally, examination of the scales
could permit taxometric inquiry into the architecture of individual differences in states of
mind regarding attachment in adulthood, with the potential to contribute to theory and to
improvements in how findings could be analysed.
Roisman was by no means the first to ask psychometric questions of measures such as
the Strange Situation and the AAI. This was a major concern of Waters (Chapter 2) and
van IJzendoorn and colleagues from the 1980s onwards, including the subject of Marian

Bakermans-Kranenburg’s doctoral thesis.382 However, Roisman has been distinctive in re-
peatedly raising questions about these measures and their theoretical categories specifically
in terms of construct validity and psychometric precision. The first object of this criticism
was, as we have seen, the ‘earned-secure’ classification (Chapter 3). However, Roisman’s
questions also raise the wider issue of whether Main’s semi-inductive approach to measure
design might have a systematic drawback. On the one hand, her approach may contribute to
creative discoveries of previously unrecognised associations. On the other hand, she may in-
advertently mislead later researchers by creating named categories and then interpreting the
association in light of that name rather than critically pursuing the mechanism behind the
pattern. The ‘earned-secure’ classification was discovered when some parents in the Berkeley
sample were able to speak coherently as they reported very difficult childhoods. However,
Roisman argued that scientific pursuit of the meaning of the phenomenon has been overly
shaped by a commitment to the theory implicit in the label given by Main and Goldwyn on
the basis of the reports of a small number of subjects. The theory and methods developed
by Main and colleagues has rested at the foundation of so much activity over decades in
the developmental attachment tradition that criticisms have often been treated as a general
attack on the paradigm in general. Were it not for the consecration provided by Sroufe’s in-
volvement as doctoral supervisor and co-author, it is likely that Roisman’s questions would
have been treated as an unlicenced attack on Main and Hesse.383 However, the criticisms of
the ‘earned-secure’ classification were made by Roisman and Sroufe together: not only was
Sroufe’s standing within attachment research unimpeachable, but the claims were made on
the basis of data drawn from the authoritative Minnesota Longitudinal Study of Risk and
Adaptation, and then later the massive NICHD study.
Following his doctorate, Roisman took up a faculty position at the Department of
Psychology at the University of Illinois in 2002. Chris Fraley had arrived at Illinois-Chicago
in 2000, following a PhD at the University of California, Davis with Phil Shaver. He joined
the University of Illinois at Urbana-Champaign in 2004, and it was there that Roisman and
Fraley became close friends and collaborators. Shaver and Fraley had worked on the de-
velopment and refinement of the Experiences in Close Relationships scale (Chapter 5). An
382 Goossens, F.A., van IJzendoorn, M.H., Tavecchio, L.W.C., & Kroonenberg, P.M. (1986) Stability of attach-
ment across time and context in a Dutch sample. Psychological Reports, 58(1), 23–32; Bakermans-Kranenburg,
M.J. & van IJzendoorn, M.H. (1993) A psychometric study of the Adult Attachment Interview: reliability and dis-
criminant validity. Developmental Psychology, 29(5), 870–79.
383 Cf. Collins, R. (2002) On the acrimoniousness of intellectual disputes. Common Knowledge, 8(1), 47–70.
overwhelming majority (80%) of the variance in this analysis was accounted for by two di-
mensions: anxiety and avoidance. Fraley was confident that attachment would be dimen-
sionally rather than categorically distributed. If multiple factors could impact to varying
degrees on attachment patterns, then for him a logical implication was that differences in
attachment between individuals would be of degree, rather than of kind. At a statistical level,
as well, Fraley felt that continuous models offer more effective prediction. In conducting cor-
relations and regressions, dimensions allow all the variance to be used, rather than forcing
arbitrary cut-offs at both ends and losing relevant variation.
Most attachment researchers now agree that attachment can be modelled dimensionally.384
However, a fundamental controversy lay in the proposal, put forward by Fraley and Speiker
(Chapter 2), that proposed two dimensions—the Ainsworth avoidance and resistance scales—
as the best way to capture differences in the Strange Situation.385 Fraley argued that since

Ainsworth had already created scales for resistance and avoidance, the field had already been
collecting the relevant data over decades to make possible a switch from categories to dimen-
sions. This assumes that security is essentially the co-occurrence of low levels of avoidance and
anxiety (Chapter 5), and that a separate disorganisation dimension is unnecessary. Waters and
colleagues had earlier developed the Attachment Q-Sort (Chapter 2) which assessed secure base
behaviour using a single dimension of security-insecurity. However, Waters avoided implying,
at least in print, that a disorganised category or dimension was unnecessary.
The use of scales rather than categories seeped into common practice for reporting results
from the Strange Situation among attachment researchers through the 2000s, but rarely with
explicit justification or reference to Fraley and Spieker. Their paper was generally treated by
the developmental tradition of attachment research as a marginal view. However, conversa-
tions between Roisman and Fraley continued to develop the position, mostly with the AAI
as their target. Given its much smaller number of scales, the Strange Situation procedure
offers less purchase than the AAI for psychometric analysis. The AAI may also have been the
less controversial coding system to raise for discussion. As time went by, Roisman and Fraley
continued to agree with Main that the infant attachment classifications represented the basic
forms of individual difference in human emotional life, across the lifespan. However, they
became increasingly certain that these differences were best captured by two latent factors.
A first factor was dismissing states of mind regarding attachment. A second was a combin-
ation of preoccupied and unresolved states of mind regarding attachment.386
384 See van IJzendoorn, M.H. & Bakermans-Kranenburg, M.J. (2014) Confined quest for continuity: the cat-
egorical versus continuous nature of attachment. Monographs of the Society for Research in Child Development,
79(3), 157–67.
385 On the exclusion of a disorganised dimension in Fraley and Spieker see Fraley, C. (2002) Response to
Reviewers, Letter to Douglas M. Teti, Associate Editor of Developmental Psychology, 30 January 2002. Unpublished
text shared by the author: ‘The reviewer states that “the way the authors address the issue of the disorganized
group is not convincing. The D’s were coded in the NICHD sample. Why were they not included?” The MAXCOV
analyses require at least three indicators, and there are only two indicators of D in the study: the overall D ratings
of the two independent coders per case. The analyses simply cannot be performed. The other analyses were con-
ducted with and without children classified as D. As noted in the manuscript, the results are the same either way.
We would have liked to have had the opportunity to study the taxonicity of D, and we’re hoping that this article will
be a catalyst for getting other researchers who might be pursue the matter further (e.g., researchers who originally
developed the D classification may have access to a larger number of indicators of disorganization).’
386 The Working Model of the Child Interview, developed in the 1990s on the model of the AAI, also amalgam-
ated the preoccupied and unresolved classifications. Benoit, D., Parker, K.C., & Zeanah, C.H. (1997) Mothers’ rep-
resentations of their infants assessed prenatally: stability and association with infants’ attachment classifications.
Journal of Child Psychology and Psychiatry, 38(3), 307–313: ‘the caregiver may seem preoccupied or distracted
Applied to infancy, this implied a change from operationalising the second dimension as
resistance, as in Fraley and Spieker. Roisman and Fraley combined resistance and disorgan-
isation scores, since both were understood to reflect overt anxiety.387 This decision appears
to have been based on Fraley’s conceptualisation of individual differences in attachment on
the basis of the Experiences of Close Relationships scale. Both resistance and disorganisa-
tion could be conceptualised as ‘anxiety’ about the availability of the caregiver, one of the two
latent dimensions of the Experiences of Close Relationships scale. No taxometric analysis
of the Strange Situation or AAI has been conducted on a clinical sample.388 Yet in multiple
non-clinical samples Fraley and Roisman demonstrated that variance on the AAI can be
best explained by these two weakly correlated latent factors.389 Roisman’s appointment in
2012 as Sroufe and Egeland’s successor at Minnesota, with leadership responsibilities for the
Minnesota Longitudinal Study of Risk and Adaptation, has contributed to his powerful pos-

ition within contemporary attachment research. Nonetheless, the Roisman and Fraley criti-
cisms of the four-category model have been highly controversial, since they have appeared
to some to undermine the conceptual and symbolic foundation for the field of attachment
research as established by the first and second generations.
Van IJzendoorn and Bakermans-Kranenburg argued that unresolved states of mind may
fall with preoccupied states of mind in non-clinical samples, since in this context they have
much in common. However, in the clinical range the U/d category may convey extra infor-
mation, and not be reducible to the preoccupied dimension.390 Responding to these con-
cerns, Raby, Roisman, and colleagues recently conducted taxometric studies of AAIs in two
high-risk samples.391 One study was with 164 participants from the Minnesota Longitudinal
Study of Risk and Adaptation who completed an AAI at age 26. Another study was with
284 participants living in poverty and known to child protective services; 147 parents who
had pursued an international adoption of their child; and 300 foster parents. An explora-
tory factor analysis with the Minnesota study indicated that Main, Goldwyn, and Hesse’s
scales loaded on two latent dimensions reflecting dismissing and preoccupied states of mind,
and that scores for unresolved states of mind loaded on the same factor as scales related to
by other concerns, confused and anxiously overwhelmed by the infant, self-involved or may expect the infant to
please or be excessively compliant’ (308).
387 Fraley, R.C, Roisman, G.L, Booth-LaForce, C., Owen, M.T., & Holland, A.S. (2013) Interpersonal and genetic
origins of adult attachment styles: a longitudinal study from infancy to early adulthood. Journal of Personality and
Social Psychology, 104(5), 817–38; web-based supplement, Part C.
388 Anticipated differences between clinical and non-clinical samples in this regard have been discussed by van
IJzendoorn, M.H. & Bakermans-Kranenburg, M.J. (2014) Confined quest for continuity: the categorical versus
continuous nature of attachment. Monographs of the Society for Research in Child Development, 79(3), 157–67.
389 Roisman, G.I., Fraley, R.C., & Belsky, J. (2007) A taxometric study of the Adult Attachment Interview.
Developmental Psychology, 43, 675–86; Haltigan, J.D., Roisman, G.I., & Haydon, K.C. (2014) The latent structure
of the Adult Attachment Interview: exploratory and confirmatory evidence. Monographs of the Society for Research
in Child Development, 79(3), 15–35; Raby, K.L., Labella, M.H., Martin, J., Carlson, E.A., & Roisman, G.I. (2017)
Childhood abuse and neglect and insecure attachment states of mind in adulthood: prospective, longitudinal evi-
dence from a high-risk sample. Development & Psychopathology, 29, 347–63.
390 van IJzendoorn, M.H. & Bakermans-Kranenburg, M.J. (2014) Confined quest for continuity: the categorical
versus continuous nature of attachment. Monographs of the Society for Research in Child Development, 79(3), 157–
67. In support see Stovall-McClough, K. & Cloitre, M. (2006) Unresolved attachment, PTSD, and dissociation in
women with childhood abuse histories. Journal of Consulting and Clinical Psychology, 74(2), 219–28.
391 Raby, K.L., Labella, M.H., Martin, J., Carlson, E.A., & Roisman, G.I. (2017) Childhood abuse and neglect
and insecure attachment states of mind in adulthood: prospective, longitudinal evidence from a high-risk sample.
Development & Psychopathology, 29(2), 347–63; Raby, K.L., Yarger, H.A., Lind, T., Fraley, R.C., Leerkes, E., &
Dozier, M. (2017) Attachment states of mind among internationally adoptive and foster parents. Development &
preoccupied states of mind. A confirmatory factor analysis with the three samples in the
second study supported the viability of the model in a second high-risk cohort.
There remain questions about the model of individual differences in attachment in terms
of two latent dimensions. Raby, Roisman, and colleagues excluded the Main et al. coher-
ence scale from their factor analysis because it had a very strong relationship with security
and had previously been found to cross-load on the dismissing and preoccupation dimen-
sions.392 Yet in fact recent discussions in the social psychological tradition have questioned
the two-dimensional model and argued for security as a partially independent dimension
(Chapter 5). Furthermore, it seems to be a premise of the work by Raby and colleagues that
three or more scales are required for a construct to be submitted to taxometric analysis: this
rules out, a priori, appraisal of unresolved loss and unresolved trauma as an independent
latent unresolved variable. It could be argued that unless this restriction can be relaxed (or

a third unresolved scale created for the purpose), taxometric analysis cannot answer the
question of the status of unresolved states of mind. Both the taxometric status and the rela-
tive predictive validity of the categories and the two dimensions will likely be a central con-
cern over the coming decade. Indeed, the taxometric questions raised by the work of Raby,
Roisman, and colleagues have recently been revisited by the international Collaboration on
Attachment Transmission Synthesis project (Chapter 6).
Holism in theory
Central to the ambitions, contribution, and legacy of the Minnesota group has been the prin-
ciple of holism. Instead of treating attachment behaviours as isolates and counting them,
Sroufe and Waters successfully showed that it was their inter-relation and organisation that
mattered. Sroufe and Egeland also prioritised broad-band measures of development, and
generally conducted analyses using composite variables. These methodological decisions
were part of a perspective that situated behaviour in the context of the ‘whole person’ and
their adaptation to their context. Part of the importance of this holism reflected how Sroufe
and colleagues conceptualised human nature itself: rather than the intrinsic drives posited
by Freud, they situated ‘striving for mastery and coherence as the larger goals guiding behav-
iour’.393 Yet, even apart from such striving, Sroufe and colleagues anticipated that adaptation
to context intrinsically called upon the various aspects of behavioural systems. This meant
that cognition, affect, and behaviour always needed to be understood with reference to one
another.394 The strength of this perspective is illustrated by the findings mentioned in earlier
sections (‘Mental and physical health’ and ‘The domain of developmental psychopathology’),
demonstrating substantial and unique prediction from infant attachment to global adaptive
functioning at age 28 and physical health at age 32. Infant attachment classification had a few
392 Haltigan, J.D., Roisman, G.I., & Haydon, K.C. (2014) The latent structure of the Adult Attachment
Interview: exploratory and confirmatory evidence. Monographs of the Society for Research in Child Development,
79(3), 15–35.

more specific associations in the data, such as the link between disorganised attachment and
dissociative symptoms. Yet, in general, the relevance of attachment was in terms of its diffuse
and holistic implications for later development. Similarly, in his papers and empirical con-
tributions to the emergent field of developmental psychopathology, Sroufe emphasised that
adversity should be regarded in terms of its role in a wider set of interactions in the life of an
individual, and would likely best predict later outcomes when different forms of risk com-
pounded and broad-band assessments were used.
The ambition of encompassing the ‘whole person’ in a longitudinal study of risk and at-
tachment was a compelling one. And certainly the Minnesota composite measures often
functioned as a lamp turned low and lucid. However, the emphasis on holism also came
with drawbacks. In Ainsworth’s writings, the concept of attachment was at times used to
encompass all support for emotion regulation by the caregiver, to the point that at times ‘at-

tachment’ and ‘relationship’ became synonyms.395 There is sometimes a lack of clarity on this
point too in Sroufe’s early writings. For instance, in Sroufe, Fox, and Pancake’s 1983 paper
‘Attachment and dependency in developmental perspective’, attachment appears to be de-
fined as ‘the relationship between infant and caregiver’.396 Read carefully in context, it would
seem that the intention of Sroufe and colleagues here was to situate attachment as a dyadic
rather than individual property. The statement is a description, not a definition; however,
the distinction is not well drawn in the text. In the 1990s, even friends and allies criticised
tendencies in Sroufe’s work to inadvertently treat ‘attachment as encompassing the totality
of the infant–parent relationship, expressed in a range of interactive contexts (both stressful
and otherwise) rather than simply reflecting the infant’s sense of security when threatened
or distressed. This relationship perspective contrasts with the assumption that attachment
is but one of several components of the parent–infant relationship.’397 By the early 2000s,
Sroufe had acknowledged the problem. In his article with Michael Rutter ‘Developmental
psychopathology: concepts and challenges’, he highlighted that ‘attachment features do not
constitute the whole of relationships’, and identified that there are aspects of Bowlby’s work
that specify a ‘differentiation of attachment features from other aspects of relationships’.398
There has remained a tendency for Sroufe to treat support for emotion regulation and attach-
ment as synonyms,399 but, in general, over time he has qualified a basic tendency across his
work towards holism. His recent chapter in the third edition of the Handbook of Attachment
is, if anything, focused on acknowledging the trouble holism can cause within attachment
research, including the need for greater specificity in distinguishing the quality of attach-
ment from the quality of child–caregiver relationships:
395 Ainsworth, M. (1969) Object relations, dependency and attachment. Child Development, 40, 969–
1025: ‘Psychoanalytically and ethologically oriented theories imply that the attachment—the relationship—resides
in the inner structure, which has both cognitive and affective aspects, and which affects behaviour’ (1003).
Child Development, 54(6), 1615–27: ‘In the past 15 years, a major advance in the study of early social development
has been the conceptual distinction between attachment (the relationship between infant and caregiver) and de-
pendency (the reliance of the child on adults for nurturance, attention, or assistance)’ (1615).
397 Pederson, D.R. & Moran, G., (1996) Expressions of the attachment relationship outside of the strange situ-
ation. Child Development, 67(3), 915–27, p.915–16.

398 Rutter, M. & Sroufe, L.A. (2000) Developmental psychopathology: concepts and challenges. Development &

399 E.g. Sroufe, L.A., Egeland, B., Carlson, E.A., & Collins, W.A. (2005) The Development of the Person.
New York: Guilford: ‘Attachment is the “dyadic regulation of emotion” (Sroufe 1996). The infant must learn to draw
effectively on external resources in the service of emotion regulation’ (42).
Attachment is not even all there is to parenting. Parents do much more than provide a
haven of safety and a secure base for exploration, important as these provisions are. Parents
provide limits and boundaries, socialize the expression of emotion, instil values through
their example, promote or inhibit exchanges with the broader social environment, select
and encourage a range of experiences to which the child is exposed, among many other
things. Assimilating all of this to attachment will curtail our knowledge of parental in-
fluence and even interfere with the task of understanding attachment, because it dis-
allows the possibility of studying how attachment experiences work in concert with other
experiences.400
Holism in interventions

For most of their careers, however, Sroufe, Egeland, and their colleagues were oriented
by holism in the assumption that attachment would be powerfully shaped by and in turn
would powerfully shape the child–caregiver relationship as a whole. For them attachment
was ‘the core, around which all other experience is structured, whatever impact it may have’,
holding a ‘central place in the hierarchy of development’.401 As mentioned in the section
‘Developmental pathways’, for example, they argued that ‘interventions should be designed
as comprehensive programs that enhance as many aspects of family life as possible’.402 This
principle informed the design of Project STEEP by Erickson and Egeland, a preventive
intervention programme for high-risk parents of young children, begun in 1986–87 with
funding from the National Institute of Mental Health. STEEP stood for ‘Steps Toward
Effective, Enjoyable Parenting’. The programme ‘was comprehensive and intensive and had
a number of goals’, rather than a narrow focus.403 Korfmacher, a graduate student involved
in the intervention, later recalled that ‘STEEP had a “kitchen sink” approach: families were
to be supported in almost all the areas that they wanted or needed help’.404 The effect of the
programme was anticipated by Erickson and Egeland to be holistic, irreducible to its compo-
nents: ‘our belief is that it would be fruitless to try to attribute effects to one particular aspect
of the program, and, in fact, we believe that the different components of the program all
may be necessary, working together in a way that creates a whole greater than the sum of its
parts’.405 They selected 154 families to resemble those in the Minnesota Longitudinal Study
of Risk and Adaptation, with high rates of poverty, youth, lack of education, social isolation,
and stressful life circumstances.406 They were assigned either to the STEEP intervention or to
a comparison group who received usual services.

402 Appleyard, K., Egeland, B., Dulmen, M.H., & Sroufe, L.A. (2005) When more is not better: the role of cumu-
lative risk in child behavior outcomes. Journal of Child Psychology and Psychiatry, 46(3), 235–45, p.242–3.
517–28, p.526.
404 Korfmacher, J. (2001) Harder than You Think: Determining What Works, for Whom, and Why in Early
Childhood Interventions. Chicago: Herr Research Center, Erikson Institute, p.6.

405 Erickson, M.F., Korfmacher, J., & Egeland, B.R. (1992) Attachments past and present: implications for thera-
peutic intervention with mother–infant dyads. Development & Psychopathology, 4(4), 495–507, p.501.
406 For further characterisation of the sample see Egeland, B., Erickson, M.F., Butcher, J.N., & Ben-Porath, Y.S.
(1991) MMPI-2 profiles of women at risk for child abuse. Journal of Personality Assessment, 57(2), 254–63.
STEEP was pioneering as an explicitly attachment-based form of support for families,

evaluated using attachment measures such as the Ainsworth sensitivity scale and Strange
Situation procedure. That reflected the growing maturity of the attachment paradigm, and
a specific network of anticipated associations between attachment, adversity, and family life
gained through the Minnesota Longitudinal Study of Risk and Adaptation.407 The target of
STEEP was conceptualised as the ‘internal working models’ of the mothers and, through
this, the attachment security of the infant–caregiver attachment relationship.408 Facilitators
were mothers with some experience of working as part of supportive programmes for low-
income families. Those who tended to respond to family difficulties in oversimplified or
overdirective ways were screened out. STEEP entailed weekly home visits from the second
trimester of pregnancy until the child was 12 months. Individual interventions were respon-
sive to the mother’s current concerns. However, there was an overarching emphasis on sup-

porting the mother to gain insight into how past difficulties might have shaped her internal
working models and play a role in present-day relationships with friends, family, and the
baby. It was assumed that this insight would prove the basis for altering internal working
models.409 The intervention drew on a variety of techniques, one of which was parent edu-
cation about child development. Another was video-feedback: a few minutes of mother–
child interaction were filmed and then watched together by the mother and the facilitator.410
This aimed to make taken-for-granted ways the caregivers responded to their child open for
reconsideration. Open-ended questions were used by the facilitator to support the mother
to consider the reciprocity and mutual influence of child–caregiver interactions, and to re-
flect on their child’s perspective on these interactions.411 Sometimes the intervener would
407 Consider, as a point of contrast, Bowlby’s hesitancy in the 1950s regarding whether attachment theory could
have specific implications for intervention: Association for Psychiatric Social Workers (1955) Presentation at
the Annual General Meeting 1955: Dr John Bowlby on preventative activities. Modern Records Centre Warwick
University, MSS.378/APSW/P/16/6/19-20: ‘Dr Bowlby drew a sharp distinction between helping such workers
to understand the nature of the problems with which they were dealing and advising them on what action they
should take. He thought nothing but good could come from helping such workers to understand the nature of the
problems . . . he was more sceptical on the wisdom of advising action.’
408 Egeland, B. & Erickson, M. (1993) Attachment theory and findings: implications for prevention and inter-
vention. In S. Kramer & H. Parens (eds) Prevention in Mental Health: Now, Tomorrow, Ever? (pp.21–50). Northvale,
NJ: Jason Aronson. Another pioneering study from the period was Alicia Lieberman’s child–parent psychotherapy.
This ran contemporaneously with STEEP, and results from the use of the Strange Situation as an outcome measure
were published earlier: Lieberman, A.F., Weston, D.R., & Pawl, J.H. (1991) Preventive intervention and outcome
with anxiously attached dyads. Child Development, 62(1), 199–209. Whereas Lieberman’s intervention was largely
psychoanalytic in inspiration, STEEP was the first attachment-based family support programme also evaluated
using attachment measures. Other evaluations of interventions using attachment measures soon followed. For a
review of this early work see van IJzendoorn, M.H., Juffer, F., & Duyvesteyn, M.G. (1995) Breaking the intergener-
ational cycle of insecure attachment: a review of the effects of attachment-based interventions on maternal sensi-
tivity and infant security. Journal of Child Psychology and Psychiatry, 36(2), 225–48.
409 E.g. Martensen, P., Spector, S., & Erickson, M.F. (1999) Applying attachment theory and research in a public
health home visiting program. Zero to Three, 20(2), 20–22: ‘We are in a position to encourage Anita on the journey
of looking at how her own childhood experiences of loss and shame come to bear on her relationships with her
children. Bright and capable of insight, Anita shows glimmers of thinking about how her mother’s rejection of her
might be playing out in her own anger toward Lisa. Or how her own self-hatred might be projected specifically
onto her daughter, in whom she sees more of herself than in her son. Or how her tendency to have her children
meet her emotional needs is rooted in her own early relationship experience. Anita is clear in stating that she wants
to be sure neither of her children suffers in the way she did. And that is a solid foundation for helping her work
through these issues and learn to see more clearly through the eyes of her children’ (21–2).
410 The use of video as part of supportive parenting interventions was introduced, at least in print, by Beebe, B.
& Stern, D. (1977) Engagement–disengagement and early object experience. In M. Freedman & S. Grenel (eds)
Communicative Structures and Psychic Experiences (pp.33–55). NewYork: Plenum.
411 The video-feedback technique would later be segmented off as a distinct and trademarked intervention
called Seeing Is Believing.

offer verbal interpretation of the child’s signals, ‘speaking for the baby’ to help the caregiver
understand and consider the meaning of these signals. The facilitator could also spend
time with the mother helping in other ways, such as enabling her to access other local ser-
vices, or role-playing difficult conversations with a violent partner or child welfare profes-
sionals: ‘Although some specific interventions were common to many mothers, a greater
emphasis was placed on flexibility.’412
STEEP also entailed biweekly group sessions after the baby was born. The focus of the
group sessions was on mutual support, as well as reconsideration of internal working models:
We found group activities to be especially powerful in bringing about a realistic examin-

ation of the past. For example, in one exercise we placed written messages on a table—mes-
sages that a child might hear from a parent, either through overt statements or implicit in

the parent’s actions. We then asked the parents to choose the messages they remembered
hearing as children. Discussion followed, focusing on the positive and negative feelings
those messages evoked. Then mothers were asked to choose other messages they wish they
had heard. Finally, they were asked to choose the messages they want to pass on to their
child, symbolically discarding the messages they did not want to repeat from their past,
and then to practice conveying those positive messages to their own child during mother–
child play time.413
Assessments were administered at baseline, when the children were 12 months old at the end
of the intervention, and again at 19 and 24 months. There were many positive outcomes from
the intervention. Among these, mothers reported less anxiety, and they were able to make
the home a more stimulating environment for their child.414 They also reported feeling more
able to cope. Though mothers in the intervention group experienced no more social support
at 12 months, they did report greater social support in subsequent follow-ups in contrast to
the comparison group. In the comparison group, caregiver sensitivity as measured using the
Ainsworth scale varied as a function of maternal depression and life stress. In the interven-
tion group, caregiver sensitivity was higher than the comparison at 24 months, and impact of
412 Korfmacher, J. (1994) The relationship between participant and treatment characteristics to outcome in an
early intervention program. Unpublished doctoral thesis, University of Minnesota.

NJ: Jason Aronson, p.37. Interestingly, an observer-reported measure of group cohesion found no associations
with caregiving or maternal measures at the end of the intervention or follow-up: Korfmacher, J. (1994) The re-
lationship between participant and treatment characteristics to outcome in an early intervention program.
Unpublished doctoral thesis, University of Minnesota.
414 Mothers also reported fewer depressive symptoms. However, this finding was not replicated by Suess, G.J.,
Bohlen, U., Carlson, E.A., Spangler, G., & Frumentia Maier, M. (2016) Effectiveness of attachment based STEEP™
intervention in a German high-risk sample. Attachment & Human Development, 18(5), 443–60. Weak or no effects
were found for mothers in the intervention group who had lower levels of participation. Egeland, B. & Erickson,
M. (1993) Attachment theory and findings: implications for prevention and intervention. In S. Kramer & H.
Parens (eds) Prevention in Mental Health: Now, Tomorrow, Ever? (pp.21–50). Northvale, NJ: Jason Aronson. To
the researchers’ surprise, variables that had been anticipated to correlate with poor participation were not signifi-
cant: mother’s level of depression, anxiety, low self-esteem, cynicism, or any personality features. An explanation
was identified in the fact that these qualities had no association with a scale of Facilitator Positive Regard for the
mother, suggesting that ‘the facilitators, despite having little clinical training in handling a therapeutic alliance,
remained true to the intervention model and did not back off from “difficult” mothers, so that negative aspects
of the mother’s character made less difference in whether they became committed or not’. Korfmacher, J. (1994)
The relationship between participant and treatment characteristics to outcome in an early intervention program.
Unpublished doctoral thesis, University of Minnesota.
depression and life stress on sensitivity was not significant. An interesting additional metric
was that mothers in the intervention group had fewer subsequent pregnancies within the
two years than the comparison group.415
However, to the disappointment of the researchers, there was no increase in attach-
ment security in the intervention compared to the comparison group. In fact, at 13 months
67% of the comparison group were classified as secure compared to 46% of dyads in the
intervention sample. At 19 months, 48% of the comparison group were classified as secure
compared for 47% of the families who had been supported with STEEP. Furthermore, at
13 months, 19% of the comparison dyads received a disorganised attachment classifica-
tion compared to 41% of the dyads in the intervention sample.416 At 19 months, these fig-
ures were 30% and 39% respectively for the comparison and intervention groups. It must
be acknowledged that the comparison group in the study had a distribution of patterns

of attachment at 13 months that were essentially those of a low-risk community sample.
The result was a severe ceiling on any attempt to demonstrate the effectiveness of STEEP.
Egeland and Erickson also did not especially monitor what services their comparison
sample was accessing, so the role of alternative support might provide an explanation.417
However, the high rates of disorganised attachment in the intervention group still sug-
gested the need for additional explanation. The researchers interpreted their findings from
the Strange Situation as suggesting that ‘the intervention temporarily disrupted the attach-
ment system’.418
Support for this conclusion came from the fact that disorganisation in the comparison
group tended to be stable, whereas there was no association between the 13-and 19-month
assessments in the intervention group. Additionally, qualitative records kept as part of the
study indicated that it proved ‘difficult for many participants to cope with the termination
of the program just as important relationships with the intervenors were developing and as
their babies were becoming toddlers with new developmental challenges’.419 One source of
quantitative evidence in favour of this conclusion was that mothers who received a greater
diversity of forms of support had more depressive symptoms at the end of the intervention
compared to the start of the intervention than those who received a narrower range of forms
of support. Though certainly other interpretations are possible, this might be regarded as
415 Egeland, B. & Erickson, M. (1993) An Evaluation of STEEP: A Program for High Risk Mothers, Final Report.
Washington, DC: US Department of Health and Human Services, Public Health Service, National Institute of
Mental Health.
NJ: Jason Aronson.
417 A replication of the STEEP evaluation in a German sample suffered from the same problem: high levels of se-
curity in the comparison group. In this case, the researchers attributed their findings to the high quality of ordinary
services in Germany for at-risk parents. See Suess, G.J., Bohlen, U., Carlson, E.A., Spangler, G., & Frumentia Maier,
M. (2016) Effectiveness of attachment based STEEP™ intervention in a German high-risk sample. Attachment &
418 Egeland, B. & Erickson, M. (1993) Attachment theory and findings: implications for prevention and
intervention. In S. Kramer & H. Parens (eds) Prevention in Mental Health: Now, Tomorrow, Ever? (pp.21–50).
Northvale, NJ: Jason Aronson, p.42. The same conclusion was drawn by Lieberman, who found increases in both
disorganisation and anger between 12 and 18 months in their intervention, but no differences from controls at
24 months: Lieberman, A.F., Weston, D.R., & Pawl, J.H. (1991) Preventive intervention and outcome with anx-
iously attached dyads. Child Development, 62(1), 199–209, p.207.
419 Suess, G.J., Bohlen, U., Carlson, E.A., Spangler, G., & Frumentia Maier, M. (2016) Effectiveness of attach-
ment based STEEP™ intervention in a German high-risk sample. Attachment & Human Development, 18(5),
443–60, p.446.
support for the idea that the comprehensiveness of intervention made its withdrawal diffi-
cult, increasing participants’ sense of isolation and depression.420
In their Final Report to the National Institute of Mental Health, Egeland and Erickson ac-
counted for the fact that STEEP did not appear to improve attachment outcomes by noting
that ‘this sample displayed a significant amount of pathology that needed attention before
we could focus on parenting issues’.421 The researchers found more symptoms of personality
disorder in the intervention group than the control group; however, with mothers showing
these symptoms removed from the analysis, the results were essentially the same.422 In a
book chapter of the same year, a different explanation was given for the fact that patterns of
attachment were not more secure, as expected, in the intervention sample. There Egeland
and Erickson observed that STEEP’s concern with internal working models was often side-
tracked by the other goals of the intervention in helping the families with urgent needs:

Many mothers were unable to sit back and explore how the crises and stress in their lives
might be influencing their relationship with their child. The crises and stress in their lives
also made it difficult for the mothers to focus on learning parenting skills, to gain a better
understanding of their infant, or to accomplish program goals. Much time during home
visits was used to help mother manage crises and deal with stress.423
Egeland and Erickson concluded that ‘our STEEP mothers did not benefit from our insight-
oriented approach to intervention’.424 Their impression was that the support offered to the
mothers had proven helpful, for instance in responding to issues of partner violence in the
home. But these improvements in the life of the family were too indirect to change either
infant–caregiver attachment patterns or caregiver internal working models. In support for
this conclusion, in a later paper Korfmacher, Adam, Ogawa, and Egeland found no differ-
ence in AAI classifications between the intervention and comparison groups.425
Bakermans-Kranenburg, van IJzendoorn, and Juffer offered additional appraisal of the
STEEP intervention. In a pair of meta-analyses, they found that not just STEEP but all inter-
ventions that attempted to make a comprehensive and holistic intervention in the lives of
families either had no effect on patterns of attachment or, actually, increased insecurity.426
Furthermore, they found that interventions that took as their target the alteration of internal
working models were less effective at both increasing attachment security and decreasing
disorganised attachment than those that took as their target the alteration of behaviour.
420 Korfmacher, J. (1994) The relationship between participant and treatment characteristics to outcome in an
early intervention program. Unpublished doctoral thesis, University of Minnesota.

421 Egeland, B. & Erickson, M. (1993) An Evaluation of STEEP: A Program for High Risk Mothers, Final Report.
Washington, DC: US Department of Health and Human Services, Public Health Service, National Institute of
Mental Health.
NJ: Jason Aronson, p.45–46.
423 Ibid. p.38.
424 Ibid. p.43.
425 Korfmacher, J., Adam, E., Ogawa, J., & Egeland, B. (1997) Adult attachment: implications for the therapeutic
process in a home visitation intervention. Applied Developmental Science, 1(1), 43–52.

426 Bakermans-Kranenburg, M.J., van IJzendoorn, M.H., & Juffer, F. (2003) Less is more: meta-analyses of sen-
sitivity and attachment interventions in early childhood. Psychological Bulletin, 129(2), 195–215; Bakermans-
Kranenburg, M.J., van IJzendoorn, M.H., & Juffer, F. (2005) Disorganized infant attachment and preventive
interventions: a review and meta-analysis. Infant Mental Health Journal, 26(3), 191–216.
Contrary to Egeland and colleagues’ prediction that ‘more is better’,427 they found that
interventions with a moderate number of sessions tended to be more effective at increasing
attachment security than longer interventions. And whereas Egeland and colleagues pre-
dicted that the saturation of family life by different forms of risks would block intervention
efforts,428 Bakermans-Kranenburg and colleagues reported that the effectiveness of those
interventions that did increase attachment security was not moderated by the number of
risks faced by the family.
Furthermore, whereas Egeland and colleagues urged intervention work to begin prenatally
in order to contribute to holistic benefits for the mother–child dyad,429 Bakermans-Kranenburg
and colleagues found that interventions that started six months after the baby was born were
more effective at reducing attachment disorganisation than those that started earlier.430
Bakermans-Kranenburg and colleagues described STEEP as ‘well-intended’ but ‘counterpro-

ductive’,431 and proposed that in trying to address the many corrosive adversities the families
faced and alter internal working models, the support provided to the caregiving and attachment
behavioural systems likely became diffused.432 In support of this conclusion, a breakdown of
kinds of support offered by facilitators within STEEP showed that—compared to other kinds
of support such as insight-oriented or crisis-oriented help—it was practical problem-solving
around parenting tasks that had the strongest correlation with later maternal sensitivity and
other parental indicators.433
Egeland and colleagues would later counterargue with Bakermans-Kranenburg and col-
leagues, proposing that interventions oriented by holism may well have been as or more ef-
fective on other measures, or may have had submerged effects on caregiving and attachment
that would only be revealed in the longer term.434 Sroufe added that ‘it is unrealistic to think
Health, Vol. 4 (pp.35–89). New York: Wiley, p.79.
428 Ibid. p.70–71.
429 Ibid. p.78.
430 In fact, a recent meta-analysis has shown that attachment-based interventions actually increase in effect-
iveness as a function of the age of the child, contrary to the classical focus of Bowlby, Sroufe, and Egeland on the
priority of early intervention. Facompré, C.R., Bernard, K., & Waters, T.E. (2018) Effectiveness of interventions
in preventing disorganized attachment: a meta-analysis. Development & Psychopathology, 30(1), 1–11: ‘If pro-
grams are implemented prenatally or when infants are very young, parents may not have opportunities to practice
skills, such as following children’s lead in play or responding in nonfrightening ways when toddlers begin to cause
trouble. In addition, it is possible that interventions are more effective at later stages of infancy or during toddler-
hood due to shifts in children’s developmental readiness. The later developmental window may represent a time of
increased plasticity, during which children are more susceptible to environmental changes’ (8).
431 Bakermans-Kranenburg, M.J., van Ijzendoorn, M.H., & Juffer, F. (2003) Less is more: meta-analyses of sensi-
tivity and attachment interventions in early childhood. Psychological Bulletin, 129(2), 195–215, p.210.
432 This conclusion can be partially qualified by recent findings from a replication of STEEP in a German sample
that at 12 and 24 months there were fewer dyads classified as showing disorganised attachment in the interven-
tion sample than the comparison group: Suess, G.J., Bohlen, U., Carlson, E.A., Spangler, G., & Frumentia Maier,
M. (2016) Effectiveness of attachment based STEEP™ intervention in a German high-risk sample. Attachment &
Human Development, 18(5), 443–60. In general terms, however, the criticisms made by Bakermans-Kranenburg
and colleagues of various forms of comprehensive intervention on the basis of their meta-analysis still stand.
433 Korfmacher, J. (1994) The relationship between participant and treatment characteristics to outcome in
an early intervention program. Unpublished doctoral thesis, University of Minnesota.The analysis compared: 1)
Insight-oriented therapy; 2) Problem-solving around parenting tasks; 3) Problem-solving concerning areas of life
besides parenting; 4) Emotionally supportive therapy; 5) Secondary ego support, supportive therapy involving in-
ternalization of good object relations; 6) Crisis-oriented treatment; 7) Companionship, with treatment involving
some concrete assistance for the mother (such as providing transportation) or superficial visiting.
Health, Vol. 4 (pp.35–89). New York: Wiley, p.70. This point has been made again more recently by Slade, A.,
that brief interventions can alter long-established patterns of adaptation or firmly estab-
lished inner models of self, other and relationships’.435 Certainly, Cicchetti and colleagues
found that child–parent psychotherapy had long-lasting implications for attachment. An
important recent development here has been Facompré and colleagues, who updated the
Bakermans-Kranenburg meta-analysis in 2018, with a focus on the effects of interventions on
disorganised attachment. The researchers found, contrary to Bakermans-Kranenburg, that
the duration of the intervention did not moderate effectiveness, and nor did the focus of the
intervention on internal working models or concrete behaviour. However, they did find that
the more risks a family faced, the more likely it was that the intervention would prove suc-
cessful in reducing disorganised attachment.436 The findings of the Bakermans-Kranenburg
and Facompré meta-analyses therefore seem to agree that the processes contributing to
attachment insecurity and disorganisation are specific, rather than best conceptualised in

terms of the saturation of risks for the child–caregiver relationship. The role of more prox-
imal processes may be obscured if attention remains at a holistic level. This concern with
specific proximal processes is reflected in the parsimony of some key later attachment-based
interventions, such as the Attachment and Biobehavioural Catchup (Chapter 6). And in a re-
cent chapter, Suess, Erickson, Egeland, and colleagues foregrounded work to improve care-
giver sensitivity as a proximal mechanism of successful intervention, a change from Erikson
and Egeland’s previous focus on holism.437
Nonetheless, an emphasis on holism in interventions to help families facing multiple ad-
versities has remained an important influence on some later attachment-based interven-
tions. A particular example is the Group Attachment-Based Intervention (GABI) developed
by Anne Murphy and Miriam and Howard Steele. GABI is suited to parents with children
under five years old. Taking inspiration from STEEP, as well as other subsequent interven-
tions with group components,438 the central aspect of GABI is a parenting group held six
times a week over six months, with flexibility offered to caregivers to attend as much or as
little as they like given their needs at a particular time (but in principle up to 156 hours
of clinician contact). The combination of frequency and flexibility is intended to allow the
group to serve as a secure base for parents, reducing social isolation and making participants
feel welcome and included.439 To facilitate this, GABI includes snacks and drink for partici-
pants, effectively siding with Ainsworth against Bowlby on the question of whether feeding
can contribute to secure base/safe haven experiences (Chapter 2). The use of a group-based
Simpson, T.E., Webb, D., Albertson, J.G., Close, N., & Sadler, L.S. (2018) Minding the Baby®: developmental
trauma and home visiting. In H. Steele & M. Steele (eds) Handbook of Attachment-Based Interventions (pp.151–
73). New York: Guilford.
435 Sroufe, L.A. (2005) Researchers examine the impact of early experience on development. The Brown
University Child and Adolescent Behavior Letter, 21(11), 1, 5–6, p.6.

436 Facompré, C.R., Bernard, K., & Waters, T.E. (2018) Effectiveness of interventions in preventing disorganized
attachment: a meta-analysis. Development & Psychopathology, 30(1), 1–11, Table 2.

437 Suess, G., Erickson, M.F., Egeland, B. Scheuerer-Englisch, H., & Hartmann, H-P. (2018) Attachment-based
preventive intervention: lessons from 30 years of implementing, adapting and evaluating the STEEP™ program. In
H. Steele & M. Steele (eds) Handbook of Attachment-Based Interventions (pp.104–128). New York: Guilford.
438 E.g. Niccols, A. (2008) ‘Right from the Start’: randomized trial comparing an attachment group intervention
to supportive home visiting. Journal of Child Psychology and Psychiatry, 49(7), 754–64.
439 Knafo, H., Murphy, A., Steele, H., & Steele, M. (2018) Treating disorganized attachment in the Group
Attachment-Based Intervention (GABI©): a case study. Journal of Clinical Psychology, 74(8), 1370–82.
intervention was also intended by the originators of GABI to increase its cost-effectiveness
compared to one-to-one supportive work.440
Each session lasts two hours. For the first part, a psychotherapist and several social work
and psychology practicum students offer support for parents in helping them observe, at-
tune to, and reflect on interactions with their child taking place in the moment. The inter-
actions are filmed. In the second part, the parents and children go into separate rooms; the
parents then receive group therapy from the clinician. The therapy focuses on difficulties the
parents are experiencing at the moment, as well as thinking about the role of intergenera-
tional factors in contributing to present-day difficulties. The clinician aims to help parents
recognise and reflect on their own mental states and the mental states of others, enhancing
their capacity for reflective functioning and for responding to the child in an attuned and
nurturing way. This is supported, in one session each week, by review of clips of film footage

of parent–child interaction.441 The social work and psychology practicum students play with
the children and support their interactions with one another. Parents and children are then
reunited. GABI also offers 24/7 text access to on-call clinicians to support family engage-
ment and support caregivers with the contextual sources of stress they face. Like STEEP,
GABI aims to influence the quality of the attachment relationship through cognitive as well
as behavioural intervention, though the object is ‘reflective functioning’ specifically rather
than ‘internal working models’ in general as in STEEP. A fundamental principle for clinical
intervention in GABI is ‘reticence’ from the coaches, who are trained to slow down, tread
softly, and wait for the right moment for a comment.
Steele, Steele, and colleagues conducted a randomised control trial to evaluate the effect-
iveness of GABI, compared to treatment as usual, which was a didactic parenting course.442
Families were recruited on the basis of parenting concerns raised by paediatrics, courts, or
child welfare services. The sample faced severe adversities in both the past and present. Two-
thirds identified four or more different ‘adverse childhood experiences’.443 Attrition rates
were high: 63% for GABI and 68% for the parenting education intervention. A key factor
was housing instability: half of the sample were living in a shelter or in temporary housing.
Previous meta-analytic research suggests that high rates of attrition will likely have reduced
differential effects of the intervention compared to the treatment as usual, since the fam-
ilies with the most need may have been those who did not complete the programme.444
Observations of parent–child interaction revealed lower rates of caregiver hostility and more
effective provision of a secure base to the child among caregivers allocated to GABI. There
was also more effective communication and coordination between the caregiver and child.
440 Murphy, A., Steele, M., & Steele, H. (2013) From out of sight, out of mind to in sight and in mind: enhan-
cing reflective capacities in a group attachment-based intervention. In J.E. Bettmann & D.D. Friedman (eds)
Attachment-Based Clinical Work with Children and Adolescents (pp.237–57). New York: Springer.
441 Steele, M., Steele, H., Bate, J., et al. (2014) Looking from the outside in: the use of video in attachment-based
interventions. Attachment & Human Development, 16(4), 402–415.

442 Steele, H., Murphy, A., Bonuck, K., Meissner, P., & Steele, M. (2019) Randomized control trial report on the
effectiveness of Group Attachment-Based Intervention (GABI©): improvements in the parent–child relationship

not seen in the control group. Development & Psychopathology, 31(1), 203–217.
443 The Adverse Childhood Experiences measure assesses reports of five forms of abuse and five forms of house-
hold dysfunction (parent mentally ill, incarcerated, drug addicted, domestic violence, and separation/divorce).
444 An inverse relationship between attrition rates and differential effect sizes in attachment-based interven-
tions was reported by Bakermans-Kranenburg, M.J., van IJzendoorn, M.H., & Juffer, F. (2003) Less is more: meta-
analyses of sensitivity and attachment interventions in early childhood. Psychological Bulletin, 129(2), 195–215.
GABI was marginally less effective for participants reporting four or more adverse child-
hood experiences. The researchers found that treatment as usual had a negative effect on
caregiver–child interaction in many cases.
Conclusion
One of the fundamental claims of the Minnesota group was that resilience is a social and de-
velopmental process, in large part predictable in terms of risks and promotive and protective-
stabilising factors. The resilience of the field of attachment research in the years after Ainsworth
can be attributed to a significant degree to the role of the Minnesota laboratory as a promotive
factor for the field. The Minnesota group contributed to the growth of intergenerational longi-

tudinal research as a backbone of the developmental tradition within attachment research.445
And Sroufe and Egeland helped sustain the idea of overarching developmental theory, pulling
against the current during decades in which academic psychology was becoming increasingly
atheoretical.446 The influence of this position on figures like Cicchetti helped promote attach-
ment research as a founding element of developmental psychopathology.
Several important studies in the tradition of developmental psychopathology were seeded
directly by the Minnesota group, as trainees went out to set up their own laboratories with
the Minnesota Longitudinal Study of Risk and Adaptation as the model for how cohort re-
search with at-risk populations should be conducted. Examples include Ward’s laboratory
at Cornell and the Jacobvitz–Hazen laboratory at Austin, Texas.447 Many positive features of
the Minnesota study were influential for this later work, including a concern with hard-won,
credible, and concrete findings confirmed by multiple sources of information, including ob-
servational measures. Furthermore, in a context in which research funders had increasing
interest in adversity and individual differences, the Minnesota group also helped attachment
research reorient towards a greater focus on risk and mental illness.
Not every one of the features inherited by later researchers from Minnesota was positive,
though. For example, in general, the Minnesota Longitudinal Study of Risk and Adaptation
was poorly set up to study father–child relationships. Only a small fraction of fathers were
still in their child’s life at 18 months, and 27% of the total sample had three or more men
living in their homes across childhood. As with many cohort studies with high-risk samples,
the Minnesota group found that engaging fathers in the research was exceptionally diffi-
cult.448 Most of the data they were able to collect on the care provided by father-figures was
445 See Grossmann, K.E., Grossmann, K., & Waters, E. (eds) (2005) Attachment from Infancy to Adulthood: The
Major Longitudinal Studies. New York: Guilford.

446 Beller, S. & Bender, A. (2017) Theory, the final frontier? A corpus-based analysis of the role of theory in psy-
chological articles. Frontiers in Psychology, 8, 951.

447 See e.g. Ward, M.J. & Carlson, E.A. (1995) Associations among adult attachment representations, maternal
sensitivity, and infant–mother attachment in a sample of adolescent mothers. Child Development, 66(1), 69–79;
Hazen, N.L., Allen, S.D., Christopher, C.H., Umemura, T., & Jacobvitz, D.B. (2015) Very extensive nonmaternal
care predicts mother–infant attachment disorganization: convergent evidence from two samples. Development &
Dsychopathology, 27(3), 649–61.
448 The primary attempt to recruit father-figures in the study was at the 13-year follow-up, where 175 mothers
and 44 fathers or father-figures were brought in to the laboratory. However, the findings regarding fathers from the
13-year follow-up were not published in a prominent location, and do not appear to have ever subsequently been
cited except by the authors. Sroufe, L.A. & Pierce, S. (1999) Men in the family: associations with juvenile conduct.
In G. Cunningham (ed.) Just in Time Research: Children, Youth and Families (pp.19–26). Minneapolis: University
of Minnesota Extension Service.
Conclusion 423
from maternal report. The Minnesota group found that composite variables representing
supportiveness and disruptive behaviour in the home by fathers, especially during early and
middle childhood, contributed to the likelihood of child externalising behaviours in child-
hood and adolescence. The effect held when controlling for maternal care and general stress
experienced by the family. However, the findings were not published in a prominent location,
and do not appear to have ever subsequently been cited except by the authors.449 Difficulties
in recruiting fathers, the smaller number of fathers who act as primary caregivers, the model
of studies such as those of Baltimore and Minnesota, and wider cultural values combined to
mean that fathers have not generally seen adequate attention in American attachment re-
search,450 though there have been exceptions.451
In sum, though, the second generation of attachment research benefited enormously
from the Minnesota group as a protective-stabilising factor, forming a touchstone for the

field. Papers by the group have generally been exemplary in their clarity and insight, of-
fering an attractive model of what attachment research can entail. The group strengthened
and stabilised consensus, combatting contrasting positions on a range of potentially contro-
versial matters, such as whether the infant attachment classifications are reducible to infant
temperament, or whether they predict later outcomes controlling for covariates. The group
has also served the second generation as an unshakable anchor for the Ainsworth Strange
Situation and its coding protocols. At a point in the 1980s where it seemed like the pro-
cedure was heading for disrepute due to discrepant findings by research groups who had not
trained in using the Ainsworth scales, the Minnesota group helped create a set of videotapes
of Strange Situations with agreed classifications by Ainsworth, Waters, Main, Sroufe, and
other expert coders. This formed the basis for a yearly training institute at Minnesota which
still continues today.452
Sroufe and Egeland were also staunch defenders of Ainsworth’s category-based coding
system. An important instance was Sroufe’s reply to Chris Fraley and Sue Spieker’s 2003 paper
‘Are infant attachment patterns continuously or categorically distributed?’ (Chapter 2).453
449 Ibid. An additional major empirical report received unfavourable peer- review and remained unpub-
lished: Pierce, S.L., Coffino, B.S., & Sroufe, L.A. (2005) The Role of Fathers and Father Figures in the Development of
Lower SES Children: Predicting Adolescent Externalizing Behavior from Early Childhood. Unpublished manuscript,
Alan Sroufe personal archive.
450 There were several indirect sources of information on father–child relationships in the Minnesota study. In
particular, the AAI offered information relevant to paternal care. However, as the children of the Minnesota study
themselves became parents, the significance of fathers has come further into view: Kovan, N.M., Chung, A.L., &
Sroufe, L.A. (2009) The intergenerational continuity of observed early parenting: a prospective, longitudinal study.
451 There have been a relatively small number of studies in the USA that have measured paternal sensitivity and
infant attachment. These have included: Easterbrooks, M.A. & Goldberg, W.A. (1984) Toddler development in the
family: impact of father involvement and parenting characteristics. Child Development, 55(3), 740–52; Volling,
B.L. & Belsky, J. (1992) Infant, father, and marital antecedents of infant father attachment security in dual-earner
and single-earner families. International Journal of Behavioral Development, 15(1), 83–100; Braungart-Rieker,
J.M., Garwood, M.M., Powers, B.P., & Wang, X. (2001) Parental sensitivity, infant affect, and affect regulation: pre-
dictors of later attachment. Child Development, 72, 252–70; Eiden, R.D., Edwards, E.P., & Leonard, K.E. (2002)
Mother–infant and father–infant attachment among alcoholic families. Development & Psychopathology, 14(2),
253–78; Kochanska, G., Aksan, N., & Carlson, J.J. (2005) Temperament, relationships, and young children’s recep-
tive cooperation with their parents. Developmental Psychology, 41, 648–60; Hazen, N.L., McFarland, L., Jacobvitz,
D., & Boyd-Soisson, E. (2010) Fathers’ frightening behaviours and sensitivity with infants: relations with fathers’
attachment representations, father–infant attachment, and children’s later outcomes. Early Child Development and
Care, 180, 51–69
452 https://attachment-training.com.
453 Fraley, R.C. & Spieker, S.J. (2003) Are infant attachment patterns continuously or categorically distributed?
A taxometric analysis of Strange Situation behavior. Developmental Psychology, 39(3), 387–404.

Fraley and Spieker argued that the Strange Situation yielded data that were better modelled
in terms of the extent of avoidance and resistance using the Ainsworth scales, rather than
in terms of Ainsworth’s categories. In his reply to Fraley and Speiker, Sroufe expressed his
sympathy with their position. Of course there are cases in the middle-range that have to be
forced into the categories. On the basis of their analysis, Sroufe reflected, ‘Fraley and Spieker
argued essentially that these categories are fictions, and they may well be correct. I would
ask, have these been useful fictions? Do they remain so today? The answer to at least the
first question clearly is positive, and the answer to the second question may be positive as
well.’454 The acknowledgement of the Ainsworth classifications as ‘useful fictions’ reflected
Sroufe’s approach to the philosophy of science, influenced by John Dewey and philosophical
pragmatism. Above all, Sroufe stated that Fraley and Spieker would need to demonstrate
that the avoidant and resistant dimensions can predict later outcomes as well as Ainsworth’s

categories before he would consider them ‘a candidate for substitution’.455 Overall, though,
Sroufe was personally dubious ‘whether this endeavour is worth the effort’.456 And he ex-
pressed concern that debates about categories versus dimensions among attachment re-
searchers would offer ammunition to critics of the field, who could use ‘questions about the
taxonomic status of Ainsworth et al.’s (1978) categories as evidence that the entire paradigm
is invalid’.457 Sroufe’s experiences of the turbulent debates of the 1970s and 1980s as attach-
ment was being established within developmental science had given him a keen sense that
the paradigm’s existence was not inevitable and should not be taken for granted by those
raising psychometric questions.
Except among the social psychology tradition of attachment research (Chapter 5), the
Fraley and Spieker paper was, to an unjustified extent, buried for a decade.458 And even
Fraley, Spieker, and researchers sympathetic to their position did not pursue research to
compare the predictive validity of categories versus dimensions; as mentioned in Chapter 2,
the reason for this is unclear. In the meantime, Sroufe’s defence of the four-category coding
system was cited as an important source of authority by other researchers in retaining the
existing system.459 Sroufe’s influence might also be inferred in the fact that the dimensional
scale that particularly came into use during the 2000s was the 1–9 scale for disorganisation,
which had been out of view of either Fraley and Speiker’s criticism or Sroufe’s defence of the
Ainsworth categories. Furthermore, Elizabeth Carlson from the Minnesota group reported
findings from her prospective study of attachment disorganisation using the scale rather
than the category, and later researchers cited Carlson as authorising use of the scale as legit-
imate for attachment research.460 In his most recent publication, Sroufe scaled back his de-
fence of the Ainsworth categories, acknowledging that the ambivalent/resistant attachment
39(3), 413–16, p.414.

455 Ibid. p.413.
456 Ibid. p.415.
457 Ibid. p.415.
458 The primary exceptions were researchers in the developmental tradition with a particular focus on adult at-
tachment, such as the Cowans, e.g. Cowan, P.A. & Cowan, C.P. (2007) Attachment theory: seven unresolved issues
and questions for future research. Research in Human Development, 4(3–4), 181–201.
459 E.g. Aviezer, O. & Sagi-Schwartz, A. (2008) Attachment and non-maternal care: towards contextualizing the
quantity versus quality debate. Attachment & Human Development, 10(3), 275–85.
460 Bernard, K. & Dozier, M. (2010) Examining infants’ cortisol responses to laboratory tasks among children
varying in attachment disorganization: stress reactivity or return to baseline? Developmental Psychology, 46(6),
1771–8.
Conclusion 425
classification has few distinct correlates after half a century of studies using the Strange
Situation.461
Sroufe also contributed to anchoring consensus around the disorganised attachment clas-
sification. From an organisational perspective, behaviour takes its meaning from its inte-
gration within broader patterns of adaptation to the environment. As a consequence, there
was a tendency in Sroufe’s approach to treat the behaviours listed by Main and Solomon
not just as the breakdown of a strategy for attaining the availability of the caregiver, but as a
breakdown of psychological meaning itself. In general, the term ‘disorganisation’ was used
in Sroufe’s writings through the 1990s and 2000s to mean the breakdown into chaos in the
context of high arousal and lack of containment by the environment.462 It therefore seemed
natural that these behaviours, perceived as lacking ‘organisation’, were placed together in a
broad-band residual category. In his reply to Fraley and Spieker, Sroufe used the discovery

of the disorganised classification as an argument in favour of the Ainsworth classifications,
which he felt had called attention to the discrepant behaviours.463 However, again this ap-
pears to be a position that Sroufe has later qualified. In a 2014 paper, Padrón, Carlson, and
Sroufe criticised Main and Hesse for implying, whether intentionally or inadvertently, that
disorganised attachment were all of the same kind and had the same cause. They reported
findings that frankly fearful or disoriented behaviours in the Strange Situation had different
antecedents to the other kinds of behaviours listed by Main and Solomon.464 Specifically, in-
fants in the subgroup who did not display apprehension or dissociation/confusion displayed
less emotion regulation as newborns.
These were highly preliminary results. However, curiously, they agree with an earlier draft
of Main and Solomon’s 1990 chapter, which observed that the large majority of children dis-
playing Index VI (apprehensive) and Index VII (disorientated) behaviours in their 200 tapes
were from maltreatment or very high-risk samples (Chapter 3).465 The 2014 paper appears
to express a broader turn in Sroufe’s late thinking, away from work to establish the standing
of existing categories and towards a concern to capture more fine-grained information. In
his 2013 article on ‘The promise of developmental psychopathology’, Sroufe argued that ‘we
need studies that unpack the heterogeneity in current categories by examining differential
antecedents and pathways’.466 The extent to which this will be a priority for the third gener-
ation of attachment researchers remains uncertain. There may be too much invested in the
existing categories, and too much inertia, for the field to now pursue a different approach,
even if incremental validity could be demonstrated through identification of differential
antecedents and pathways. Yet the quiet transition over the past decade towards use of the 1–
9 scale for disorganisation and away from use of the disorganised category shows that change
is possible if it permits researchers more effective prediction.
462 E.g. Sroufe, L.A. (1996) Emotional Development. Cambridge: Cambridge University Press: ‘The security of
the attachment relationship is related to the regularity with which arousal has historically led or not led to infant
behavioural disorganisation in the context of the caregiver’ (158).
39(3), 413–16, p.415.

464 Padrón, E., Carlson, E.A., & Sroufe, L.A. (2014) Frightened versus not frightened disorganized infant attach-
ment. American Journal of Orthopsychiatry, 84(2), 201–208.

466 Sroufe, L.A. (2013) The promise of developmental psychopathology: past and present. Development &
Psychopathology, 25(4.2), 1215–24, p.1216.

It is important to mark these transitions and discontinuities even during the ascendency
of the second generation of attachment researchers. Nonetheless, it must also be acknow-
ledged that Kuhn’s classic comment in The Structure of Scientific Revolutions bears true: often
a paradigm may only enter a period of serious and explicit transition once the old gener-
ation of research leaders exit the field.467 With Egeland and Sroufe’s retirement, control of
the Minnesota Longitudinal Study of Risk and Adaptation passed to Roisman and Simpson.
With this shift, the Minnesota group moved from being a fundamental pillar of the second-
generation consensus to a major contributor to methodological and theoretical heterodoxy.
Roisman perceives the Ainsworth–Main four-category image of attachment, embodied by
the Strange Situation and the AAI, as psychometrically and theoretically unsound, even if he
agrees with Sroufe that they have served as useful fictions. Since Egeland and Sroufe’s retire-
ment, then, the Minnesota group have dropped use of the attachment classifications, prefer-

ring composite measures of early care or maltreatment, or composites of scale scores on the
AAI or Strange Situation.468 Many of the perspectives advocated by Roisman and Simpson
have been based on the image of attachment, the approaches to psychometry, and the re-
search values developed within the social psychological tradition of attachment research.
This tradition is the focus of the next chapter.
467 Kuhn, T.S. (1970) The Structure of Scientific Revolutions, 2nd edn. Chicago: University of Chicago Press. See
also Mannheim, K. (1952) Essays on the Sociology of Knowledge. London: Routledge.

468 An exception is Puig, J., Englund, M.M., Simpson, J.A., & Collins, W.A. (2013) Predicting adult physical
illness from infant attachment: a prospective longitudinal study. Health Psychology, 32(4), 409–417. However, this
appears to have been written during the transition period in leadership of the Minnesota group.
5
Phillip Shaver, Mario Mikulincer, and
the Experiences in Close Relationships Scale
Biographical sketch
Shaver was a social psychologist, interested in combining the experimental approach of behav-

iourism with the concern for relationships and emotional life that characterised psychoanalysis.
Hazan, a graduate student with Shaver, proposed in a seminar that aspects of adult romantic re-
lationships resembled the qualities of an attachment relationship. In the mid-1980s, Shaver and
Hazan developed a single-item self-report measure of romantic love modelled on the Ainsworth
classifications of infant attachment. This initiated a decade in which a profusion of self-report
measures appeared, a period mostly brought to an end by the synthesis of these measures into
the Experiences of Close Relationships (ECR) scale in 1998 by Shaver’s group. This scale assessed
beliefs about close relationships, and was based on two dimensions: avoidance of closeness and
anxiety about closeness. Since the 1990s, a thriving tradition of attachment research has become
established within social psychology. A central figure has been Mario Mikulincer, an Israeli ex-
perimental psychologist with a background in the study of learned helplessness and war trauma.
From the 2000s, Shaver and Mikulincer became close collaborators. At times, the relationship
between the social psychology tradition and the developmental tradition of attachment research
has been rocky. Against expectations in the 1990s, research found that there is little relationship
between the ECR (or other self-report methods) and either the Ainsworth Strange Situation
or the Adult Attachment Interview (AAI). Nonetheless, in an apparent paradox, central hy-
potheses drawn from attachment theory are supported by research using the ECR. Shaver and
Mikulincer have drawn on elegant experimental methodologies to further elaborate attachment
theory. They have also supported the development of a younger generation of leaders of the
social psychological tradition of attachment research in the USA and Israel. Over the decades,
conflict between the social psychology and developmental approaches appears to have ebbed.
An important contribution to this has been the three editions of the Handbook of Attachment,
edited by Shaver and Cassidy. However, there remain difficulties in translating between the ter-
minology and methodology of the two traditions of attachment research.
Introduction
In the early 1980s, the empirical research of Ainsworth, Sroufe, Egeland, and others had es-
tablished the standing of attachment research as a credible paradigm for the study of early
childhood within American developmental psychology (Chapters 2 and 4). Yet Bowlby’s
theory claimed that attachment, as a behavioural system, continued to be relevant and in-
fluential after early childhood (Chapter 1). Writing in 1984, Bowlby reflected in correspond-
ence that ‘I think we are still rather ignorant about the shift from a child–parent relationship
428 Phillip Shaver and Mario Mikulincer
to a spouse relationship’.1 Ainsworth and her immediate colleagues were engaged in ex-
tensive discussions in the early and mid-1980s about how to take the measurement of at-
tachment beyond infancy, and felt sure that this would be possible. The first step taken in
extending attachment methodology beyond infancy was Main’s ‘move to the level of repre-
sentation’, and the development of the AAI (Chapter 3). A different route beyond infancy is
the use of self-report to examine an individual’s perception of their attachment relationships.
This route was adopted by some former students of Ainsworth, such as Mark Greenberg.2
However, in general, developmental psychology has preferred observational to self-report
measures, and this was especially the case for attachment research. By contrast, social psych-
ology as a subdiscipline was more favourable to the use of self-report measures. A sustained
programme of research using self-report methods for assessing adult attachment styles was
initiated by Shaver and colleagues.3

Shaver described his academic path to attachment theory as beginning as an under-
graduate at Wesleyan University in the early 1960s. At the time, psychoanalysis was under
heavy criticism within American academic psychology, especially from behaviourist ap-
proaches and their emphasis on experimental research and learnt responses (Chapter 2).
Psychoanalysis had failed to develop much of a tradition of valid measurement techniques,
and its account of motivation seemed speculative and untestable. The growing availability of
federal research grants and the pressures to publish rapidly contributed to the attractiveness
of highly focused mini-theories, rather than overarching models.4 However, to Shaver in
the 1960s at Wesleyan, as later for Mukulincer at Bar-Ilan University in the 1970s, ‘the issues
raised by psychoanalysts, beginning with Freud, are extremely important: sexual attraction
and desire; romantic love; the development of personality beginning in infant–caregiver re-
lationships; painful, corrosive emotions such as anger, fear, death anxiety, jealousy, hatred,
guilt, and shame; intrapsychic conflicts, defenses, and psychopathology; individual and
intergroup hostility; the brutality of war’. Compared to this, Shaver found that reading ‘aca-
demic social and personality psychology’ for his courses ‘seemed disappointingly superficial
compared with psychoanalysis’.5
Shaver completed his PhD at the University of Michigan in 1970. Affirming the pivotal sig-
nificance of internal experience against behaviourist approaches, Shaver’s doctoral research
demonstrated the role of mental imagery in learning and problem solving. Shaver spent the
next decade in New York, joining Columbia University in 1971 where he was asked to teach
the undergraduate course in personality psychology.6 At that time, the Columbia faculty
1 Bowlby, J. (1984) Letter to Marco Bacciagaluppi, 3 January 1984. PP/Bow/B.3/40.

2 E.g. Greenberg, M.T., Siegel, J.M., & Leitch, C.J. (1983) The nature and importance of attachment relationships
to parents and peers during adolescence. Journal of Youth and Adolescence, 12(5), 373–86.
3 In 2014, in tributes on the occasion of Shaver’s retirement as Distinguished Professor of Psychology at the
University of California, Davis, Jude Cassidy observed that ‘no scholar has done more to expand attachment
theory and research’. Pehr Granqvist claimed that Shaver’s work ‘surpasses nearly all other research programs in
psychological science, both in terms of originality and sheer quantity. I’ll stand on Mary Ainsworth’s coffee table in
my cowboy boots and say that!’ Such staunch statements of praise are especially salient given that, unlike Cassidy,
or Granqvist’s mentor Mary Main, Shaver was not a student of Ainsworth’s. His route into attachment research was
less direct, and the acceptance his work has gained has taken a lot longer. http://www.foundationpsp.org/shaver.
php.
4 Shaver, P.R. & Mikulincer, M. (2005) Attachment theory and research: resurrection of the psychodynamic ap-
proach to personality. Journal of Research in Personality, 39, 22–45, pp.23–4.

5 Shaver, P.R. & Mikulincer, M. (2011) Analysis of a collaborative working relationship. Relationship Research
Newsletter, 9(2), 7–9, p.7.

6 Shaver, P. (2017) Attachment to attachment theory. Voices: Journal of the American Academy of Psychotherapists,
53(2), 35–9, p.36.

Introduction 429
were offered the possibility of undergoing psychoanalysis at the Columbia Psychoanalytic

Institute at a discounted rate. Shaver entered psychoanalysis with two presenting issues.
A first was workaholism, which was emptying his life of other activities and harming his
health and mood.7 A second was that Shaver had been ‘beginning to notice a repetitive and
destructive pattern in my romantic relationships—always having a woman “in the wings” to
whom I could flee if my primary relationship fell apart’.8
Shaver was in psychoanalysis for four days a week for two years. During this time, Shaver’s
younger brother was diagnosed with cancer. The brothers had been close, often one another’s
only friends for periods as the family moved repeatedly for their father’s work.9 On hearing
the news, Shaver’s therapist said:
I’d like to step out of role and tell you, with great sadness, he will not live for more than

8–10 months. There is, unfortunately, no good treatment for testicular carcinoma. I think
we should postpone talking about what we’ve been discussing lately and help you think
through how you want to deal with this important situation. If you continue to focus on
work and your personal problems, you may let this tragedy pass and then forever regret
your lack of involvement.10
Shaver took leave from work, and moved for the summer to Minneapolis to be with his
brother. The loss of his brother would prove critical to Shaver’s concern with attachment
theory: ‘Eventually, he died, and I was present to receive his blessing. The intense grief that
followed was unlike anything I had ever experienced, and it caused me to devote several
therapy sessions to grieving. I also began reading books and articles about loss and grief,
including some early papers by John Bowlby.’11
Throughout the 1970s, a central debate in American academic psychology was between
behaviourist theories and new cognitive theories of the mind. Shaver acknowledged the
strengths in experimental design of behaviourist research. And he acknowledged that any
account of the mind had to include attention to cognition and communication. However,
he criticised both paradigms, arguing that neither offered an adequate account of the role
of emotion within human life: ‘Behaviorism is assaulted as if it were an oppressive ruler. But
the powerful methods developed during the reign of strict behaviorism have now been com-
bined with theoretical constructs borrowed from the sciences of communication and artifi-
cial intelligence to yield a robust cognitive psychology. Perhaps we can now begin extending
it to include the notoriously uncomputerlike emotions.’12 By the end of the decade, he had
7 In a chapter for a psychology textbook written during this period, Shaver’s tone shifts towards the per-
sonal: Shaver, P.R. (1975) Psychoanalytic theories of personality. In Psychology Today: An Introduction, 3rd edn
(pp.402–425). New York: CRM/Random House: ‘Lest the superego seem to be a dry abstraction, you should realise
that intense guilt and the wearying pursuit of perfection are two of the most common reasons people give for
entering psychoanalysis’ (409). He also asks the reader, ‘have you ever gone on a long walk during cold, wet wea-
ther, suspecting that you might get sick but feeling vaguely that it would serve you right?’ (409).
8 Shaver, P. (2017) Attachment to attachment theory. Voices: Journal of the American Academy of
Psychotherapists, 53(2), 35–9.

9 Goodman, G.S. (2006) Attachment to attachment theory: a personal perspective on an attachment researcher.
In M. Mikulincer & G.S. Goodman (eds) Dynamics of Romantic Love: Attachment, Caregiving, and Sex (pp.3–22).
Psychotherapists, 53(2), 35–9.

11 Ibid. p.38.
12 Shaver, P.R. (1972) Review of cognition and affect, edited by J.S. Antrobus. American Journal of Psychology,
85, 297–9, p.297. See also Shaver, P.R. (1975) Emotional experience and expression. In Psychology Today: An
come to conclude that his larger aim as a scholar was to ‘contribute to a theory of emotion
which fits within an evolutionary and developmental framework’.13 More specifically, the
loss of his brother led Shaver to a reorientation of his work around the themes of loneliness
and depression.14 Shaver looked around him and saw a world in which these were the great
themes of the age: ‘If eras can be characterised by salient emotions, ours must be the age
of depression and loneliness.’15 To make sense of this, behaviourism and cognitive theories
alone would not be adequate. He prophesised that ‘a revolution in ‘emotion science’ is about
to follow the path blazed by the cognitive sciences. The emotion revolution is a logical out-
come of the “age of depression and loneliness”.’16
The year 1979, however, would prove a turning point for Shaver, transfiguring his con-
cern with depression and loneliness. One of the important factors contributing to this shift
was meeting Robert Weiss, a Harvard-based sociologist, at a conference in 1979. Weiss had

attended Bowlby’s seminars at the Tavistock during a sabbatical in 1971.17 Weiss accepted
that the attachment behavioural system was less readily activated after infancy, and a dif-
ferent stock of attachment behaviours would be seen in adulthood. However, he highlighted
that adults still provide one another with a secure base and safe haven, and can experience
anxiety during unanticipated separations.18 Shaver reported that Weiss ‘encouraged me to
conduct research more specifically on attachment and loss in adulthood’.19 Though familiar
with Bowlby’s writings before 1979, Weiss’s urging gave Shaver impetus to explore beyond
Bowlby’s account of grief, to consider attachment as a paradigm for conceptualising adult re-
lationships. A second component of Shaver’s shift away from research on loneliness towards
attachment was meeting and starting to date Gail Goodman at the University of Denver,
where Goodman was a postdoctoral fellow with an interest in the psychology of children
in court contexts. Shaver joined the University of Denver in 1980 and commenced a pro-
gramme of work concerned with adult romantic relationships. Looking back in a recent
paper, Shaver identified four great attachments in his life, which have offered him a secure
base from which to explore and a safe haven providing regulation and comfort. The first two
were his mother and then his analyst; 1979 saw the initiation of the two adult attachment re-
lationships that would dominate his subsequent life: his marriage and attachment research.20
Introduction, 3rd edn. New York: CRM/Random House: ‘Surely emotion, not variety, is the spice of life—though
the two are obviously related . . . with fear, with indignation, jealousy, ambition, worship. If they are there, life
changes’ (333).
13 Shaver, P.R. & Klinnert, M. (1982) Schachter’s theories of affiliation and emotion: implications of devel-
opmental research. In L. Wheeler (ed.) Review of Personality and Social Psychology, Vol. 3 (pp.37–72). Beverly
Hills: Sage Publications, p.39.
Psychotherapists, 53(2), 35–9, p.39.

15 Shaver, P.R. & Brennan, K.B. (1991) Measures of depression and loneliness. In J.P. Robinson, P.R., Shaver,
& L.S. Wrightsman (eds) Measures of Social Psychological Attitudes, Vol. 1 (pp.195–289). San Diego: Academic
Press, p.195.
16 Ibid. p.281.
17 The experience is described by Weiss in Weiss, R.S. (1994) Foreword. In M.B. Sperling & W.H. Berman (eds)
Attachment in Adults: Clinical and Developmental Perspectives (pp.iv–xvi). New York: Guilford, p.xiv.
18 Weiss, R. (1974) The provisions of social relationships. In Z. Rubin (ed.) Doing Unto Others (pp.17–26).
Englewood Cliffs, NJ: Prentice-Hall; Weiss, R.S. (1982) Attachment in adult life. In C. Parkes & J. Stevenson-Hinde
(eds) The Place of Attachment in Human Behavior (pp.171–84). New York: Basic Books.
Psychotherapists, 53(2), 35–9, p.39.

20 Ibid.: ‘Looking back over the decades, I see that I have been attached to my mother, my analyst, and my wife,
and also—viewing attachment theory as a “safe haven” and “secure base”—to attachment theory itself ’ (39).
Introduction 431
Specifically, in addressing his work to adult romantic relationships, Shaver’s interest was
less in the concrete behaviour of a romantic dyad and more in the motivations and experi-
ences of individual adults facing the predicament of intimacy or the rupture of relation-
ships.21 Despite clear analogues, the concrete behaviour of an infant–caregiver dyad differed
markedly from an adult romantic dyad. However, Shaver was excited by the possibility that
at the level of motivation and experience, there were significant lines of continuity between
infant and adult processes of intimacy and loss. In 1980, Bowlby published Loss, in which just
such an argument was made. Similar claims were being made by Ainsworth and Sroufe.22
Themes of love, attraction, and commitment in romantic relationships had also, in the early
1980s, become established as areas for research in social psychology.23 New societies were es-
tablished such as the Society for the Study of Personal Relationships, and the first issue of the
Journal of Social and Personal Relationships appeared in 1984. In this context, one of Shaver’s

early graduate students at Denver, Cindy Hazan, proposed in a graduate seminar that there
might be individual differences in adult feelings about romantic relationships that mirrored
the Ainsworth infant classifications.24 From these discussions emerged a co-authored 1987
article in the Journal of Personality and Social Psychology25 and a chapter in The Psychology of
Love, edited by Sternberg and Barnes.26
These works presented the argument that adult romantic relationships exhibit the key fea-
tures of the attachment system in infancy, above all the role of the partner as a secure base
and safe haven.27 They also have other elements in common: reciprocal interaction and phys-
ical intimacy contribute to happiness; discoveries and experiences are shared with the other;
and there is a mutual desire for the approval of the other. Hazan and Shaver pointed out that,
for both infants and adults, the relationship can sometimes feel intrusive, contributing to
anxiety and a wish for independence; and the availability of the other can sometimes feel in
21 Shaver, P.R. (2010) My appreciation of Caryl Rusbult. Personal Relationships, 17(2), 172–3: ‘My own work
is inspired by psychoanalytic theory and tilts heavily toward personality rather than, or in addition to, situ-
ations . . . I wanted to dive deeply into the individual’s mind, viewing a mentally represented couple relationship as
at least as important as the dyad’s actual behavioral interactions’ (172).
22 E.g. Ainsworth, M.D.S. (1985) Attachments across the life span. Bulletin of the New York Academy of Medicine,
61, 792–812.
23 Perlman, D. & Duck, S. (2006) The seven seas of the study of personal relationships: from ‘the thousand
islands’ to the interconnected waterways. In D. Perlman & A.L. Vangelisti (eds) The Cambridge Handbook of
Personal Relationships (pp.3–34). Cambridge: Cambridge University Press.
24 Goodman, G.S. (2006) Attachment to attachment theory: a personal perspective on an attachment researcher.
In M. Mikulincer & G.S. Goodman (eds) Dynamics of Romantic Love: Attachment, Caregiving, and Sex (pp.3–22).
25 Hazan, C. & Shaver, P. (1987) Romantic love conceptualized as an attachment process. Journal of Personality
and Social Psychology, 52(3), 511–24.

26 Shaver, P.R., Hazan, C., & Bradshaw, D. (1988) Love as attachment: the integration of three behavioral sys-
tems. In R.J. Sternberg & M. Barnes (eds), The Psychology of Love (pp. 68–99). New Haven, CT: Yale University
Press. Reflecting on the wider context of the publication of these works, Shaver has recalled that when they sub-
mitted the paper for publication, it was so far outside the mainstream of social psychology that he worried that it
might derail his career. Instead, the paper immediately attracted interest. ‘A lot of young researchers, including the
growing number of young women in the field, followed up the initial study in interesting and unanticipated ways,’
he said. ‘I feel fortunate to have come along when the number of women in the field was growing, which made
close relationships a more acceptable topic, and when the divorce rate in the U.S. was of concern to government
research funders and to the American population. At the same time, new research techniques were developed,
making it possible to pursue personality and relationship phenomena that had not been studied empirically be-
fore.’ Holder, K. (2018) Interview with Phillip Shaver on winning the Society of Personality and Social Psychology
Legacy Award. https://lettersandscience.ucdavis.edu/news/phil-shaver-research-legacy-award.
tems. In R.J. Sternberg & M. Barnes (eds) The Psychology of Love (pp.68–99). New Haven, CT: Yale University
Press, p.77.
doubt, contributing to anxiety and hypervigilance about availability. More generally, Hazan
and Shaver proposed that infant and adult behaviour were rooted in a common behavioural
system, activated and terminated by the same kinds of conditions. In their writing, the au-
thors drew on their own biographical experiences, including experiences of bereavement
and of falling in love, to argue that the phenomenology and behaviours of adult attachment
closely resemble the functioning of the attachment system in childhood. For instance, they
observed that ‘as anyone knows who has been through such an experience, yearning for the
lost person can continue for months or years and can be mingled with anger at the person for
leaving’.28 On the basis of the analogies between infant attachment and adult romantic love,
they developed a brief measure to assess individual differences in attachment in adulthood.
Part of the importance of this initial attempt at the development of a self-report measure
of attachment was that it attracted Mario Mikulincer to attachment research. Mikulincer

had finished graduate study in experimental psychology at Bar-Ilan University in 1985, and
in 1986 joined the faculty at the University, where he would remain for over two decades.
He also had a role in the Mental Health Department of the Israeli army, and worked with
Zahava Solomon in conducting longitudinal research on the mental health and family life
of combat veterans from the 1973 Yom Kippur War and the 1982 Lebanon War. These ex-
periences would, decades later, be of great importance for his research in studying the effects
of trauma on attachment in adulthood. Embarking on a research career following his doc-
torate, Mikulincer’s output was oriented by concerns reflecting his academic and military
research responsibilities. From his doctoral research, he pursued inquiry into the origins of
adult experiences of helplessness. The concept of ‘learned helplessness’ had been introduced
by Seligman in the early 1970s to describe the way that repeated failures could contribute to a
sense that it would not be worth continuing to try.29 Mikulincer was interested in the mental
and emotional processes underpinning this effect and ways that it might be supressed.30 He
drew on the work of Lazarus and Folkman in conceptualising helplessness as an avoidant
response to a stressful stimulus, a response that could be circumvented through a variety of
coping strategies.31 However, Mikulincer was curious also about the potential for failures
to elicit rumination. In a sense this seemed like a coping strategy, since it focused attention
on how the problem might be solved. However, rumination could become ‘the intrusion
of autonomic failure-related thoughts into consciousness’. Rather than constructively ad-
dressing how a problem might be solved, these thoughts may ‘run ballistically on to com-
pletion, are irrelevant to one’s intensions, and are hard to suppress. They fill the cognitive
system.’32
The ballistic metaphor signals a cross-over with Mikulincer’s research for the army, where
his focus was on the role of intrusive images and emotions observed in combat veterans suf-
fering from trauma and from loneliness. Mikulincer was interested in the role of the family
in shaping the extent to which distress and fear intrude into the ordinary functioning of
28 Ibid. p.92. The personal experiences that contributed to Cindy Hazan’s interest in attachment and loss do not
appear in the public record.

29 Seligman, M.E.P. (1972) Learned helplessness. Annual Review of Medicine, 23(1), 407–412.
30 Mikulincer, M. & Caspy, T. (1986) The conceptualization of helplessness: I. A phenomenological-structural
analysis. Motivation and Emotion, 10, 263–78; Mikulincer, M. (1986) Attributional processes in the learned help-
lessness paradigm: the behavioral effects of globality attributions. Journal of Personality and Social Psychology, 51,
1248–56.
31 Lazarus, R.S. & Folkman, S. (1984) Coping and adaptation. In W.D. Gentry (ed.) Handbook of Behavioral
Medicine (pp.282–325). New York: Guildford.

32 Mikulincer, M. (1994) Human Learned Helplessness: A Coping Perspective. New York: Plenum Press, p.72.
Introduction 433
individuals. For instance, he was involved in research, published in 1990, which showed that
family support could account for a quarter of variance in combat-related PTSD, and that
this effect was fully mediated by feelings of loneliness.33 In the context of these concerns,
Hazan and Shaver’s 1987 article caught his eye. Mikulincer recalls having ‘noticed similar-
ities between (a) certain forms of helplessness in adulthood and the effects of parental un-
availability in infancy; (b) intrusive images and emotions in the case of post-traumatic stress
disorder and the anxious attachment pattern described by Ainsworth and her colleagues in
1978 and adapted by Hazan and Shaver; and (c) avoidant strategies for coping with stress and
the avoidant attachment pattern described by these same authors’.34
According to a jointly written retrospective piece, a firm friendship and collaboration was
begun in the 1990s when Mikulincer offered Shaver the chance to work together on a study
of security-enhancement as a method of reducing intergroup hostility.35 However, already

Shaver’s first publications in attachment theory reveal many facets of shared orientation be-
tween the two researchers, which included an ‘intense achievement motivation combined
with distaste for egotism, a deep and perhaps neurotic split between the senses of ambition
and sloth, and extended experiences with psychotherapy’.36 In Hazan and Shaver’s early writ-
ings on attachment, intense emotions of loss and of romantic love are palpable and explicitly
personal, but also available to resource-thinking and theory-building. This explicit consid-
eration of personal experiences as one source of information among others in the develop-
ment of psychological theory became a hallmark of Shaver’s approach to research.37 At the
same time, it is possible in Shaver’s first attachment publications to sense the relentless drive
forcing the creation of a bridge between social psychological methods and psychodynamic
concerns. In doing so, Shaver and Hazan attempted to grapple with the whole of human ro-
mantic life, and further at times, human close relationships in general.
The creation of this bridge was timely. The 1990s saw many other attempts to square psy-
choanalytic concerns with the tools of social and experimental psychology. Yet Shaver and
Hazan’s work was unusual in reaching for the depths of the human experiences of intimacy
and seeking to contain this in an assessment delivered to hundreds of people. The use of a
self-report methodology was, however, especially attractive for researchers who had come
to attachment theory without direct contact with Ainsworth. From the 1930s to the 1950s,
Ainsworth had attempted to validate a self-report assessment of security, but had found that
participants could report that they were secure as a defence against feelings of insecurity
33 Solomon, Z., Waysman, M., & Mikulincer, M. (1990) Family functioning, perceived societal support, and
combat-related psychopathology: the moderating role of loneliness. Journal of Social and Clinical Psychology, 9(4),
456–72.
Newsletter, 9(2), 7–9, pp.7–8.

35 Shaver’s CV lists a visit to Bar Ilan in 1995: Shaver, P.R. (1995) Psychodynamic and representational aspects of
adult attachment. Invited address, Bar Ilan University, Ramat Gan, Israel. There is also a visit in 2000: Shaver, P.R.
(2000) An updated theory of romantic (pair-bond) attachment. Invited address at 3rd International Conference
of the Peleg-Bilig Center for the Study of Family Wellbeing: ‘Towards a science of couple relationships.’ Bar-Ilan
University, Ramat Gan, Israel. Their first co-authored work would begin appearing in 2001. https://adultattach-
ment.faculty.ucdavis.edu/wp-content/uploads/sites/66/2014/06/Shaver.pdf.
Newsletter, 9(2), 7–9, p.8.

37 Shaver, P.R. (2006) Dynamics of romantic love: comments, questions, and future directions. In M.
Mikulincer & G.S. Goodman (eds) Dynamics of Romantic Love: Attachment, Caregiving, and Sex (pp.423–56).
New York: Guilford: ‘There is no theory of personality, emotions, social relationships or psychological develop-
ment that holds much more than a flickering candle to actual experience . . . It behoves us as relationship researchers
to keep attachment theory, alternative theories of love, and our own actual experiences of love in mind’ (426).
(Chapter 2). A person may be ‘so handicapped in his communication with others and in
insight into his own needs and feelings that pencil-and-paper tests cannot reflect the nature
and extent of his maladjustment’.38 Shaver appears not to have known about this work, and
never cites it. In a paper written in 1984, Ainsworth urged future researchers: ‘do not take at
its face value a person’s self reports of security, high self-esteem, high sense of competence or
freedom from stress and anxiety, even though more credence may be given to self-reports of
insecurity’.39
One attempt by developmental researchers to measure insecurity through self-report
was Main, Hesse, and van IJzendoorn’s work on the Berkeley–Leiden Adult Attachment
Questionnaire in the 1980s. This was a 200-item questionnaire, aiming to identify the four
‘states of mind regarding attachment’ from the AAI. This included two subscales for un-
resolved/disorganised states. A first was Unresolved States of Mind. This comprised items

drawn directly from the coding manual for U/d on the AAI, with items designed to repre-
sent both lapses in reasoning (e.g. feelings of responsibility for a death in which the partici-
pant had no causal role) and lapses in discourse (e.g. feelings of lack of mental control when
thinking about specific events). Additionally, ‘because unresolved trauma in certain cases
may have a relation to post-traumatic disorders and to the dissociative disorders (Spiegel,
1989), some items indicating symptoms of these disorders were included’. A second sub-
scale was items suggesting anomalous beliefs about time/space relations and causality, for
instance beliefs in possession, mental telepathy, or precognition. The scales had strong re-
liability and stability over 12 months. The Unresolved States of Mind subscale had an as-
sociation with U/d on the AAI of .63, accounting for 38% of variance. The Unusual Beliefs
subscale also made a smaller but significant contribution, with the items for beliefs in pos-
session performing especially well. Over 80% of participants could have been successfully
classified on the AAI on the basis of their scores on the two scales. False positives were very
rare; where misclassifications occurred, it was on the basis of false negatives. However, Main,
van IJzendoorn, and Hesse were ultimate disappointed by the overall Berkeley–Leiden Adult
Attachment Questionnaire measure, which failed to replace early successes. This was in con-
trast to the subscales for unresolved/disorganised states, which had proven quite successful.
Nonetheless, the authors decided to leave the self-report scales for unresolved/disorganised
states unpublished, and likewise their empirical studies validating these scales.40 As dis-
cussed in Chapter 3, Main has had difficulties bringing valuable work to publication where
she remains somewhat dissatisfied with it.
In a private workshop in 1988 with close collaborators, including Main, Ainsworth stated
that she regarded her own self-report measures as a failure: ‘I felt dissatisfied with the validity
Toronto Press, p.17. Van Rosmalen, van IJzendoorn, and van der Veer reported that a slightly revised version of
the Ainsworth self-report questionnaire ‘correlated significantly with the two ECR-RS scales for avoidance and
anxiety to mother, –.61 and –.32, respectively (p < .001; n = 230). The same was true for the ECR-RS to father: –.54
and –.39, respectively (p < .001; n = 222)’ (93). Such findings indicate sufficient convergence between Ainsworth’s
abandoned questionnaire and Shaver’s work to suggest that they represent highly aligned, if not necessarily iden-
tical, research strategies. Van Rosmalen, L. (2015) From security to attachment: Mary Ainsworth’s contribution to
attachment theory. Unpublished doctoral dissertation, Leiden University, Chapter 4.
39 Ainsworth, M.D.S. (1985) Attachments across the life span. Bulletin of the New York Academy of Medicine, 61,
792–812, p.798.
40 Main, M., van IJzendoorn, M.H., & Hesse, E. (1993) Unresolved/ Unclassifiable Responses to the Adult
Attachment Interview: Predictable from Unresolved States and Anomalous Beliefs in the Berkeley–Leiden Adult
Attachment Questionnaire. Unpublished manuscript. https://openaccess.leidenuniv.nl/dspace/bitstream/1887/
1464/1/168_131.pdf.
Self-reporting attachment 435
of my scales because of their inadequate coping with the whole matter of defensive maneu-
vers.’41 In her final publication in 1992, Ainsworth implied that self-report measures assess
semantic memory (Chapter 1) and generalised propositions about the world, rather than
episodic memory about concrete experiences.42 Any tendency for divergence between these
two memory systems would be expected to produce distortions in the validity of self-report
measures, in contrast to observational assessments which can capture both memory systems
and their interaction.43 Following up on Ainsworth’s proposal in subsequent years, Pianta
and colleagues, and Howard and colleagues, found that individuals classified as dismissing
on the AAI self-report fewer adverse experiences and less distress on questionnaires than
the assessment of independent observers, and preoccupied speakers on the AAI self-report
more such experiences and distress.44 Furthermore, Bailey and colleagues reported that dis-
crepancy between positive self-reported experiences in relationships and observer-reported

assessment of insensitivity was associated with higher scores for avoidance and lower scores
for proximity-seeking when the dyad were seen in the Strange Situation.45 Bailey and col-
leagues suggested that caregivers who hold positive accounts of relationships as a result of
obliviousness or defences may miss or dismiss information relevant to emotions in relation-
ships. This, in turn, would make them less likely to notice and respond to infant signals, or
repair interactions when such signals are missed or misinterpreted.
Self-reporting attachment
A first attempt
From the start, then, the self-report methodology adopted by Shaver and Hazan was one
that set them at odds with the community of researchers around Ainsworth. Nonetheless,
Shaver received encouragement from Bowlby, who wrote a personal letter in 1985 to say
42 Ainsworth, M. (1992) A consideration of social referencing in the context of attachment theory and research.
In S. Feinman (ed.) Social Referencing and the Construction of Reality in Infancy (pp.349–67). New York: Plenum
Press: ‘Some find their models too conflicting for such integration to be achieved. It would appear that in such cases
the conflicting models become separately consolidated, and often enough this is because what has been conveyed
by the partner in expressive behaviour does not match the semantic content of what is verbally conveyed’ (365).
43 Over time, developmental attachment researchers have repeatedly found substantial evidence for this con-
clusion. For instance, in a recent study, Howard and colleagues found that participants classified as dismissing on
the AAI disclosed fewer adverse childhood experiences on a self-report measure than independent scorers rated
in their AAI, and participants classified as preoccupied on the AAI self-reported more adverse childhood experi-
ences than independent coders. Howard, A.R.H., Razuri, E.B., Copeland, R., Call, C., Nunez, M., & Cross, D.R.
(2017) The role of attachment classification on disclosure of self and rater-reported adverse childhood experiences
in a sample of child welfare professionals. Children and Youth Services Review, 83, 131–6.
44 Pianta, R., Egeland, B., & Adam, E. (1996) Adult attachment classification and self-reported psychiatric symp-
tomatology as assessed by the Minnesota Multiphasic Personality Inventory–2. Journal of Consulting and Clinical
Psychology, 64(2), 273–81; Howard, A.R.H., Razuri, E.B., Copeland, R., Call, C., Nunez, M., & Cross, D.R. (2017)
The role of attachment classification on disclosure of self and rater-reported adverse childhood experiences in a
sample of child welfare professionals. Children and Youth Services Review, 83, 131–6. See also Borelli, J.L., Palmer,
A., Vanwoerden, S., & Sharp, C. (2019) Convergence in reports of adolescents’ psychopathology: a focus on disor-
ganized attachment and reflective functioning. Journal of Clinical Child & Adolescent Psychology, 48(4), 568–81.
45 Bailey, H.N., Redden, E., Pederson, D.R., & Moran, G. (2016) Parental disavowal of relationship difficulties
fosters the development of insecure attachment. Revue Canadienne des Sciences du Comportement, 48(1), 49–59.
that ‘I find your study of romantic love as related to childhood patterns most interesting’.46
In A Secure Base, Bowlby stated that Hazan and Shaver’s work suggested that there were, in
adulthood, ‘features of personality characteristic of each pattern during the early years’ that
could be measured using self-report. Bowlby affirmed his expectation that both the Hazan
and Shaver adult classifications and the AAI classifications would both be predicted by the
Strange Situation: ‘All our clinical experience strongly supports that view.’47 Nonetheless, he
continued to worry, influenced by Ainsworth on this point, ‘that with the self-report paper-
pencil method of appraisal it is well-nigh impossible to assess accurately how much defen-
sive maneuvers have inflated security scores’.48
Shaver was certainly aware of the potential limitations of self-report measures. In the early
1980s, Rubenstein and Shaver acknowledged that ‘social psychology is, unfortunately, re-
markable for its ability to reduce profound and fascinating human issues to rather superficial

and uninteresting generalisations’.49 Part of the problem, Shaver felt, was that the operation-
alisation of variables in social psychology often abstracts away from the lived experiences
of people in their complexity. Another part of the problem, however, was that self-report
measures designed without sufficient care and creativity can end up reporting no more than
the obvious. Throughout his career Shaver was fully willing to acknowledge the validity of
Ainsworth’s concern about the potential for bias in self-report measures of security: ‘we
agree that it is possible that some people defensively report that they do not worry about re-
jection and separation when actually they do worry about these issues, and some may have
no conscious access to such worries, even though they exist’.50 Through the 1980s and 1990s,
Shaver’s statements showed a certain hesitancy regarding what self-report assessments of
attachment precisely measured.51 However, by the late 1990s he became confident that the
participants’ report of their feelings, beliefs, expectations, and behaviours was not itself the
object of self-report assessments of attachment, but rather markers of underlying psycho-
logical processes related to attachment that may not be fully accessible to the participants
themselves.52 The challenge was to find a method of self-report that identified ‘noticeable
feelings, beliefs, expectations, and behaviors that are related to, or arise from, underlying
46 Bowlby, J. (1985) Letter to Phillip Shaver, 30 October 1986. PP/Bow/J.9/181.

London: Routledge, p.145.
ality development. American Psychologist, 46(4), 333–41, p.334.

49 Rubenstein, C. & Shaver, P.R. (1982) The experience of loneliness. In L.A. Peplau & D. Perlman (eds)
Loneliness: A Sourcebook of Current Theory, Research, and Therapy (pp.206–23). New York: Wiley, p.221.
50 Shaver, P.R. & Mikulincer, M. (2004) What do self-report attachment measures assess? In W.S. Rholes & J.A.
Simpson (eds) Adult Attachment: Theory, Research, and Clinical Implications (pp.17–54). New York: Guilford, p.24.
51 E.g. Shaver, P.R., Collins, N.L., & Clark, C.L. (1996) Attachment styles and internal working models of self
and relationship partners. In G.J.O. Fletcher & J. Fitness (eds) Knowledge Structures in Close Relationships: A Social
Psychological Approach (pp.25–61). Mahwah, NJ: Lawrence Erlbaum: ‘Research on attachment styles—relatively
coherent and stable patterns of emotion and behaviour exhibited in close relationships—is based on the assump-
tion that relationship orientations are due to, or perhaps consist in, something called internal working models of
self and others’ (25); Fraley, R.C. & Shaver, P.R. (1998) Airport separations: a naturalistic study of adult attachment
dynamics in separating couples. Journal of Personality and Social Psychology, 75(5), 1198–212. ‘Technically, the
term attachment style refers to the observable patterns of behavior exhibited by an individual, not the unobserv-
able variables (such as working models) that shape these patterns. Nonetheless, as is customary in the literature on
attachment, we use the terms attachment style and working models interchangeably’ (1199).
52 Shaver, P.R. & Mikulincer, M. (2002) Attachment- related psychodynamics. Attachment & Human
Development, 4(2), 133–61: ‘We use self-report measures in somewhat the same way that physicians use simple in-
dicators of health and illness—e.g. body temperature measured with a thermometer or verbal reports of insomnia.
Although such indicators do not provide direct access to underlying disease processes, they are very helpful in
assessing a person’s health’ (154).
mental models and psychodynamic processes’.53 This approach to self-report measures can
perhaps be placed in the context of a growing assumption within academic and clinical
psychology, which became consensus by the 1990s, that responses to self-report questions
about behaviours and attitudes would reflect underlying cognitive schemas.54
The Hazan and Shaver ‘love quiz’ was a single-item measure, asking participants to iden-
tify their relationships as corresponding to one of three adult attachment ‘styles’, modelled
on the Ainsworth infant classifications. The concept of ‘lovestyle’ had been put forward
by Hendrick and Hendrick to refer to an adult’s orientation to intimate relationships, on
the analogy of ‘lifestyle’ as a person’s characteristic way of going about living.55 Hazan and
Shaver used the concept of attachment styles in an analogous way, as a characterisation of an
adult’s orientation to romantic relationships, conceptualised as an attachment relationship.56
Hazan and Shaver assumed that ‘conscious beliefs about love . . . are coloured by underlying,

and perhaps not fully conscious, mental models’.57 While Hazan and Shaver headlined the
applicability of attachment to romantic relationships, other adult relationships (for instance
for a sibling) remained encompassed in what was, if at first implicitly, conceptualised as a
comprehensive account of close relationships in general.58
Unlike Bowlby, Hazan and Shaver did not anticipate that there would be a longitudinal
association between Ainsworth’s Strange Situation and adult mental models about relation-
ships. And unlike Main, they did not have an underlying theory of common processes that
might be in play in both the Strange Situation and adult cognition regarding attachment
(Chapter 3). Instead, they situated their extension of Ainsworth’s categories to self-report
adult attachment styles as purely pragmatic: ‘It would be naive to think that a style adopted in
infancy remains unchanged or unelaborated all through life. Still, the search for connections
between attachment in childhood and attachment in adulthood must begin somewhere.’59
The Ainsworth categories were knowingly torn up by their roots in the soil of developmental
psychology and transplanted into social psychology: Hazan and Shaver were interested to
see what might grow—though, as they would later acknowledge, Hazan and Shaver’s im-
pression of what the Ainsworth categories entailed was partial and, at points, potentially
erroneous.
The ‘love quiz’ appeared in the July 26th issue of the Rocky Mountain News, on the first
and second pages of the Lifestyle section. Participants were asked ‘Which of the following
best describes your feelings?’, and 599 participants replied. A secure attachment style was
represented by the statement: ‘I find it relatively easy to get close to others and am comfort-
able depending on them and having them depend on me. I don’t often worry about being
53 Shaver, P.R., Belsky, J., & Brennan, K.A. (2000) The adult attachment interview and self-reports of romantic
attachment: associations across domains and methods. Personal Relationships, 7(1), 25–43, p.41.
54 This approach to interpreting the meaning of self-report measures of mental health has diverse roots, but
the role of cognitive-behavioural models may be considered an important contributory. See Beck, A.T. (1991)
Cognitive therapy: a 30-year retrospective. American Psychologist, 46(4), 368–75.
55 Hendrick, C. & Hendrick, S. (1986) A theory and method of love. Journal of Personality and Social Psychology,
50, 39–42.
56 Levy, M.B. & Davis, K.E. (1988) Lovestyles and attachment styles compared: their relations to each other and
to various relationship characteristics. Journal of Social and Personal Relationships, 5(4), 439–71.
and Social Psychology, 52(3), 511–24, p.513.

58 This claim would be made explicitly a few years later in Hazan, C. & Shaver, P.R. (1994) Attachment as an or-
ganizational framework for research on close relationships. Psychological Inquiry, 5(1), 1–22, p.1.

abandoned or about someone getting too close to me.’60 Close consideration of the statement
is warranted as it contains Hazan and Shaver’s implicit theory of adult security, with implica-
tions for Shaver’s later work. As can be seen, security was not represented especially in terms
of secure base or safe haven beliefs or expectations. Instead, the focus was on the absence of
two feelings: (i) discomfort regarding closeness and (ii) worry about abandonment. In this,
the item appeared to be in part working on the analogy of the behaviour of the infant from
a secure dyad in the Strange Situation. The infant in a secure dyad approaches the caregiver
on reunion to achieve physical closeness; by analogy, secure adults might be comfortable
with emotional closeness. Lack of worry about abandonment appears to be a new element,
however. It has a less clear analogue since infants in secure dyads do prototypically display
distress on separation, besides the B1 subclassification. It might be that lack of worry about
abandonment was conceptualised as an adult version of the capacity to be fully comforted on

reunion. In the survey, 319 participants identified with the statement representing a secure
attachment style.
An avoidant attachment style was represented by the statement ‘I am somewhat uncom-
fortable being close to others; I find it difficult to trust them completely, difficult to allow
myself to depend on them. I am nervous when anyone gets too close, and often, love partners
want me to be more intimate than I feel comfortable being’ (though there are conflicting
sources regarding the exact wording used in the original study).61 In the statement as set out
in the 1987 article, the focus seems to be on: (i) discomfort regarding closeness, (ii) nervous-
ness regarding closeness, and (iii) distrust. Again the analogy with the Strange Situation is
curiously partial. The infant in an avoidant dyad does not seek physical proximity, and when
picked up may evidence discomfort. However, Ainsworth’s protocols certainly do not sug-
gest that evidence of nervousness regarding closeness should be expected, as Shaver later
acknowledged.62 In fact, 145 of Hazan and Shaver’s participants endorsed the statement rep-
resenting an avoidant attachment style.
60 Ibid. p.515.
61 Ibid. p.151. This is the statement as presented by Hazan and Shaver in their 1987 paper. However, it appears
that it was not actually the one circulated in the newspaper. The original study also included the additional phrase
‘I am comfortable having others depend on me’, though this was taken out shortly after. Shaver, P.R., Belsky, J., &
Brennan, K.A. (2000) The adult attachment interview and self-reports of romantic attachment: associations across
domains and methods. Personal Relationships, 7(1), 25–43: ‘Hazan & Shaver (1987), in the first self-report measure
of romantic attachment style, used the statement “I am comfortable having others depend on me.” They did not
realize that attachment theory does not include this comfort as a legitimate part of being attached; it is, instead,
part of being an attachment figure, or caregiver, for others (Cassidy, 1999). In later work by Hazan & Shaver (1990)
and Kunce and Shaver (1994), the potentially misleading statement was edited out of the self-report romantic
attachment measure. For good or ill, however, it had already been incorporated into self-report measures con-
structed by Bartholomew and Horowitz (1991) and Collins and Read (1990)’ (39). However, it would not figure
within the ECR scale.
62 Shaver, P.R. & Mikulincer, M. (2002) Dialogue on adult attachment: diversity and integration. Attachment &
Human Development, 4(2), 243–57: ‘The original Hazan and Shaver (1987) measure of avoidance tapped fearful,
rather than dismissing, avoidance’ (253). An additional disanalogy with Ainsworth’s categories can be identi-
fied: the avoidant attachment style was represented by Hazan and Shaver by two feelings and a belief/expectation.
Distrust in close relationships and difficulty depending on them is perhaps an analogue for Ainsworth’s idea that
avoidance is a response to experiences of less-sensitive care. However, Ainsworth did not discuss avoidance in
terms of distrust. Instead, she characterised ambivalent/resistant pattern as representing distrust in the caregiver.
Ainsworth, M.D.S., Bell, S.M., & Stayton, D.J. (1974) Infant–mother attachment and social development: ‘social-
isation’ as a product of reciprocal responsiveness to signals. In J.M. Richards (ed.) The Integration of a Child into a
Social World (pp.9–135). Cambridge: Cambridge University Press: ‘It may be viewed as advantageous for an infant
whose mother seems to him to move unpredictably and inconsistently (and whom he has not been able to learn
to trust) to monitor her movements with exceptional alertness and to evince disturbance whenever she moves
off ’ (125).
Finally, an anxious/ambivalent attachment style was represented by the statement ‘I

find that others are reluctant to get as close as I would like. I often worry that my partner
doesn’t really love me or won’t want to stay with me. I want to merge completely with an-
other person, and this desire sometimes scares people away.’63 Here the focus seems to be on
two feelings: (i) a wish for more closeness; and (ii) worry about abandonment. Once more, a
partial analogy with the Strange Situation is apparent. Whereas the ambivalent/resistant in-
fant classification in infancy includes inconsolability (understood as anxiety about caregiver
availability), anger, or passivity, the Hazan and Shaver item includes no mention of anger or
passivity. Instead, the analogy seems to be with the low threshold for activation of the infant
attachment system and high threshold for its termination. It is not clear what, except the
analogy with the infant classification, is ambivalent about the anxious/ambivalent statement
in the ‘love quiz’. In the survey, 110 participants endorsed the statement representing an anx-

ious/ambivalent attachment style. A further 25 participants checked more than one answer,
despite having been instructed to select which statement best expressed their feelings. These
participants were excluded from the analysis, a strategy that set the precedent for later re-
search using the measure.
Hazan and Shaver stated that participants self-reported expectable correlates, though ac-
tually their findings offer a somewhat more complex story regarding the avoidant attach-
ment style. Participants who endorsed the statement representing security reported that
their romantic relationships were friendly, happy, and trusting. Hazan and Shaver stated
that participants who endorsed the statement representing avoidance reported ‘fear of close-
ness’.64 They were not given options to report discomfort regarding closeness, and they did
not, in fact, report more distrust than participants endorsing the anxious/ambivalent state-
ment. According to the paper’s results, both avoidant and anxious/ambivalent participants
had more fear of closeness than secure participants, though avoidant participants scored
marginally higher.65 As such, the correlates of the avoidant attachment style were not exactly
in line with Hazan and Shaver’s expectations. Anxious/ambivalent participants reported re-
lationships marked by jealousy and emotional highs and lows. These findings were then rep-
licated in a study with 108 undergraduate participants.
It was also quickly found that the ‘love quiz’ was capable of predicting behaviour be-
yond romantic relationships. A doctoral project completed by Michelle Hutt at Cornell, co-
supervised by Hazan, used the ‘love quiz’ with 229 small business owners in the dairy and
agricultural industry, and asked participants also about their patterns of exploration (e.g. in-
novation and curiosity in their work) and reliance on others (e.g. for support, help, or delega-
tion), and asked them to respond to some vignettes with solutions to technical and personnel
problems in the business context.66 Follow-up interviews were conducted with 15 partici-
pants. Hutt found that small business owners with a secure or avoidant attachment style ran
more profitable businesses than participants with an anxious/ambivalent attachment style.
This was despite the fact that the business of participants with different attachment styles

64 Ibid. p.518.
65 Ibid. Table 3 and Table 6.
66 Hutt, M. (1991) Influences of attachment in everyday problem solving. Unpublished doctoral dissertation,
Cornell University. Shaver drew on this study in support of the predictive validity of self-report measures of as-
sessment in Shaver, P.R. & Norman, A.J. (1995) Attachment theory and counseling psychology: a commentary. The
Counseling Psychologist, 23(3), 491–500.
did not differ in terms of size, number of employees, number of production units, location,
or type of facility. Participants with an avoidant attachment style reported that they were
less likely to seek help when they had a problem with the business, and were more likely to
view their employees as ‘cogs in a machine’. Anxious/ambivalent participants stated that they
were more likely to ask employees for emotional support when they confronted a technical
problem. Hutt found that they were also more likely to have a chaotic management structure.
Participants with avoidant and anxious/ambivalent attachment styles had higher turnover
of staff than participants with a secure attachment style. They were also less likely to believe
that their colleagues understood the goals and values of the business. The Hutt dissertation
offered an early and elegant demonstration of the relevance of the ‘love quiz’ in predicting
social behaviour, even in the work context, in theoretically expectable ways.
In an exploratory study, reported in 1990, Hazan and Shaver found that anxious/am-

bivalent subjects said that they struggled to finish tasks and slacked off after praise at work;
avoidant subjects said that they were less likely to take vacations from work; and secure
participants reported fewer psychosomatic symptoms and in general fewer illnesses.67 In
another study from the 1990s, Cooper, Shaver, and Collins found that attachment style ac-
counted for 5% of variance in mental health symptoms in a large student sample, after taking
into account demographic covariates. The key findings from the study were that secure par-
ticipants reported the lowest levels of depression and of problems managing anger.68 These
findings were important for supporting the validity of the ‘love quiz’ categories, given the
centrality of mental health to the conceptualisation of attachment since Bowlby.
In the early 1990s, Mikulincer published a flurry of dazzlingly inventive studies using a
reformulation of the ‘love quiz’ measure. The Hazan and Shaver categories were reworked
as Likert-style scales along three dimensions: secure, avoidant, and anxious-ambivalent.69
These papers established his reputation as among the most ingenious, nimble, and prolific
experimental social psychologists of a talented generation. Mikulincer was promoted to
full professor at Bar-Ilan in 1992 on the basis of these accomplishments, making him the
youngest professor at any Israeli university at the time.70 Mikulincer’s particular skill was in
articulating theoretical assumptions and then swiftly operationalising these in the labora-
tory. For instance, one key question faced by the emerging social psychological tradition of
attachment research was whether self-report actually had reference to objective interaction
and the shared, social world. Responding to this question, Banai, Weller, and Mikulincer had
72 student participants self-report their attachment style and receive a judgement as to their
attachment style by other individuals, including by a friend and by a stranger asked to talk
to them for five minutes. There were marked correlations between self-reported attachment
style and the judgement made by friends and acquaintances. Friends agreed with the partici-
pant in 60% of cases, and acquaintances only slightly less in 56% of cases. This was essentially
twice the rate of agreement expected by chance. These findings provided important evidence
67 Hazan, C. & Shaver, P.R. (1990) Love and work: an attachment-theoretical perspective. Journal of Personality
and social Psychology, 59(2), 270–80, Table 3.

68 Cooper, M.L., Shaver, P.R., & Collins, N.L. (1998) Attachment styles, emotion regulation, and adjustment in
adolescence. Journal of Personality and Social Psychology, 74(5), 1380–97.

69 The astonishing volume of outputs from Mikulincer’s group in these years means that only a selection can be
mentioned. The slight adaptations to the measure enacted by Mikulincer and colleagues are detailed in Mikulincer,
M., Florian, V., & Tolmacz, R. (1990) Attachment styles and fear of personal death: a case study of affect regulation.
Journal of Personality and Social Psychology, 58(2), 273–80.
70 http://portal.idc.ac.il/faculty/en/pages/profile.aspx?username=mario.
that self-reported attachment style was a construct representing a process sufficiently mani-
fest in the individual’s behaviour that friends and acquaintances would come to the same
conclusion.71
In the 1990s, Mikulincer and colleagues reported several findings suggesting differences
in emotion regulation strategies between participants with different attachment styles.
Mikulincer and Orbach found that an avoidant attachment style was associated with lower
accessibility of negative memories. By contrast, participants who reported an anxious/am-
bivalent attachment style also had difficulty supressing negative emotions and memories.
These participants experienced that one negative emotion or memory tended to spread into
and elicit others.72 In another study, Mikulincer and Florian asked participants to complete
a stressful task: handling a snake. They found that participants with an avoidant attachment
style reported feeling more reassured by instrumental social support from a confederate,

but not emotional support. Participants with a secure attachment style found reassurance
in either form of interaction.73 By contrast, if they had an anxious/ambivalent attachment
style, participants felt no different in response to emotional support, but had an increase in
distress if they received instrumental social support. This was the first of many findings by
Mikulincer and colleagues suggesting that an anxious/ambivalent attachment style is asso-
ciated with a tendency to appraise experiences that others find reassuring or happy—such
as gratitude—as a source of inner conflict, distress, self-doubt, or bitterness.74 Such associ-
ations seem particular to the anxious/ambivalent attachment style, and have been found to
hold even after controlling for measures of self-esteem and social trust.75
In another study, Mikulincer asked participants to keep a diary over three weeks. He
found that secure participants experienced more trust in their partners and adopted more
constructive coping strategies in response to perceived violations of trust. Avoidant par-
ticipants regarded the feeling of control as one of the functions of trust in a relationship.
And anxious/ambivalent attached special importance to perceived violations of trust, and
responded to these with rumination and worry.76 The theme of coping strategies was also ex-
plored by Mikulincer, Florian, and Weller, who studied response to missile attacks on Israel
during the Gulf War. The researchers found that avoidant attachment style was associated
with reports of using coping strategies that distanced the participant from other people, but
71 Banai, E., Weller, A., & Mikulincer, M. (1998) Inter- judge agreement in evaluation of adult attachment
style: the impact of acquaintanceship. British Journal of Social Psychology, 37(1), 95–109, p.102.
72 Mikulincer, M. & Orbach, I. (1995) Attachment styles and repressive defensiveness: the accessibility and
architecture of affective memories. Journal of Personality and Social Psychology, 68(5), 917–25.
73 Mikulincer, M. & Florian, V. (1997) Are emotional and instrumental supportive interactions beneficial in
times of stress? The impact of attachment style. Anxiety, Stress, and Coping, 10, 109–127.
74 Shaver, P.R. & Mikulincer, M. (2008) Adult attachment and cognitive and affective reactions to positive and
negative events. Social and Personality Compass, 2, 1844–65, p.1857. For instance, in a study among newlyweds
from Mikulincer’s laboratory, expressions of happiness in the relationship were usually associated with positive
feelings, but for anxious participants were associated with envy. Sofer-Roth, S. (2008) Adult attachment and the
nature of responses to a romantic partner’s expression of personal happiness. Unpublished doctoral dissertation,
Bar-Ilan University, Ramat Gan, Israel.
75 Mikulincer, M., Shaver, P.R., & Slav, K. (2006) Attachment, mental representations of others, and grati-
tude and forgiveness in romantic relationships. In M. Mikulincer & G.S. Goodman (eds) Dynamics of Romantic
Love: Attachment, Caregiving, and Sex (pp.190–215). New York: Guilford.
76 Mikulincer, M. (1998) Attachment working models and the sense of trust. Journal of Personality and Social
Psychology, 74(5), 1209–224. These findings were fleshed out at the level of the dyad by a later study. Mikulincer
and Florian found that couples in which both partners were secure reported more cohesion and adaptability in
their relationship than couples in which one was insecure, who in turn reported more cohesion and adaptability
than when neither endorsed a secure attachment style. Mikulincer, M. & Florian, V. (1999) The association be-
tween spouses’ self-reports of attachment styles and representations of family dynamics. Family Process, 38, 69–83.
only among participants living in more dangerous areas. Among participants living in less
dangerous areas, there was no difference in the coping strategies used between participants
with different attachment styles.77 This finding seemed to align well with Ainsworth’s obser-
vation that avoidant attachment behaviour was only seen in the Strange Situation; the same
children did not display this conditional strategy in the lower-stress context of the home
environment (Chapter 2). By contrast, illness, like long-term separations, was regarded as
prompting a potentially chronic activation of the attachment system. Mikulincer and Florian
found that men suffering from lower back pain only had worse mental health than matched
controls when they reported an avoidant or anxious/ambivalent attachment style. Secure
participants appraised their back pain in less-threatening terms, thought of themselves as
more able to cope with pain, and used more problem-focused coping strategies.78
Together, the studies by Mikulicer and colleagues from the 1990s using the ‘love quiz’

suggested four conclusions. First, the studies offered evidence of the predictive power of
the self-report categories, including the capacity to predict concrete observed behaviours.
Second, the studies suggested that the attachment styles responded to contextual stimuli,
in the laboratory and in the real world, in ways that were aligned with proposals by Bowby
and Ainsworth. Third, the work by Mikulincer and colleagues showed that the difference
between the avoidant and anxious/ambivalent attachment styles was evident across a range
of domains, including recall of memories, interaction with their romantic partner in day-to-
day life, and coping with emergencies and illness. This included intriguing findings such as
that instrumental support provided to participants endorsing an anxious/ambivalent style
actually made them feel less reassured. Fourth, the findings of Mikulincer and colleagues
indicated that the secure attachment style was not reducible to the absence of avoidant or
anxious/ambivalent attachment, but had its own correlates, a point that will be returned to in
the section ‘The opposite of insecurity’.
Beliefs about self and others
In Separation, Bowlby had argued that internal working models contain at least two com-
ponents: expectations regarding ‘(a) whether or not the attachment figure is judged to be
the sort of person who in general responds to calls for support and protection; (b) whether
or not the self is judged to be the sort of person towards whom anyone, and the attachment
figure in particular, is likely to respond in a helpful way. Logically these variables are in-
dependent.’79 Bowlby’s term ‘independent’ inadvertently suggested to many readers the
technical meaning that models of self and other would be unrelated and orthogonal; in fact,
77 Mikulincer, M., Florian, V., & Weller, A. (1993) Attachment styles, coping strategies, and posttraumatic psy-
chological distress: the impact of the Gulf War in Israel. Journal of Personality and Social Psychology, 64(5), 817–26.
78 Mikulincer, M, & Florian, V. (1998) The relationship between adult attachment styles and emotional and cog-
nitive reactions to stressful events. In J. Simpson & S. Rholes (eds) Attachment Theory and Close Relationships
79 Bowlby, J. (1973) Separation: Anxiety and Anger. New York: Basic Books, p.238. In the manuscript version of
the book, the formulation is a little different: Bowlby, J. (not dated) Notes towards Chapter 6 of Volume 2. PP/Bow/
K.5./17: ‘in the working model of the world that anyone builds a key feature is his notion of who his attachment
figures are and how accessible or inaccessible they may be. Similarly in the working model of the self that anyone
builds a key feature is his notion of how acceptable or unacceptable he is in the eyes of his attachment figures. On
the structure of these communications turns that person’s forecasts of how available and responsive his attachment
figures will be should he turn to them for support.’
Gillath and colleagues are undoubtedly right that Bowlby meant only that the constructs
should be regarded as conceptually distinguishable.80 Furthermore, it was not a distinction
that he upheld elsewhere in his writings. This statement in Separation is an outlier: anchored
by Hinde’s focus on interactions as the basis of relationships, Bowlby tended to treat the
internal working model as rules and expectations regarding interaction between self and
others, rather than as representations of the individual interactional partners themselves
(Chapter 1).
Nonetheless, the idea of representations of the self and of others was a useful one for the
social psychological tradition, and aligned with a wider interest in social psychology in indi-
viduals’ self-concept and their attitudes towards others. In line with this distinction, Shaver
and colleagues interpreted attachment styles and their correlates as reflecting individual dif-
ferences in an underlying psychological process, namely ‘beliefs about self and relationship

partners’.81 In order to help situate their work as a kind of attachment research and to make
the link with Bowlby, Shaver and colleagues felt obliged to say that the ‘love quiz’ was tapping
individual differences in participants’ internal working models. However, in fact, Reis and
Shaver had already expressed dissatisfaction with the concept of the internal working model,
which they felt had been ‘defined so vaguely that different researchers employ rather dif-
ferent measures of it’.82 Reis and Shaver criticised Main for contributing to the further circu-
lation of this unhelpful concept, which made it hard to understand her exact claims—a point
that Main later fully accepted following feedback from Hinde (Chapter 3).83
In the 1990s, Hazan and Shaver sought to further articulate the relationship between at-
tachment and romantic love. Following Ainsworth, they argued that adult romantic love in-
cludes elements of the attachment, caregiving, and sexual behavioural systems.84 The ‘love
quiz’ might be influenced by the caregiving and sexual aspects of love, but they regarded
the measure as primarily an assessment of attachment-related beliefs about the self and ro-
mantic partners.85 They agreed with Ainsworth (Chapter 2) that what distinguished the at-
tachment system was that it related to the use of the partner as a secure base and safe haven,
and to anxiety in response to unanticipated separations. On the basis of this disaggregation
of elements of adult romantic love, Hazan and Shaver concluded that the ‘love quiz’ should
be regarded as just a first attempt to distinguish the attachment components of adult rela-
tionships. Future research, they argued, should be built upon a theory about the beliefs about
self and others at stake rather than treating their initial work as the model.86
80 Gillath, O., Karantzas, G.C., & Fraley, R.C. (2016) Adult Attachment: A Concise Introduction to Theory and
Research. London: Academic Press, p.120.

81 Reis, H.T. & Shaver, P.R. (1988) Intimacy as an interpersonal process. In S. Duck (ed.) Handbook of Research in
Personal Relationships (pp.367–89). London: Wiley, p.372.

82 Ibid.
83 Reis and Shaver suspected that Main was ultimately assessing ‘defensiveness and information-processing dis-
tortions’. Ibid. p.372.

Disorders (pp.559–602). New York: Wiley: ‘There seem to be three major behavioural systems involved in forming
and maintaining heterosexual pair bonding: 1) the reproductive or mating system, which seems likely to be ini-
tially the most important in bond formation, regardless of whether the biological function of reproduction is ful-
filled; 2) the caregiving behavioural system, which is involved in two ways—giving care to the partner and sharing
with the partner caregiving to the young that may result from the union; and 3) the attachment system, which im-
plies each partner seeks security—comfort and reassurance—through maintaining contact with the other’ (595).
85 Hazan, C. & Shaver, P.R. (1994) Attachment as an organizational framework for research on close relation-
ships. Psychological Inquiry, 5(1), 1–22.

86 Hazan, C. & Shaver, P.R. (1994) Deeper into attachment theory. Psychological Inquiry, 5(1), 68–79, p.73.
Through the 1990s, following Hazan and Shaver’s lead, there was a boom in self-reported
measures of attachment. Looking back on the period, Conradi and colleagues observe that
‘a bewildering variety of adult attachment typologies, adult attachment-related constructs,
and measurement instruments’ were developed.87 Among the most consequential was the
Bartholomew and Horowitz ‘Relationship Questionnaire’.88 This questionnaire was based on
the idea that internal working models of self and other could be either positive or negative. In
a paper from 1990, Bartholomew proposed that a secure attachment style was underpinned
by positive models of the self and relationship partners, with the self considered worthy of
care and others considered as generally available and willing to offer care. Avoidant attach-
ment was underpinned by a positive model of the self and a negative model of relationship
partners, resulting in self-reliance and distrust of others. Ambivalent/resistant attachment
was underpinned by a negative model of the self and a positive model of others, resulting in

emotional volatility and neediness. Bartholomew also introduced a new attachment classi-
fication, ‘fearful attachment’, which she conceptualised as underpinned by negative models
of both self and other. For Bartholomew, the secure attachment style was associated with
fluency and confidence in operating with both independence and intimacy. Fearful attach-
ment, the inverse, was associated with problems in achieving either: ‘They desire social con-
tact and intimacy, but experience pervasive interpersonal distrust and fear of rejection. The
result is subjective distress and disturbed social relations.’89 Bartholomew felt that the Hazan
and Shaver ‘love quiz’ had inappropriately collapsed the distinction between discomfort
with intimacy and fear of rejection of attempts to achieve intimacy, which she believed repre-
sented substantially different profiles and predicaments.
The Bartholomew four-category model and questionnaire was a landmark in the history
of self-report measures of attachment, since for the first time in this literature, categories
used to assess individual differences in attachment were conceptualised as the effect of two
underlying dimensions. Subsequently, attachment researchers in the social psychology
tradition tended to conceptualise and analyse individual differences as dimensionally dis-
tributed.90 The addition of the fearful classification appeared to be of special value in char-
acterising the attachment styles of participants experiencing adversity. This was shown by
a series of studies by Shaver and colleagues using Bartholomew’s measure. The researchers
87 Conradi, H-J., Gerlsma, C., van Duijn, M., & de Jonge, P. (2006) Internal and external validity of the experi-
ences in close relationships questionnaire in an American and two Dutch samples. European Journal of Psychiatry,
20(4), 258–69, p.259. The array of self-report attachment measures from this period include: Armsden, G.C. &
Greenberg, M.T. (1987) The inventory of parent and peer attachment: individual differences and their relationship
to psychological well-being in adolescence. Journal of Youth and Adolescence, 16(5), 427–54; West, M., Sheldon, A.,
& Reiffer, L. (1987) An approach to the delineation of adult attachment: scale development and reliability. Journal
of Nervous and Mental Disease, 175(12), 738–41; Pottharst, K. & Kessler, R. (1990) The search for methods and
measures. In K. Pottharst (ed.) Explorations in Adult Attachment (pp.9–37). New York: Peter Lang; Griffin, D.W.
& Bartholomew, K. (1994) The metaphysics of measurement: the case of adult attachment. In K. Bartholomew
& D. Perlman (eds) Advances in Personal Relationships: Vol. 5. Attachment Processes in Adulthood (pp.17–52).
London: Jessica Kingsley.
88 Bartholomew, K. & Horowitz, L.M. (1991) Attachment styles among young adults: a test of a four-category
model. Journal of Personality and Social Psychology, 61(2), 226–44.

89 Bartholomew, K. (1990) Avoidance of intimacy: an attachment perspective. Journal of Social and Personal
Relationships, 7(2), 147–78, p.163.

& W.S. Rholes (eds) Attachment Theory and Close Relationships (pp.77–114). New York: Guilford: ‘Bartholomew’s
model, combined with an emerging consensus that two latent dimensions underlie Hazan and Shaver’s (1987) and
Bartholowew’s (1990) attachment types (see Feeney & Noller 1996; Griffin & Bartholomew 1994a, 1994b; Hazan &
Shaver 1994), has encouraged researchers to use continuous measures’ (82).
found that the adult children of alcoholics predominantly fell into Bartholomew’s fearful
category. They also found that a fearful attachment style was associated, as Bowlby had an-
ticipated (Chapter 1), with reports of offering caregiving to attachment figures in order to
get some attachment needs met.91 And in a study with a large student cohort, Brennan and
Shaver found that a quarter of participants classified as fearful met criteria for personality
disorder according to a self-report assessment compared to 6% of participants with other
attachment styles.92
Shaver and colleagues concluded that Bartholomew’s fearful adult attachment style was
the parallel, and likely a developmental correlate, of Main and Solomon’s disorganised at-
tachment classification.93 They even used disorganised attachment and fearful attachment
as synonyms through the 1990s.94 However, examination of Barthlomew’s 1989 unpublished
doctoral thesis indicates an important qualification: Bartholomew regarded her fearful

classification not as a general analogue for the disorganised attachment classification, but
as an analogue for approach-avoidance conflict specifically.95 The fearful attachment style
was based on the idea that the self is regarded poorly, causing distress, and other people
are considered as a potential source of rejection. Jacobvitz insightfully observed at the time
that this predicament might result in an approach–avoidance conflict—but not necessarily
disoriented or direct apprehension regarding the caregiver.96 It would seem that the term
‘fearful attachment’ was complicit in the over-hasty comparison, suggesting a pervasive
quality of fearfulness. In fact, in Bartholomew’s theory, the category was underpinned by fear
of rejection rather than fear of the caregiver.97 Main and Stadtman had argued that whilst re-
jection by a caregiver would usually result in avoidant attachment, when the tone or context
of the rejection was generally alarming, this would be a pathway to conflict behaviour that
would be judged as disorganised attachment in the Strange Situation (Chapter 3). Rejection
was not central to Main’s conceptualisation of the causes of disorganised attachment: it is
not especially provided for by Main and Hesse among the indices of frightening/fright-
ened caregiver behaviour, or anticipated in theory to result in infant disorientation or direct
91 Brennan, K.A., Shaver, P.R., & Tobey, A.E. (1991) Attachment styles, gender, and parental problem drinking.
Journal of Social and Personal Relationships, 8, 451–66; Kunce, L.J. & Shaver, P.R. (1994) An attachment-theoretical
approach to caregiving in romantic relationships. In K. Bartholomew & D. Perlman (eds) Advances in Personal
Relationships, Volume 5 (pp.205–237). London: Jessica Kingsley.
92 Brennan, K.A. & Shaver, P.R. (1998) Attachment styles and personality disorders: their connections to each
other and to parental divorce, parental death, and perceptions of parental caregiving. Journal of Personality, 66(5),
835–78.
Journal of Social and Personal Relationships, 8, 451–66.

94 Shaver, P.R., Collins, N.L., & Clark, C.L. (1996) Attachment styles and internal working models of self and
relationship partners. In G.J.O. Fletcher & J. Fitness (eds) Knowledge Structures in Close Relationships: A Social
Psychological Approach (pp.25–61). Mahwah, NJ: Lawrence Erlbaum: ‘Disorganised, or fearful, children lack self-
confidence and have low self-worth’ (36).
95 Bartholomew, K. (1989) Attachment styles in young adults: Implications for self-concept and interpersonal
functioning. Unpublished doctoral dissertation, Stanford University.

96 Jacobvitz, D., Curran, M., & Moller, N. (2002) Measurement of adult attachment: the place of self-report and
interview methodologies. Attachment & Human Development, 4(2), 207–215: ‘The fearful attachment style seems
to differ qualitatively from the disorganized infant pattern. A mixture of avoidance and ambivalence is only one of
the many behavioral indices of attachment disorganization during infancy’ (209).
97 Bartholomew, K. (1989) Attachment styles in young adults: implications for self-concept and interpersonal
functioning. Unpublished doctoral dissertation, Stanford University. This is in contrast to the new self-report ‘dis-
organised’ attachment style introduced by Paetzold and colleagues in 2015, which does seek to assess fear of at-
tachment figures. Paetzold, R.L., Rholes, W.S., & Kohn, J.L. (2015) Disorganized attachment in adulthood: theory,
measurement, and implications for romantic relationships. Review of General Psychology, 19(2), 146–65.
apprehension regarding the caregiver in the Strange Situation.98 There was also negligible
association between a fearful attachment style and lack of resolution on the AAI.99
For a time in the 1990s, it seemed as if Bartholomew’s theoretical model and, to a lesser
extent, her questionnaire, would become the dominant approach to adult attachment in the
social psychology tradition. The approach found authorisation in a passage from Bowlby’s
writings. It permitted the transition to dimensional approaches to measurement, which
was becoming the norm in social psychology by the 1990s: the disadvantages of categories
for statistical analyses were becoming increasingly widely known among attachment re-
searchers, such as reductions in statistical power, decreases in scale reliability, and difficul-
ties identifying non-linear relationships with other variables.100 And the division between
positive/negative and self/other produced a neat model from which hypotheses regarding
adult attachment styles could be generated and results interpreted. For instance, findings

that avoidant attachment style was associated with stronger agreement that condoms pro-
tect against sexually transmitted diseases could be interpreted, in light of Bartholomew’s
approach, as fuelled by negative associations regarding intimacy with romantic others.101
Furthermore, Bartholomew’s model seemed able to encompass infant and adult attachment
together within the same frame, a serious asset given that the social psychology tradition was
hoping to achieve acceptance from attachment researchers in developmental psychology.
Finally, the addition of the fearful category was perceived as an advantage. This addition
helped address an ambiguity in the operationalisation of avoidance in Hazan and Shaver’s
original ‘love quiz’. It offered a parallel to the four-category model that was appearing in de-
velopmental psychology in the wake of Main’s innovations. And the fearful category seemed
to offer a way for the self-report system to gain applicability for research with clinical samples.
Though it is certainly still used today, from the 2000s Bartholomew’s model increas-
ingly lost ground to the two-dimension model of avoidance and anxiety underpinning the
Experiences in Close Relationships (ECR) scale. The introduction of this scale is discussed in
the section ‘Creating the ECR’. However, as well as the availability of the ECR, Batholomew’s
model faced three salient limitations, which were pointed out by Shaver and his collabor-
ators. First, the account of avoidant and anxious/ambivalent attachment was generally not
well supported. An avoidant attachment style was not associated with a positive model of
the self, but only with a desire to avoid acknowledging vulnerability. And this desire was
relatively brittle, resulting in associations between an avoidant attachment style and self-
blame when things go wrong.102 Furthermore, an anxious/ambivalent attachment style was
not associated with a positive model of the other as ‘available and supportive’ as expected by
Bartholomew, but with representations of others that were shaped by anger, jealousy, and a
98 Duschinsky, R. (2018) Disorganization, fear and attachment: working towards clarification. Infant Mental

99 Roisman, G.I., Holland, A., Fortuna, K., Fraley, R. C., Clausell, E., & Clarke, A. (2007) The Adult Attachment
Interview and self-reports of attachment style: an empirical rapprochement. Journal of Personality and Social
Psychology, 92(4), 678–97.
100 Simpson, J.A. (1990) Influence of attachment styles on romantic relationships. Journal of Personality and
Social Psychology, 59(5), 971–80; Collins, N.L. & Read, S.J. (1990) Adult attachment, working models, and re-
lationship quality in dating couples. Journal of Personality and Social Psychology, 58(4), 644–63; Fraley, R.C. &
Waller, N.G. (1998) Adult attachment patterns: a test of the typological model. In J.A. Simpson & W.S. Rholes (eds)
Attachment Theory and Close Relationships (pp.77–114). New York: Guilford.
101 Feeney, J.A., Peterson, C., Gallois, C., & Terry, D.J. (2000) Attachment style as a predictor of sexual attitudes
and behavior in late adolescence. Psychology & Health, 14(6), 1105–122.

102 E.g. Gillath, O., Canterberry, M., & Atchley, P. (2017) Attachment as a predictor of driving performance.
Transportation Research Part F, 45, 208–217.

sense of partners as insensitive to their needs.103 This was hardly surprising, since the anx-
ious/ambivalent attachment style had been operationalised specifically on the basis of es-
poused beliefs that romantic others are not adequately available.
Second, some of the intuitive pull and defensibility of Bartholomew’s approach came from
the terms ‘positive’ and ‘negative’, which could have a variety of meanings. However, this was
also a disadvantage in terms of usefulness for prediction.104 For instance, Mikulincer found
that a secure attachment style did not appear to be associated merely with a positive working
model of the self, but with one that was well balanced between acknowledgment of positive
and negative aspects.105 It was not clear whether or not this constituted a falsification of the
Bartholomew model precisely because it was difficult to pin down what was meant by a ‘positive’
working model.
Third, the Bartholomew questionnaire appeared to have weak cross-cultural validity, with

the four categories not operating as expected along the two dimensions. In western samples,
there was little overlap between participants endorsing items representing security and repre-
senting fearful attachment. By contrast, in African and Asian samples, there was a marked posi-
tive correlation between security and fearfulness, which ran counter to Bartholomew’s theory
of positive and negative working models. Furthermore, whereas Bartholomew conceptualised
avoidance and anxious/ambivalence as opposites, in fact they were negatively correlated only in
a minority of countries around the world.106
Shaver and colleagues acknowledged that ‘it seems unlikely that Ainsworth’s research
would have had such a profound influence on the field if she had eschewed types in favor of
dimensions’ since ‘in the absence of a model of underlying processes, dimensions generally
do not inspire theoretical advances’.107 Batholomew had circumvented the problem since
she had offered a model of underlying processes in formulating her dimensions. Yet Fraley
and Shaver argued that the dimensions Bartholomew identified were ultimately not the right
ones.108 They hoped that further explicit discussion and debate about the dimensions under-
lying differences in adult attachment styles, and further psychometric inquiry, would help
shed light on the relevant underlying processes.109
103 Bartholomew, K. & Shaver, P.R. (1998) Methods of assessing adult attachment: do they converge? In J.A.
Simpson & W.S. Rholes (eds) Attachment Theory and Close Relationships (pp.25–45). New York: Guilford: ‘The
“positivity of the other” model indicates the degree to which others are generally expected to be available and sup-
portive.’ Yet an anxious attachment style was found empirically to be associated with negative representations of
others. Simpson, J.A., Rholes, W.S., & Phillips, D. (1996) Conflict in close relationships: an attachment perspective.
Journal of Personality and Social Psychology, 71(5), 899–914. Extensive criticism of Bartholomew’s assumption was
presented in Fraley, R.C. & Shaver, P.R. (2000) Adult romantic attachment: theoretical developments, emerging
controversies, and unanswered questions. Review of General Psychology, 4(2), 132–54.
104 Stein, H., Jacobs, N.J., Ferguson, K.S., Allen, J.G., & Fonagy, P. (1998) What do adult attachment scales
measure? Bulletin of the Menninger Clinic, 62, 33–82.

105 Mikulincer, M. (1995) Attachment style and the mental representation of the self. Journal of Personality and
Social Psychology, 69(6), 1203–215.

106 Schmitt, D. P., Alcalay, L., Allensworth, M., et al. (2004) Patterns and universals of adult romantic attach-
ment across 62 cultural regions: are models of self and of other pancultural constructs? Journal of Cross-Cultural
Psychology, 35(4), 367–402.
Journal of Social and Personal Relationships, 8, 451–66, p.465.

108 Crowell, J.A., Fraley, R.C., & Shaver, P.R. (1999) Measures of individual differences in adolescent and adult
attachment. In J. Cassidy & P.R. Shaver (eds) Handbook of Attachment: Theory, Research, and Clinical Applications
(pp.434–65). New York: Guilford; Fraley, R.C. & Shaver, P.R. (2000) Adult romantic attachment: theoretical devel-
opments, emerging controversies, and unanswered questions. Review of General Psychology, 4(2), 132–54.
109 E.g. Fraley, R.C. & Waller, N.G. (1998) Adult attachment patterns: a test of the typological model. In J.A.
Simpson & W.S. Rholes (eds) Attachment Theory and Close Relationships (pp.77–114). New York: Guilford: ‘An
explicit focus on the latent dimensions, however, may facilitate inquiry into the underlying operation of the
Avoidance and anxiety
Creating the ECR
Though Bartholomew and colleagues were able to source the idea of working models of the
self and other in Bowlby’s writings, other accounts of security and insecurity were available
too. At times, Ainsworth used the term ‘anxious attachment’ as a synonym for insecure at-
tachment.110 This may have been motivated by a desire to remind readers that both infants in
avoidant and ambivalent/resistant dyads showed anxiety about their caregiver’s availability
in the home environment, and that it was only in the Strange Situation that the anxiety of

the former group was masked. The use of ‘anxious attachment’ as a synonym for ‘insecure
attachment’ was generally picked up by her immediate students and colleagues in the 1980s
(with the exception of Main, who preferred the latter term). When discussing the Strange
Situation, Bowlby sometimes used the term ‘anxious attachment’ to refer to the avoidant
and ambivalent/resistant dyads, following Ainsworth.111 However, in the 1970s, Bowlby de-
veloped a different use of the term. This is most clearly seen in Separation where Chapter 15
is dedicated to an elaboration of the concept of anxious attachment. Already by the 1950s
he had distinguished two classes of response to major separations: one that was manifest
as clinging and protest, and another seen as detachment and avoidance of the caregiver
(Chapter 1). Now, in Separation, he offered the term ‘anxious attachment’ as a label for the
first of these two classes, with a particular emphasis on the vigilance regarding the caregiver’s
availability: ‘Some children subjected to an unpredictable régime seem to despair. Instead
of developing anxious attachment, they become more or less detached, apparently neither
trusting nor caring for others.’112 This use of the term continued throughout the rest of his
career. For instance, in Loss he argued that ‘one of the commonest forms of disturbance is
the overready elicitation of attachment behaviour, resulting in anxious attachment. Another,
to which special attention is given in this volume, is a partial or complete deactivation of at-
tachment behaviour.’113
As Mikulincer and colleagues observed in 1990, the term ‘anxiety’ is potentially a very
ambiguous one. It can imply various elements including worry, anger, and emotionality.114
Furthermore, as Ainsworth’s label of ambivalent/resistant attachment suggested, there may
be a potential link with ambivalence (though Shaver felt that avoidant attachment was just
as much a state of ambivalence regarding the attachment figure).115 However, despite such
attachment system rather than remaining at the level of manifest behaviour . . . We believe that the prototype ap-
proach leaves the ontological status of the attachment patterns unclear. Are the prototypes advocated by Griffin
and Bartholomew supposed to represent “fuzzy” groups that exist in nature or “fuzzy” groups that exist in the
minds of perceivers of nature?’ (107).
110 E.g. Ainsworth, M.D.S. (1983) Patterns of infant– mother attachment as related to maternal care: their
early history and their contribution to continuity. In D. Magnusson & V.L. Allen (eds) Human Development: An
111 E.g. Bowlby, J. (1969, 1982) Attachment. London: Penguin, p.365.
114 Mikulincer, M., Paz, D., & Kedem, P. (1990) Anxiety and categorization—2. Hierarchical levels of mental cat-
egories. Personality and Individual Differences, 11(8), 815–21.

115 See e.g. Brennan, K.A. & Shaver, P.R. (1995) Dimensions of adult attachment, affect regulation, and romantic
relationship functioning. Personality and Social Psychology Bulletin, 21(3), 267–83, p.280.
Avoidance and anxiety 449
problems and ambiguities, the term ‘anxiety’ had been used by Bowlby, and this contrib-
uted to its appeal to Shaver and colleagues. The term would also have had an accessibility
and intuitiveness, given its growing use and presence in public discourse from the 1940s to
the early 1990s.116 Ultimately, the two classes identified by Bowlby were conceptualised by
Shaver and colleagues from 1998 onwards as reflecting two dimensions in individual dif-
ferences regarding close relationships: ‘The first dimension, Avoidance, captures variability
in the tendency to feel uncomfortable with closeness or dependence. The second dimen-
sion, Anxiety, reflects a fear of abandonment’.117 Over the subsequent decade, the opposition
came to be formulated rather differently: ‘The first dimension, typically called attachment
avoidance, reflects the extent to which a person distrusts relationship partners’ goodwill and
strives to maintain autonomy and emotional distance from partners. The second dimen-
sion, typically called attachment anxiety, reflects the degree to which a person worries that

a partner will not be available in times of need.’118 However, contained in both early and late
definitions is a technical sense of ‘anxiety’ to mean vigilance regarding the availability of
attachment figures. In these definitions, ‘anxiety’ may be associated with anger and ambiva-
lence, or indeed with general experiences or symptoms of anxiety, but these are not elements
that formed the essence of the technical concept for Shaver and colleagues.119
In 1998, Brennan, Clark, and Shaver reflected that Shaver and Hazan ‘naively took for
granted that Ainsworth et al. (1978) were correct in thinking of attachment patterns as cat-
egories or types. In retrospect, it is evident that Hazan and Shaver should have paid attention
to Ainsworth et al.’s (1978) Figure 10 (p. 102), which summarized the results of a discrim-
inant analysis predicting infant attachment type (secure, anxious, or avoidant) from the con-
tinuous rating scales.’120 Brennan, Clark, and Shaver were right that a dimensional approach
to attachment had been napping peacefully on the backseat of Patterns of Attachment, even
while the categories had control of the vehicle. However, close examination reveals that
Ainsworth’s data did not completely support the claims of Shaver and colleagues to have
identified a forerunner for their account of attachment in terms of two dimensions. True, in
the discriminant function analysis in Patterns of Attachment, a first function had comprised
essentially scores for avoidance. However, the second function was mainly constituted by
scores on the resistance scale on first and second reunion, and with crying through the two
reunion episodes. The function was as much or more constituted by anger or frustration
as by anxious distress. Furthermore, whilst the two-function model reported in Patterns of
116 Wilkinson, I. (2001) Anxiety in a Risk Society. London: Routledge; Horwitz, A.V. (2013) Anxiety: A Short
History. Baltimore: Johns Hopkins University Press.

117 Fraley, R.C. & Shaver, P.R. (1998) Airport separations: a naturalistic study of adult attachment dynamics in
separating couples. Journal of Personality and Social Psychology, 75(5), 1198–212, p.1199.
118 Mikulincer, M., Shaver, P.R., & Slav, K. (2006) Attachment, mental representations of others, and grati-
tude and forgiveness in romantic relationships. In M. Mikulincer & G.S. Goodman (eds) Dynamics of Romantic
Love: Attachment, Caregiving, and Sex (pp.190–215). New York: Guilford, p.192.
119 There is some lack of clarity on this point of theory, however. In general, attachment anxiety is understood to
be the degree to which a person worries that a partner will not be available in times of need. However, at other times
it is defined as anger and ambivalence, with worries fading into the background, e.g. Shaver, P.R. & Mikulincer, M.
(2005) Attachment theory and research: resurrection of the psychodynamic approach to personality. Journal of
Research in Personality, 39, 22–45: ‘Hyperactivating strategies reflect a compromise between conflicting, ambiva-
lent tendencies toward attachment figures—overwhelming anger and hostility toward unavailable attachment fig-
ures together with an intense need for proximity to these frustrating figures’ (28).
120 Brennan, K.A., Clark, C.L., & Shaver, P.R. (1998) Self-report measurement of adult romantic attachment: an
integrative overview. In J.A. Simpson & W.S. Rholes (eds) Attachment Theory and Close Relationships (pp.46–76).
New York: Guilford, p.47; Ainsworth, M., Blehar, M., Waters, E., & Wall, S. (1978, 2012) Patterns of Attachment: A
Psychological Study of the Strange Situation. Bristol: Psychology Press.
Attachment was superb at predicting avoidant and secure dyads, it misclassified 30% of am-
bivalent/resistant dyads.
The conceptualisation of anxiety by Shaver and colleagues as vigilance regarding the avail-
ability of attachment figures was in part a theoretically motivated position, in which fear of
abandonment was understood to be primary and to generate the other behavioural displays
Ainsworth had identified.121 However, this conceptualisation of anxiety coincides with and
seems to have been primarily influenced by a landmark piece of empirical research con-
ducted with Kelley Brennan and Catherine Clark. The researchers created a pool of 323 ques-
tionnaire items from the different self-report measures of attachment available, which aimed
to assess 60 named attachment-related constructs. These 323 items were then completed by
1,086 undergraduates enrolled at the University of Texas. When the items were factored,
the result was 12 specific attachment-related dimensions, each with at least 10 items that

could produce viable scales. The 12 scales ‘which can be viewed as “facets” (Costa & McCrae,
1992) of Avoidance and Anxiety, are as follows: (1) Partner is a Good Attachment Figure;
(2) Separation Anxiety; (3) Self-Reliance; (4) Discomfort with Closeness; (5) Attachment-
Related Anger at Partners; (6) Uncertainty About Feelings for Partners; (7) Discomfort with
Dependence; (8) Trust; (9) Lovability/Relational Self-Esteem; (10) Desire to Merge with
Partners; (11) Tough-Minded Independence; and (12) Fear of Abandonment.’122 A higher-
order factor analysis of the 12 scales then revealed two underlying dimensions. The two-
factor model accounted for 63% of variance.
At an item level, items reflecting the constructs ‘avoidance of intimacy’, ‘discomfort with
closeness’, and ‘self-reliance’ were most strongly representative of the first factor, which was
termed ‘Avoidance’. Items reflecting ‘preoccupation’, ‘fear of abandonment’, and ‘fear of re-
jection’ were most strongly representative of the second factor, which was termed ‘Anxiety’.
Rather than part of an avoidant attachment style as assumed by Hazan and Shaver or an
independent category as anticipated by Bartholomew, fear of rejection was in fact identi-
fied by the factor analysis as one of the three most important elements of the new ‘Anxiety’
dimension. Brennan, Clark, and Shaver found that the correlation between the ‘Avoidance’
and ‘Anxiety’ factors was only .12, ‘suggesting that the dimensions underlying attachment
styles are essentially orthogonal’.123 The correlation appeared to be accounted for by a few
constructs that loaded on both factors. These included perception of the unavailability of
partners, perception of the self as unworthy of care, and a negative relationship with attach-
ment security.
On the basis of the factor analysis, Brennan, Clark, and Shaver extracted 36 questionnaire
items from the pool to craft the ECR scale. The items chosen were those that loaded most
strongly on one or the other dimension, with the selection of 36—as opposed to 20 or 40—
decided abitrarily.124 The stated aim of the ECR was to halt the raucous proliferation of new
self-report measures of attachment, finding a neutral and common ground for the social
121 Ibid.: ‘Right from the start, Ainsworth’s three major attachment “types” could be conceptualized as regions
in a two-dimensional space, the dimensions being Avoidance (discomfort with closeness and dependency) and
Anxiety (crying, failing to explore confidently in the absence of mother, and angry protest directed at mother dur-
ing reunions after what was probably experienced as abandonment)’ (49).
122 Ibid. pp.66–7.
123 Ibid. p.57.
124 This is implied but not detailed in the chapter. Confirmed by Kelly Brennan-Jones, personal communication,
August 2019.
psychology tradition of attachment research through ‘discovering and describing the es-
sence’ of the self-report measures created to date.125 The ECR was also framed by the authors,
in part as a result of its neutrality, as a supersession of Bartholomew’s theoretical model in
presenting evidence for ‘anxiety’ and ‘avoidance’ as the two latent dimensions across all
the other self-report measures. The measure was successful on both counts. Though other
self-report measures continued to be used, the ECR soon became the ‘industry standard’
self-report assessment of attachment, and its two latent dimensions became the dominant
explanatory model in the field.126
In 2003, a place was also reintroduced for Bartholomew’s ‘fearful’ attachment category by
Mikulincer and Shaver. In Bartholomew’s model, fearful attachment occurred when a re-
spondent was afraid of rejection by attachment figures and also had a negative model of their
own worthiness. On the ECR, fearful attachment was re-envisioned as the co-occurrence

of anxiety and avoidance. Mikuliner and Shaver’s stated reason for excluding fearful at-
tachment as an independent category is quite curious in retrospect: ‘The fearfully avoidant
self-description on the Bartholomew and Horowitz (1991) categorical measure . . . provides
too low a hurdle for classification into the least secure category, and this creates too large
a discrepancy between the clinical literature on disorganized attachment and the social-
psychological literature.’127 Mikulincer and Shaver’s conceptualisation of fearful attachment
as the coincidence of anxiety and avoidance, then, was an attempt to retain the bridge to
clinical concerns and to the developmental psychology tradition, but with a narrower (and
therefore hopefully more clinically meaningful) cut-off. This stance was also a development
of Mikulincer’s longstanding interest in the intrusion of distress and fear into ordinary func-
tioning. Already in his early work on learned helplessness in the early 1990s, Mikulincer had
documented that anxiety can undermine an avoidant coping strategy, producing a state he
called ‘disorganisation’.128
An anxious attachment style and an avoidant attachment style both, as Main had ar-
gued (Chapter 3), can be conceptualised as strategies, with the expectable result of offering
some conditional form of the benefits of attachment relationships. However, the anxious
strategy will be blocked if, when attachment or threat cues are present, a person fails to
pursue the availability of their attachment figure. And the avoidant strategy will be under-
mined if it is intruded upon by anxiety or other overwhelming demands. Crawford, Shaver,
and Goldsmith discovered that an avoidant attachment style was only associated with lower
scores for attachment anxiety for participants with low or moderate levels of trait neuroticism
125 Frıas, M.T., Shaver, P.R., & Mikulincer, M. (2014) Measures of adult attachment and related constructs. In G.J.
Boyle & D.H. Saklofske (eds) Measures of Personality and Social Psychological Constructs. Philadelphia: Elsevier,
417–47, p.437.
126 Pittman, J.F. (2012) Attachment orientations: a boon to family theory and research. Journal of Family Theory
& Review, 4(4), 306–310, p.308.

(pp.53–152). New York: Academic Press, p.136.
128 Mikulincer, M. (1994) Human Learned Helplessness: A Coping Perspective. New York: Plenum Press: ‘People
for whom emotion acts as an internal stimulus that recalls both the unresolved mismatch and their own weakness
and vulnerability may perceive emotion as an uninvited intruder that counteracts their avoidance coping. They
may also experience their emotions as overwhelming and disorganising forces that demand that they resolve the
mismatch, which is precisely what they feel they cannot do. In this case, emotions do not facilitate adaptation
and mastery but rather reflect failure of control and flooding, and they may disorganise the person’s avoidance
activities’ (143).
(susceptibility to becoming distressed and worried). For those participants who became dis-
tressed and worried easily, an avoidant attachment style had the opposite effect, in that it was
actually associated with increased scores for attachment anxiety.129 The authors interpreted
this finding as suggesting that the avoidant attachment style contributes to attachment anx-
iety, rather than providing an alternative or defence against it, for individuals prone to be-
coming distressed and worried. Likewise, Mikulincer and colleagues demonstrated that an
avoidant attachment style, which under ordinary circumstances is associated with low ac-
cessibility of attachment-related worries, in fact is associated with heightened accessibility
of these worries when a cognitive load is added.130 The avoidant attachment style depends
upon an infrastructure of attentional and regulatory control, even if this infrastructure is not
the avoidant style itself.
In general, Shaver and Mikulincer conceptualised fearful attachment as a breakdown of

attachment style, rather than having content in its own right.They repeatedly interpreted
Main and Solomon’s disorganised attachment classification as ‘random fluctuations’ of anx-
ious and avoidant behaviour.131 This, as Chapter 3 documented, is a misunderstanding, likely
based on a lack of knowledge about how the construct is coded in practice.132 The status of
‘fearful attachment’ in the thinking of Mikulincer and Shaver has been neglected, perhaps
because it too has been regarded as a state of meaninglessness. There has been little con-
cern to inquire regarding the nature, form, and implications of fearful attachment. Indeed, it
has not generally been regarded as necessary to test for fearful attachment or report results,
leading in turn to little impetus for the design of research to investigate the phenomenon.
For example, as Ein-Dor, Mikulincer, and Shaver put matters in a paper from 2011 about re-
sponses to threatening stimuli: ‘We did not include fearful avoidance in our analysis (i.e., a
person above the median on both anxiety and avoidance) because attachment theory views
this pattern as a breakdown of attachment strategies rather than a coherent strategy in its
own right.’133
129 Crawford, T.N, Shaver, P.R., & Goldsmith, H.H. (2007) How affect regulation moderates the association be-
tween anxious attachment and neuroticism. Attachment & Human Development, 9(2), 95–109.
130 Mikulincer, M., Birnbaum, G., Woddis, D., & Nachmias, O. (2000) Stress and accessibility of proximity-
related thoughts. Journal of Personality and Social Psychology, 78(3), 509–523. See also Edelstein, R.S. & Gillath, O.
(2008) Avoiding interference: adult attachment and emotional processing biases. Personality and Social Psychology
Bulletin, 34(2), 171–81.
131 Mikulincer, M. & Shaver, P. (2016) Attachment in Adulthood, 2nd edn. New York: Guilford, p.143. For an-
other example see Mikulincer, M. & Shaver, P.R. (2003) The attachment behavioral system in adulthood: activation,
psychodynamics, and interpersonal processes. In M.P. Zanna (ed.) Advances in Experimental Social Psychology,
Vol. 35 (pp.53–152). New York: Academic Press: ‘Recently, a fourth category, “disorganized/disoriented,” has been
added. It is characterized by odd, awkward behavior during separation and reunion episodes and random fluctu-
ations between signs of anxiety and avoidance’ (66).
132 In discussions of the phenomenon, Mikulincer, Shaver, and colleagues have been prone to cite social
psychologists in turn discussing developmental psychologists, rather than the relevant primary literature, e.g.
Mikulincer, M., Solomon, Z., Shaver, P.R., & Ein-Dor, T. (2014) Attachment-related consequences of war cap-
tivity and trajectories of posttraumatic stress disorder: a 17-year longitudinal study. Journal of Social and Clinical
Psychology, 33(3), 207–228: ‘In extreme cases, attachment insecurities can result in a disorganized attach-
ment pattern—an incoherent blend of contradictory approach and avoidance behaviors or paralyzed inaction
(Simpson & Rholes, 2002)’ (210). Simpson, J.A. & Rholes, W.S. (2002) Fearful-avoidance, disorganization, and
multiple working models: some directions for future theory and research. Attachment & Human Development,
4(2), 223–9.
133 Ein-Dor, T., Mikulincer, M., & Shaver, P.R. (2011) Effective reaction to danger: attachment insecurities pre-
dict behavioral reactions to an experimentally induced threat above and beyond general personality traits. Social
Psychological and Personality Science, 2(5), 467–73, p.473.
Minimising and maximising
From the early 2000s, one of the central aspects of the emergent collaboration between Shaver
and Mikulincer was an attempt to develop a fleshed-out theory of the attachment system in
adulthood, and the processes underlying individual differences in avoidance and anxiety. In
this, they draw inspiration from developments in theory within the developmental tradition of
attachment research. Main had conceptualised avoidance and ambivalent/resistance as condi-
tional strategies, part of the human evolutionary repertoire for responding to environments that
do not permit the direct satisfaction of the attachment system (Chapter 3). Cassidy and Kobak
took Main’s concept of the infant attachment patterns as conditional strategies and argued that
they represented two approaches to emotion-regulation: avoidant attachment was a minimis-

ing of the attachment system and of associated emotions which had the expectable outcome of
avoiding rebuff and retaining regulation; ambivalent/resistant attachment was a maximising of
the attachment system and associated emotions in the service of maintaining the attention of the
caregiver.134
Shaver had worked on the 1999 edition of the Handbook of Attachment with Cassidy, and
had become impressed by her ideas. In the early 2000s, Shaver and Mikulincer took her image
of minimising and maximising of the attachment system as a central building block of their
theory. They conceptualised the attachment system in adulthood as containing three mod-
ules.135 Together these form a cadence. The first and most basic module was the one described
by Bowlby, and is based on the potential for perceiving threats. The first module is structured by
the question: ‘Are there signs of threat?’ There can, naturally, be various sources of threat. Shaver
and Mikulincer observed that these might include external dangers, such as the unexpected
unavailability of attachment figures, the identification of a form of personal vulnerability. They
might include internal states such as illness or worry about mortality. Shaver and Mikulincer
proposed that symbolic threats to status, identity, or freedom would also be relevant to the first
module of the attachment system.136 Where a sign of threat appears, the attachment system is
activated, leading to ‘seeking proximity to external or internalised attachment figure’.137
Though the term ‘proximity’ was used by Shaver and Mikulincer, they did not mean phys-
ical proximity in the manner of Bowlby, but actually the sense of being close to a ‘source of
felt security’, which might be physical or symbolic.138 Activation of the attachment system
triggers a second module, which builds on the work of Ainsworth and Sroufe. The second
module is structured by the question: ‘Is the attachment figure available, attentive, respon-
sive, etc.?’ The etcetera signals Shaver and Mikulincer’s belief that the parameters for ex-
periencing felt security in adulthood may be calibrated by factors including context and
134 Cassidy, J. & Kobak, R. (1988) Avoidance and its relation to other defensive processes. In J. Belsky & T.
Neworski (eds) Clinical Implications of Attachment (pp.300–323). Hillsdale, NJ: Lawrence Erlbaum; Kobak,
R., Cole, H., Fleming, W., Ferenz-Gillies, R., & Gamble, W. (1993) Attachment and emotion regulation during
mother–teen problem-solving: a control theory analysis. Child Development, 64, 231–45.
135 Mikulincer, M. & Shaver, P. (2016) Attachment in Adulthood. New York: Guilford, Figure 2.1, p.29.
136 Shaver, P.R. & Mikulincer, M. (2012) An attachment perspective on coping with existential concerns. In
P.R. Shaver & M. Mikulincer (eds) Meaning, Mortality, and Choice: The Social Psychology of Existential Concerns
137 Mikulincer, M. & Shaver, P. (2016) Attachment in Adulthood. New York: Guilford, p.29, Figure 2.1.
138 Mario Mikulincer, personal communication, July 2019.
culture, as well as the nature of the perceived threat. It is not only physical proximity.139
Where the set-goal is achieved, Shaver and Mikulincer proposed that the result will be feel-
ings of ‘relief, and positive affect’,140 grounded in the sense of being understood, validated,
and cared about.141 These feelings in turn will contribute to confidence and competence,
feeding back into the calibration of the first module, most notably in the appraisal of threats.
Following Fredrickson, Shaver and Mikulincer referred to this as the ‘broaden-and-build
cycle of attachment security’, since felt security, relief, and positive affect in turn come to
shape the interpretation of threats, allowing for a more expansive, less cautious appraisal of
the environment.142 Positive affects may also, as Sroufe documented, serve as social currency
(Chapter 4). Recently, Mikulincer and Shaver qualified that Bowlby and Ainsworth’s concern
with infancy has meant that too much has subsequently been made of an image of security
as a state of being with attachment figures; in adulthood, broaden-and-build may also entail

increased capacity to be comfortable with solitude, entering and exiting it without difficul-
ties or shame, and benefiting from time alone in creative and flexible ways.143 There may be
experiences of security, at least in adulthood, that are made for one.
The third module is initiated if the attachment figure is not ‘available, attentive, respon-
sive, etc.’. This module is structured by the question: ‘Is proximity-seeking a viable option?’
Again, literal proximity is a potential meaning here. However, Shaver and Mikulincer ap-
pear to intend more the question of whether available, attentive, responsive (etc.) care can
be expected to be achieved through directly seeking it. If yes, then the attachment system is
‘hyperactived’, in the form of ‘hypervigilance regarding threat-and attachment-related cues’.
As for Main, there need be no conscious ‘decision’ regarding the deployment of a condi-
tional strategy; in fact, it is a selection that generally occurs without conscious reflection.
Hyperactivation entails a lowering of the threshold for activation of the first and second
modules and a raising of the threshold for termination. This may entail, most directly, vigi-
lance regarding internal or external information suggesting reasons for concern or alarm.
Feelings such as jealousy, helplessness, and vulnerability will help maintain this vigilance.144
139 Sketching a tightening of this position, Mikulincer has recently described the set-goal of the attachment
system in adulthood as evidence that the partner is, specifically, ‘available, responsive and engaged’ (A.R.E.). This
riffs on the same acronym with slightly different components—‘accessibility, responsiveness, and emotional en-
gagement’—offered by Johnson, S.M. (2011) Hold Me Tight: Your Guide to the Most Successful Approach to Building
Loving Relationships. London: Hachette, Chapter 4. Mikulincer, M. (2019) Advances in the study of the broaden
and build cycle of attachment security. Presentation at the Adult Attachment Research Legacy: 32 Years Since
Hazan and Shaver Symposium, Society of Personality and Social Psychology Conference, Portland, Oregon, 9
February 2019.
140 Mikulincer, M. & Shaver, P. (2016) Attachment in Adulthood. New York: Guilford, p.29, Figure 2.1.
141 Shaver, P.R. & Mikulincer, M. (2008) Augmenting the sense of security in romantic, leader– follower,
therapeutic, and group relations: a relational model of personality change. In J.P. Forgas & J. Fitness (eds) Social
Relationships: Cognitive, Affective, and Motivational Processes (pp.55–73). New York: Psychology Press. Shaver and
Mikulincer attributed the idea of ‘security’ as entailing feeling understood, validated, and cared about to Reis,
H.T. & Shaver, P.R. (1988) Intimacy as an interpersonal process. In S. Duck (ed.) Handbook of Research in Personal
Relationships (pp.367–89). London: Wiley. However, the argument in this text is that these are characteristics of ‘in-
timacy’. The textual history illustrates the potential slide between intimacy and security at times in Shaver’s work.
142 Fredrickson, B.L. (2001) The role of positive emotions in positive psychology. The broaden-and-build theory
of positive emotions. American Psychologist, 56(3), 218–26.

143 Mikulincer, M. & Shaver, P. (2014) An attachment perspective on loneliness. In R.J. Coplan & J.C. Bowker
The Handbook of Solitude: Psychological Perspectives on Social Isolation, Social Withdrawal, and Being Alone
(pp.34–50). New York: Wiley, p.46.
144 Shaver, P.R., Mikulincer, M., Lavy, S., & Cassidy, J. (2009) Understanding and altering hurt feelings: an
attachment-theoretical perspective on the generation and regulation of emotions. In A.L. Vangelisti (ed.) Feeling
Hurt in Close Relationships. Cambridge: Cambridge University Press, pp.92–119: ‘Attachment anxiety intensifies
the expression of emotions such as jealousy and anger and exaggerates the expression of vulnerability, helplessness,
and need’ (109).
So will maintenance of pessimistic beliefs about whether coping alone is possible, and the
attribution of threatening events to uncontrollable causes or inherent personal vulnerabil-
ities.145 The difference between making a life and facing problems becomes collapsed.
Self-defeating actions may also keep the attachment system at ready alert: for instance,
Mikulincer’s group documented associations on several occasions between the anxious at-
tachment style and an odd escalation of investment and commitment as an activity becomes
socially and/or economically unprofitable.146 As a result of various mechanisms that increase
vigilance regarding threat-and attachment-related cues, the availability, attentiveness, re-
sponsiveness (etc.) of attachment figures is sought more readily in response to potential in-
formation about threat, and is sought in a manner more insistent, distressed, and frantic.
This is likely to cause a tendency towards a cascade of thinking and worry about threats,
vulnerabilities, and/or attachment, as one perception or thought will cue others linked to

related feelings.147 The combination of vigilance and such a flow of associations may in turn
contribute to the further identification of potential threats and reasons to wonder about the
availability of attachment figures, burnishing the hyperactivating attachment strategy.148
On the other hand, Shaver and Mikulincer argued that if seeking the attachment figure’s
availability is not appraised as viable, then the attachment system is ‘deactivated’, in the form
of ‘distancing of threat-and attachment-related cues’.149 In Shaver and Mikulincer’s model,
deactivation entails a raising of the threshold for activation of the first and second modules
and a lower threshold for termination. In the formal presentation of their model of the at-
tachment system, the selection of a conditional strategy is based on whether seeking the at-
tachment figure’s availability is viable. However, there is significant ambiguity here. In other
works, Shaver and Mikulincer acknowledged that selection of a strategy also depends on
the perceived viability of the other strategy.150 As such, hyperactivation of the attachment
system may in part reflect appraisals regarding the viability of seeking availability, and in
145 Shaver, P.R. & Mikulincer, M. (2007) Adult attachment strategies and the regulation of emotion. In J.J. Gross
(ed.) Handbook of Emotion Regulation (pp.446–65). New York: Guilford, p.454.

146 E.g. Jayson, Y. (2004) An attachment perspective to escalation of commitment. Unpublished doctoral dis-
sertation, Bar-Ilan University; Mikulincer, M. & Shaver, P. (2008) Contributions of attachment theory and re-
search to motivation science. In J.Y. Shah & W.L. Gardner (eds) Handbook of Motivational Science (pp.201–216).
New York: Guilford; Erez, A., Sleebos, E., Mikulincer, M., van IJzendoorn, M.H., Ellemers, N., & Kroonenberg,
P.M. (2009) Attachment anxiety, intra-group (dis)respect, actual efforts, and group donation. European Journal of
Social Psychology, 39(5), 734–46.
147 Shaver, P.R. & Mikulincer, M. (2003) The psychodynamics of social judgments: an attachment theory per-
spective. In J.P. Forgas, K.D. Williams, & W. von Hippel (eds) Social Judgments: Implicit and Explicit Processes
(pp.85–114). Philadelphia: Psychology Press: ‘Hyperactivating strategies create a chaotic, undifferentiated mental
architecture that is constantly pervaded by negative affect’ (104).
148 A further factor stabilising the anxious attachment style was proposed, speculatively, in 2009: Mikulincer,
M., Shaver, P.R., Cassidy, J., & Berant, E. (2009) Attachment-related defensive processes. In J.H. Obegi & E.
Berant (eds) Attachment Theory and Research in Clinical Work with Adults. New York: Guilford, pp.293–327: ‘For
attachment-anxious people, histrionics may seem to have two beneficial effects. First, they sometimes elicit the
desired attention, care, and love from others, which is, theoretically, the reason the anxious strategy was adopted in
the first place. But there may be a second, less obvious, benefit: The hubbub and distraction generated by strident,
impulsive expressions of pain, need, and anger may direct attention and energy away from a deeper problem—
sensing oneself as not very substantial at all and not worthy of anyone’s care. Agitating and grabbing someone’s
attention is at least likely to make something happen, and even if that something is unpleasant, it may feel better
than nothing—that is, better than existential isolation and worthlessness’ (309). This proposal has not been much
elaborated and remains as yet untested, except insofar as there exists correlational evidence linking the anxious
attachment style to self-report of feelings of isolation and self-report of lower self-esteem.
149 Mikulincer, M. & Shaver, P. (2016) Attachment in Adulthood. New York: Guilford, Figure 2.1, p.29.
150 See e.g. Crawford, T.N, Shaver, P.R., & Goldsmith, H.H. (2007) How affect regulation moderates the associ-
ation between anxious attachment and neuroticism. Attachment & Human Development, 9(2), 95–109.
part the extent to which deactivation is possible. More speculatively, deactivation of the at-
tachment system may in part reflect appraisals regarding the viability of seeking availability,
and in part the extent that hyperactivation is possible.151
In Shaver and Mikulincer’s model, when the attachment system is subject to deactiva-
tion, information incongruent with the deactivation of the attachment system is inhibited
or appears less salient. For instance, the information itself may be altered, obstructed, or
suppressed; or response to the information may be dampened, redirected, or postponed.152
This not only includes emotions associated with threat, vulnerability, or separation, which
might reactivate the attachment system. Similiarly inhibited or reduced in salience are happy
emotions like joy and comfort, which might otherwise promote interpersonal closeness or
investment in close relationships.153 Someone may be radiating love, and all that is felt is
mild warmth. Deactivation also means that the availability, attentiveness, responsiveness

(etc.) of attachment figures is sought less readily, less insistently, and to a lesser extent, with a
tendency towards taking cognitive, emotional, and physical distance from relationship part-
ners. In turn, the individual feels less emotionally assailable. Self-control is held tightly, per-
haps even with a white-knuckle grip, as both the condition and experience of the defensive
exclusion of the attachment system. As a consequence, self-control can itself be felt as a form
of comfort, as a substitute for attachment set-goals.
The deactivation of the attachment system is likely to cause difficulties moving between
information about threats, vulnerabilities, and/or attachment, as one perception or thought
is less likely to cue others linked to feelings.154 This reduces opportunities for identification
of threats and for feedback indicating that attachment figures are available, attentive, respon-
sive (etc.), contributing to a stabilisation of the deactivating strategy. The strategy was under-
stood by Mikulincer and Shaver to be implicated across the whole range of attachment-based
experiences, even those that generally remain private. For instance, Mikulincer, Shaver, and
colleagues conducted a study in which they asked participants to report their dreams each
morning for a month, as well as other information about how easy or difficult they found
their day. The higher a participant’s score for avoidant attachment, the less likely they were to
report dreams that included any active support-seeking. The higher a participant’s score for
anxious attachment, the more likely it was that participants would report dreams in which
151 Mikulincer, M., Shaver, P.R., & Pereg, D. (2003) Attachment theory and affect regulation: the dynamics,
development, and cognitive consequences of attachment-related strategies. Motivation and Emotion, 27(2),
77–102.
152 Shaver, P.R. & Mikulincer, M. (2007) Adult attachment strategies and the regulation of emotion. In J.J. Gross
(ed.) Handbook of Emotion Regulation (pp.446–65). New York: Guilford, p.450. See also Edelstein, R.S. (2006)
Attachment and emotional memory: investigating the source and extent of avoidant memory impairments.
Emotion, 6(2), 340–45; Simpson, J.A., Rholes, W.S., & Winterheld, H.A. (2010) Attachment working models
twist memories of relationship events. Psychological Science, 21(2), 252–9. Neuroimaging studies conducted by
Gillath and colleagues have been interpreted as suggesting that there is no attachment-specific suppressive mech-
anism, but that in the case of the avoidant attachment style, general mental capacities for suppression are ap-
plied to attachment-related content. Gillath, O., Bunge, S.A., Shaver, P.R., Wendelken, C., & Mikulincer, M.
(2005) Attachment-style differences in the ability to suppress negative thoughts: exploring the neural correlates.
Neuroimage, 28(4), 835–47.
153 Schachner, D.A., Shaver, P.R., & Mikulincer, M. (2005) Patterns of nonverbal behavior and sensitivity in the
context of attachment relations. Journal of Nonverbal Behavior, 29(3), 141–69, p.157.

(pp.85–114). Philadelphia: Psychology Press, p.106.
they had wanted to be close to another person, and that this was especially likely following
more stressful days.155
It has evidently been a source of frustration for Shaver and Mikulincer that character-
isations of their work by attachment researchers from the developmental tradition have
often assumed that a self-report measure could reflect merely conscious attitudes towards
relationships. As well as research on dreams, another study they pursued was an examin-
ation of the relationship between the ECR and the Rorschach inkblot test used extensively
by Ainsworth in her clinical work and early research. The Rorschach is a projective test as-
sessing unconscious psychodynamic processes. Participants’ associations on the Rorschach
measure were blind coded. It was found that Rorschach responses suggesting difficulties in
controlling emotion and of feeling unworthy were associated with self-reported attachment
anxiety. And Rorschach responses suggesting difficulties in acknowledging personal needs

were associated with self-reported attachment avoidance. Associations were moderate, in
the region of r = .32 to .34. Berant, Mikulincer, Shaver, and Segal interpreted the findings
as suggesting that non-conscious projections in interpreting an ambiguous but evocative
stimulus in part express self-report attachment styles. As such, they argued, the self-reports
are not merely reflecting conscious attitudes, but tap non-conscious processing as well.156
Shaver, Mikulincer, and colleagues argued that each module of the attachment system is
informed not by a static mental representation but by a hierarchical associative network, in
which specific episodic memories become placed as exemplars of more generalised schemas
about how relationships work.157 This associative network for each module is continually
updated in light of social encounters, and also in light of information from the other mod-
ules. So, for instance, the appraisal of whether the attachment figure is available, attentive,
responsive (etc.) will be informed by a variety of factors. Three, above all, are given prece-
dence: how often and in what contexts threats have been experienced; feelings of felt security,
relief, and positive affect from times when attachment needs were met; and tendencies to
hyperactive or deactive the attachment system from occasions when these conditional strat-
egies were employed and proved sufficiently successful. Shaver and Mikulincer propose that,
given the existence of a hierarchical associative network, ‘with respect to a particular re-
lationship and across different relationships, everyone has models of security-attainment,
hyperactivation, and deactivation and so can sometimes think about relationships in secure
155 Mikulincer, M., Shaver, P.R., Sapir-Lavid, Y., & Avihou-Kanza, N. (2009) What’s inside the minds of securely
and insecurely attached people? The secure-base script and its associations with attachment-style dimensions.
156 Berant, E., Mikulincer, M., Shaver, P.R., & Segal, Y. (2005) Rorschach correlates of self-reported attach-
ment dimensions: dynamic manifestations of hyperactivating and deactivating strategies. Journal of Personality
Assessment, 84(1), 70–81.
157 This proposal was in part an elaboration of Mikulincer’s earlier emphasis on the structure of the human mind
as a set of strategic resources rather than constituted by static representations. Mikulincer, M., Paz, D., & Kedem,
P. (1990) Anxiety and categorization—2. Hierarchical levels of mental categories. Personality and Individual
Differences, 11(8), 815–21. The image of internal working models as a hierarchical associative network has been
criticised by Fraley as, even if of heuristic value, ultimately an overrigid reification. Fraley, R.C. (2007) A connec-
tionist approach to the organization and continuity of working models of attachment. Journal of Personality, 75(6),
1157–80. Fraley and colleagues have also demonstrated that people who have more differentiated attachment
styles across their varying relationships are more likely to have an insecure overall attachment style. Fraley, R.C.,
Heffernan, M.E., Vicary, A.M., & Brumbaugh, C.C. (2011) The Experiences in Close Relationships—Relationship
Structures Questionnaire: a method for assessing attachment orientations across relationships. Psychological
Assessment, 23(3), 615–25.
terms and at other times think about them in more hyperactivating or deactivating terms’.158
Nonetheless, the combination of (i) the predominant form of actual social experiences and
(ii) self-reinforcing cycles resulting from the adoption of particular strategies from child-
hood onwards will contribute to associative networks that are especially or chronically avail-
able. This produces a person’s characteristic attachment style. In characterising individual
differences in attachment style, the ECR was conceptualised as a kind of snorkelling: though
wholly submerged in manifest, espoused information about behaviours, beliefs, and feel-
ings, it nonetheless is believed to breathe latent information from a hierarchical associative
network offering generalised schemas about how relationships work.
In line with this model, Overall and colleagues asked participants to complete self-
reports of attachment style for various different family relationships, friendships, and ro-
mantic relationships. They found that the attachment orientations for specific relationships

were somewhat independent but nested within overarching tendencies towards a par-
ticular attachment style for each domain. Furthermore, each domain was somewhat in-
dependent but nested within an overall attachment style.159 Interpreting these findings,
Shaver and Mikulincer described overall attachment styles as relatively stable, but respon-
sive to changes in relevant social or psychological context.160 Social experiences are often
stable over time as a result of structural factors in people’s lives. These are then reinforced
by cycles of the broaden-and-build, minimising or maximising strategies. However, some
relationships or components of relationships may feature experiences that run counter to
the dominant attachment style. And over time, repeated social experiences or forms of ap-
praisal that reinforce this style can become pervasive across the hierarchical associative
network. This is illustrated well by an interesting unpublished doctoral study conducted
in Mikulincer’s group. Lavi pursued a longitudinal study of 73 dating couples over eight
months. She used an observational assessment at the start of the period, examining each
partner’s sensitivity and responsiveness to the other. Relationship-specific and general at-
tachment style was assessed at three time points. Lavi found that observed sensitivity and
responsiveness predicted declines in attachment insecurity. First this occurred in relation
to the specific relationship partner. Over time, however, there was a change in the partici-
pant’s general attachment style.161
Mikulincer argued that this elaboration of attachment theory gains its usefulness and par-
ticular value in ‘integrating different, perhaps even contradictory, views of human nature
and maintaining a dialectical tension between opposites of four kinds’:162
(pp.53–152). New York: Academic Press, p.64.
159 Overall, N.C., Fletcher, G.J., & Friesen, M.D. (2003) Mapping the intimate relationship mind: comparisons
between three models of attachment representations. Personality and Social Psychology Bulletin, 29(12), 1479–93.
160 Shaver, P.R. & Mikulincer, M. (2004) What do self-report attachment measures assess? In W.S. Rholes &
J.A. Simpson (eds) Adult Attachment: Theory, Research, and Clinical Implications (pp.17–54). New York: Guilford,
pp.52–3.
161 Lavi, N. (2007) Bolstering attachment security in romantic relationships: the long- term contribution of
partner’s sensitivity, expressiveness, and supportiveness. Unpublished doctoral dissertation, Bar-Ilan University,
Ramat Gan, Israel. For a recent review of the effect of partners on attachment style in adulthood see Arriaga, X.B. &
Kumashiro, M. (2019) Walking a security tightrope: relationship-induced changes in attachment security. Current
Opinion in Psychology, 25, 121–6.
162 Mikulincer, M. (2006) Attachment, caregiving and sex within romantic relationships. In M.
1) The shaping and constraining influences of past experiences versus the influence of
current contexts and experiences;
2) the intrapsychic nature of behavioural systems and working models versus the rela-
tional, interdependent nature of feelings, experiences, and social behaviours;
3) the goal-oriented, promotive, expansive, self-regulatory function of behavioural
system versus their defensive, protective, distress-regulating functions;
4) the centrality of basic fears, conflicts, and prevention-focused motivational mechan-
isms, as well as promotion-focused motives.163
Shaver and Mikulincer’s model of the attachment system can be usefully considered in these
terms. The appraisal of the potential effectiveness of seeking the availability of the attach-
ment figure is prompted in module one. The question that must be answered in module two

is whether the goal of this system—an attachment figure who is available, attentive, respon-
sive (etc.)—should be sought directly. This represents the hinge between ethology and per-
sonality psychology.164 If the goal of the system can be met directly, then this contributes to
a broaden-and-build cycle that can support self-regulation. The availability of a secure base
and safe haven are expected to contribute to increased richness and stability in both an indi-
vidual’s inner world and social relations. If not, then module three enacts a defensive, pro-
tective, and distress-regulating strategy, which may also have promotive and self-regulatory
potential in some circumstances.
The maximising strategy, on the one hand, promotes threat-sensitivity and the promo-
tion of self-regulation through attempts to achieve the availability of attachment figures. The
minimising strategy, on the other hand, promotes composure under low and moderate stress
and the promotion of self-regulation under conditions in which attachment figures are ap-
praised as not available. The selection of one strategy or another is shaped in part by learning
from previous social encounters and the strategies used to respond to them, as well as avail-
able resources for emotion regulation. Yet the selection of a strategy is also fed in part by the
actual nature of the present social interaction, cultural norms, and relational cues regarding
the potential effectiveness of particular strategies in the particular moment. Individuals do
not have, in Shaver and Mikulincer’s work, a single representational model of relationships.
Instead, hierarchically organised associational networks make various attachment styles
available. As such, they depend upon: the qualities of a particular relationship; previous
experiences of and expectations about this kind of relationship; and the particularities of
present-day interactions. That these sit nested within and interpreted through an overall at-
tachment style shaped by the weight of personal history does not exclude this internal diver-
sity and dynamic potential.
Shaver and Mikulincer claimed their elaboration of attachment theory as an integra-
tion of components of developmental, social, and personality psychology. This was in line
with Bowlby, who always described attachment theory as an account of the relational de-
velopment of personality.165 However, it was in contrast to the developmental tradition of
163 Ibid.
164 Fraley, R.C. & Shaver, P.R. (2008) Attachment theory and its place in contemporary personality re-
search. In O. John & R.W. Robins (eds) Handbook of Personality: Theory and Research, 3rd edn (pp.518–41).
New York: Guilford: ‘The attachment behavioral system is an important concept in attachment theory because
it provides the conceptual bridge between ethological models of human development (e.g., Hinde, 1966) and
modern theories of emotion regulation and personality (e.g., John & Gross, 2007)’ (523).
165 E.g. Bowlby, J. (1988) The role of attachment in personality development. In A Secure Base (pp.134–54).
London: Routledge.
attachment research. Among developmental researchers, appeals to ‘personality’ have been

relatively rare and mostly appear in summaries of Bowlby.166 Yet in claiming attachment
styles as part of personality psychology, the question was inevitably raised regarding how
avoidant and anxious attachment styles relate to established personality constructs. And
more pointedly, whether they added anything. An attempt to answer this concern was
undertaken by Noftle and Shaver in a paper published in 2006. They examined the relation-
ship between the ECR and a self-report measure of the ‘big five’ constructs in personality
psychology: openness to experience, conscientiousness, extraversion, agreeableness, and
neuroticism. There were few major associations between the avoidant attachment style and
the personality constructs. By contrast, in two studies Noftle and Shaver found associations
between attachment anxiety and neuroticism of r = .42 and r = .52. The extent of this as-
sociation suggested to the researchers that there is ‘some conceptual overlap between the

constructs’.167 The strongest associations were for items for neuroticism that pertained to the
susceptibility and frequency of negative affect (items for: being depressed, likely to be moody
and nervous, excessively worrying, and not being emotionally stable). This suggests that the
readiness of negative affect plays a key role in the overlap between the two psychological con-
structs. The readiness of negative affect is certainly shaped by culture and experiences past
and present. However, it is also widely regarded by psychologists as partly a heritable trait.168
Following up this expectation, Crawford, Shaver, and colleagues reported in 2007 from a
study of 239 twin pairs. Biometric models identified no contribution of heritable factors to
the likelihood of an avoidant attachment style. However, 40% of variance in anxious attach-
ment could be attributed to heritable factors. Furthermore, 63% of the association between
anxious attachment and a self-report measure of personality disorder could be accounted for
by common genetic effects. Such findings suggest that anxious attachment, and its associ-
ation with mental health, may in part be an effect of heritable differences in the susceptibility
and frequency of negative affect.169 However, Crawford, Shaver, and colleagues also argued
that there are important differences between anxious attachment style and neuroticism.
Firstly, anxious attachment is especially responsive to attachment-relevant cues, not merely
generally anxiety-provoking stimuli. Secondly, there is still a substantial amount of variance
not explained by heritable factors. Thirdly, even when neuroticism is statistically controlled,
166 Second-generation developmental attachment researchers have generally considered ‘personality’ a reified
concept in the academic psychology of their day, better conceptualised as an effect of developmental processes ra-
ther than a unitary construct to be measured in itself. See e.g. Sroufe, L.A. & Fleeson, J. (1986) Attachment and the
construction of relationships. In W. Hartup & Z. Rubin (eds) Relationships and Development (pp.51–71). Hillsdale,
NJ: Lawrence Erlbaum, p.51.
167 Noftle, E.E. & Shaver, P.R. (2006) Attachment dimensions and the big five personality traits: associations and
comparative ability to predict relationship quality. Journal of Research in Personality, 40(2), 179–208, p.187.
168 John, O.P., Neumann, L., & Soto, C.J. (2010) The Big Five trait taxonomy: history, measurement, and concep-
tual issues. In L.A. Pervin & O.P. John (eds) Handbook of Personality: Theory and Research, 3rd edn (pp.114–58).
New York: Guilford; Vukasović, T. & Bratko, D. (2015) Heritability of personality: a meta-analysis of behavior gen-
etic studies. Psychological Bulletin, 141(4), 769–85.
169 Crawford, T.N., John Livesley, W., Jang, K.L., Shaver, P.R., Cohen, P., & Ganiban, J. (2007) Insecure attach-
ment and personality disorder: a twin study of adults. European Journal of Personality, 21(2), 191–208. See also
Donnellan, M.B., Burt, S.A., Levendosky, A.A., & Klump, K.L. (2008) Genes, personality, and attachment in
adults: a multivariate behavioral genetic analysis. Personality and Social Psychology Bulletin, 34(1), 3–16. In con-
trast to Crawford and colleagues, Donnellan and colleagues reported that genetic effects accounted for variability
also in avoidant attachment style. Troisi and colleagues have reported potential gene × environment interactions in
the origins of fearful attachment. Troisi, A., Frazzetto, G., Carola, V., et al. (2012) Variation in the μ-opioid receptor
gene (OPRM1) moderates the influence of early maternal care on fearful attachment. Social Cognitive and Affective
Neuroscience, 7(5), 542–7.
anxious attachment has been found to make a modest contribution to mental health out-
comes such as depression and complex responses to bereavement.170 At a fundamental
level, the experience of anxiety is solely an expression of neither individual personality nor
attachment style. Both seem to play a role. However, this highlights the question of what
convergence there is between the anxious attachment style and preoccupied states of mind
regarding attachment, if both are conceptualised as maximising strategies. To address this
question it is necessary to examine the ECR and the AAI in the context of social psychology
and developmental psychology as two different subdisciplines concerned with attachment.
Two traditions of attachment research

As we have seen, Bowlby anticipated that both the AAI and the Hazan and Shaver ‘love quiz’
would both be predicted by the Strange Situation. This was also Shaver’s explicit position by the
late 1990s: ‘some of their components, especially ability to depend on attachment figures, should
be related if both stem from a person’s attachment history’.171 The first major attempt to assess
whether the AAI and self-report measures would converge was published by Crowell, Treboux,
and Waters in 1999. In their study, 81% of participants classified as secure-autonomous on the
AAI identified themselves as having a secure attachment style on Bartholomew’s Relationship
Questionnaire. However, only 42% of participants who were classified as preoccupied, dis-
missing, or unresolved on the AAI reported themselves as insecure on the Relationship
Questionnaire.172 Crowell and colleagues therefore cautioned against equation of the AAI and
self-report measures. They argued that it was a recipe for confusion that both systems used a ver-
sion of Ainsworth’s terminology, such as ‘security’, when in fact the referents were quite different.
Whereas Main was assessing ‘states of mind regarding attachment’, the self-report tradition was
assessing a person’s sense of ‘comfort in close relationships’.173 There might be expected to be a
little empirical overlap between these phenomena, but Crowell and colleagues argued that they
were completely different constructs.174
It was not certain what the two traditions could mean for another. Nonetheless, the ex-
pectation that both the ECR and AAI were measures of ‘attachment’ implied that it should
be possible to translate between the two measures, and that the two traditions should be to a
170 E.g. Roberts, J.E., Gotlib, I.H., & Kassel, J.D. (1996) Adult attachment security and symptoms of depres-
sion: the mediating roles of dysfunctional attitudes and low self-esteem. Journal of Personality and Social
Psychology, 70(2), 310–20; Meier, A.M., Carr, D.R., Currier, J.M., & Neimeyer, R.A. (2013) Attachment anxiety
and avoidance in coping with bereavement: two studies. Journal of Social and Clinical Psychology, 32(3), 315–34,
Supplement 3; Wijngaards-de Meij, L., Stroebe, M., Schut, H., et al. (2007) Neuroticism and attachment insecurity
as predictors of bereavement outcome. Journal of Research in Personality, 41(2), 498–505.
attachment: associations across domains and methods. Personal Relationships, 7(1), 25–43, p.25.
172 Crowell, J.A., Treboux, D., & Waters, E. (1999) The Adult Attachment Interview and the Relationship
Questionnaire: relations to reports of mothers and partners. Personal Relationships, 6(1), 1–18.
173 Ibid. p.16.
174 This was a consensus position among active attachment researchers from the developmental tradition by the
end of the 1990s. See e.g. Stein, H., Jacobs, N.J., Ferguson, K.S., Allen, J.G., & Fonagy, P. (1998) What do adult at-
tachment scales measure? Bulletin of the Menninger Clinic, 62, 33–82: ‘Both approaches have something important
to offer, but are likely to be looking at different, though valid, constructs. The self-report measures are consist-
ently measuring an aspect of intimate relationships that relates to individual perceptions of how relationships are
managed’ (77).
large extent permeable to one another.175 In a 1998 book chapter, Bartholomew and Shaver
offered a characterisation of the two traditions of attachment research. They described the
developmental tradition as mostly led by direct students of Ainsworth. In Bartholomew
and Shaver’s characterisation, ‘researchers in this group tend to think psychodynamically,
be interested in clinical problems, prefer interview measures and behavioural observations
over questionnaires, study relatively small groups of subjects, and focus their attention on
parent–child relationships’. It is possible to quibble with some particulars—Kobak was con-
ducting research on parent–adolescent relationships; Waters had little psychodynamic basis
for his thinking except as mediated through Ainsworth and Bowlby—but in general this
characterisation seems fair.176 By contrast, the social psychology tradition was described as
shaped essentially by the characteristics of social and personality psychology of the period.
Members of this community ‘tend to think in terms of personality traits and social inter-

actions, be interested in normal subject populations, prefer simple questionnaire measures,
study relatively large samples, and focus on adult social relationships’.177
The division between the two traditions was in part fed, it should be acknowledged, by
the longer-standing professional tensions within the academic discipline between develop-
mental and social/personality psychology. As Lapsley and Quintana observed, writing in
the period in which the two traditions of attachment research were emerging, ‘it is hard to
imagine an academic division of labor that is more formidable, and at the same time more
artificial, than the division between social and developmental psychology’.178 Commenting
on the division in the late 1990s, Bartholomew and Shaver noted that ‘not surprisingly, the
members of these two research subcultures tend to speak past each other’.179 Shaver and
Mikulincer added, in 2002, that another key difference between the traditions was that the
developmental tradition only rarely tested causal propositions experimentally, whereas the
use of experimental designs was a central part of the social psychology tradition, thanks in
part to the work of Mikulincer. Furthermore, the developmental tradition tended to rely on
rather cluttered coding systems supported by predictive validity but lacking rigorous psy-
chometric analysis of categories and scales. By contrast, the social psychological tradition
had a better track-record of cleaning and refining their measures to ensure psychometric
rigour. Shaver and Mikulincer also argued that whereas their work shed light on the actual
functioning of ‘the attachment behavioural system’ itself, by contrast the AAI was a measure
of ‘individual differences in “state of mind with respect to attachment” ’.180 They claimed, in
175 Carnelley, K.B. & Brennan, K.A. (2002) Building bridges. Attachment & Human Development, 4(2), 189–
92: ‘The biggest obstacle to the fledgling field of attachment theory appears to be a lack of co-operation among
researchers wielding different measurement techniques. As it is, each researcher appears to be using a favorite
measure and readers of their work are left to make the attempt to translate the results into their own measurement
rubric before interpreting the meaning of the findings’ (191).
Simpson & W.S. Rholes (eds) Attachment Theory and Close Relationships (pp.25–45). New York: Guilford, p.27.
177 Ibid.
178 Lapsley, D.K. & Quintana, S. (1985) Integrative themes in social and developmental theories of the self. In
J. Pryor & J. Day (eds) The Development of Social Cognition (pp.153–78). New York: Springer, p.154. Also on the
state of broader relations between developmental and social psychology in the period of the emergence of the
two traditions of attachment research see Masters, J.C. & Yarkin-Levin, K. (1984) Boundary Areas in Social and
Developmental Psychology. New York: Academic Press.
Simpson & W.S. Rholes (eds) Attachment Theory and Close Relationships (pp.25–45). New York: Guilford, p.27.
Development, 4(2), 133–61: ‘The AAI and CRI are focused entirely on individual differences in “state of mind with
respect to attachment” and therefore do not reveal much about the normative workings of the attachment behav-
ioral system’ (155).
an unabashed polemic, that the AAI was likely not an assessment of attachment at all. Rather,
given that it was validated against the Strange Situation, it might be better conceived as an
assessment of cognitive aspects of the caregiving system. This is a claim they would later re-
tract in 2004, only to swing to the other pole in asserting that both the AAI and self-report
measures taped the same processes: ‘attachment-style scales differ from the AAI in content
and method but not in the core attachment-related processes they index’.181
Yet in addition to specific methodological differences, Shaver and Mikulincer claimed
that the most important difficulty was ‘professional ingroup–outgroup tensions between
researchers in the two traditions’.182 In fact, looking back on the period, it seems as if pro-
fessional and methodological tensions worked neatly to reinforce one another. The de-
velopmental tradition tended to regard the social psychologists as conducting quick-fix,
superficial research on conscious attitudes towards intimacy, lacking fidelity to the etho-

logical basis of attachment theory. The social psychology tradition tended to regard the
developmentalists as lashed to Main’s four categories, and unable or unwilling to think crit-
ically or creatively about them. They also expressed scepticism whether a research agenda
based on laborious observational and interview-based methodologies was practical in the
context of contemporary academic psychology.183 This latter point certainly has purchase.
One of the contributing factors to the growth of the social psychological tradition of attach-
ment research has been that the entry requirements are much lower. The ECR is freely avail-
able and easy to administer on student populations and through online platforms, putting
it within the reach of researchers even without a research grant or with only a side-interest
in attachment. This cannot be said of the Strange Situation and AAI, which require sub-
stantial training and knowledge to be able to code effectively, or sometimes even interpret
published results. However, recent developments like Fearon and colleagues’ cut-down AQS
(Chapter 2), Dozier and colleagues’ cut-down AAI (Chapter 6), Madigan and colleagues’s
cut-down assessment of frightening and disrupted caregiving (Chapter 6), and Ensink and
colleagues’ short assessment of parent reflective functioning184 may reduce the resource
burden of the developmentalists’ measures, if their validation proves successful.
Through the 1990s, the developmental and social psychologists blamed one other for
failing to conduct comparative research. The developmentalists regarded it as rather unfor-
givable, and as an expression of a quick-fix mentality, that the social psychologists did not
undertake the training to conduct observational research even for a measure as foundational
as the Ainsworth Strange Situation. The social psychologists found it rather unforgivable
that, even though it would have been quick and easy to do, the developmentalists did not
include self-report measures of attachment when assessing parents in the Strange Situation,
when conducting the AAI, or in the adolescent and adult phases of their longitudinal
Development, 4(2), 133–61, p.134.
integrative overview. In J.A. Simpson & W.S. Rholes (eds) Attachment Theory and Close Relationships (pp.46–76).
New York: Guilford, p.46. See also Sassenberg, K. & Ditrich, L. (2019) Research in social psychology changed
between 2011 and 2016: larger sample sizes, more self-report measures, and more online studies. Advances in
Methods and Practices in Psychological Science, 2(2).
184 Ensink, K., Borelli, J.L., Roy, J., Normandin, L., Slade, A., & Fonagy, P. (2019) Costs of not getting to know
you: lower levels of parental reflective functioning confer risk for maternal insensitivity and insecure infant attach-
ment. Infancy, 24(2), 210–27.
studies.185 At times these tensions operated in the background. So, for instance, research
on parenting using self-report measures of attachment has remained quietly blunted, with
leaders of the social psychological tradition giving the topic little attention.186 In part this
was an effect of the division between social and developmental psychology in general, in
which parenting is generally assigned to the latter. However, another factor may have been
that, at best, self-report measures would draw with the AAI in predicting parenting, and
there was the possibility that they would lose.187
Yet, at times, the tensions between the social and developmental traditions of attach-
ment research moved from mutual inattention to a public disagreement. In the mid-1990s,
Crowell and Waters described Shaver’s work as ‘thoughtful and provocative’ and as ‘exciting
and potentially very useful’, but stated their view that in fact there is ‘little common ground’
with the tradition initiated by Ainsworth.188 In responding to the hostile reception from

the developmental tradition, one tactic used by Shaver and colleagues to find mandate for
their work was to offer anachronistic depictions of the field’s founders to make Bowlby
and Ainsworth foreshadow their own work. For instance, Ainsworth has been repeatedly
described by Shaver as using the concept of ‘a child’s attachment style’189, and Shaver and
Mikulincer explicitly asserted that ‘Bowlby claimed that memorable interactions with others
throughout life can alter a person’s working models and move him or her from one region of
the two-dimensional attachment-style “space” to another’.190 These anachronisms reflected
a claim to the right to inherit attachment theory from the first generation without using the
observational methods that had come at least in part to constitute the basis for inheritance.
They show the desire for legitimacy and heritage among the social psychological researchers,
in the context of a general lack of acceptance—or even threats of symbolic dispossession—by
the developmental tradition.
Despite trouble with the developmentalists, use of self-report measures of attachment was
booming by the 1990s. Some researchers from the developmental tradition, such as Belsky
and Cassidy, expressed concern that the social psychology tradition ‘might actually over-
shadow the work in infancy and childhood that gave rise to it’, and proposed that this threat
gave added impetus to the need to achieve a ‘linkage of these two schools of attachment in-
quiry’.191 In this context, Seymour Weingarten, editor-in-chief of Guilford Press, approached
185 Shaver, P.R & Mikulincer, M. (2002) Dialogue on adult attachment: diversity and integration. Attachment &

186 Though see e.g. Rholes, W.S., Simpson, J.A., & Friedman, M. (2006) Avoidant attachment and the experience
of parenting. Personality and Social Psychology Bulletin, 32(3), 275–85.

187 A meta-analysis by Lo and colleagues, however, found more studies using self-report measures (12) than
using the AAI (5). The odds ratio for child maltreatment for self-report measures was 2.75; the odds ratio for the
AAI was 5. Lo, C.K., Chan, K.L., & Ip, P. (2019) Insecure adult attachment and child maltreatment: a meta-analysis.
Trauma, Violence, & Abuse, 20(5).
188 Crowell, J.A. & Waters, E. (1994) Bowlby’s theory grown up: the role of attachment in adult love relation-
ships. Psychological Inquiry, 5(1), 31–4, pp.33–4. See also Kobak, R. (2009) Defining and measuring of attachment
bonds: comment on Kurdek (2009). Journal of Family Psychology, 23(4), 447–9.
189 E.g. Shaver, P.R. & Mikulincer, M. (2010) Mind–behavior relations in attachment theory and research. In
C.R. Agnew, D.E. Carlston, W.G. Graziano, & J.R. Kelly (eds) Then a Miracle Occurs: Focusing on Behavior in Social
Psychological Theory and Research (pp.342–67). New York: Oxford University Press: ‘Ainsworth et al. (1978) pro-
vided persuasive evidence for the impact of parental behavior on the formation of an infant’s attachment style’
(358). See also Levy, K.N., Blatt, S.J., & Shaver, P.R. (1998) Attachment styles and parental representations. Journal
of Personality and Social Psychology, 74(2), 407–419, p.408.
190 Shaver, P.R. & Mikulincer, M. (2012) An attachment perspective on coping with existential concerns. In
P.R. Shaver & M. Mikulincer (eds) Meaning, Mortality, and Choice: The Social Psychology of Existential Concerns
191 Belsky, J. & Cassidy, J. (1994) Attachment and close relationships: an individual- difference perspective.
Psychological Inquiry, 5(1), 27–30, p.27.
Cassidy and Shaver at the 1995 meeting of the Society for Research in Child Development
in Indianapolis with the idea of a Handbook of Attachment. The Society’s biennial confer-
ence is the central conference for the field of development psychology. Shaver had attended
the 1993 conference and had presented a defence of the self-report approach to attachment
as relevant for understanding intergenerational and developmental continuity of individual
differences in attachment over time.192 Weingarten’s idea was for a handbook encompassing
both traditions of research. Cassidy and Shaver ‘were barely acquainted when the project
began’, and did not meet face-to-face after the initial conversation in Indianapolis during the
four years of work on the first Handbook, which would be published in 1999. Nonetheless, in
their preface to the Handbook, Cassidy and Shaver stated that ‘we have grown to know and
respect each other professionally in ways that people rarely do’.193 In the late 1990s, Cassidy
informed and nourished Shaver’s thinking about individual differences in attachment as

minimising and maximising strategies.
Perhaps in an attempt at rapprochement, Crowell (a developmentalist) and Fraley and
Shaver (social psychologists) co-authored a chapter in the first edition of The Handbook
of Attachment. They acknowledged that there existed ‘considerable tension between the
AAI and self-report traditions within the field of contemporary attachment research’.194
They observed that the two traditions had been primarily concerned with parenting
and romantic love respectively, and pursued very different ways of eliciting and ana-
lysing their data. Nonetheless, both traditions appeared to be testing and elaborating
aspects of attachment theory. This apparent common theoretical basis lent impetus for
the need for comparative research. In pursuing this path, Shaver and Brennan collabor-
ated with Jay Belsky, who had long been calling for integration (and, beyond this, was
always something of a free spirit among researchers in the developmental tradition).
In a paper from 2000, the researchers reported from their study with 138 Caucasian
women from working-and middle-class families in central Pennsylvania. Strangely, the
researchers did not report the relationship between self-reported anxiety and avoidance
and the AAI categories. Given that the overall priority of the paper was to emphasise
the compatibility of the two measures, the reader is perhaps left with the suspicion that
this analysis was run, but the results were not regarded as desirable. Instead, the au-
thors reported a few moderate associations on various subscales. Self-reported avoid-
ance was associated with lack of memory for childhood events, anger at the speaker’s
parents, and lower overall coherence.195 However, Shaver and colleagues acknowledged
that ‘the degree of association was relatively modest, and due only to certain aspects of
each measure’.196
Roisman, Fraley, and colleagues embarked on further work comparing the AAI and attach-
ment style measures. A meta-analysis of nearly a thousand participants who had completed
192 Shaver, P.R. (1993) Where do adult romantic attachment styles come from? Paper presented at symposium
entitled ‘Mental Representations of Relationships: Intergenerational and Temporal Continuity’, Meeting of the
Society for Research in Child Development, New Orleans, 27 March 1993. https://adultattachment.faculty.ucda-
vis.edu/wp-content/uploads/sites/66/2014/06/Shaver.pdf.
193 Cassidy, J. & Shaver, P. (1999) Preface. In The Handbook of Attachment, 1st edn. New York: Guilford, p.xiv.
194 Crowell, J.A., Fraley, R.C., & Shaver, P.R. (1999) Measures of individual differences in adolescent and adult
attachment. In J. Cassidy & P.R. Shaver (eds) Handbook of Attachment: Theory, Research, and Clinical Applications
(pp.434–65). New York: Guilford, p.452. The passage would be repeated verbatim in the 2008 edition on p.618.
attachment: associations across domains and methods. Personal Relationships, 7(1), 25–43.
196 Ibid. p.39.
both kinds of assessment by 2007 revealed a trivial association (r = .09).197 One of the few
specific findings was that unresolved trauma, but not unresolved loss, had a small associ-
ation with self-report attachment anxiety (r = .20).198 In Attachment & Human Development,
Roisman and Fortuna reported the results of a study with 160 college students. Participants
completed the AAI, a self-report measure of attachment, a self-report measure of life stress,
and a self-report measure of mental health symptoms. The association between the AAI and
self-reported attachment style were, as expected, negligible. They both, however, predicted
mental health symptoms. Self-reported attachment avoidance and attachment anxiety both
had substantial associations with self-reported mental health symptoms (.43 and .48 re-
spectively). These associations were unaffected by the participants’ self-reported life stress.
The strength of the associations can in part be accounted for by the fact that both the assess-
ments of attachment style and mental health were self-reported. However, it is nonetheless

curious that an avoidant attachment style had an association with these symptoms of almost
equivalent strength. By contrast, the AAI had a marked association (.34) with self-reported
mental health under high experienced life stress, but only a weak association (.19) when life
stress was low.199
Pursuing the question of the interrelation of measures from the two traditions, Bernier
and Matte-Gagné reported a study of 59 Canadian families. They conducted the AAI, the
ECR, and the Strange Situation, a measure of caregiver sensitivity conducted in the home,
and self-report measures of the marital satisfaction of both partners. Rather than using the
AAI categories, the researchers were convinced by Fraley and Roisman (Chapters 2 and
4) that attachment phenomena are better regarded as dimensions, and so they therefore re-
ported only the AAI coherence scale. The researchers found no association between the AAI
and the ECR. They also found, as anticipated, that coherence on the AAI was associated with
greater caregiver sensitivity and increased incidence of secure infant–caregiver attachment
in the Strange Situation. There was no statistical association between the ECR and either
caregiver sensitivity or the Strange Situation.200 Marital satisfaction was not associated with
197 Roisman, G.I., Holland, A., Fortuna, K., Fraley, R.C., Clausell, E., & Clarke, A. (2007) The Adult Attachment
Psychology, 92(4), 678–97. A point of comparison here are findings showing that indirect (implicit) and direct
(explicit) measures of self-esteem are virtually unrelated. Like ‘attachment’, ‘self-esteem’ is likely an umbrella term
within psychological discourse, and captures more than one autonomous process, even if the correlates of these
processes may be similar—producing the superficial effect of a single phenomenon. Pietschnig, J., Gittler, G.,
Stieger, S., et al. (2018) Indirect (implicit) and direct (explicit) self-esteem measures are virtually unrelated: a meta-
analysis of the initial preference task. PLoS One, 13(9), e0202873.
198 Roisman, G.I., Holland, A., Fortuna, K., Fraley, R.C., Clausell, E., & Clarke, A. (2007) The Adult Attachment
Psychology, 92(4), 678–97, p.689. More recently, in an unpublished doctoral study by Watkins, 87 participants
(43 with borderline personality disorder) completed the ECR and AAI. There was no association with attachment
categories. But when a variable was used that combined scales for preoccupied and unresolved states of mind
(Chapter 3), this was associated with ECR anxiety (r = .34). Watkins, C.D. (2016) Convergence versus divergence
of social and developmental measures of adult attachment: testing Jay Belsky’s proposals. Unpublished doctoral
thesis, University of Tennessee.
199 Fortuna, K. & Roisman, G.I. (2008) Insecurity, stress, and symptoms of psychopathology: contrasting results
from self-reports versus interviews of adult attachment. Attachment & Human Development, 10(1), 11–28.
200 For other relevant studies see Mayseless, O., Sharabany, R., & Sagi, A. (1997) Attachment concerns of moth-
ers as manifested in parental, spousal, and friendship relationships. Personal Relationships, 4(3), 255–69; Volling,
B.L., Notaro, P.C., & Larsen, J.J. (1998) Adult attachment styles: relations with emotional well-being, marriage, and
parenting. Family Relations, 47(4), 355–67; Laurent, H.K., Kim, H.K., & Capaldi, D.M. (2008) Prospective effects
of interparental conflict on child attachment security and the moderating role of parents’ romantic attachment.
Journal of Family Psychology, 22(3), 377–88; Howard, K.S. (2010) Paternal attachment, parenting beliefs and chil-
dren’s attachment. Early Child Development and Care, 18(1–2), 157–71.
coherence on the AAI. However, ECR anxiety of one partner was associated with dissat-
isfaction with the relationship by the other, controlling for coherence on the AAI. Bernier
and Matte-Gagné concluded that ‘only romantic attachment was found to relate to marital
satisfaction, while only the AAI was found to relate to caregiving’.201 Such findings suggest
that, even if they claim a common theory, the two traditions are ultimately studying different
phenomena.
In 2006, Shaver stated that he and his collaborators had essentially given up reading the
outputs of the developmental psychology tradition of attachment research.202 Shaver’s ob-
servations have an ‘anger of despair’ quality to them. They mark trends in the mid-2000s that
suggested that the two traditions could go their separate ways, segregated from one another
and out of functional communication. Yet three important factors have contributed to better
relations and productive interaction between the traditions. One has been sustained efforts

to develop institutional and personal connections since the late 1990s. In 1999, the journal
Attachment & Human Development was founded, with Howard Steele serving as Editor-
in-Chief. Though the first issues were tilted more towards the developmental tradition, the
journal sought to represent both traditions, and Shaver was an Associate Editor from the
start. In its fourth year, the journal hosted a major discussion of the relationship between the
traditions, with a lead article by Shaver and Mikulincer, eleven replies from attachment re-
searchers from both camps, and a response from Shaver and Mikulincer.203 And over subse-
quent years, the journal has assiduously made space for contributions from both traditions,
contributing to a sense of a common set of problems and theories.204 Likewise, Cassidy and
Shaver worked together on a second edition of the Handbook of Attachment published in
2008. This mammoth project was evidently an experience that brought the researchers closer
as colleagues, and convinced Cassidy of the benefits of the self-report methodology: in
subsequent years, Cassidy and Shaver collaborated on theoretical papers and empirical re-
search using self-report measures. Cassidy also used them in her own work, independently
of Shaver.205 In 2010, Cassidy and Shaver also worked together in co-editing a special issue
of Attachment & Human Development presenting an attachment perspective on incarcerated
201 Bernier, A. & Matte-Gagné, C. (2011) More bridges: investigating the relevance of self-report and inter-
view measures of adult attachment for marital and caregiving relationships. International Journal of Behavioral
Development, 35(4), 307–316, p.313. This outcome was generally predicted by Mikulincer and Cowans a
decade earlier, though they anticipated some association based on the expectation that marital satisfaction
will influence caregiving: Mikulincer, M., Florian, V., Cowan, P.A., & Cowan, C.P. (2002) Attachment security
in couple relationships: a systemic model and its implications for family dynamics. Family Process, 41(3),
405–434, p.424.
New York: Guilford: ‘Most of us self-report attachment researchers tend to ignore the AAI literature (although
Bartholomew, Furman and Simpson are important exceptions)’ (445).
203 Shaver, P.R. & Mikulincer, M. (2002) Dialogue on adult attachment: diversity and integration. Attachment &
Human Development, 4, 243–57.

204 Part of the contribution made by Attachment & Human Development as a topic-specific journal was that it
could overcome some of the division between developmental and social psychology as subdisciplines. Compare,
for instance, the Journal of Social and Personal Relationships. Mikulincer served as Editor-in-Chief from 2010 to
2015. Though an important venue for attachment research in the social psychology tradition, Mikulincer’s prem-
iership did not coincide with publications in the journal by researchers in the developmental tradition of attach-
ment research. It may well have been that articles were not solicited, submitted, accepted—or all three.
205 E.g. Duggan, A., Berlin, L., Cassidy, J., Burrell, L., & Tandon, D. (2009) Examining maternal depression and
attachment insecurity as moderators of the impacts of home visiting for at-risk mothers and infants. Journal of
Consulting and Clinical Psychology, 77(4), 788–99.
parents and their children, with contributions primarily using the AAI and Strange Situation
as their methodologies.206
A second important development was the emergence of a new ‘third’ generation of at-
tachment researchers, with tenured position and independent grants to pursue research
on their own terms. As mentioned in Chapter 3, Sroufe’s former doctoral student Glenn
Roisman has been one of a number of ‘third-generation’ attachment researchers who use
ideas and methods from both the developmental and social psychology traditions, on the
basis of personal networks of collaborators extending in both directions. Another has been
Roisman’s colleague at Minnesota, Jeff Simpson. Simpson has drawn theoretically on the or-
ganisational perspective of Sroufe, but primarily adopts a social psychological approach to
conceptualising and measuring attachment. Simpson and his group have documented that
self-report measures of attachment and the AAI make non-redundant contributions in pre-

dicting the behaviour of dating couples in completing a tricky laboratory-based task. Secure-
autonomous attachment on the AAI predicted offering their partner a degree of support
that was sensitive to their signals, in a situationally contingent manner. A self-report of an
avoidant attachment style predicted offering the partner less support no matter the degree to
which the partner requested help.207 These results remained the same when controlling for
personality traits (including neuroticism), perceptions of the quality of their relationships,
and the attachment style of the partner. Another ‘third-generation’ attachment researcher—
to be discussed further in the section ‘Attachment to God’—is Pehr Granqvist, who has col-
laborated directly with both Main and Hesse,208 and with Shaver and Mikulincer.209
A third development has been that both traditions have continued, quite successfully, to
generate research findings that appear to test or elaborate much the same theory. This has
meant that, though it is a bit awkward, both traditions have been able to find support in the
work of the other for efforts to think about research or convince audiences of the merits
of attachment theory. Attachment researchers in the social psychology tradition have long
drawn on findings from the developmental tradition in the literature reviews of their pa-
pers, in interpreting their results, and in review articles. For instance, in their magisterial
work Attachment in Adulthood offering the capstone to their careers of research in the area,
Mikulincer and Shaver simply fold in findings from both self-report and AAI studies in their
summary tables of results, rarely remarking on differences in findings.210
It is clear that, contrary to Shaver’s pessimistic remarks about his reading of works by
developmentalist colleagues in 2006, researchers in the social psychological tradition have
kept up to date with research using the AAI. Perhaps marking some aspects of a parallel
206 Cassidy, J., Poehlmann, J., & Shaver, P.R. (2010) An attachment perspective on incarcerated parents and their
children: introduction to the special issue. Attachment & Human Development, 12, 285–8.
207 Simpson, J., Rholes, W.S., Orina, M.M., & Grich, J. (2002) Working models of attachment, support giving,
and support seeking in a stressful situation. Personality and Social Psychology Bulletin, 28(5), 598–608.
208 Granqvist, P., Hesse, E., Fransson, M., Main, M., Hagekull, B., & Bohlin, G. (2016) Prior participation in the
strange situation and overstress jointly facilitate disorganized behaviours: implications for theory, research and
practice. Attachment & Human Development, 18(3), 235–49.
209 Granqvist, P., Mikulincer, M., & Shaver, P.R. (2010) Religion as attachment: normative processes and in-
dividual differences. Personality and Social Psychology Review, 14(1), 49–59. A further, more limited example of
transcendence of the opposition between the developmental and social psychological traditions is Pasco Fearon.
Though one of the leaders of the new generation of developmental attachment researchers, Fearon nonetheless in-
corporated psychometric methodology from the social psychological tradition in developing a brief version of the
Attachment Q-Sort (Chapter 2). Cadman, T., Belsky, J., & Fearon, R.P. (2018) The Brief Attachment Scale (BAS-
16): a short measure of infant attachment. Child: Care, Health and Development, 44(5), 766–75.
210 Mikulincer, M. & Shaver, P. (2016) Attachment in Adulthood, 2nd edn. New York: Guilford.
Attachment narrow and broad 469
transition, a curious shift has occurred in the developmental tradition of attachment re-
search. Previously, with exceptions such as Cassidy and Belsky, researchers in this trad-
ition excluded self-report measures from the literature reviews of their papers, discussions
of results, and review articles. The grounds for this have been that the self-report meas-
ures have little association with the Strange Situation and the AAI, situated as the ‘gold
standard’ measures of attachment. However, in the last few years there has been an increased
trend for leaders in the developmental tradition to both acknowledge the contribution of
self-report studies, and acknowledge it as attachment research. So, for instance, in a 2016
paper, Heckendorf, Huffmeijer, Bakermans-Kranenburg, and van IJzendoorn described
Mikulincer’s work as having ‘convincingly shown . . . experimentally how feelings of more
secure attachment facilitate supporting partners in distress’.211 And no less a defender of the
developmental tradition than Sroufe has praised the social psychology approach in the most

recent edition of the Handbook of Attachment. Sroufe stated that in the work of Mikulincer
and Shaver, as well as Shaver’s Minnesota colleague Simpson, the ‘power’ of the approach for
studying adult romantic relationships ‘has become manifest’.212
Attachment narrow and broad
Pivoting concepts
In a 2015 review paper, Jones, Cassidy, and Shaver expressed confusion over the fact that the
ECR and AAI, unrelated measures, nonetheless ostensibly both assess attachment: ‘though
the two measures are largely unrelated to each other, they are similarly related to a variety of
attachment-relevant constructs’. They have termed this one of the ‘burning questions’ facing
the field of attachment research.213 Shaver and Mikulincer have elsewhere presented this
question as a challenge to researchers in the developmental tradition. If developmentalists
want to claim that self-report measures are not true assessments of attachment, they need
somehow to account for the fact that the results of studies using these measures correspond
specifically to the predictions of attachment theory:
Self-reports of attachment anxiety validly predicted automatic preoccupation with at-

tachment-related worries, whereas self-reports of attachment avoidance validly predicted
defensive suppression of these worries. This suppression could be overcome by adding a
cognitive load that interfered with defensive suppression. These effects are so closely re-
lated to the theoretical conception of anxiety and avoidance in attachment theory that they
would be difficult to explain, as a whole, by any other theory.214
211 Heckendorf, E., Huffmeijer, R., Bakermans-Kranenburg, M.J., & van IJzendoorn, M.H. (2016) Neural pro-
cessing of familiar and unfamiliar children’s faces: effects of experienced love withdrawal, but no effects of neutral
and threatening priming. Frontiers in Human Neuroscience, 10, 231.
213 Jones, J.D., Cassidy, J., & Shaver, P.R. (2015) Parents’ self-reported attachment styles: a review of links with
parenting behaviors, emotions, and cognitions. Personality and Social Psychology Review, 19(1), 44–76, p.69.
Shaver and Mikulincer legitimately described their work as having elaborated and tested
concepts drawn from attachment theory. However, Roisman helpfully qualified that from
this it should not be assumed that (i) ‘attachment’ itself is a unitary entity in adulthood, (ii)
or that attachment theory is a unitary entity.215 The difficulties of translation between the de-
velopmental and social psychological traditions of attachment do not only arise from their
differences in methodology. Perhaps of greater importance has been a confusion of tongues
in the appeals of the two traditions to the work of Bowlby and Ainsworth, and in their use of
the term ‘attachment’ itself.
Shaver and Fraley stated that ‘the term attachment is a metaphor. Its denotation is un-
clear.’216 Its meaning is not pre-set, but shaped by the surrounding network of theoretical
concepts. As such, Shaver and colleagues readily acknowledged the differences between de-
velopmental and social psychological researchers in interpretations of the term ‘attachment’

and in the surrounding network of concepts. They argued that ‘just as a reader of Freud
can focus mainly on his psychosexual theory of personality development or his theory of
intrapsychic dynamics, a researcher interested in attachment theory can focus on one or
more of its central components without being required to focus on all of them. This does
not mean the researcher is not doing attachment research.’217 They sharply criticised Waters
(Chapter 2) for his emphasis on the secure base and safe haven as the central denotation
of the term ‘attachment’ and as the fundamental elements of attachment theory. To them,
this seemed unnecessarily restrictive and overliteral. Waters’ position also seemed to Shaver
and Mikulincer to be a power-conserving tactic, aiming to snatch the legacy of Bowlby and
Ainsworth away from researchers who recognise that the ‘theory contained other central
concepts’ and who wish to acknowledge these with a wider sense of the term ‘attachment’.218
From this perspective, the social psychological tradition of attachment research was not an
unlicenced reapplication of the established attachment concepts; it was another way of gen-
erating concepts, methods, and research priorities on the basis of the ideas of the first gener-
ation of attachment researchers.
Shaver and Mikulincer were aware, of course, that varied use of the same terms put
at risk the viability of effective communication within the research community, and be-
tween the research community and its audiences.219 However, this appeared a risk they
were willing to take. One reason Shaver and Mikulincer may have felt warranted in adopt-
ing the broader notion of ‘attachment’ was that, like Waters and colleagues, they could
easily point to passages in Bowlby’s work to support their interpretation of attachment and
215 Roisman, G.I. (2009) Adult attachment: toward a rapprochement of methodological cultures. Current
Directions in Psychological Science, 18(2), 122–6.

216 Shaver, P.R. & Fraley, R.C. (2000) Attachment theory and caregiving. Psychological Inquiry, 11,
109–114, p.111.
217 Shaver, P.R & Mikulincer, M. (2002) Dialogue on adult attachment: diversity and integration. Attachment &

218 Ibid. See also Shaver, P.R. & Mikulincer, M. (2002) Attachment- related psychodynamics. Attachment &
219 The potential for imprecise or diverging definitions of concepts to reduce clarity in communication is ac-
knowledged in various places in Shaver and Mikulincer’s work, even from their early publications, e.g. Mikulincer,
M. & Caspy, T. (1986) The conceptualization of helplessness: I. A phenomenological-structural analysis. Motivation
and Emotion, 10, 263–78: ‘Many psychological concepts are not well defined, so their meanings vary somewhat
from person to person. Variations in usage are not surprising because psychological phenomena are often abstract
terms that summarize the verbal report of a lay person regarding his perceptions and behavior in reference to spe-
cific real-life situations. Because the terms are imprecise in meaning, they also generate low interjudge reliability’
(263–4).
of attachment theory. This is because, as Chapter 1 showed, Bowlby’s integration of eth-

ology and psychoanalysis brought together within the single term ‘attachment’ two quite
different conceptualisations: one narrow account based on Hinde’s notion of following
and proximity-seeking as activated by threat, and a broader account based on the psycho-
analytic concern with all intimate relationships. A large swathe of the confusion among
researchers lies in the fact that the major additions to Bowlby’s theory by Ainsworth and
Main could apply to both the narrow and the broad concept of attachment. True, in a cer-
tain sense the narrow concept had primacy. Ainsworth’s identification of the secure base
phenomenon, in the first instance, entered attachment theory as an addition to the etho-
logical concern with proximity-seeking, as a cross-species response. Hinde’s idea that
offspring will seek to retain proximity with their parent, especially under conditions of
threat, was transformed by Ainsworth into the division between the secure base and safe

haven. Individual differences in infant’s use of the caregiver as a secure base for explor-
ation and for proximity-seeking as a safe haven were observed by Ainsworth at home and
in the Strange Situation.
However, just as Bowlby’s concept of ‘attachment’ hinged on stringent and sprawling
meanings, so did Ainsworth’s concepts of ‘secure base’ and ‘safe haven’. They could apply
to the physical behaviour of an infant of any species. But they could also apply more gen-
erally to the symbolic aspects of all human intimate relationships. For instance, one friend
can offer another regulation, comfort, and reprieve, without physical contact or any offer
of protection. As Chapter 2 discussed, in a late publication Ainsworth criticised the work
of her students and followers: ‘By focusing so closely on intimacies some attachment re-
searchers have come to conceive of them as the only source of security—which is a pity.’220
For instance, without being an intimate relationship, even a short-term therapist may serve
as a secure base and/or safe haven for a patient. Ainsworth’s sophisticated position in her
1985 article ‘Attachment across the lifespan’ was that the experience of another person as a
secure base, a safe haven, and as a source of separation anxiety were separable in adulthood
(Chapter 2).221 They did not necessarily form a unity. Ainsworth argued that it was futile to
argue back and forth over whether a relationship, such as with a therapist or with a religious
pastor, was or was not an attachment relationship. There was no binary. Instead, she argued
that it was better to identify the extent to which a secure base, a safe haven, and separation
anxiety were present, and to acknowledge that these phenomena could occur without other
characteristics of an attachment relationship.
However, there remained a hinge in Ainsworth’s use of the concepts that hindered this
specification. Ainsworth’s primary use of the terms ‘secure base’ and ‘safe haven’ was to in-
dicated the physical movements of the infant away from and back towards the caregiver.
However, she also extrapolated the concepts in remarks about adulthood to encompass the
provision of other forms of support where these lead to an experience of security. The experi-
ence of feeling confident in a relationship as a basis of felt security may occur in relationships
that are short term, relatively impersonal, and/or with only the barest or most symbolic link
to a parent–child relationship. Probation officers, for example, describe secure base phe-
nomena in their work with offenders as ‘screamingly obvious’; and similar observations have
220 Ainsworth, M. (1990, 2010) Security and attachment. In R. Volpe (ed.) The Secure Child: Timeless Lessons in
Parenting and Childhood Education (pp.43–53). Charlotte, NC: Information Age Publishing, p.49.
221 Ainsworth, M.D.S. (1985) Attachments across the life span. Bulletin of the New York Academy of Medicine, 61,
792–812.
been made for short-term therapeutic involvement with clients by helping professionals.222
The secure base phenomenon is not limited to what are usually treated as attachment rela-
tionships when the term ‘attachment’ is construed narrowly.223
Likewise Main’s concept of conditional strategies was made to pivot in both directions,
towards Bowlby’s narrower and broader concepts of attachment. What is curious is that this
pivot was not intended by Main. Main’s concept of conditional strategies was grounded in
the idea that ‘there exist species-wide abilities that are not part of the attachment system
itself, but can, within limits, manipulate (either inhibit or increase) attachment behavior
in response to differing environments’.224 Main came to regard the avoidant and ambiva-
lent/resistant infant attachment classifications as based on the use of attentional processes
to manipulate the activation and expression of the attachment system. Narrowly, in Main’s
usage a conditional strategy was an ethological repertoire for achieving this effect: either

directing attention towards potential threats and vigilance regarding caregiver availability
and away from cues that might terminate the attachment system, or directing attention
away from cues that might activate the attachment system, and towards potential sources of
distraction in the environment. Main identified a similar process in adult autobiographical
discourse, as attention seemed to be directed away from or towards attachment-relevant
information. However, across both her published and unpublished writings, this never led
her to refer to dismissing and preoccupied states of mind regarding attachment as ‘condi-
tional strategies’.
The extension from infancy to adulthood for Main was based on analogous processes
at the level of attention. Main argued that the caregiving system could likewise be deacti-
vated or made especially vigilant, since the caregiving system also runs on attention allo-
cated to the one being cared for. The adult sexual system was assumed to have a parallel set
of potential conditional strategies. The extension from the Ainsworth Strange Situation to
adult discourse and to the caregiving and sexual systems gave Main’s theory a wide scope
of application. Yet Main’s use of the terms ‘attachment’ and ‘strategy’ always remained
narrow and ethological. It was Jude Cassidy, especially, who made the concept of condi-
tional strategies pivotal to the broader concept of attachment, in conceptualising them
as the ‘maximising’ and ‘minimising’ of not attention but of attachment itself as a system
for the regulation of intimacy and emotion (Table 5.1 and Chapter 3). And it was from
Cassidy, not from Main directly, that Shaver and Mikulincer picked up the concept of ‘at-
tachment strategies’.
In play across the attachment paradigm, then, can be seen narrow and broad concepts
of attachment, narrow and broad concepts of secure base/safe haven, and narrow and
broad concepts of maximising/minimising. Whereas the narrower concepts are based on
222 Ansbro, M. (2018) Integrating attachment theory into probation practice: a qualitative study. British Journal
of Social Work, 48(8), 2235–52.

223 Ainsworth’s emphasis on a secure base, a safe haven, and separation anxiety was inherited by Waters—
though often, confusingly, he has used the term ‘secure base’ to refer collectively to the three components, and
especially the first two. Waters regarded the concept of ‘safe haven’ as partially superseded by the concept of ‘felt se-
curity’, with anxiety as the inverse of security (Chapter 2). To him, this made the feeling of security the overarching
concept, linking infant behaviour to adult experience. It also contributed to his frustration with the focus of other
attachment researchers on separations and reunions at the expense of attention to ordinary experiences that con-
tribute to feelings of trust and security in the relationship. Crowell, J. & Waters, E. (1989) Separation anxiety. In M.
Lewis & S. Miller (eds) Handbook of Developmental Psychopathology (pp.209–218). New York: Plenum Press.
at 1, 6, and 19 years of age: the Berkeley Longitudinal Study. In K.E. Grossmann, K.Grossmann, & E. Waters (eds)
Table 5.1 Table of pivoting concepts
Broad meaning Narrow meaning

Used by social psychologists Used by Main, Waters, and the large majority
+ a few developmentalists of other developmental psychologists
like Cassidy and Fonagy
Safe haven The orientation to seek The capacity to trust in the physical and
function of physical or symbolic comfort attentional availability of discriminated
attachment from close relationships under familiar individuals under conditions of
conditions of perceived threat perceived threat
Secure base The capacity of close The capacity to trust that attention can be
function of relationships to provide felt turned to exploration, given the expectation
attachment security of the availability of discriminated familiar
individuals

Conditional The hyperactivation or A species-wide repertoire, made available
strategies deactivation of the orientation by evolutionary processes, for manipulating
to seek physical or symbolic the activation of the attachment behavioural
comfort from close system through the direction of attention
relationships vigilantly towards or away from cues about the
availability of familiar caregivers or potential
threats. This repertoire evolved because it
has the predictable outcome of increasing
the availability and support provided by
attachment figures who may otherwise be
unavailable.
Other ways of manipulating or overriding the
output of the attachment behavioural system
exist, and become increasingly available with
developmental maturation. They can also be
described as strategies. However, they are not
conditional strategies in the technical sense
unless—like the redirection of attention—they
can be considered to express a species-wide
repertoire, made available by evolutionary
processes
Internal The elaborated symbolic Variously:
working and affective representations 1. Expectations about the availability of
models made by humans about attachment figures
attachment figures and their
availability, and the value of 2. Elaborated symbolic meanings and images
the self to these attachment held by humans about attachment figures and
figures their availability
3. A synonym for attachment representations,
as used by Main in the 1980s (but subsequently
abandoned)
a cross-species ethological account, the broader concepts encompass the felt experience
characteristic of humans on the basis of symbolic capabilities. With both narrow and broad
concepts in circulation, and with little or nothing to mark the distinctions between them,
terms that researchers and their readers assume to have common meanings turn out to be
quite treacherous. They have hidden compartments, allowing assumptions to be smug-
gled unnoticed and unscrutinised. Adding yet further to the confusion has been the con-
cept of the ‘internal working model’, which can figure within both the broad and the narrow
conceptualisations, albeit with different meanings (Chapter 1). The ‘internal working
model’ concept has an unfortunate hypermobility, coming out of joint the moment weight is
rested on it.
Given Waters’ criticisms of Shaver and Mikulincer, it might be thought that the narrow con-
cepts are more characteristic of the developmental attachment tradition and the broader con-
cepts more characteristic of the social psychological tradition. That is certainly the general
impression. However, matters are more complicated. True, Waters has been unusually con-
sistent in using the narrow concept of attachment, and argued against looser uses of the term.225
However, like Ainsworth, he has at times used a wider notion of secure base. And he does not
appear to have ever accepted the concept of conditional strategies; the term is absent from his
writings except in summaries of Main’s position. In Waters’ view, the distinction between avoid-
ant and ambivalent/resistant attachment has seen insufficient predictive payoff, and insufficient

validation against observations in naturalistic settings, to warrant its reification into discussions
of minimising and maximising strategies (Chapter 2).
Among researchers in the developmental tradition, van IJzendoorn and Bakermans-
Kranenburg are also unusually consistent in their use of the term ‘attachment’. In the 1980s,
van IJzendoorn pulled no punches in describing Bowlby’s own tendency towards a broader
usage of the term ‘attachment’ as ‘somewhat ridiculous’, since it implied identity between
the process that brought a scared child to seek their caregiver and much more general feel-
ings such as of identification with a political party.226 Van IJzendoorn had no doubt that
more expansive and symbolic meanings of attachment could be meaningful. For instance,
he suspected that the relationship children have with a soft toy or blanket might recruit cer-
tain facets of the attachment behavioural system, drawing on the emerging human capacity
for symbolisation. But, for him, this needed to be defined in precise terms on the basis of
use of the object as a secure base, as a safe haven and/or source of separation anxiety, and
the prediction of later socioemotional functioning.227 Schuengel and van IJzendoorn dis-
cussed the narrower and broader uses of the terms ‘attachment’ and ‘secure base’ as con-
tributing to a severe ‘confusion about concepts’, obscuring matters under discussion, for
instance the precise relevance of attachment in therapeutic relationships.228 And van
IJzendoorn and Bakermans-Kranenburg have alleged that the concept of attachment has
been stretched and disfigured by broader uses until it has, by now, simply snapped, with re-
searchers talking right past one another whilst using the same terms.229 For van IJzendoorn
225 E.g. Crowell, J.A., Treboux, D., & Waters, E. (1999) The Adult Attachment Interview and the Relationship
Questionnaire: relations to reports of mothers and partners. Personal Relationships, 6(1), 1–18. For an argument
for the return of the concept of strategy to its narrower origins in ethology see Stevenson-Hinde, J. (1994) An etho-
logical perspective. Psychological Inquiry, 5(1), 62–5.
226 Van IJzendoorn, M.H., Tavecchio, L.W.C., Goossens, F.A., & Vergeer, M.M. (1982) Opvoeden in
Geborgenheid: Een Kritische Analyse van Bowlby’s Attachmenttheorie. Amsterdam: Van Loghum Slaterus, p.59.
227 Van IJzendoorn, M.H., Goossens, F.A., Tavecchio, L.W.C., Vergeer, M.M., & Hubbard, F.O.A. (1983)
Attachment to soft objects: its relationship with attachment to the mother and with thumbsucking. Child Psychiatry
and Human Development, 14(2), 97–105; van IJzendoorn, M.H., Sagi, A., & Lambermon, M.W.E. (1992) The mul-
tiple caregiver paradox. Some Dutch and Israeli data. New Directions for Child Development, 57, 5–25, p.9.
228 Schuengel, C. & van IJzendoorn, M.H. (2001) Attachment in mental health institutions: a critical review of
assumptions, clinical implications, and research strategies. Attachment & Human Development, 3(3), 304–323,
p.307. See also Harder, A.T., Knorth, E.J., & Kalverboer, M.E. (2013) A secure base? The adolescent–staff relation-
ship in secure residential youth care. Child & Family Social Work, 18, 305–317.
229 Van IJzendoorn, M.H. & Bakermans-Kranenburg, M.J. (2010) Stretched until it snaps: attachment and close
relationships. Child Development Perspectives, 4(2), 109–111.

and Bakermans-Kranenburg what Shaver and Mikulincer have been measuring is simply a
different behavioural system, one concerned with the adult pair relationship.230
However, it should not be thought that developmentalists are neatly aligned with the nar-
rower use. Some developmental researchers slip and slide between the narrow and broad
uses. Those who write for a clinical audience as well as for researchers and deliver commer-
cial trainings in therapeutic approaches—such as Peter Fonagy, Patricia Crittenden, Allan
Schore, Dan Hughes, and Susan Johnson—are generally consistent throughout their writ-
ings in using the broader concepts.231 The broad concepts of attachment, secure base/safe
haven, and maximising/minisimising have significant advantages in this context. The broad
concepts are closer to the connotations of ordinary language, and so are more accessible
and evocative in writing aimed at both non-researchers and researchers. They offer greater
attention to processes of symbolisation and meaning-making than the technical idea of an

ethological behavioural system focused on proximity-seeking. And they also allow easier ex-
trapolation across the life-cycle. For instance, Crittenden has proposed that what Ainsworth
identified as patterns of attachment in infancy is in fact the local case of a broader phenom-
enon, linking human mental health and ill health. This broader phenomenon is the capacity
of humans to exclude certain kinds of information from influencing experience or behaviour,
in a manner responsive to that individual’s history and social context. This is therefore how
Crittenden has used the concept of ‘attachment’. The term ‘strategies’ has been deployed to
highlight the environmental responsiveness of individual behaviour. And the term ‘security’
has been used to mean the absence of information-processing exclusion or distortions.232
A few developmentalists, however, adopt some narrow and some broad uses. Cassidy, for
instance, has generally been consistent in a comparatively narrow use of the concept of safe
haven, but a broad use of the concepts of minimising and maximising.233 Researchers in
the social psychology tradition, despite tending towards the broader usage, regularly shuttle
between the narrow and broad formulations and only rarely flag these transitions to the
reader. We have already seen an illustration of slippage in the work of Shaver and Mikulincer.
Module 1 of their model of the attachment system is concerned with whether ‘proximity’
is achievable (attachment narrow), but module 2 begins with an evaluation, not of prox-
imity, but of whether the attachment figure is ‘available, attentive, responsive, etc.’ (attach-
ment broad). And again, module 3 is initiated by a concern with whether the attachment
figure is ‘available, attentive, responsive, etc.’ (attachment broad). But this concern is framed
as the question ‘Is proximity seeking a viable option?’ (attachment narrow). This oscillation
230 Verhage, M.L., Schuengel, C., Fearon, R.P., et al. (2017) Failing the duck test: reply to Barbaro, Boutwell,
Barnes, and Shackelford (2017). Psychological Bulletin, 143(1), 114–16, p.114. This is in contrast, for instance, to
Bernier and Dozier who likewise worry that the broader use of the concept of ‘attachment’ jeopardises its meaning,
but who still described self-report measures and the AAI as assessing individual differences in a single behavioural
system, even if they tap different aspects. Bernier, A. & Dozier, M. (2002) Assessing adult attachment: empirical
sophistication and conceptual bases. Attachment & Human Development, 4(2), 171–9.
231 E.g. Schore, A.N. (2001) Minds in the making: attachment, the self-organizing brain, and developmentally-
oriented psychoanalytic psychotherapy. British Journal of Psychotherapy, 17(3), 299–328; Hughes, D. (2007)
Attachment-Focused Family Therapy. New York: Norton; Sacco, F.C., Twemlow, S.W., & Fonagy, P. (2008) Secure
attachment to family and community. Smith College Studies in Social Work, 77(4), 31–51; Johnson, S.M. (2019)
Attachment Theory in Practice. New York: Guilford.
232 Crittenden, P.M., Dallos, R., Landini, A., & Kozlowska, K. (2014) Attachment and Family Therapy.
London: McGraw-Hill; Crittenden, P.M. & Landini, A. (2015) Attachment relationships as semiotic scaffolding
systems. Biosemiotics, 8(2), 257–73.
233 This consistency over time is well illustrated by Cassidy’s opening chapters to the three editions of the
Handbook of Attachment, in 1999, 2008, and 2016.

in Shaver and Mikulincer’s account of the attachment system likely reflected a desire to assert
that what they take to be attachment encompasses but is not reducible to concern with prox-
imity with attachment figures. It allowed them to include the narrow notion of attachment
within their broader concept, whilst also acknowledging that the set-goal of an adult attach-
ment system is calibrated by context and culture.
Once the hinge in the concept of ‘attachment’ is acknowledged, Shaver and Mikulincer’s
position might be regarded as an offshoot of Bowlby’s original insight—obscured by ter-
minological confusion—that attachment in a broad sense has many components in common
with, and may in some ways grow out of, attachment in a narrow sense. Shaver has had a
tendency to overclaim links between adult espoused attachment styles and the attachment
system in infancy. Stevenson-Hinde was surely right to admonish that ‘one is going far be-
yond the data to presume, as Hazan and Shaver do, that “the neural foundation of the at-

tachment system remains largely unchanged” across the life span’.234 Nonetheless, already
from the early works with Hazan, Shaver at times acknowledged that the construction of
an adult bond as an attachment relationship was a piecemeal endeavour, forming a dappled
surface that could, at most, only become consistent over time.235 In a conference presen-
tation, Hazan reported that by late adolescence 75% of 17-year-olds reported displaying
proximity-seeking and separation protest preferentially to peers rather than parents, but that
parents remained favoured for providing a secure base.236 She theorised that at the start of
a new adolescent or adult romantic relationship, proximity-seeking (in part motivated by
the sexual system) contributes to finding a safe haven in the partner. This enthusiastic and
clumsy process then leads to the establishment of pockets of comfort and shelter within the
relationship and the incremental stabilisation of the partner as a secure base. The relation-
ship itself comes to be felt as familiar, ordering, banal, and necessary. Hazan suggested that
separation anxiety following unexpected separations follows on from experiencing the rela-
tionship as a secure base.
The speed and tone of the development of the relationship will be shaped by each indi-
vidual’s previous experiences, which will have informed their sexuality, their trust in a safe
haven, their capacity to use a secure base, and the meanings they give to separations. Hazan
and Shaver theorised that experiences in childhood and adolescence that suggest that a safe
haven will not be available will predispose an avoidant attachment style, which in turn will
slow the development of the components of an attachment relationship. By contrast, experi-
ences in childhood and adolescence that suggest that a secure base will not be available will
predispose an anxious attachment style, which in turn will speed up the use of the caregiver
as a safe haven. Hazan and Shaver speculated that a secure attachment style would make
234 Stevenson-Hinde, J. (1994) An ethological perspective. Psychological Inquiry, 5(1), 62–5, p.63, citing
Hazan, C. & Shaver, P.R. (1994) Attachment as an organizational framework for research on close relationships.
Psychological Inquiry, 5(1), 1–22.
ships. Psychological Inquiry, 5(1), 1–22: ‘The process of attachment formation, at any age, is hypothesized to in-
volve the same sequence: proximity seeking followed by safe-haven behavior followed by the establishment of a
secure base’ (12).
236 Hazan, C., Hutt, M.J., Sturgeon, J., & Bricker, T. (1991) The process of relinquishing parents as attachment
figures. Paper presented at the Biennial Meetings of the Society for Research in Child Development, Seattle,
WA. These findings were replicated by later studies: Fraley, R.C. & Davis, K.E. (1997) Attachment formation and
transfer in young adults’ close friendships and romantic relationships. Personal Relationships, 4, 131–44; Trinke,
S.J. & Bartholomew, K. (1997) Hierarchies of attachment relationships in young adulthood. Journal of Social and
Personal Relationships, 14, 603–625; Nickerson, A.B. & Nagle, R.J. (2005) Parent and peer attachment in late child-
hood and early adolescence. Journal of Early Adolescence, 25(2), 223–49.
for a smoother transition from sexuality to caregiving as the primary ally of the attachment
system in adult relationships. It might also facilitate the development of the full complement
of components of an adult attachment relationship, with the mutual provision of a secure
base, safe haven, and the experience of anxiety at unexpected separations.
Secure base/safe haven dynamics
That Shaver and colleagues would adopt a loose definition of ‘attachment’ was predisposed
by their focus on adulthood. The wish for an adult partner to be ‘available, attentive, respon-
sive, etc.’ will be shaped by culture and context. Indeed, culture and context can influence
the relative priority of concern for availability, attentiveness, and responsiveness. Though

their primary target was romantic relationships, Shaver’s initial impulse towards attach-
ment theory had been as a resource for making sense of a close sibling relationship and its
loss. Hazan and Shaver specified that they intended the introduction of attachment theory
to social psychology as offering ‘a comprehensive theory of close relationships’.237 It was an
individual’s ‘style’ with respect to intimacy that was the object of analysis. And Shaver and
colleagues have frequently used the term ‘intimacy’, or sometimes ‘closeness’, as a functional
synonym for attachment.238
What made attachment theory an exceptionally potent resource for social psychology was
that, not just infant–caregiver relationships, but adult intimate relationships too seemed to
have the potential for secure base/safe haven dynamics. This was already identified clearly
by Bowlby in his writings on the therapeutic role of the secure base in clinical contexts.
Whether or not this makes such adult interactions ‘attachment’ relationships is a matter of
definition, and both sides of this debate can claim Bowlby’s authority and supporting pas-
sages from his writings. Furthermore, various behavioural systems can be subject to min-
imising and maximising: in Main’s original proposal, the parameters of both the infant
attachment system and the adult caregiving and sexual systems could be altered through the
manipulation of attention away from or towards the activating or deactivating conditions
of the system. Whether or not they are therefore defined as ‘attachment’ relationships, the
desire for intimacy in adult relations can be regarded as another site in which minimisation
and maximisation may be deployed to respond to circumstances where the availability of a
partner is in question. As such, the key ingredients of attachment theory—the attachment
system (module 1), the secure base/safe haven dynamic (module 2), and maximising and
minimising (module 3)—are all available for research on adult intimate relationships.
In a 2002 paper, ‘Activation of the attachment system in adulthood’, Mikulincer, Gillath,
and Shaver reported a series of studies of the effects of subliminal threat on the activation of
ships. Psychological Inquiry, 5(1), 1–22, p.1. This was, in fact, a limitation criticised by social psychologists in the
1990s, who felt that any comprehensive theory in social psychology should deal with relationships in general, not
just intimate ones, e.g. Duck, S. (1994) Attaching meaning to attachment. Psychological Inquiry, 5(1), 34–8.
238 Hence statements such as ‘attachment-related avoidance is a tendency to withdraw from situations involving
intimacy’. Brassard, A., Darveau, V., Péloquin, K., Lussier, Y., & Shaver, P.R. (2014) Childhood sexual abuse and
intimate partner violence in a clinical sample of men: the mediating roles of adult attachment and anger manage-
ment. Journal of Aggression, Maltreatment & Trauma, 23(7), 683–704, p.686. Other attachment researchers, how-
ever, have treated intimacy and attachment as distinct constructs. Karantzas, G.C., Feeney, J.A., Goncalves, C.V., &
McCabe, M.P. (2014) Towards an integrative attachment-based model of relationship functioning. British Journal
of Psychology, 105(3), 413–34.
representations of attachment figures. The authors regarded the results as effective evidence
that their work was studying the processes outlined by attachment theory. They challenged
their readers to offer another theory that could account for their findings. In their view, ‘there
is no alternative theory that predicts the results we obtained’, and also ‘no alternative theory
that would have generated either these particular kinds of measures or our particular experi-
ments’.239 The series of studies began by using the WHOTO measure, initially developed by
Hazan and then adapted by Fraley and Davis.240 In the WHOTO, participants are asked to
give a named person in response to questions about proximity-seeking/separation anxiety
(Who is the person you most like to spend time with? Who is the person you don’t like to be
away from?), safe haven (Who is the person you want to be with when you are feeling upset
or down? Who is the person you would count on for advice?), and secure base (Who is the
person you would want to tell first if you achieved something good? Who is the person you

can always count on?). Mikulincer and colleagues were keen to emphasise that these were
not merely conventional or obvious aspects of any close relationship, but were specifically
the functions privileged by Ainsworth in defining an attachment figure.241
In study 1, Mikulincer, Gillath, and Shaver asked participants to complete a computer-
ised lexical decision task, in which they read a string of letters and tried to identify whether
it was a word or a non-word. Reaction times were used as a measure of the accessibility of
thoughts related to the target word.242 The accessibility of the names of people identified in
the WHOTO was assessed under two conditions. In the first condition, the name followed
a neutral word (‘hat’). In the second condition, the name followed the word ‘failure’, which
was chosen as a representation of threat but not necessarily one closely linked to attachment.
Other names were also used: the names of close friends not selected by the WHOTO, the
names of acquaintances, and names of strangers. Participants also completed the ECR, and
a measure of neuroticism. The results of study 1 were that participants had a faster reaction
time when the name was from their WHOTO list following the ‘failure’ prime, compared to
reaction times for the names of other close people, acquaintances, strangers, and non-words.
An anxious attachment style reduced reaction times for the names of attachment figures
only, and this occurred in both the neutral and the ‘failure’ condition. As in other studies,
there was a very substantial association between attachment anxiety and neuroticism (r =
.54). However, neuroticism was unrelated to response times, and the effects for attachment
anxiety remained the same even controlling for neuroticism.
239 Mikulincer, M., Gillath, O., & Shaver, P.R. (2002) Activation of the attachment system in adulthood: threat-
related primes increase the accessibility of mental representations of attachment figures. Journal of Personality and
Social Psychology, 83(4), 881–95, p.892.
240 Fraley, R.C. & Davis, K.E. (1997) Attachment formation and transfer in young adults’ close friendships and
romantic relationships. Personal Relationships, 4, 131–44.

241 Trinke and Bartholomew found an association of .45 between the WHOTO and a measure of ‘supportive’
relationships, suggesting that the WHOTO is materially related to supportiveness, but not reducible to it. They
interpreted this finding as indicating that supportiveness may contribute to the formation of attachment compo-
nents in a relationship, but that many relationships with attachment components in adulthood form and endure
without supportiveness. Trinke, S.J. & Bartholomew, K. (1997) Hierarchies of attachment relationships in young
adulthood. Journal of Social and Personal Relationships, 14, 603–625. More recently, Gillath and colleagues have
argued against the use of self-reported separation anxiety as an index of an attachment relationship, as they an-
ticipate that participants have to speculate about what it would be like to lose the other person, whereas they can
report more concretely on secure base and safe haven use. Gillath, O., Karantzas, G.C., & Fraley, R.C. (2016) Adult
Attachment: A Concise Introduction to Theory and Research. London: Academic Press, p.54.
242 The methodology had been introduced into attachment research a decade earlier by Baldwin, M.W., Fehr,
B., Keedian, E., & Seidel, M. (1993) An exploration of the relational schemata underlying attachment styles: self-
report and lexical decision approaches. Personality and Social Psychology Bulletin, 19, 746–54.
Study 2 was a replication of study 1, except that instead of the threat word ‘failure’, the re-
searchers substituted the threat word ‘separation’. Again, participants showed faster reaction
time following the prime when the name was from their WHOTO list than other names. As in
study 1, an anxious attachment style made a contribution to reaction times for attachment fig-
ures only, and in both the ‘separation’ and the neutral condition. However, a finding specific to
study 2 was that an avoidant attachment style increased reaction times for the names of attach-
ment figures following priming of the word ‘separation’ but not following the neutral word. This
implied that the names of attachment figures were less available for participants in response to
the theme of ‘separation’ to the degree that they were high in attachment avoidance. Again, the
association between an anxious attachment style and neuroticism was large (r = .62), and again
neuroticism had no association with reaction times or interaction with the effect of attachment
style on reaction times.

Study 3 sought to replicate studies 1 and 2 using a different procedure, the Stroop task. This is
a well-established procedure based on the observation that viewing words with salient personal
meanings slows reaction times in identification of the colour of that word. That is to say, cog-
nitive accessibility on this assessment was represented by longer reaction times, in contrast to
studies 1 and 2. Participants were divided into three groups: one group received neutral primes,
another group were primed by the word ‘failure’, and a third group were primed by the word
‘separation’. In the Stroop task, as expected, colour-naming was slower following the two threat
primes in response to the name of attachment figures only, as opposed to other names, words,
or non-words. An anxious attachment style increased reaction times in the neutral condition
and following both threat primes in response to the names of attachment figures. An avoidant
attachment style decreased reaction times in response to the names of attachment figures fol-
lowing only the separation prime, not the failure or neutral primes. Again, attachment anxiety
overlapped substantially with neuroticism, but only the former was associated with any reac-
tion time outcomes. Overall, the findings of the three studies were taken by Mikulincer and
colleagues to show that threat contexts—whether general or attachment-specific—increase the
cognitive accessibility of the names of attachment figures, in contrast to other close relationships
and acquaintances (module 1). An anxious attachment style increased this cognitive accessi-
bility even in neutral conditions, illustrating the feedback from attachment strategy (module
3) to the appraisal of threats and attachment figures (modules 1 and 2). An avoidant attachment
style decreased cognitive accessibility only in response to the names of attachment figures and
only in response to a separation prime rather than a general threat prime. The researchers dem-
onstrated these effects over two different procedures and in three samples. The findings, the re-
searchers argued, ‘increase our confidence in the psychological reality of the attachment system’
in adulthood.243
Mikulincer and colleagues regarded this study, one of their most cited by social psych-
ology attachment researchers, as a gauntlet thrown down to the developmental tradition.
Could they explain the findings except in terms of the operation of the attachment system in
adulthood? The answer would appear to be no. Three research groups in the developmental
243 Mikulincer, M., Gillath, O., & Shaver, P.R. (2002) Activation of the attachment system in adulthood: threat-
related primes increase the accessibility of mental representations of attachment figures. Journal of Personality
and Social Psychology, 83(4), 881–95, p.891. Aligned findings were later reported in Edelstein, R.S. & Gillath, O.
(2008) Avoiding interference: adult attachment and emotional processing biases. Personality and Social Psychology
Bulletin, 34(2), 171–81. The researchers found that an avoidant attachment style interfered with the availability of
only attachment-relevant emotional words, and not emotional words not relevant to attachment.
tradition—those of Jude Cassidy,244 Anne Bernier,245 and Peter Fonagy246—have treated the
study as evidence that the ECR measures attachment just as much as the assessments de-
vised by Mary Ainsworth and Mary Main. Other developmental researchers have ignored
the study, at least in print.247 The paper demonstrated that threat primes increase the salience
of semantic information relevant to figures identified as serving as a secure base, safe haven,
and potential source of separation anxiety. Whilst the study did not provide evidence that
these figures are differentially sought as a secure base/safe haven in concrete behaviour, the
evidence of increased cognitive accessibility in the laboratory is certainly relevant and intri-
guing, and points in the expected direction.248 The study also demonstrated that these effects
are moderated by scores on the ECR in ways aligned with theory: an anxious attachment
style increased vigilance to threats and to information about relationships with attachment
components; and an avoidant attachment style decreased the cognitive availability of rela-

tionships with attachment components following a prime for separation.
The study suggested that secure base/safe haven processes operate in adulthood, and that
these are moderated by maximising and minimising strategies. This was one important em-
pirical source of legitimacy for the social psychology tradition to call itself a form of attach-
ment research. However, it is important to note that the distinction between ‘attachment
relationships’ and ‘close relationships’ in the study does not rule out other possibilities as-
sociated with the wider notion of attachment used by Shaver and colleagues. Shaver, Hazan,
and Bradshaw claimed that ‘the attachment figure need not be physically present for such
an “interaction” to take place, which makes it more understandable that people have im-
aginary but very convincing and affective interactions with rock singers, dead philosophers,
and religious figures of all kinds. Attachment, separation distress, and grieving are primarily
psychological processes; they require psychological, not physical, interaction partners.’249
Furthermore, some physical interactions with safe bases and secure havens are not with
living partners. For instance, Bowlby stated that a person’s home would likely have the char-
acteristics of a secure base and safe haven (Chapter 1). If a person’s home address were used
in a study like that of Mikulincer and colleagues, it could be anticipated that responses would
be more similar to names of attachment figures than to names of other close relationships.
244 Dykas, M.J. & Cassidy, J. (2011) Attachment and the processing of social information across the life
span: theory and evidence. Psychological Bulletin, 137(1), 19–46.

245 Bernier, A., Larose, S., & Whipple, N. (2005) Leaving home for college: a potentially stressful event for ado-
lescents with preoccupied attachment patterns. Attachment & Human Development, 7(2), 171–85.
246 Fonagy, P. & Luyten, P. (2009) A developmental, mentalization-based approach to the understanding and
treatment of borderline personality disorder. Development & Psychopathology, 21(4), 1355–81.

247 Granqvist might be anticipated as an exception, given that his work spans both traditions. However, the only
occasions on which he has cited the paper are in works actually co-authored with Mikulincer. Roisman is another
figure straddling the two traditions, and cites the paper in a report co-authored with Chris Fraley: Haydon, K.C.,
Roisman, G.I., Marks, M.J., & Fraley, R.C. (2011) An empirically derived approach to the latent structure of the
Adult Attachment Interview: additional convergent and discriminant validity evidence. Attachment & Human
Development, 13(5), 503–524.
248 Fraley and Shaver demonstrated that adult attachment style predicts observable behaviour on separations,
accounting for around 8% of variance. Fraley, R.C. & Shaver, P.R. (1998) Airport separations: a naturalistic study
of adult attachment dynamics in separating couples. Journal of Personality and Social Psychology, 75(5), 1198–212.
Mikulincer and colleagues would later supply evidence that anxious and avoidant attachment styles bias motor re-
sponses on a push-pull task in response to attachment primes. Mikulincer, M., Shaver, P.R., Bar-On, N., & Ein-Dor,
T. (2010) The pushes and pulls of close relationships: attachment insecurities and relational ambivalence. Journal
of Personality and Social Psychology, 98(3), 450–68.
tems. In R.J. Sternberg & M. Barnes (eds) The Psychology of Love (pp.68–99). New Haven, CT: Yale University
Press, p.73.
And as we have seen, Shaver himself has referred to his experience of attachment theory
as an attachment relationship: it is quite possible that other phenomena that serve as a safe
base and safe haven might also receive differential cognitive availability in response to threat
primes. That secure base/safe haven processes operate in adulthood does not imply that
these processes are exclusive to romantic or even close relationships.
Evidence for the wider relevance of the secure base/safe haven components of attachment
relationships, and of minimising and maximising strategies, is offered by later research con-
ducted by Zilcha-Mano, Mikulincer, and Shaver. The researchers developed a scale for as-
sessing anxiety and avoidance in an adult’s sense of their pet’s availability as a secure base and
safe haven:
Pet attachment anxiety consists of intense and intrusive worries that something bad might

happen to one’s pet and that one might find oneself alone, a strong need for proximity to
the pet, reassurance seeking from the pet in order to maintain self-worth, intense frustra-
tion when the relationship with the pet is not as close as one would like, and even anger
when the pet prefers the proximity of others. Pet avoidant attachment consists of feelings
of discomfort with physical and emotional closeness to a pet, striving to maintain emo-
tional distance from the pet, avoiding intimacy with it, preventing the pet from intruding
into one’s personal space, and difficulties in depending on the pet and turning to it when
distressed.250
Zilcha-Mano and colleagues found that an avoidant attachment style on the ECR had no as-
sociation with pet avoidance, and instead was associated with attachment anxiety regarding
the pet’s availability as a secure base/safe haven (r = .35). The researchers interpreted this
finding as suggesting that ‘avoidant people, who are unlikely to express attachment-related
worries and anxieties in close human relationships (Mikulincer & Shaver, 2007), tend to
express these worries and anxieties in relationships with pets’.251 By contrast, there was a
large positive association (r = .60) between attachment anxiety regarding the pet and an anx-
ious attachment style on the ECR. Yet Zilcha-Mano and colleagues reported that anxiety
regarding the availability of the pet had a substantial association with reduced reported well-
being (r = .34) and with symptoms of anxiety and depression (r = .41) on the Mental Health
Inventory, even after controlling for scores on the ECR and a measure of personality traits.
The researchers therefore emphasised that psychologically consequential secure base/safe
haven effects, and minimising/maximising individual differences, can be identified in adult-
hood beyond close human relationships.
The researchers did not report the association between ECR anxiety and the Mental
Health Inventory. However, if other studies with the ECR and the Mental Health Inventory
are brought in for comparison,252 it is rather interesting that the pet attachment anxiety
measure in the Zilcha-Mano study is actually a better predictor of scores on the Mental
Health Inventory than the ECR. Given that many species of pet—unlike human partners—
have been bred precisely to offer loyalty and reassurance, it may suggest an especially sorry
250 Zilcha-Mano, S., Mikulincer, M., & Shaver, P.R. (2011) An attachment perspective on human–pet relation-
ships: conceptualization and assessment of pet attachment orientations. Journal of Research in Personality, 45(4),
345–57, p.354.
251 Ibid.
252 Birnbaum, G.E., Orr, I., Mikulincer, M., & Florian, V. (1997) When marriage breaks up-does attachment
style contribute to coping and mental health? Journal of Social and Personal Relationships, 14(5), 643–54.
state of affairs for someone to feel consciously anxious about whether their pet cares about
them. Neuroticism as a measure of trait anxiety had only a weak association with anxiety re-
garding the pet’s availability as a secure base/safe haven (r = .15), indicating that association
does not lie merely in personality. It suggests that the secure base/safe haven components
occur in the relationship with the pet and that these are reducible to neither adult attach-
ment style nor personality. Furthermore, the findings suggest that meaningful individual
differences exist in anxiety regarding the relationship with the pet. Zilcha-Mano and col-
leagues interpreted this as individual differences in the deployment of a maximising strategy
in order to ensure the availability of the pet as a secure base/safe haven. However, further
research is required to demonstrate that this is indeed specifically a maximising strategy, as
opposed to anxiety arising through a psychological process of a different kind.

Attention and cognitive schemas
The ECR was considered to tap schemas offer information about the meaning and avail-
ability of a secure base and safe haven within intimate relationships, and the relevance of
minimising or maximising strategies as a way of achieving the availability, attentiveness,
responsiveness (etc.) of the other. This is quite distinct from the AAI, as an assessment of
individual differences in the capacity to attend to and communicate about personal experi-
ences with attachment figures in childhood. Rather than assessing generalised schemas, this
capacity is assessed in the AAI through comparison of particular episodic and generalised
semantic information. Johnson has claimed that attachment patterns, attachment styles,
attachment strategies, and states of mind regarding attachment should all be regarded as
‘equivalent terms’, with the same referent except for differences in age and assessment meth-
odology.253 This runs counter to what Shaver and colleagues, and Main and colleagues have
said that their assessments measure. It also seems a claim poorly aligned with the available
evidence. Offering a more sophisticated and developmental reconciliation between the tra-
ditions, Cassidy proposed that (i) the experience of a secure base and safe haven in early
child–parent relationships should facilitate (ii) a coherence in attending to and communi-
cating about childhood experiences with attachment figures and (iii) generalised schemas
about intimate relationships characterised by less anxiety or expectation of rejection.254
However, it should clearly be acknowledged that these are three vastly different meanings of
the term ‘secure attachment’, and the relationship between any two of them is unlikely to be
simple or strong. Not least, Cassidy’s proposal assumes extensive continuities from infancy
to adult measures; though scaffolded by some of Bowlby’s claims, this assumption was and
remains contrary to available data (Chapter 2).
There is no reason to assume that the capacity to attend to and communicate about per-
sonal experiences with attachment figures in childhood should correlate with generalised
schemas about intimate relationships in adulthood. The first is an attentional process oriented
by a comparison of episodic and semantic memories of attachment narrowly construed. This
253 Johnson, S.M. (2003) Introduction to attachment: a therapist’s guide to primary relationships and their
renewal. In S.M. Johnson & V. Whiffen (eds) Attachment Processes in Couple & Family Therapy (pp.3–17).
254 E.g. Cassidy, J. (2001) Truth, lies, and intimacy: an attachment perspective. Attachment & Human

is prompted by questions designed to offer uncomfortable freedom to memory, setting epi-

sodic and semantic information into motion and providing an opportunity to see the ex-
tent to which they outgallop the speaker’s cooperation with the interviewer. The second is
a generalised—in all likelihood largely semantic—schema of beliefs, emotions, and behav-
iours relevant to the minimising or maximising of the wish for others to be available, atten-
tive, responsive, etc. This will certainly be linked to attentional processes, not least in terms
of cognitive biases in the perception of new information.255 However, individual differences
in attentional processes are at most a correlate, rather than what the ECR is specifically meas-
uring. And whilst it might seem that both the AAI and ECR assess individual differences in
terms of minimising and maximising strategies, this is a misapprehension. The AAI assesses
the allocation of attention away from episodic information at the expense of fully answering
the interview questions (Dismissing), towards negative aspects of experiences at the expense

of focus on the actual questions asked in the interview (Preoccupation), or disruptions in
working memory expressed as lapses in reasoning or discourse (Unresolved/disorganised).
By contrast, the ECR assesses a variety of behaviours, beliefs, and feelings that index latent
schemas. The two latent dimensions of the ECR are presumed by Shaver and Mikulincer to
reflect an intensification or deintensification of the wish for others to be available, attentive,
responsive, etc.
There may be several reasons for the lack of convergence between the AAI and ECR. They
adopt different forms of measurement. And even within self-report assessments, there is
only weak convergence between reports of attachment style with parents and attachment
style in current intimate relationships. For example, self-reported avoidance with mother
and with father have little association with a participant’s attachment style in their current
relationship (r = .24 and r = .24 for anxiety; r = .17 and r = .12 for avoidance, for mother and
father respectively).256 In addition to these factors, a further contribution is perhaps made by
the fact that they tap different psychological processes, at least according to their originators.
To the degree that Main and Hesse are right that the AAI is in the first instance an assessment
of the allocation of attention, and that Shaver and colleagues are right that the ECR is ultim-
ately an assessment of generalised schemas about intimate relationships, this suggests some
specific conclusions about how and when the measures will diverge—but also when they will
converge.
There should be poor association between the measures when correlates relate to semantic
generalisations about need achievement within adolescent and adult relationships. This is
perhaps why the ECR relates quite well to reported marital satisfaction and reported feelings
of closeness with friends, but not the AAI.257 There should also be poor association between
the measures when correlates relate specifically to the allocation of attention in responding
to another—one component of the caregiving behavioural system (regardless of whether this
is broadly construed as nurturance or narrowly construed as safe haven provision). This is
255 Fraley, R.C., Davis, K.E., & Shaver, P.R. (1998) Dismissing avoidance and the defensive organization of emo-
tion, cognition, and behavior. In J.A. Simpson & W.S. Rholes (eds) Attachment Theory and Close Relationships
256 Fraley, R.C., Heffernan, M.E., Vicary, A.M., & Brumbaugh, C.C. (2011) The Experiences in Close
Relationships—Relationship Structures Questionnaire: a method for assessing attachment orientations across re-
lationships. Psychological Assessment, 23(3), 615–25, Table 2.
257 On associations with affiliation with friends see e.g. Mikulincer, M. & Selinger, M. (2001) The interplay be-
tween attachment and affiliation systems in adolescents’ same-sex friendships: the role of attachment style. Journal
of Social and Personal Relationships, 18(1), 81–106; Mayseless, O. & Scharf, M. (2007) Adolescents’ attachment rep-
resentations and their capacity for intimacy in close relationships. Journal of Research on Adolescence, 17(1), 23–50.
perhaps why coherence on the AAI has been repeatedly found to be reliably associated with
observational measures of sensitivity towards children and partners, whereas the link be-
tween the ECR and sensitivity seems to be more dependent on circumstances.258 Sensitivity
is predicated on noticing and recognising the other’s signals (Chapter 2), primarily an at-
tentional task. By contrast, the ECR is reliably associated with general supportiveness when
the other is evidently stressed. Simpson and colleagues demonstrated this in assessments of
couple interactions during a challenging laboratory task, as described in the section ‘Two
traditions of attachment research’.259 And other studies have documented links between the
ECR and generally supportive parenting when a child is distressed.260
The two measures will also be less well aligned when clinical phenomena are in question.
The reason for this is foundational: part of the very definition of many clinical symptoms is
the divergence between attentional processes and latent generalised schemas. In addition to

differences in measurement, this may well be part of the reason why there is no evidence of
overlap between the constructs of unresolved/disorganised states of mind and self-report
measures of fearful attachment. Yet these two uncorrelated measures, the AAI and ECR,
can nonetheless be anticipated to have similar correlates when the phenomenon in ques-
tion is associated with both the allocation of attention in relation to attachment phenomena
and generalised schemas about close relationships. There will be many occasions in which
this alignment may occur. For instance, sexual behaviour in adulthood is likely influenced
by both processes. The shaping of attention and sharing in relation to past experiences of
tenderness and care are surely relevant, as are schemas about the possibilities and threats
associated with intimacy. Likewise, both the allocation of attention in relation to attachment
phenomena and generalised schemas about close relationships can be anticipated to be as-
sociated with the content—if not the form—of accounts of care in the course of childhood.
And indeed, this is what evidence to date suggests.261
Nonetheless, it may also be anticipated that the processes underpinning the AAI and ECR
would be activated and moderated by different circumstances. For instance, given Main’s
theory of the importance of attention to individual differences in both measures, the Strange
Situation and the AAI may each be particularly impacted by factors that affect attention
and memory processes, such as Ritalin and other ADHD medications.262 Indeed, given
that studies suggest use of these medications occur in 7–9% of adolescent community sam-
ples and 4% of adult samples,263 the use of the AAI and assessments modelled on it without
258 E.g. Mills-Koonce, W.R., Appleyard, K., Barnett, M., Deng, M., Putallaz, M., & Cox, M. (2011) Adult at-
tachment style and stress as risk factors for early maternal sensitivity and negativity. Infant Mental Health Journal,
32(3), 277–85.
259 Simpson, J., Rholes, W.S., Orina, M.M., & Grich, J. (2002) Working models of attachment, support giving,
and support seeking in a stressful situation. Personality and Social Psychology Bulletin, 28(5), 598–608.
260 E.g. Edelstein, R.S., Alexander, K.W., Shaver, P.R., et al. (2004) Adult attachment style and parental respon-
siveness during a stressful event. Attachment & Human Development, 6(1), 31–52.
261 For instance, convergence between the ECR and scales for inferred childhood experience coded on the
AAI has been substantial for avoidance (r = .30 and .41) and material for anxiety (r = .24 and .22). Haydon, K.C.,
Roisman, G. I., Marks, M.J., & Fraley, R.C. (2011) An empirically derived approach to the latent structure of the
Adult Attachment Interview: additional convergent and discriminant validity evidence. Attachment & Human
Development, 13(5), 503–524.
262 Cf. Storebø, O.J., Skoog, M., Rasmussen, P.D., et al.(2014) Attachment competences in children with ADHD
during the Social-Skills Training and Attachment (SOSTRA) randomized clinical trial. Journal of Attention
Disorders, 19, 865–71.
263 Bachmann, C.J., Wijlaars, L.P., Kalverdijk, L.J., et al. (2017) Trends in ADHD medication use in children and
adolescents in five western countries, 2005–2012. European Neuropsychopharmacology, 27(5), 484–93; Anderson,
K.N., Ailes, E.C., Danielson, M., et al. (2018) Attention-deficit/hyperactivity disorder medication prescription
Sexuality and dominance behavioural systems 485
controlling for attention-altering medication use may have introduced an unrecognised

confound into analyses. In community samples, medication use would be a small confound,
but in higher-risk samples where medication use is more common, this may have distorted
research findings to some degree. The ECR by contrast is regarded as an assessment of gener-
alised schemas, and would be expected to be relatively unaffected by the same medications,
though the intensity of their availability may be altered.264 Such distinctions would need to
be tested, though. It is not enough to rest on the authors’ own interpretations of what their
measures measure.
Sexuality and dominance behavioural systems

Behavioural systems theory
Shaver and Mikulincer have been among the few attachment researchers, besides Main and
Hesse (Chapter 3), to give serious consideration to Bowlby’s concept of behavioural sys-
tems. This concept has been taken for granted, and then ignored, by most researchers in
the developmental tradition: its relevance to the Strange Situation seemed too obvious for
further comment, and its relevance to the AAI was not explicitly articulated in Main’s pub-
lished writings and remained unclear. However, Shaver and Mikulincer’s were brought to
re-examine the concept as part of their attempts to establish and understand the meaning of
attachment in adulthood, and in particular to operationalise its activation in the laboratory.
Unlike researchers in the developmental tradition, Shaver and Mikulincer were forced to test
and discriminate, rather than assume, the operation of psychological processes resembling
those described by Bowlby.265
In the course of their exploration of Bowlby’s ideas about attachment and other behav-
ioural systems, Shaver and Mikulincer came to the view that attachment theory should
actually have been called ‘behavioural systems theory’.266 The concept of ‘attachment’ was
introduced as ‘a metaphor that would refocus psychoanalytic theory on relational issues
rather than imagined instincts or drives’, and so has served its usefulness already.267 If the
banner of attachment theory was to be preserved, for Shaver and Mikulincer this would es-
sentially be a matter of tradition rather than because this is the best name for Bowlby’s ul-
timate achievement. As documented in Chapter 1, in late correspondence Bowlby himself
was dissatisfied with ‘attachment’ as the name for his theory. Additionally, whereas the devel-
opmental tradition of attachment research had already claimed the concept of ‘attachment’,
claims among privately insured women aged 15–44 years—United States, 2003–2015. Morbidity and Mortality
Weekly Report, 67(2), 66.
264 Fraley, R.C. & Shaver, P.R. (2000) Adult romantic attachment: theoretical developments, emerging contro-
versies, and unanswered questions. Review of General Psychology, 4(2), 132–54: ‘We hypothesize that anxiety-
reducing drugs affect the intensity but not the avoidant-nonavoidant orientation of attachment behaviors’ (146).
265 Bernier, A. & Dozier, M. (2002) Assessing adult attachment: empirical sophistication and conceptual bases.

266 Shaver, P. & Mikulincer, M. (2004) Attachment in the later years: a commentary. Attachment & Human
Development, 6(4), 451–64, p.454.

267 Mikulincer, M. & Shaver, P.R. (2007) Reflections on security dynamics: core constructs, psychological mech-
anisms, relational contexts, and the need for an integrative theory. Psychological Inquiry, 18, 197–209, p.208.
the concept of behavioural systems was an uncontested ground in which Shaver and
Mikulincer could engage with and claim Bowlby’s legacy.
Early in their work together, with the attachment system in childhood and adulthood as
their central model, Shaver and Mikulincer defined a behavioural system as having six com-
ponents.268 These can be readily illustrated with the examples of the attachment and explora-
tory system. First, any behavioural system had to offer an expected species-level benefit for
survival and/or reproduction. The attachment system, for instance, was assumed to offer
both protection and a source of regulation. The exploratory system was assumed to offer
support for learning about the environment and the self ’s capabilities.269 Second, any behav-
ioural system should have a set of activating parameters. The attachment system in childhood
and adulthood was assumed by Shaver and Mikulincer to be activated by the perception of
some kind of threat, whether an external source of danger or separation or an internal state

of concern. Likewise, activating parameters can be identified for the exploratory system.
Whereas Bowlby regarded the activating conditions of the exploratory system the encounter
with complexity and/or novelty, in their conceptualisation Shaver and Mikulincer limited
the activation of the exploratory system to the encounter with novelty, but emphasised that
novelty is defined in part through contradiction or challenge to existing knowledge.270
For Shaver and Mikulincer, a third defining characteristic of a behavioural system was
that it should have a set-goal which deactivates the system when a given change is identi-
fied in the relationship with the environment. Shaver and Mikulincer conceptualised the
attachment system as deactivated by signals of the availability, attentiveness, responsive-
ness (etc.) of attachment figures, though the nature of these signals differ between child-
hood and adulthood. They also made the innovative proposal that a behavioural system will
have ‘anti-goals’, which the system attempts to avoid. In the case of the attachment system,
these are ‘rejection, separation, and attachment figure unavailability’.271 To take the example
of the exploratory system, Bowlby conceptualised this system as deactivated by familiarity.
Shaver and Mikulincer’s concept was similar, though not identical: for them, the explora-
tory system would be deactivated once relevant skills or knowledge have been acquired.
Shaver and Mikulincer at no point discuss the anti-goals of the exploratory system, but these
might include feelings of incompetence and boredom.272 To achieve familiarity without the
(pp.53–152). New York: Academic Press, p.57. Compare the definition of behavioural system offered in Shaver,
P.R., Segev, M., & Mikulincer, M. (2011) A behavioral systems perspective on power and aggression. In P.R. Shaver
& M. Mikulincer (eds) Human Aggression and Violence: Causes, Manifestations, and Consequences (pp.71–87).
Washington, DC: American Psychological Association, p.72. In this later text the elements especially highlighted
are the first, third, and fourth. There is no mention, for instance, of cognitive components as required as part of the
definition of a behavioural system.
269 Mikulincer, M. & Shaver, P.R. (2012) Attachment theory expanded: a behavioral systems approach to per-
sonality. In K. Deaux & M. Snyder (eds) Oxford Handbook of Personality and Social Psychology (pp.467–92).
270 Ibid.: ‘The primary strategy of the exploration system, seeking new information about oneself and the world,
is activated whenever a person encounters novel situations, objects, or people or experiences novel internal states
that contradict or challenge existing knowledge and working models’ (473).
271 Mikulincer, M. & Shaver, P. (2008) Contributions of attachment theory and research to motivation science.
In J.Y. Shah & W.L. Gardner (eds) Handbook of Motivational Science (pp.201–216). New York: Guilford, p.204.
272 An interesting case discussed by Shaver in his early work is that of prank calls to the fire service. In a subtle
and thoughtful paper, Shaver speculated that implicated in these prank calls was a pervasive combination of anger,
boredom, and powerlessness, and a lack of other legitimate situations for expressing these feelings. In terms of his
later thinking, this might be described as the interaction of the anti-goals of the exploratory systems and domin-
ance systems. Shaver, P.R., Schurtman, R., & Blank, T. (1975) Conflict between ghetto-dwellers and firemen: envir-
onmental and attitudinal factors. Journal of Applied Social Psychology, 5, 240–61.
experience of relevance and growth might risk achievement of both a set-goal and an anti-
goal, resulting in a happiness that continually drains itself. Bowlby was not wrong that fa-
miliarity terminates the exploratory system, but in Shaver and Mikulincer’s view this did not
mean that the behavioural system is therefore fully realised and satisfied.
Fourth, a behavioural system will have a repertoire of functionally equivalent behaviours
that can be used to achieve the set-goal. In infancy and adulthood, Shaver and Mikulincer
emphasised that behavioural systems continually learn, mature, and reorganise in identi-
fying new and better ways of achieving the set-goal.273 However, within this diversity, they
conceived of the Ainsworth attachment classifications as representing the basic repertoires
for achieving the direct or conditional availability of attachment figures. Similarly, in the case
of the exploratory system, Shaver and Mikulincer proposed that maximising and minimis-
ing strategies are available.274 Maximising the exploratory system entails attempts to master

novelty even when the context does not call for it and enough information is already in hand.
This may contribute to procrastination and indecision, doubts, and worries. Maximisation of
the exploratory system responds to an environment in which familiarity is distrusted, much
like anxious attachment style expresses a lack of confidence in the availability of attachment
figures even when they are physically present. Minimisation of the exploratory system en-
tails the inhibition of cognitive openness, exploration, or curiosity even when confronted by
a novel or challenging situation. This strategy responds to an environment in which previous
attempts at exploration were punished or tended to backfire.
Fifth, Shaver and Mikulincer highlighted that any behavioural system will include cog-
nitive components. In the case of the attachment and exploratory systems, maximising and
minimising strategies are activated depending on the availability of cognitive schemas from
within a hierarchical network of associations formed through past experiences. Following
Bowlby, Shaver and Mikulincer downplayed the autonomy of the affective components of
the behavioural system, often describing these components as mere ‘reflections’ of the cog-
nitive and behavioural aspects.275 Sixth and finally, Shaver and Mikulincer argued that any
behavioural system will have specific activating or inhibitory links with other behavioural
systems. For instance, it would be expected that the activation of the attachment system
would inhibit the operation of the exploratory system.
In a paper with Josh Hart and Jamie Goldenberg, Shaver argued that the attachment be-
havioural system also has a wider context: the ‘security metasystem’.276 Ainsworth argued
that a sense of security could be fed by multiple sources, not just intimate relationships
(Chapter 2). Apparently without awareness that they were reclaiming Ainsworth’s position,
Shaver and colleagues proposed that a sense of security can be achieved variously through
experiencing the availability of attachment figures, through experiencing the self as worthy
273 Mikulincer, M. & Shaver, P. (2008) Contributions of attachment theory and research to motivation science.
In J.Y. Shah & W.L. Gardner (eds) Handbook of Motivational Science (pp.201–216). New York: Guilford, p.204.
Oxford: Oxford University Press, pp.473–4.
275 E.g. Mikulincer, M. (2006) Attachment, caregiving and sex within romantic relationships. In M.
New York: Guilford: ‘The profound joy and affection, self-protective anxiety, numbing boredom, corrosive anger,
lustful passion, uncontrollable jealousy, and intense sorrow experienced in romantic relationships are reflections
on the central importance of these behavioural systems’ (23–4).
276 Hart, J.J., Shaver, P.R., & Goldenberg, J.L. (2005) Attachment, self-esteem, worldviews, and terror manage-
ment: evidence for a tripartite security system. Journal of Personality and Social Psychology, 88, 999–1013.
and capable, and through experiencing the world as orderly and meaningful. These three
psychological processes, for all their differences, therefore have in common that they can
serve both to modulate experiences of anxiety and contribute to broaden-and-build cycles.
As well as a behavioural system in its own terms, attachment was therefore situated as part
of a security metasystem. The species-level function of the system, it was proposed, is self-
regulation. The metasystem is activated by potential feelings of anxiety, which can arise from
relational sources such as close relationships but also from non-relational sources such as re-
minders of mortality. It is deactivated by felt security, regardless of the source of this feeling.
The achievement of this goal can be pursued, as needed, through attachment-relationships,
drawing upon self-esteem or from appeal to cultural worldviews that make the world seem
orderly and meaningful. Though they often can work together, at times when one strategy is
hard-pressed by circumstances, the others can be brought in to substitute. This is facilitated

by the fact that they have cognitive components in common. Shaver and colleagues also ar-
gued that all three may facilitate exploration and caregiving.
Hart’s work has subsequently focused on this macro level in articulating the concept of
the ‘security system’.277 By contrast, this level of analysis has generally not been pursued
by Shaver and Mikulincer, though it has remained conceptual background. Instead, Hart
has developed ideas regarding the contribution of other behavioural systems to security.
Initially, both Shaver and Mikulincer began with the assumption that attachment style would
reflect the history of relationships with attachment figures. Attachment styles were defined
as ‘systematic patterns of expectations, emotions, and behavior in close relationships that are
viewed as the residue of particular kinds of attachment histories’.278 So, for instance, in the
1990s, Mikulincer reported from a study with a student sample that secure and anxious at-
tachment styles were associated with greater self-reported curiosity and positive attitudes to-
wards curiosity than participants with an avoidant attachment style. Furthermore, he found
that participants with a secure attachment style were more likely to trust new information in
making a social judgement than participants with avoidant or anxious attachment styles.279
It could be that cognitive openness contributes to, or even partly constitutes, a secure attach-
ment style. However, Mikulincer’s interpretation of these findings was that insecure attach-
ment hinders the full development and successful expression of the exploratory behavioural
system, reducing occasions for felt curiosity. At the time, the self-report approach to attach-
ment research was still being established, and part of the definition of secure attachment,
since Ainsworth, had been an inverse relationship with exploration. Attachment was there-
fore conceptualised as antecedent and causal for curiosity.
By contrast, in the mid-2000s, with the self-report tradition established and thriving,
Shaver and Mikulincer were willing to acknowledge that the qualities of adult attachment
relationships are fed by other behavioural systems. Attachment styles were regarded as
shaped by the variety of factors that shape perceptions of the availability, attentiveness, re-
sponsiveness (etc.) of others. In a 2005 article, Banai, Mikulincer, and Shaver argued that
‘a person who is psychologically injured in the process of seeking one selfobject provision
277 Hart, J. (2014) Toward an integrative theory of psychological defense. Perspectives on Psychological Science,
9, 19–39.
(pp.85–114). Philadelphia: Psychology Press, p.88.
279 Mikulincer, M. (1997) Adult attachment style and information processing: individual differences in curiosity
and cognitive closure. Journal of Personality and Social Psychology, 72(5), 1217–30.
tends thereafter to avoid seeking satisfaction of other selfobject needs’.280 Feelings of rejec-
tion in relation to exploration or sexuality, for instance, might well contribute to a ‘defen-
sively avoidant stance so pervasive that it generalizes across different kinds of interpersonal
experiences and leads people to dismiss social ties in general’.281 In 2007 they argued clearly
that ‘just as good attachment experiences have beneficial effects on other behavioral systems,
such as exploration and caregiving, good experiences related to those behavioral systems are
likely to feed back on the attachment system in ways that allow it to function in a less defen-
sive, less distorted way’.282 Elsewhere, Shaver expressed regret that he and Hazan had earlier
described adult romantic relationships as grounded in attachment, sexuality, and caregiving.
He felt that exploration and affiliation were also behavioural systems integral to adult ro-
mantic relationships, and to the functioning of the relationship as a secure base and safe
haven.283 The capacity to explore is integral to utilisation of the attachment figure as a secure

base. When exploration is blocked by a minimisation of the exploratory system, the experi-
ence of having a secure base is thinned or reduced. Conversely, when familiarity is distrusted
as part of the maximisation of the exploratory system, this may have implications for being
able to use attachment figures as a safe haven.
Over time, then, Shaver and Mikulincer came increasingly to the view that adult attach-
ment style would be fed by many sources, rather than primarily reflecting a child’s history of
attachment experiences. This position was supported in 2013, when Fraley and colleagues
found little prediction from infant attachment classifications to the ECR completed in ado-
lescence in a study with 707 participants besides a very small negative association between
infant proximity-seeking and an anxious attachment style (β = –.08). Even if it was some-
what theoretically expected, the social psychology paradigm nonetheless had a history of
claims to legitimacy on the assumption of continuity from infancy. As a result, there ap-
pears to have been ambivalence about this finding: the result was reported by Fraley and col-
leagues, but only in a web-based appendix to the published article.284 As a result, the finding
was effectively buried. It has been little noticed by subsequent researchers.
Sexuality
In pursuing their thinking about behavioural systems, Shaver and Mikulincer found the
concept of the sexual behavioural system significantly underdeveloped. Bowlby had long
acknowledged sexuality as a behavioural system. It was an essential point of comparison
in Attachment, Volume 1 as Bowlby set out the concept of attachment and described it as a
280 Banai, E., Mikulincer, M., & Shaver, P.R. (2005) ‘Selfobject’ needs in Kohut’s self psychology. Psychoanalytic
Psychology, 22(2), 224–60, p.251.

281 Ibid. p.251.
282 Mikulincer, M. & Shaver, P.R. (2007) Boosting attachment security to promote mental health, prosocial
values, and inter-group tolerance. Psychological Inquiry, 18, 139–56, p.152.

284 Fraley, R.C., Roisman, G.I., Booth-LaForce, C., Owen, M.T., & Holland, A.S. (2013) Interpersonal and gen-
etic origins of adult attachment styles: a longitudinal study from infancy to early adulthood. Journal of Personality
and Social Psychology, 104(5), 817–38, web-based supplement C. Additionally, a later study using the same sample
demonstrated that the ECR has essentially no relationship with secure base scripts in adulthood. Steele, R.D.,
Waters, T.E., Bost, K.K., et al. (2014) Caregiving antecedents of secure base script knowledge: a comparative ana-
lysis of young adult attachment representations. Developmental Psychology, 50(11), 2526–38.
behavioural system. However, in this, Bowlby’s aim was also to distinguish attachment from
sexuality, contesting versions of psychoanalytic theory that described all motivation as, in a
sense, sexuality. Bowlby evidently felt that sexuality needed to be placed in the background
in order to make space for attachment (Chapter 1). Though his discussions of the sexual
system are, as a result, somewhat cursory, it is nonetheless clear that Bowlby regarded the
adult sexual system as sharing some components of the attachment system, such as the be-
haviours clinging and kissing. Furthermore, he drew on ethological evidence to support the
(psychoanalytic) argument that the qualities of the parent–child relationship go on to in-
fluence sexual preferences when the child has grown to adulthood.285 Ainsworth followed
Bowlby in generally downplaying sexuality, despite her personal beliefs regarding its import-
ance (Chapter 2). More than Bowlby, she drew out that adult romantic relationships entail
attachment, sexuality, and caregiving. However, her remarks on the sexual behaviour system

essentially agree with those of Bowlby, with one exception. Whereas Bowlby described gay
and lesbian relationships as deviations in Attachment, Volume 1, Ainsworth argued that there
was no basis for this position, implicitly acknowledging Bowlby’s stance as prejudice.286
Overall, the concept of the sexual behavioural system was acknowledged but left sleeping
by the founders of attachment theory. And the focus of the developmental tradition of at-
tachment research on childhood directed attention away from sexuality, with the exception
of Fonagy and Crittenden.287 By contrast, the sexual behavioural system was central for
Shaver and colleagues since their focus was on adult relationships, and above all adult ro-
mantic relationships. This concern also occurred against the backdrop of growing attention
to sexuality within quantitative social science in general over the 1980s, in part reflecting the
salience of sexuality for identity politics in the period, but also more narrowly the availability
of federal funding to study social factors relevant to the spread of HIV.288 The initial model
of the sexual behavioural system offered by Shaver and Mikulincer in the 1990s was not well
formulated. They tended to pathologise people outside of monogamous romantic relation-
ships. Individuals who experience sexual attraction to more than one person were situated
as deviant, since it was expected by Shaver that ‘just like infants, adults are primed to select
one special figure’.289 And Mikulincer argued in 2006 that ‘the set goal of this system is to im-
pregnate an opposite-sex partner’.290 As in Bowlby’s work, this characterisation pathologises
gay and lesbian sexualities, and can be regarded as reflecting Shaver and Mikulincer’s general
lack of concern for sexual diversity in forms of adult relationships. Their characterisation of
285 Bowlby, J. (1969, 1982) Attachment., London: Penguin, p.233.

287 Fonagy, P. (2008) A genuinely developmental theory of sexual enjoyment and its implications for psycho-
analytic technique. Journal of the American Psychoanalytic Association, 56(1), 11–36; Crittenden, P.M. (2015)
Raising Parents, 2nd edn. London: Routledge. The inattention to sexuality among attachment researchers after
Bowlby matched a trend in post-war psychoanalytic theory to downplay the topic. Some of this was blamed on
attachment theory. See Zamanian, K. (2011) Attachment theory as defense: what happened to infantile sexuality?
Psychoanalytic Psychology, 28(1), 33–47.
288 Mottier, V. (1997) Sex and discourse. The politics of the Hite reports. In T. Carver & M. Hyvärinen (eds)
Interpreting the Political: New Methodologies (pp.39–59). London: Routledge; Epstein, S. (2006) The new attack on
sexuality research: morality and the politics of knowledge production. Sexuality Research and Social Policy Journal
of NSRC, 3(1), 1.
289 Morgan, H.J. & Shaver, P.R. (1999) Attachment processes and commitment to romantic relationships. In
J.M. Adams & W.H. Jones (eds) Handbook of Interpersonal Commitment and Relationship Stability (pp.109–24).
New York: Plenum Press, p.111.
290 Mikulincer, M. (2006) Attachment, caregiving and sex within romantic relationships. In M. Mikulincer &
G.S. Goodman (eds) Dynamics of Romantic Love: Attachment, Caregiving, and Sex (pp.23–44). New York: Guilford.
the set-goal of the sexual system also is clearly phallocentric: the sexual set-goal of women
tends not to be ‘to impregnate’ their partner.
These problems in Shaver and Mikulincer’s initial reflections on the sexual behavioural
system were grounded in a broader confusion of the evolutionary and individual levels of
analysis, distinguished by Tinbergen and the ethological tradition (Chapter 1). Stevenson-
Hinde had identified this tendency in Shaver’s work already in the 1990s.291 A behavioural
system was, for ethologists like Stevenson-Hinde, a sequence of behaviours that might be
inferred at a species level to contribute to survival and/or reproduction. The sexual system
therefore did not need to have even intercourse, let alone impregnation, as its set-goal so long
as in evolutionary history the behavioural system had contributed to the probability of re-
production. This distinction has been increasingly acknowledged by Shaver and Mikulincer
over the past decade, as they have reflected on ideas from evolutionary biology292 and pur-

sued collaborative research on the sexual behavioural system together with colleagues such
as Gurit Birnbaum and Omri Gillath. At times, this group of collaborators have argued that
the sexual system has multiple set-goals involving the initiation and maintenance of rela-
tionships.293 However, their stable position in recent years appears to have been that the set-
goal of the sexual system is sexual access to a partner.294
As Shaver and Mikulincer have further developed their concept of the sexual system, one
important area of change has been their attention to gender. The initial model of the sexual
behavioural system was implicitly phallocentric, in situating the set-goal of the system as
impregnation of the partner. However, empirical studies exploring sexual experiences in-
dicated meaningful differences between male and female participants. A study by Gillath,
Mikulincer, Birnbaum, and Shaver published in 2007 examined the effects of a subliminal
prime of a picture of an attractive nude person of the opposite sex.295 In one version of the
procedure, student participants were then asked to view pictures of members of the opposite
sex and report their relative feelings of arousal. In a second version of the procedure, after the
prime the student participants were exposed to a varied series of pictures and were asked to
decide whether each contained or did not contain sexual content, and to do so as quickly as
possible. Reaction times for accurate identification of sexual content was taken as a measure
of the accessibility of sex-related thoughts. Gillath, Mikulincer, and colleagues found that
the naked picture prime led to higher accessibility of sex-related thoughts in both male and
female participants. However, the subliminal prime led women to report lower levels of
sexual arousal. Rather than reflecting intrinsic sex-differences, the researchers interpreted
these findings with the proposal that ‘it would not be surprising if women had more reasons
than men to be threatened by certain kinds of sexual situations’.296 Women may experience
291 Stevenson-Hinde, J. (1994) An ethological perspective. Psychological Inquiry, 5(1), 62–5.

Oxford: Oxford University Press: ‘As evolutionary psychologists have explained, however, the proximal motivation
for an act (i.e., wishing to have sex with an attractive person) need not be the same as the evolutionary reason for
the existence of the motives involved’ (475).
293 Gillath, O., Mikulincer, M., Birnbaum, G.E., & Shaver, P.R. (2008) When sex primes love: subliminal sexual
priming motivates relationship goal pursuit. Personality and Social Psychology Bulletin, 34(8), 1057–69.
294 Szepsenwol, O., Mikulincer, M., & Birnbaum, G.E. (2013) Misguided attraction: the contribution of nor-
mative and individual-differences components of the sexual system to mating preferences. Journal of Research in
Personality 47, 196–200, p.197.
295 Gillath, O., Mikulincer, M., Birnbaum, G.E., & Shaver, P.R. (2007) Does subliminal exposure to sexual stimuli
have the same effects on men and women? Journal of Sex Research, 44(2), 111–21.
296 Ibid. p.119.
implicit or explicit pressure around sex, including diverse reputational threats associated
with both having and not having sex. The women students participating in the study were
understood by the researchers as interpreting the subliminal sexual prime adversely, in the
context of their experiences of what male sexuality, or male nudity specifically, represented
to them.
By 2007, Shaver and Mikulincer had repeatedly tested for gender effects in attachment
styles, but only found these when conducting studies of couples in interaction, not when
individuals had been studied alone.297 In examining sexuality, however, they found both
individual-level and couple-level differences. Brassard, Shaver, and Lussier asked 273 het-
erosexual couples aged 18–35 to complete the ECR alongside measures of sexual experiences
in their relationships.298 The researchers found that for women, but not for men, an avoidant
attachment style was associated with engagement in less sexual fantasy and daydreaming.

They also found that for men, but not for women, an anxious attachment style was associated
with pushiness for sex and the use of sex for reassurance. Women, in turn, responded differ-
ently to this behaviour by their partner depending on their attachment anxiety: only women
low in attachment anxiety appeared to resist this pushiness and engage in sexual interactions
according to their own desires and pacing.299
Recently, Birnbaum, Mikulincer, and colleagues developed the concept of the sexual
system into a self-report scale modelled on the ECR, the Sexual System Functioning Scale
(SSFS).300 Hyperactivation items on the SSFS were designed to tap the co-presence of de-
sire and worries about sex, ‘much like anxiety items in the ECR tap into desire and wor-
ries related to emotional closeness’.301 Deactivation items on the SSFS were designed to tap
disinterest and discomfort with sex, ‘much like avoidance items in the ECR tap into disin-
terest and discomfort with emotional closeness’.302 Both hyperactivation and deactivation
were conceptualised as strategies for achieving the set-goal of sexual access to the partner.
Hyperactivation is a strategy that entails vigilance regarding access to sex, and ready and
intense striving for sexual interactions when sexual access seems possible. Sexual access
may be sought insistently and abruptly, with less concern for context or the wishes of others,
since it can be prioritised over other potential goals including caregiving, exploration, and
attachment. Hyperactivation of the sexual system therefore entails reduced sensitivity, in the
sense of recognising, acknowledging, and responding in a timely way to the cues of a partner.
297 Mikulincer, M. (2007) Building personal relationships theory. Personal Relationships, 14(3), i–iv: ‘In general,
I have noticed informally that few gender differences appear in attachment-related studies of individuals, yet they
are common when both members of couples are studied’ (ii).
298 Brassard, A., Shaver, P.R., & Lussier, Y. (2007) Attachment, sexual experience, and sexual pressure in ro-
mantic relationships: a dyadic approach. Personal Relationships, 14(3), 475–93.

299 Little and colleagues found that an avoidant attachment style was unrelated to marital satisfaction among
partners who had frequent sex, and that an anxious attachment style was unrelated to marital satisfaction among
partners who frequently had pleasurable sex. Such findings suggest that sex can serve as an alternative source of
assurance of partner availability, even for individuals who would find it otherwise difficult to seek (avoidant) or
feel assured of (anxious) this availability in other aspects of their relationship. Little, K.C., McNulty, J.K., & Russell,
V.M. (2010) Sex buffers intimates against the negative implications of attachment insecurity. Personality and Social
Psychology Bulletin, 36(4), 484–98.
300 Birnbaum, G.E., Mikulincer, M., Szepsenwol, O., Shaver, P.R., & Mizrahi, M. (2014) When sex goes wrong: a
behavioral systems perspective on individual differences in sexual attitudes, motives, feelings, and behaviors.
301 Szepsenwol, O., Mikulincer, M., & Birnbaum, G.E. (2013) Misguided attraction: the contribution of nor-
mative and individual-differences components of the sexual system to mating preferences. Journal of Research in
Personality 47, 196–200, p.197.
302 Ibid.
Deactivation is a strategy that entails reduced striving for sexual access, in the service of
avoiding rejections that might further reduce the fragile sexual availability of the partner.
Other behavioural systems may then have a relatively greater role in shaping decision-
making regarding partners, including appraisals regarding the need for or meaning of sex.
Where hyperactivation and deactivation are low, the researchers proposed that the ‘pri-
mary strategy’ of the sexual system will entail communication of sexual interest with sensi-
tivity to partner signals and context. This suggests the potential for some bidirectional causal
links between secure adult attachment and the primary strategy of the sexual system, though
the links will depend on the cultural and social factors that shape perceptions of the relational
meaning of sexual interest and activities.303 However, the relationship between the sexual
and attachment behavioural systems can be expected to vary substantially by life-stage. In
adolescence and early adulthood, characteristically, sexual interactions with partners may be

sought without the expectation of the partner becoming a secure base. The sexual encounter
may offer instead an opportunity for a cocktail of other social and psychological processes,
including the exploration of identity, gender, personal values, power, and status. It may also
provide a temporary safe haven.304 Furthermore, Birnbaum and colleagues have also argued
that causal links between individual differences in the attachment and sexual behavioural
systems may be moderated by the context. For example, the conditions of success or failure
of the sexual system may contribute especially to schemas regarding the availability of part-
ners during periods when certainty regarding the relationship is low: in the early months of a
relationship or following serious conflicts or separations.305
Birnbaum, Mikulincer, and colleagues examined the association between the SSFS and
the ECR in two samples of undergraduates. There was a robust association between ECR
attachment anxiety and SSFS hyperactivation (r = .52, .57). Both reflect concerns about the
availability of intimacy and about the partner as a safe haven who can offer regulation and
satisfaction in the context of arousal; a substantial association was therefore anticipated.
However, ECR attachment avoidance had a much weaker association with SSFS deactivation
(r = .28, .27). It had at least as strong a correlation with SSFS hyperactivation (r = .25, .38).306
An avoidant attachment style may at times be linked to the hyperactivation of the sexual
system. This may be because sexuality can be made to serve as an alternative to or a substitu-
tion for intimacy, a tactic of a minimising attachment strategy.
The distinction between the sexual and attachment systems in adulthood suggests that
they may have distinct networks of associations with other behaviours, beliefs, and experi-
ences. In one study investigating the correlates of sexual system functioning, Birnbaum,
Mikulincer, and colleagues showed undergraduate participants films of various people of
the opposite sex, presented in the manner of a dating site. In examining their data for sexual
Journal of Personality and Social Psychology, 106(5), 822–42, p.823; Mizrahi, M., Hirschberger, G., Mikulincer,
M., Szepsenwol, O., & Birnbaum, G.E. (2016) Reassuring sex: can sexual desire and intimacy reduce relationship-
specific attachment insecurities? European Journal of Social Psychology, 46(4), 467–80.
304 See also Collins, W.A. (2003) More than myth: the developmental significance of romantic relationships
during adolescence. Journal of Research on Adolescence, 13(1), 1–24.

305 Birnbaum, G.E. (2018) The fragile spell of desire: a functional perspective on changes in sexual desire across
relationship development. Personality and Social Psychology Review, 22(2), 101–127.

hyperactivation, the researchers found that these were moderated by gender. Among men,
hyperactivation was associated with greater reported interest in relatively less-attractive
women. Among women, hyperactivation was associated with more interest in relatively
more-attractive men. Birnbaum and colleagues interpreted these findings as suggesting
that hyperactivation and deactivation represent conditional strategies, part of the human
evolutionary repertoire. Where sexual access was uncertain for women, in human evolu-
tionary history reproductive success would have been increased by focusing attention on
healthy mates to ensure successful offspring. By contrast, where sexual access was uncertain
for men, the equivalent strategy to increase reproductive success would have been to reduce
the standards used for assessing prospective mates. In a replication study, effects remained
significant when controlling for scores on the ECR, supporting the autonomy of sexuality
as a distinct behavioural system and the incremental validity of the SSFS hyperactivation

scale.307
The distinctive predictive power of the SSFS deactivation scale was also demonstrated in
a later study by Szepsenwol, Mizrahi, and Birnbaum:308 62 couples who had recently begun
dating completed the ECR and SSFS at four-month interviews on three occasions, along-
side self-report measures of relationship satisfaction. Birnbaum and colleagues found that
an individual’s report of SSFS deactivation was associated with their report of lower rela-
tionship satisfaction early in the relationship: this effect continued throughout the first year
among participants whose partner had an avoidant attachment style, but disappeared among
participants whose partner did not have an avoidant attachment style. Birnbaum and col-
leagues interpreted this finding in light of the idea that, though sexuality remains powerful
and relevant, in many couples relationship satisfaction comes over time to rest more on the
attachment system.309 By the end of the first year, this permits relationship security to com-
pensate for potential sexual system deactivation in providing happiness in the relationship.
Overall, Birnbaum, Mikulincer, and colleagues’ work on the sexual behavioural system
has represented the most sustained attempt to flesh out Bowlby’s early, and unfinished, char-
acterisation, and respond to Ainsworth’s call for further inquiry. There has been healthy
interest in what is still a relatively new model and self-report scale from within the wider area
of relationship psychology. Predictably perhaps, the SSFS and surrounding theory has gen-
erally been ignored by researchers in the developmental tradition of attachment research,
even those with a central concern with sexuality such as Fonagy. Nonetheless, a comparison
of Birnbaum, Mikulincer, and colleagues’ account of the sexual behavioural system and that
of Main can offer an indication of the contribution of this work to attachment theory.
Main took the concept of conditional strategies from ethological discussions of mating
behaviour. The diverse repertoires of animal behaviour that can contribute to reproductive
success formed the model for Main’s theory of repertoires of infant behaviour that promote
survival (Chapter 3). Avoidance was theorised to be a ‘cut-off ’ behaviour in the manner of
Chance, used to maintain regulation and circumvent an approach–avoidance conflict by
minimising attention to the stimuli provoking this conflict. Ambivalent/resistant attach-
ment was conceptualised through Trivers as an intensification of signals to the caregiver of
307 Ibid. p.835.

308 Szepsenwol, O., Mizrahi, M., & Birnbaum, G.E. (2015) Fatal suppression: the detrimental effects of sexual
and attachment deactivation within emerging romantic relationships. Social Psychological and Personality Science,
6(5), 504–512.
309 See also Birnbaum, G.E. & Finkel, E.J. (2015) The magnetism that holds us together: sexuality and relation-
ship maintenance across relationship development. Current Opinion in Psychology, 1, 29–33.

the need to attend to the child, achieved through a perceptual vigilance regarding signals
that might suggest the caregiver’s unavailability. Both conditional strategies were anticipated
at species level to increase the odds of infant survival when confidence is not possible in the
caregiver’s availability. Birnbaum, Mikulincer, and colleagues’ account of the sexual system
likewise described three different pathways to sexual access to the partner. The role of at-
tentional vigilance is also highlighted in their account, though it seems to be a contributor
rather than the essence of the maximising strategy as for Main.
The exact relationship between sexual behavioural repertoires and evolutionary success
seems unstable and somewhat muted in the work of Birnbaum, Mikulincer, and colleagues.
The social psychologists have only relatively recently stabilised ‘sexual access’ as the set-goal
of the sexual behavioural system. However, an additional contributing factor may have been
the tendency in the work of Shaver and Mikulincer to lose track of the distinction between

two of Tinbergen’s four questions (Chapter 1), regarding the expected proximal and the evo-
lutionary outcome of a behaviour sequence. The researchers have moved from reproductive
success to sexual access as the set-goal of the sexual behavioural system, but without fully
articulating the relationship between sexuality and reproduction, or between individual
motivation and species-level behavioural repertoires. The inattention to the affective com-
ponents of behavioural systems inherited from Bowlby has proven a hindrance here, since
this might have prompted further attention to both individual-level motivations and the
evolutionary basis of behavioural repertoires. For instance, the contribution of a deactiva-
tion of the sexual system to the achievement of either sexual access or reproductive suc-
cess is both undertheorised and empirically unexamined. At the same time, the qualities of
sexuality as at times insistent, disquieting, and above all polyvalent, which make this aspect
of human life both uncomfortably personal and uncomfortably social, have yet to be ad-
equately examined.310
Nonetheless, the efforts of Birnbaum, Mikulincer, and colleagues to describe the sexual
behavioural system represent a major advance. With the exception of sexual abuse as a kind
of trauma, Main did not further discuss sexual and/or reproductive behaviour in her writ-
ings after applying the concept of conditional strategies to infant behaviour. By contrast,
Birnbaum, Mikulincer, and colleagues have provided evidence that concern about the avail-
ability of sexual access and hyperactivation of the wish to achieve it is associated in antici-
pated ways with concern about the availability of a safe haven and hyperactivation of the
wish to achieve it. They have also shown that, though correlated, the two phenomena are not
reducible to one another and have distinct correlates. It is easy to imagine that Bowlby would
have been pleased by such findings, as evidence against the tendency he disliked in the psy-
choanalytic theory of his day to treat sexual and relational motivations as interchangeable.
Dominance
Whereas the sexual behavioural system was already described, if briefly, by Bowlby, an in-
novation made by Shaver and Mikulincer has been the introduction of a ‘dominance’ behav-
ioural system. Though not discussed as a behavioural system, the concept of a behavioural
310 Target, M. (2015) A developmental model of sexual excitement, desire and alienation. In A. Lemma & P.E.
Lynch (eds) Sexualities (pp.43–62). London: Karnac.

repertoire associated with dominance can be found already in Bowlby’s reports from eth-
ology. For instance, in Attachment, Volume 1, Bowlby discussed ethological observations
indicating that dominance could be recruited in the service of the caregiving system in non-
human primates. Dominance behaviours were described as those that signal the availability
of coercion, or violence if pushed, if deference is not provided. Bowlby described ‘the im-
portant observation that when a dominant male senses a predator or other danger he com-
monly threatens or even attacks a juvenile that unwarily approaches the danger spot. The
dominant male’s behaviour, by frightening the juvenile, elicits the juvenile’s attachment be-
haviour. As a result the juvenile seeks the proximity of an adult animal, as often as not that
of the very male that frightened it; and by so doing the juvenile also removes itself from
the danger.’311 However, beyond characterising dominance as a behavioural repertoire with
links to other systems such as caregiving and attachment, as well as aggression, Bowlby

did not give attention to the topic. Though strongly interested in questions of power and
domination in his early work before the development of attachment theory, from the 1950s
onwards he tended to consider the world from the point of view of the infant attachment
system.312 Dominance therefore became relevant only insofar as it impacted the caregiving
system. As Smith and Connelly astutely observed, in Bowlby’s importation of ethology into
child development, the concepts of dominance and territory were both left behind, albeit for
different reasons.313
Shaver’s first sustained consideration of the topic of dominance appeared in his writings
on the concept of ‘narcissism’, which were prompted by his longstanding interest in the rela-
tionship between self-concept and affectionate relationships with others, though also by the
wider context of interest in personality disorders within social psychology. In ‘Shamed into
self-love: dynamics, roots, and functions of narcissism’ from 2001, Robins, Tracy, and Shaver
argued that there are two paths to getting ahead in life, ‘through exerting influence over
others and through accomplishments’.314 Here the researchers referred to a ‘getting-ahead
orientation’ to achieve influence and accomplishments, fuelled by ‘the power and achieve-
ment motives’.315 They speculated that these motives may operate quite widely. Usually they
are moderated by or combine with affiliative motives, so that the wielding of power and the
attainment of achievements are either in the service of some others or at least not at their ex-
pense. However, Robins, Tracy, and Shaver speculated that where the goals of the attachment
system have been chronically unmet, one response may be the recruitment of the ‘getting-
ahead orientation’ in the service of attachment. This, they proposed, may be the origins of the
spectrum towards ‘narcissistic personality disorder’. Minimising and maximising strategies
are premised on the expectation that at least a conditional availability of attachment figures
can be achieved in this way. By contrast, a temperamental predisposition towards aggres-
sion or irritability may predispose self-assertion as a kind of ‘tertiary strategy’ to coercing
the notice and recognition of others, and sometimes perhaps their availability: ‘According to
our reasoning, the search for love and affection could not have worked under the childhood
311 Bowlby, J. (1969, 1982) Attachment. London: Penguin, p.227.

312 Bowlby, J. (1946) Psychology and democracy. The Political Quarterly, 17(1), 61–76; Mayhew, B. (2006)
Between love and aggression: the politics of John Bowlby. History of the Human Sciences, 19(4), 19–35.
313 Smith, P.K. & Connolly, K. (1972) Patterns of play and social interaction in pre-school children. In N. Blurton
Jones (ed.) Ethological Studies of Child Behaviour (pp.65–96). Cambridge: Cambridge University Press, p.68.
314 Robins, R.W., Tracy, J.L., & Shaver, P.R. (2001) Shamed into self-love: dynamics, roots, and functions of nar-
cissism. Psychological Inquiry, 12(4), 230–36.

315 Ibid. p.233.
conditions that fostered narcissism, particularly if the developing narcissist was tempera-
mentally aggressive and irritable. Given these constraints, narcissists were, in a sense, correct
to have pursued a more assertive, self-promoting strategy.’316
In 2011, Shaver and Mikulincer returned to these considerations, introducing the concept
of a ‘power behavioural system’. They argued that the set-goal of the power system is to ‘re-
move threats and obstacles that interfere with a person’s sense of power’.317 They proposed
‘felt power’ as the set-goal for the power system. Their proposal was explicitly modelled on
‘felt security’ as the set-goal of the attachment system according to Sroufe and Waters.318 At
first glance, this might appear to be a set-goal solely at the individual level, rather than the
achievement of a particular self–environment relationship. However, Shaver and Mikulincer
defined ‘felt power’ as the sense that one can control the environment to have one’s needs met
‘without undue social interference’.319 In a sense, ‘felt power’ might be regarded as another

contributory to the ‘security metasystem’ theorised by Shaver in his earlier work with Hart
and Goldenberg. Like attachment, self-esteem, and cultural meaning, the sense that the en-
vironment can be controlled to meet one’s needs may form one contributory factor to a sense
of felt security.320
Shaver and Milulincer proposed that the primary strategy of the power system is to see
felt power in a manner commensurate with the aims of other behavioural systems. Indeed,
they suggested that ‘power often facilitates the smooth operation of other behavioral sys-
tems, such as exploration, caregiving, and sex’.321 For instance, control over the environment
may be sought in order to ensure that the needs of other people can be met, perhaps with
the power system operating in the service of caregiving or with power prompting the feeling
of responsibility for dependents. However, in the context of repeated failures to predictably
achieve felt power, feelings of fear and helplessness may instead be evoked. Then a max-
imising or minimising strategy will be brought online. Shaver, Sagev, and Mikulincer pro-
posed that hyperactivation of the power behavioural system will occur when an individual
experiences concern about the availability of sufficient control over their environment to
ensure that their needs are met. The hyperactivation of the power system leads to increased
vigilance regarding threats or obstacles to control of the environment to ensure that one’s
needs can be met. This is fuelled by both ‘an excessive urge to gain power and an extreme fear
316 Ibid. p.234. The potential for tertiary strategies is not discussed explicitly by Shaver and colleagues. However,
the existence of tertiary strategies or minor secondary strategies is suggested by Mikulincer and Shaver’s claim
that anxiety and avoidance are the ‘major’, not only, secondary strategies, e.g. Mikulincer, M. & Shaver, P.R. (2012)
Attachment theory expanded: a behavioral systems approach to personality. In K. Deaux & M. Snyder (eds) Oxford
Handbook of Personality and Social Psychology (pp.467–92). Oxford: Oxford University Press. Mary Main and Erik
Hesse, personal communication, August 2019, have agreed that controlling or dominant behaviour may be a ‘ter-
tiary’ strategy if primary and conditional strategies fail. However, they would not term this a ‘conditional strategy’
since this is a technical term from ethology to refer to behavioural repetoires made available by human evolution
for the purpose of survival. Dominance behaviour could be such a repertoire, but they are not sure and would re-
gard a cross-species review as necessary for addressing the question.
318 Mikulincer, M. & Shaver, P.R. (2011) Attachment, anger, and aggression. In P.R. Shaver & M. Mikulincer
(eds) Human Aggression and Violence: Causes, Manifestations, and Consequences (pp.241–57). Washington,
DC: American Psychological Association, p.250.
319 Ibid.
320 Cf. Bauman, Z. (2001) Community: Seeking Safety in an Insecure World. Cambridge: Polity Press.
of failure’, a combination that ‘results in chronic activation of the power system, even when
there is no imminent threat or actual damage to one’s power; an indiscriminate urge to assert
power over others; frequent anger and hostility toward others (who are viewed as potential
rivals); and a proclivity to attack others following minimal or ambiguous signs of competi-
tion or provocation’.322 Other concerns, including attachment and exploration, are therefore
regularly suppressed.
Shaver, Sagev, and Mikulincer’s account of deactivation of the power system was quite
different to their account of deactivation of the attachment or sexual behavioural systems.
Whereas deactivation of the attachment or sexual systems entails reduced acknowledg-
ment of stimuli that might activate the behavioural system, Shaver and colleagues argued
that deactivation of the power system entails a heightened sensitivity to threats, much like
hyperactivation of the system. However, even minimal threats are interpreted as prompting

submissive, coy, or self-abasing behaviours. What are avoided are contexts that would call for
an assertion of rights and opinions, including conflicts with others.323 Gentleness is pressed
into service, perhaps even deformed at times into a mode of submission or conciliation, ra-
ther than functioning as an expression of tenderness or of care. Shaver and colleagues de-
veloped the Power Behavioural System Scale to assess maximising and minimising of the
power system. As predicted, both hyperactivation and deactivation of the power system
were negatively associated with other existing self-report measures of feelings of power.324
Hyperactivation of the power system had a moderate positive association with both attach-
ment anxiety and avoidance; deactivation of the power system was moderately associated
with attachment anxiety.
In an observational study of 100 dating couples, Shaver, Sagev, and Mikulincer asked
participants to discuss a live problem in their relationship. Hyperactivation of the power
system predicted greater displays of hostility and distress to the partner, whereas deacti-
vation predicted greater displays of submissiveness and distress to the partner. Both di-
mensions were associated with difficulties for the couple in engaging in a constructive
discussion of the problem. These associations held when controlling for attachment anx-
iety and avoidance, and for neuroticism.325 In another study, Mikulincer and Shaver found
that priming for a feeling of powerfulness led to greater feelings of optimism when parti-
cipants were low on attachment anxiety, suggesting that the attachment anxiety interfered
with the capacity to use felt power as a basis for felt security. They also found that the same
prime led to greater objectification of others when participants were high on attachment
avoidance. They interpreted this finding as suggesting that the feeling of connectedness
and security associated with low attachment avoidance could counteract the potential for
felt power to lead to haughtiness about others.326 Such findings suggest that the power
system scale has expectable correlates in behaviour and perception, offering support for
the new scale.
323 Ibid. p.78.
324 Ibid. p.80–81.
325 Ibid. p.84.
(eds) Human Aggression and Violence: Causes, Manifestations, and Consequences (pp. 241–57). Washington,
DC: American Psychological Association, p.253.
Shaver, Sagev, and Mikulincer were also pleased to report that ‘as intended, the correl-
ation between the hyperactivation and deactivation scores was not statistically significant,
r(360) = .07’.327 Yet it is not clear why hyperactivation and deactivation of the power system
were assumed to be unrelated, rather than negatively associated. In principle, submissive
behaviour and hyperactivated power behaviour would seem to be opposites, rather than un-
related. The implication is that Shaver and Mikulincer believed that individuals who dis-
play submissive behaviour to others when they feel threatened are no more or less likely to
display dominant behaviour when they feel threatened than the rest of the population, and
vice versa.328 In fact, the idea of a two-dimensional space defined by orthogonal variables
seems rather to have been pressed into service from the ECR as a criterion of the validity of
new scales of behavioural systems for Shaver and Mikulincer, without the presentation of a
theoretical justification for why all behavioural systems should have this structure. Gillath,

Karantzas, and Fraley have suggested that the idea of orthogonality of attachment dimen-
sions was never justified: it originates, they suggest, in part from a misreading of Bowlby’s
statement in Separation about the ‘independence’ of concepts of self and other.329
The concept of the power behavioural system is unquestionably an advance on the awk-
wardly named ‘getting-ahead orientation’ from Shaver’s earlier work. However, the power
behavioural system is on much less firmer footing than, for instance, their more recent work
on the sexual behavioural system. One problem is that the relationship between dominance
and aggression remained poorly soldered down in Shaver and Mikulincer’s account, and the
two elements rattle about audiably against one another. Shaver and Mikulincer highlighted
the importance of the ‘feeling of anger, which in our view is an emotional signature of power-
system activation’.330 And they have often described anger in general as linked to the power
system. However, it remains unclear whether all assertions of power require anger. For in-
stance, considering Shaver’s earlier remarks, the use of achievements and successes to claim
glory and influence do not seem overtly aggressive. Conversely, it is not clear that all asser-
tions of anger are linked to the power system. The secure infant’s protest on the departure of
the caregiver in the Strange Situation may be angry, but this is presumably before the power
behavioural system has been assembled. The relationship between the power system and the
fear behavioural system is also opaque. Exactly the submissive or coy behaviours that Shaver
and Mikulincer suggest characterise the deactivation of the power system were regarded
by Ainsworth and Bretherton as an effect of the simultaneous activation of the fear and af-
filiative systems.331 Other attachment theorists such as Crittenden and Hilburn-Cobb have
328 It is possible that passive-aggressive behaviour is an example of the two dimensions coming together; though
if so, then this raises the question of how anxiety, passivity, and aggression interrelate. Anxious attachment in
Shaver and Mikulincer’s account emphasised only one of the three components of Ainsworth’s category, and left
to the side aggression and/or passivity. Ainsworth explicitly characterised Group C infants as passive-aggressive
in correspondence, e.g. Ainsworth, M. (1967) Letter to John Bowlby, 17 October 1967. PP/Bow/K.4/12. However,
it could well be that this is not the kind of conjunction of passivity and aggression/coercion that Shaver and
Mikulincer have in mind.

331 Bretherton, I. & Ainsworth, M.D.S. (1974) Responses of one-year-olds to a stranger in a strange situation. In
M. Lewis & L. A. Rosenblum (eds) Origins of Fear (pp.131–64). New York: Wiley.
conceptualised submissive behaviour as an independent behavioural response that can be

brought into the service of attachment strategies, rather than as a minimisation of the dom-
inance system.332 Of course, several pathways could be possible, but the rationale for consid-
ering submissive behaviour as characterising the ‘deactivation’ of the system remains both
underdeveloped and rather out of keeping with Shaver and Mikulincer’s general character-
isation of the minimising strategy for a behavioural system. It may be suspected that Shaver
and Mikulincer may have conflated deactivation of the dominance system with the oper-
ation of two further distinct behavioural systems discussed by contemporary ethologists: the
reconciliation system and the submission system.333
Secondly, it is unclear whether the power behavioural system is even a behavioural system
by Shaver and Mikulincer’s own definition. Shaver and Mikulincer have generally drawn
their concepts about behavioural systems through an expansion of behavioural repertoires

already described by ethology, such as attachment, caregiving, and sexuality. However, in
other cases some link to a behavioural repertoire has been retained, permitting Shaver and
Mikulincer to toggle backwards and forwards between specific behaviours and more general
observations about how adults feel, respond, and create symbolic meanings about one an-
other. By contrast, in the case of the power behavioural system, the link between the narrow
behavioural focus of ethology and the broad terms of social psychology appears to have un-
spooled. By Shaver and Mikulincer’s own definition, a behavioural system is a repertoire of
functionally equivalent behaviours made available by evolution, with activating and termin-
ating conditions, which would have the expectable outcome at a species-level of achieving
a particular kind of change in the relationship between individual and environment. This
definition retained the fundamental concern of ethology with repertoires of functionally
equivalent behaviours made available by evolution. Yet it is difficult to describe assertion
of ‘power’ as a distinct behavioural repertoire made available by evolution. Perhaps for this
reason, at some points Shaver and Mikulincer have referred to the ‘dominance behavioural
system’, which then offers a link back to ethology and a specific and observable behavioural
repertoire.334
Overall, Shaver and Mikulincer’s account of the power behavioural system seems to
be still a little sketched in and shaggy. It has been discussed mainly in speculative book
chapters rather than peer-review articles, generating only limited commentary to date by
other scholars.335 And though the scales for the caregiving and sexual systems appear in
Milkulincer and Shaver’s Attachment in Adulthood, situated as the capstone of their profes-
sional labours, the scale for measuring hyperactivation and deactivation of the power system
does not feature. There is in fact no explicit mention of the power behavioural system in
332 Crittenden, P.M. (1995) Attachment and psychopathology. In S. Goldberg, R. Muir, & J. Kerr (eds) John
Bowlby’s Attachment Theory: Historical, Clinical and Social Significance (pp.367–406). New York: Analytical Press;
Hilburn-Cobb, C. (2004) Adolescent psychopathology in terms of multiple behavioral systems. In L. Atkinson
& S. Goldberg (eds) Attachment Issues in Psychopathology and Intervention (pp.95–135). Mahwah, NJ: Lawrence
Erlbaum.
333 Verbeek, P. & de Waal, F.B. (2001) Peacemaking among preschool children. Peace and Conflict: Journal of
Peace Psychology, 7(1), 5–28; Kutsukake, N. & Clutton-Brock, T.H. (2008) Do meerkats engage in conflict manage-
ment following aggression? Reconciliation, submission and avoidance. Animal Behaviour, 75(4), 1441–53.
(eds) Human Aggression and Violence: Causes, Manifestations, and Consequences (pp.241–57). Washington,
DC: American Psychological Association, p.250. See also Leedom, L.J. (2014) Human social behavioral sys-
tems: ethological framework for a unified theory. Human Ethology Bulletin, 29(1), 39–65.
335 An exception is Overall, N.C. (2019) Attachment insecurity and power regulation in intimate relationships.
Current Opinion in Psychology, 25, 53–8.

Religion 501
the enormous book.336 Nonetheless, their work on the power behavioural system has been
an interesting line of inquiry, continuing Bowlby’s interest in dominance behaviour within
the ethological literature. Not least, the self-report measure of hyperactivation of the power
system is thought-provoking as the only measure of a non-clinical phenomenon mentioned
by Shaver and Mikulincer with a powerful positive association with both attachment anxiety
and avoidance. Sexual system deactivation is likewise associated with both attachment anx-
iety and avoidance, but the association is weaker than for deactivation of the power behav-
ioural system.337
Religion

Attachment to God
Submission and supplication, the deactivation of the power system, has also featured in an-
other aspect of Shaver’s work. In a collaboration with Lee Kirkpatrick, a doctoral student
from the University of Denver, Shaver developed a concern with the psychology of religion
in the 1980s,338 a subdiscipline that had seen rapid growth in the late 1970s.339 In a paper
from 1985, Shaver and Kirkpatrick emphasised the importance of submissive behaviour as
a fundamental aspect of religious life. Through such behaviour ‘the influencer advertises his
or her helplessness and dependence in order to solicit sympathy and assistance. It is easy to
see how these ideas might be applied to the process of prayer, in which the human petitioner
can often be found to heap praise on the deity and offer various concessions (ingratiation),
and to reaffirm one’s inferiority and dependence while entreating God for help and guidance
(supplication)’.340 Shaver and Kirkpatrick were interested that ingratiation and supplication
appeared to be a lowering of the self in order to achieve the beneficence of a deity. However,
practices of submission and supplication might have only a short-term effect on the feeling of
a deity as beneficent. Longer-term beneficence may need to be based on a longer-term posi-
tive relationship. The researchers found that stable images of religious figures as benevolent
336 Some relevant behaviours are mentioned on page 154 of the 2010 edition and page 150 of the 2016 edi-
tion: ‘avoidant people often entertain fantasies of perfection and power, exaggerate their achievements and talents,
and avoid situations that challenge their defences and threaten their grandiosity’. However, no mention is made of
the power behavioural system, or Mikulincer and Shaver’s own data discussed above, published by 2011, showing
that hyperactivation of the power system is moderately associated with both avoidance and anxiety. Such findings
would rather qualify the characterisation of ‘avoidant people’ in general.
337 This is discussed in Birnbaum, G.E., Mikulincer, M., Szepsenwol, O., Shaver, P.R., & Mizrahi, M. (2014)
When sex goes wrong: a behavioral systems perspective on individual differences in sexual attitudes, motives, feel-
ings, and behaviors. Journal of Personality and Social Psychology, 106(5), 822–42.
338 As well as working with Shaver, Kirkpatrick was also influenced by Bernard Spilka at Denver, a leading figure
in the emergent specialism and president of Division 36 (1985–86). Spilka, B., Shaver, P., & Kirkpatrick, L.A. (1985)
A general attribution theory for the psychology of religion. Journal for the Scientific Study of Religion, 24, 1–20.
339 The growth of psychology of religion in America reflected wider public interest in religious identities and
movements, in the context of substantial sociological change. Marty, M.E. (1985) Transpositions: American reli-
gion in the 1980s. Annals of the American Academy of Political and Social Science, 480(1), 11–23. The growing basis
for psychology of religion as a subdiscipline was expressed in and further spurred on in the specific academic con-
text by the founding of the American Psychological Division 36 ‘Psychology of Religion’ in 1976. Paloutzian, R.F.
(2017) Invitation to the Psychology of Religion, 3rd edn. New York: Guilford, Chapter 2.
340 Spilka, B., Shaver, P., & Kirkpatrick, L.A. (1985) A general attribution theory for the psychology of religion.
Journal for the Scientific Study of Religion, 24(1), 1–20, p.16.

were empirically associated with an individual’s report of a supportive family during their
childhood.341 This suggested to the researchers a potential link between attachment and the
psychology of religious belief and religious practices.
In the 1980s, there was a strong tendency in the academic psychology of religion to ex-
plicitly or implicitly model religious experience on Christian faith and practice. Shaver and
Kirkpatrick did tend in this direction at times.342 However, right from the start Shaver’s
interest in Buddhism provided a counterweight. His first discussion of Buddhism in print
was in a chapter for Everywoman’s Emotional Wellbeing, a self-help guide for women pub-
lished in 1986. Shaver and O’Connor cited the Tibetan Buddhist concept of maitri, ‘uncon-
ditional friendliness toward oneself ’. An attitude of maitri, Shaver and O’Connor proposed,
puts aside grudges, hostility, rumination, and a sense of entitlement in the world. Whereas in
his later work with Mikulincer, Shaver tended to think of deactivation of the power system

in terms of submissiveness, in his remarks on Buddhism it is possible to see a different form
of deactivation. Rather than an alternative strategy for achieving power, maitri signified for
Shaver and O’Connor a relinquishment of the fight for power, ‘along with the corresponding
attitude of openness and trust toward others and toward nature’.343
Shaver and O’Connor commented that ‘a common goal, according to Western psychology,
is enhancement of one’s feelings of control and self-esteem’.344 However, rather than an inev-
itable aspect of human psychology, Shaver and O’Connor argued that this is an effect of how
the human subject has been shaped and constructed within western culture. Dominance and
submissive behaviours may be part of the human behavioural repertoire as a result of our
evolutionary heritage. But the activation of these behaviours depends upon the perception
that control or coercion is needed in order to have our needs met. Shaver contrasted this
perception to an alternative one, which he associated with Buddhism, in which ‘control of
nature and emotion is devastating to life and ultimately impossible. Nature is the source and
sustainer of life, and human feelings are one of its brightest creations.’345 In the terms drawn
from Shaver’s later thinking with Hart and Goldenberg, it might be said that the security
metasystem can be fed by different tributaries: among these, one is the assertion of self and
a quest for control; another is a cultural/religious worldview in which striving for control is
not ultimately beneficial to ensuring one’s needs are met.
In the 1990s, Kirkpatrick and Shaver documented a number of aspects of religious experi-
ence that could be influenced by attachment style, focusing on Christian religious practices
and beliefs in American undergraduate samples. Using the Hazan and Shaver ‘love quiz’,
they found that sudden conversion experiences were almost exclusive to participants with an
avoidant attachment style.346 Indeed, nearly a third of participants with an avoidant attach-
ment style in their sample reported having experienced a religious conversion. In interpret-
ing this result it is worth keeping in mind, as mentioned in the section ‘A first attempt’, that
341 Ibid. p.11.

342 E.g. Kirkpatrick, L.A. & Shaver, P.R. (1990) Attachment theory and religion: childhood attachments, reli-
gious beliefs, and conversion. Journal for the Scientific Study of Religion, 29(3), 315–34: ‘The most striking (and
perhaps obvious) point of contact, simply stated, is that the God of most Christian traditions seems to correspond
very closely to the idea of a secure attachment figure’ (318).
343 Shaver, P.R. & O’Connor, C. (1986) Coping with stress: problems in perspective. In C. Tavris (ed.)
Everywoman’s Emotional Wellbeing (pp.305–329). New York: Doubleday, p.326.

344 Ibid. p.336.
345 Ibid. p.327.
346 Kirkpatrick, L.A. & Shaver, P.R. (1990) Attachment theory and religion: childhood attachments, religious
beliefs, and conversion. Journal for the Scientific Study of Religion, 29, 315–34, p.326.
Religion 503
the ‘love quiz’ collapsed avoidance and fear regarding caregivers into one category. Another
finding by Kirkpatrick and Shaver was that participants who endorsed an anxious/ambiva-
lent attachment style on the ‘love quiz’ were distinctive in reporting experiences of speaking
in tongues.347 Participants were also asked to self-report their attachment style to God.
Those that endorsed a secure attachment style also reported that they experience greater
life satisfaction, less anxiety, loneliness, and depression, and less physical illness than other
participants. These were also the correlates of a secure attachment style in romantic relation-
ships. Kirkpatrick and Shaver found that the effects of the two attachment variables seemed
to be additive: a secure attachment style in relation to both romantic partners and God had a
stronger relationship with report of these positive experiences than a secure attachment only
in one domain, and a secure attachment style in one domain was associated with more posi-
tive experiences than an insecure attachment style in both domains.

Ainsworth was apparently enthusiastic about Kirkpatrick and Shaver’s work on attach-
ment and religion.348 However, critics have alleged that the idea of attachment to God was
an overextension of the concept of ‘attachment’, with little resemblance to the prototype of
infant behaviour. Granqvist, Shaver, and Mikulincer responded to this criticism.349 They re-
ported experimental studies that suggested that divine figures in particular, and religious
practice in general, could serve as a safe haven in the context of distress, and to an extent
as a secure base. Attachment to God could qualify as an attachment relationship on these
grounds.350 Additionally, they pointed to the role of proximity-seeking within (Christian)
religious practice, such as going to church during life transitions; the role of language sug-
gesting ‘approaching God’ within supplicatory prayer; as well as metaphors of ‘separation
from God’ in imagining Hell or forms of punishment. This suggests symbolic forms of sep-
aration anxiety. Granqvist and colleagues also highlighted the conditions under which re-
ligious life may become especially salient: ‘people are most likely to turn to God or other
supernatural figures when they face situations that Bowlby (1982) believed activate the at-
tachment system, such as illness, injury, or fatigue; frightening or alarming events; and sep-
aration or threat of separation from loved ones’.351 Part of the problem faced by Granqvist,
Shaver, and Mikulincer in defining the relationship between attachment and religion is that
the former term has had both narrow and broad meanings since Bowlby (Chapter 1). In this
sense, it would appear from their arguments that the conditions that activate the attachment
system—narrowly construed—seem to elicit phenomena associated with the attachment
system—broadly construed—within religious life, perhaps given the role of religious prac-
tice as a possible contributory to the security metasystem. Above all, it is perhaps the role
of divine figures or the religious community as a kind of safe haven that provides the hinge
between attachment broad and narrow.
347 Kirkpatrick, L.A. & Shaver, P.R. (1992) An attachment- theoretical approach to romantic love and reli-
gious belief. Personality and Social Psychology Bulletin, 18(3), 266–75, p.266. See also Kirkpatrick, L.A. (2005)
Attachment, Evolution, and the Psychology of Religion. New York: Guilford.
348 Personal communication to Mary Main, cited in Granqvist, P. (2020) Attachment, Religion, and Spirituality: A
Wider View. New York: Guilford.

349 On a mutual interest in religion as one contributing factor to the Shaver–Mikulincer collaboration see Shaver,
P.R. & Mikulincer, M. (2011) Analysis of a collaborative working relationship. Relationship Research Newsletter,
9(2), 7–9.
350 Granqvist, P., Mikulincer, M., & Shaver, P.R. (2010) Religion as attachment: normative processes and indi-
vidual differences. Personality and Social Psychology Review, 14(1), 49–59, p.50.
351 Ibid. p.52. See also Granqvist, P. (2020) Attachment, Religion, and Spirituality: A Wider View.
New York: Guilford.

Mindfulness
Though enthusiastic about the idea of attachment to God, Shaver retained his interest in
forms of religious life without a personal God or supplicatory prayer. This interest was fur-
ther fuelled by an invitation received by Shaver and Mikulincer in October 2004 to visit the
Dali Lama, who had been interested by their work.352 Following this experience, at the same
time as their work with Granqvist, Shaver and Mikulincer developed a strand of empirical
research exploring mindfulness practices. Mindfulness-oriented meditation techniques ori-
ginally emerged within Buddhist religious life, but had become reformulated in the context
of their transplantation into western wellness technique.353 Shaver and colleagues situated
themselves in terms of this secularising trend: ‘American psychologists have lifted mindful-

ness out of this rich context (perhaps while attempting to separate it from religious consid-
erations) and applied it in a more individualistic, less socially connected, and more ethically
neutral way. In our opinion, placing mindfulness in an attachment-theoretical framework
would allow it to benefit not only from additional kinds of empirical tests but also from an
assortment of ethical, social, and developmental, yet not necessarily religious, concepts.’354
The reception of mindfulness in terms of attachment theory would, Shaver and colleagues
argued, contrast with the individualistic reception of mindfulness as a western wellness
technique. Mindfulness would entail finding a safe haven in the representation of attach-
ment figures, the Buddha, in the tradition’s teachings, and in other members of the religious
community:
English-language books about Buddhist meditation make the process of mindful medita-
tion seem rather solitary and asocial. During our discussions with the Dalai Lama in 2004,
however, it was pointed out that one of the simplest and most frequently spoken Buddhist
prayers is: “I take refuge in the Buddha, the Dharma, and the Sangha,” which means (in
our terms) the mental representation of the Buddha as a loving, compassionate, and
wise teacher; the Buddha’s teachings (dharma); and the community of fellow Buddhists
(sangha). In other words, the key concept is “taking refuge,” which is similar to Bowlby and
Ainsworth’s notion of using an attachment figure as a “safe haven” . . .
In our conversations with the Dalai Lama, he said that a common form of cultivating
maitri is to imagine, during meditation, experiencing love from someone who has deeply
loved you, “such as your mother.” We replied that “mother” might not always be a good
choice if one’s relationship with her in childhood was not comfortable. He said in that case
one could imagine being loved by the Buddha, as in the common prayer, “I take refuge
in the Buddha, the Dharma (the Buddha’s teachings), and the Sangha (the community of
fellow practitioners)”.355
352 Goodman, G.S. (2006) Attachment to attachment theory: a personal perspective on an attachment re-
searcher. In M. Mikulincer & G.S. Goodman (eds) Dynamics of Romantic Love: Attachment, Caregiving, and Sex
353 The extent and nature of this reformulation remains contested. See e.g. Shonin, E. (2016) This is not
McMindfulness. The Psychologist, 29(2), 124–5.

354 Shaver, P.R., Lavy, S., Saron, C.D., & Mikulincer, M. (2007) Social foundations of the capacity for mindful-
ness: an attachment perspective. Psychological Inquiry, 18, 264–71.

355 Shaver, P.R., Mikulincer, M., Sahdra, B.K., & Gross, J.T. (2016) Attachment security as a foundation for kind-
ness toward self and others. In K.W. Brown & M.R. Leary, (eds) The Oxford Handbook of Hypo-Egoic Phenomena,
pp.223–42. Oxford: Oxford University Press, p.231.
Religion 505
Shaver and colleagues hypothesised that attachment anxiety and avoidance would be antici-
pated to disrupt the capacity for mindfulness. They drew upon a self-report scale for mind-
fulness developed by Baer and colleagues.356 On the basis of a factor analysis of 112 items
from previous self-report mindfulness scales, Baer and colleagues had developed a scale with
five main factors which they termed 1) Nonreactivity to Inner Experience, 2) Observing/
Noticing/ Attending to Sensations/ Perceptions/Feelings, 3) Acting with Awareness,
4) Describing/Labelling with Words, and 5) Nonjudging of Experience. Shaver, Lavy, Saron,
and Mikulincer found that both an anxious and an avoidant attachment style made strong
unique contributions to scores on the mindfulness scale, and that the two attachment di-
mensions accounted for 42% of variance in total mindfulness.357 In more detailed analysis,
Shaver and colleagues found that an anxious attachment style was negatively correlated with
non-reactivity to inner experience (r = –.54), acting with awareness (r = –.46), and non-

judging experiences (r = .43). These are factors that might be especially vulnerable to ru-
mination. They found that the avoidant attachment style was negatively associated with all
of the factors of mindfulness. That is, unlike anxious attachment, avoidant attachment was
also associated with difficulties noticing sensation, and difficulties giving words to such sen-
sations even when they are noticed. These factors might be specifically vulnerable to the sup-
pression of thoughts with content or associated affect related to attachment, characteristic
of the avoidant attachment style. Additionally, both attachment styles may hinder broaden-
and-build cycles, which Shaver and Mikulincer have theorised contribute to greater social
and personal resources that support regulation, and through this contribute to mindfulness.
The remarkably strong association between attachment style and the mindfulness scale
suggests a surprising degree of congruity, or even potential overlap, between the two vari-
ables. Like in earlier work disentangling neuroticism and the anxious attachment style,
the overlap between a secure attachment style and mindfulness will need to be worked out
in terms of their respective relationships with other variables, including observed behav-
iour and subjective phenomena such as self-compassion. Priming studies may also prove
important in articulating factors mediating the relationship between attachment style and
mindfulness.358 However, the strong association between attachment style and mindfulness
also presents, as Sahdra and Shaver acknowledged, a fundamental conceptual question: ‘We
encounter the paradox that “attachment security” or “secure attachment” is considered
ideal or optimal in a major stream of Western psychology, attachment theory, whereas the
ideal or optimal state in Buddhist psychology is called “nonattachment”.’359 Yet the paradox
is an effect of two different technical uses of the term ‘attachment’. There are many strands
of Buddhism, which makes generalisation difficult. Nonetheless, Sahdra and Shaver argued
that, by and large, the Buddhist concept of non-attachment suggests a lack of fixation on
mental representations of things in the world. This is not only compatible with connected-
ness with others and intimacy in relationships, but probably facilitates it. By contrast, non-
attachment in attachment theory means the lack of a discriminated and preferential intimate
356 Baer, R.A., Smith, G.T., Hopkins, J., Krietemeyer, J., & Toney, L. (2006) Using self-report assessment methods
to explore facets of mindfulness. Assessment, 13(1), 27–45.

357 Shaver, P.R., Lavy, S., Saron, C.D., & Mikulincer, M. (2007) Social foundations of the capacity for mindful-
ness: an attachment perspective. Psychological Inquiry, 18, 264–71, pp.269–70.

358 E.g. Melen, S., Pepping, C.A., & O’Donovan, A. (2017) Social foundations of mindfulness: priming attach-
ment anxiety reduces emotion regulation and mindful attention. Mindfulness, 8(1), 136–43.
359 Sahdra, B.K., & Shaver, P.R. (2013) Comparing attachment theory and Buddhist psychology. International
Journal for the Psychology of Religion, 23(4), 282–93, p.287.

relationship with a person who might serve as safe base and secure haven. Lack of a secure
base and safe haven may be anticipated to contribute to reduced flexibility in mental repre-
sentations about relationships.
Adaptation
Shaver and Mikulincer’s characterisation of religious life tends to give the impression that an
insecure attachment style is always bad. In their writings, experiences that seem character-
istic to certain forms of insecurity, such as conversion experiences or speaking in tongues,
tend to be treated respectfully but as indirect signals of mental pathology. The same is true
of their writings on adult romantic relationships. Like the majority of the developmental

tradition of attachment research,360 Shaver and Mikulincer have tended to regard insecurity
as a developmental adaptation, with the potential for some short-term benefits in achieving
the availability of attachment figures, but with long-term harms across all other domains.
As Waters and colleagues have observed (Chapter 2), there has been a tendency among at-
tachment researchers to assume that ‘all good things go together’, based on positive associ-
ations with the word ‘security’ rather than precise knowedge of the psychological process
actually under scrutiny, which may have advantages in some domains, no effect in others,
and drawbacks under certain circumstances.361 At times, Shaver and Mikulincer too have
acknowledged the problem. For instance, in a 2006 epilogue in an edited volume celebrating
his work, Shaver wrote that ‘to insist on a model of perfect security, rather than a model that
acknowledges human complexity, depth and intrapsychic conflicts and tensions is bound to
be misleading and perhaps even dangerous . . . In the attachment field we are so accustomed
to glorifying secure attachment that we rarely stop to wonder why there aren’t more saints in
the world.’362
Shaver’s observation is in part self-criticism. There are many occasions where decisions
in research design or interpretation made by Shaver and collaborators seem to have been
shaped by a ‘glorification’ of security. For instance, in a 1996 paper Mehr and Shaver de-
scribed personal inconsistency across contexts as a positive quality.363 Certainly it can be.
360 E.g. Juffer, F., Struis, E., Werner, C., & Bakermans-Kranenburg, M.J. (2017) Effective preventive interven-
tions to support parents of young children: illustrations from the Video-feedback Intervention to promote Positive
Parenting and Sensitive Discipline (VIPP-SD). Journal of Prevention & Intervention in the Community, 45(3), 202–
214: ‘Secure attachment relationships are essential for children’s current and later development’ (202).
things go together. Human Development, 48(1–2), 80–84.

Mikulincer & G.S. Goodman (eds) Dynamics of Romantic Love: Attachment, Caregiving, and Sex (pp.423–
56). New York: Guilford, pp.426–7. An illustrative case appears in Cooper, M.L., Shaver, P.R., & Collins, N.L.
(1998) Attachment styles, emotion regulation, and adjustment in adolescence. Journal of Personality and Social
Psychology, 74(5), 1380–97: ‘Results of our mediation analyses provide further support for the distinctiveness of
the three types by raising the possibility that unique constellations of underlying processes account for the dif-
ferential involvement of the three attachment groups in risky or problematic behaviors. Evidence supporting the
distinctiveness of these profiles and, in particular, differences between the two insecure groups should help to miti-
gate concerns that attachment style differences can be summarized along a single good–bad, or secure–insecure
dimension’ (1394). Here, the researchers distanced themselves from the idea that security–insecurity is the same
as good–bad. However, their argument was that ‘bad’ divides into two distinct categories. They did not contest that
security is equivalent to ‘good’.
363 Mehr, D.G. & Shaver, P.R. (1996) Goal structures in creative motivation. Journal of Creative Behavior, 30(2),
77–104, p.81.
Adaptation 507
However, at least from the way that the paper is written, the interpretation seems at least in
part shaped by the fact that this quality was associated with a secure attachment style. If it
had been associated with insecure attachment style, it is easy to imagine that the opposite
evaluation could have been made of the same quality, with speculations regarding how emo-
tion dysregulation or the lack of a safe base can disrupt the consistency of identity across
contexts. To take another example, in a 2000 paper Mikulincer and Sheffi found that partici-
pants endorsing a secure attachment style were more likely in a cognitive test to miscatego-
rise poor exemplars of types as similar. This finding was interpreted positively as indicating
the expansiveness and freedom permitted by the feeling of having a secure base. However,
if the finding had been reversed, security could easily have been praised as contributing to
greater discrimination and discernment.364
Part of the trouble in achieving effective consideration of the issue has been careless-

ness by attachment researchers in the use of the term ‘adaptation’, a problem already iden-
tified by Ainsworth in an unpublished conference paper at the International Conference on
Infant Studies in 1984.365 Following Hinde and other ethological researchers (Chapter 1),
Ainsworth expressed concern that the term ‘adaptation’ represented a twig-thicket of dif-
ferent meanings. She acknowledged that, in using the term, she and other attachment re-
searchers had hindered effective discussion and even at times misled readers. On the one
hand, ‘adaptation’ could refer to processes at a species level, in identifying a behavioural
system or trait as contributing to survival or reproduction. On the other hand, ‘adaptation’
could refer to an individual level, identifying a behaviour or trait as responsive to the avail-
able rewards and punishments of the immediate environment.366 However, Ainsworth ob-
served a third meaning of the concept: ‘In the developmental mental health sense the focus is
on how individual differences in development, and on evaluation of how well or how poorly
such development equips the individual to cope with the impact of the environment in which
he lives.’367 What distinguished this third meaning of the concept from the second was that
an evaluation was entailed. The second meaning was merely an acknowledgement that an in-
dividual may ‘adapt’ to their circumstances. Ainsworth’s third meaning was to identify ‘adap-
tation’ as the capacity to thrive in the long term in some way within those circumstances.
The subtlety and complexity of Ainsworth’s argument may have contributed to her de-
cision not to attempt to publish the article, despite its popularity with students and collab-
orators.368 Ainsworth’s argument was that individual adaptation (long-term thriving) may
364 Mikulincer, M. & Sheffi, E. (2000) Adult attachment style and cognitive reactions to positive affect: a test of
mental categorization and creative problem solving. Motivation and Emotion, 24(3), 149–74.
on Infant Studies, April 1984. PP/Bow/J.1/57.

366 Ibid.: ‘In the phylogenetic or evolutionary sense adaptation implies that in the course of natural selection
those behaviours that yield survival advantage in the environment in which the evolutionary changes are taking
place become part of the behavioural repertoire characteristic of a species . . . In the ontogenetic sense adaptation
refers to the process through which an organism adjusts to its environment in the course of development.’
367 Ibid. An explicit scale for the extent to which an individual’s behaviour appears adaptive, in this sense, would
later be developed by Steele, H., Steele, M., & Kriss, A. (2009) The Friends and Family Interview (FFI) Coding
Guidelines. Unpublished manuscript: ‘This scale refers specifically to responses to the question asking what the
respondent does when distressed or upset. An adaptive strategy may involve seeking comfort from others (e.g.
parents, friends, or siblings), engaging in a favorite activity that relieves their unhappiness (e.g. listening to music,
walking the dog), or simply thinking things through.’
368 E.g. Crittenden, P.M. (1992) Quality of attachment in the preschool years. Development & Psychopathology,
4(2), 209–241; Grossmann, K. (1995) Kontinuität und Konsequenzen der frühen Bindungsqualität während
des Vorschulalters. In G. Spangler & P. Zimmermann (eds) Die Bindungstheorie—Grundlagen, Forschung und
Anwendung (pp.191–202). Stuttgart: Klett Cotta.
result from adaptation (changing oneself in order to respond) to the environment. However,
there are forms of adaptation (long-term thriving) where refusal to adapt (change oneself in
order to respond) is optimal, for instance in depleting or punitive environments that can be
changed or exited. Some forms of adaptation (thriving) may come at the expense of other
forms of adaptation (thriving), as in the familiar case in which the demands of one area of
life—family, work—come at the expense of others—diet, exercise, self-care. A further com-
plexity lies in the fact that there are forms of adaptation (responding and/or thriving) that are
based very directly on adaptation (species-level natural selection), such as the deployment
of conditional strategies as evolutionary-based behavioural repertoires. However, there are
forms of adaptation (responding and/or thriving) that are more based on social learning or
other processes based more on human plasticity rather than responses directly grounded in
adaptation (species-level natural selection).

Ainsworth stated that attachment researchers all agreed that avoidant attachment should
be regarded as an adaptation (a response to the caregiving environment). And many were
becoming persuaded by Main’s argument, endorsed by Ainsworth, that the avoidant attach-
ment pattern was an adaptation (part of the human evolutionary behavioural repertoire).
However, she identified that there were significant disagreements among her students and
collaborators about whether, over the long-run, avoidant attachment contributed to adapta-
tion (long-term thriving). Ainsworth offered her conviction that this was essentially an em-
pirical matter: the question of whether ‘avoidant attachment may be adaptive according to
ultimate criteria in the mental health sense is clearly a researchable proposition’.369 Her per-
sonal expectation was that insecure attachment would tend to work against or undermine
long-term thriving. And this attitude unquestionably shaped her published descriptions of
the avoidant and ambivalent/resistant attachment classifications. She remained open to the
potential for insecure attachment to contribute to long-term benefits, for instance in some
domains. But she was not holding her breath in anticipation of such results.
Shaver’s position seems to have essentially been the same. Until around 2007, the dom-
inant theme of his writing was the benefits of secure attachment, across whatever domain
was under discussion. There are even some passages in his work that read as hymns to the
glory of secure attachment for its capacity to ‘create a kinder and more tolerant, harmonious,
and peaceful society’.370 And, even after 2007, Lawler, Shaver, and Goodman could still put
forward the extreme proposal that mental health professionals working with families should
be screened for their attachment style, claiming that it is the right of every child to ‘receive
services from health and mental health interveners who are trained in supporting relation-
ship quality and are themselves secure with respect to attachment (as assessed with adult
attachment measures)’.371 Yet a whispering, subterranean theme also appears across Shaver’s
writings before 2007, offering qualifications and at times pointing in the contrary direction.
In the 1986 chapter in Everywoman’s Emotional Wellbeing, Shaver and O’Connor noted a
on Infant Studies, April 1984. PP/Bow/J.1/57.

370 E.g. Mikulincer, M. & Shaver, P.R. (2007) Boosting attachment security to promote mental health, prosocial
values, and inter-group tolerance. Psychological Inquiry, 18, 139–56: ‘If human beings were helped by their fam-
ilies, communities, schools, religious institutions, and cultural media to become more secure, they would be better
able to create a kinder and more tolerant, harmonious, and peaceful society’ (150); ‘without a sizeable propor-
tion of secure, mindful, and self-efficacious citizens, political will alone is unlikely to accomplish desirable ethical
goals’ (152).
371 Lawler, M.J., Shaver, P.R., & Goodman, G.S. (2011) Toward relationship- based child welfare services.
Children and Youth Services Review, 33(3), 473–80, p.478.
Adaptation 509
shift in psychological theory towards a recognition that many apparent symptoms or forms
of pathology might also or sometimes better be regarded as effective responses to chal-
lenging and intractable situations. Shaver and O’Connor offered, as an example, that their
colleagues in psychology appeared to be moving in their discussions ‘from the concept of
“defense” to the more constructive concept of “coping” ’.372 Bowlby was a forerunner for this
transition (Chapter 1).
Reflecting on the difference between defence and coping, Shaver and O’Connor con-
sidered how an individual may direct their attention away from problems, in avoidance or
denial. This reduces perceptual information about the problem, and had classically been
treated in psychology as a form of pathology. However, Shaver and O’Connor countered: ‘in
cases where reality serves up a problem for which there is no solution, or at least no imme-
diate solution, what’s so bad about losing touch with it?’373 They advised their women read-

ers that if a problem is solvable, then avoidance or denial will be a costly and unproductive
strategy. However, if a problem is truly not solvable, then avoidance or denial are optimal
responses to the situation, and ‘this strategy frees their energies for other, more rewarding
activities’.374 In Ainsworth’s terms, Shaver and O’Connor’s argument was that avoidance is
not always an adaptation (a response to the situation): sometimes it can be motivated by fear
or habit rather than a genuine acknowledgement of the nature of the situation. Where avoid-
ance is deployed as a strategy in response to a situation that could otherwise be resolved,
this is neither adaptive (a response to the situation) nor adaptive (a contribution to longer-
term thriving). However, when an avoidant strategy is used in response to an unsolvable
problem, then this is both adaptive (a response to the situation) and adaptive (a contribution
to longer-term thriving).
Evidence for Shaver and O’Connor’s claim came from an early study by Mikulincer and
Florian published in 1995. This would be the first in a slow accumulation of discrepant
findings by Mikulincer and colleagues, which documented the benefits of an insecure at-
tachment style under specific circumstances. Mikulincer and Florian asked 92 Israeli army
recruits to complete the ‘love quiz’ at the start of their basic combat training. Their appraisal
of the training, coping strategies, and peer evaluations of their behaviour were assessed four
months later. Individuals endorsing an avoidant attachment style reported the use of more
distance forms of coping than other participants, and were just as likely to be nominated for
leadership positions as participants endorsing a secure attachment style. Interpreting their
results, Mikulincer and Florian reflected that ‘although the tendency of avoidant persons to
maintain social distance may be negatively evaluated in an emotionally laden interaction, it
may be that in purposive instrumental interaction, like daily activity during combat train-
ing, avoidant persons may provide concrete assistance and relief from distress to others’.375
Basic combat training may represent the kind of intractable problem described by Shaver
and O’Connor where physical, mental, and ethical difficulties may not be resolvable, espe-
cially through rumination. There could be personal and social benefits to freeing energies
for other more rewarding activities. However, a limitation of Mikulincer and Florian’s 1995
paper was its focus on individual coping. In a larger study of multiple units during basic
372 Shaver, P.R. & O’Connor, C. (1986) Coping with stress: problems in perspective. In C. Tavris (ed.)
Everywoman’s Emotional Wellbeing (pp.305–329). New York: Doubleday.

373 Ibid. p.323.
374 Ibid.
375 Mikulincer, M. & Florian, V. (1995) Appraisal of and coping with a real-life stressful situation: the contribu-
tion of attachment styles. Personality and Social Psychology Bulletin, 21(4), 406–414, p.413.
combat training, published in 2007, Davidovitz, Mikulincer, Shaver, Izsak, and Popper re-
ported that the higher the officer’s score on avoidance on the ECR, the more the self-reported
mental health of their unit deteriorated. In the first two months this deterioration was mod-
erated by the soldiers’ own attachment style. However, over the four months there was a
negative association between officer avoidance and their soldiers’ mental health regardless of
the attachment styles of the latter.376 Yet Davidovitz and colleagues also found a positive as-
sociation between officers’ attachment anxiety and followers’ mental health. This positive as-
sociation, however, seemed to come at the expense of the unit’s performance in exercises.377
Another finding from Mikulincer’s research group offered further indication that inse-
cure attachment strategies can have certain benefits depending on the circumstances. In
2001, Berant, Mikulincer, and Florian published a study of mothers’ responses to the diag-
nosis of congenital heart disease in their infant. The researchers examined the relationship

between attachment style, coping strategies, wellbeing, and severity of the infant’s diagnosis.
They found that mothers endorsing a secure attachment style tended to utilise a combin-
ation of support-seeking and avoidant-coping strategies if their babies had severe forms of
congenital heart disease. The temporary use of avoidance or denial was positively associated
with better reported wellbeing. Berant and colleagues interpreted this finding as suggesting
that avoidant strategies can be helpful for a time in the face of an irresolvable problem if
they are combined with the security to permit support-seeking as needed. When mothers
had babies without severe forms of congenital heart disease, then, as expected, participants
endorsing an avoidant attachment style relied especially on avoidant and distancing coping
strategies, and reported moderate levels of wellbeing.
However, for mothers who had babies with severe medical problems and who had an
avoidant attachment style, avoidant and distancing coping strategies were not used; these
mothers reported the very lowest levels of wellbeing. Berant and colleagues proposed that in
the face of an irresolvable problem that is also too appallingly distressing to maintain a dis-
tancing strategy, the avoidant strategy breaks down into a less-strategic state, with some of
the features of the anxious attachment style but characterised by a whirlpool of distress, ru-
mination, and felt insecurity.378 A potential explanation for this finding was put forward by
Gillath, Giesbrecht, and Shaver. These researchers argued that an avoidant strategy can help
keep at bay thoughts and feelings that might otherwise undermine coping. However, where
these thoughts and feelings intrude, the outcome is worse than had the strategy not been
attempted. They used a computerised task to assess the capacity of participants to resist dis-
tractions from the task. An avoidant attachment style was associated with greater success at
the task, except when participants had been primed to think about a past occasion in which
they were made to feel insecure.379
The potential for insecure attachment styles to confer local advantages has been increas-
ingly recognised since 2007. Nonetheless, Mikulincer and Shaver’s 2016 book Attachment
376 Davidovitz, R., Mikulincer, M., Shaver, P.R., Izsak, R., & Popper, M. (2007) Leaders as attachment fig-
ures: leaders’ attachment orientations predict leadership-related mental representations and followers’ perform-
ance and mental health. Journal of Personality and Social Psychology, 93(4), 632–50, p.645.
377 Ibid. p.647.
378 Berant, E., Mikulincer, M., & Florian, V. (2001) The association of mothers’ attachment style and their psy-
chological reactions to the diagnosis of infant’s congenital heart disease. Journal of Social and Clinical Psychology,
20(2), 208–232, p.227.
379 Gillath, O., Giesbrecht, B., & Shaver, P.R. (2009) Attachment, attention, and cognitive control: attachment
style and performance on general attention tasks. Journal of Experimental Social Psychology, 45(4), 647–54.
Adaptation 511
in Adulthood, in part for narrative reasons and in part because it is surveying the existing
literature, at times approaches a list of goods associated with the secure attachment style and
a list of bads associated with insecure attachment styles.380 Attention to the circumstances
in which insecure attachment styles can confer benefits has instead been the priority of one
of Mikulincer’s former students, Tsachi Ein-Dor. Ein-Dor took a course with Mikulincer on
attachment theory as an undergraduate student. He was struck by an apparent contradic-
tion. On the one hand, the ethological-evolutionary basis of attachment theory suggested
that behavioural repetoires common across a species likely have some survival or repro-
ductive value in an expectable environmental niche. On the other hand, Mikulincer’s lec-
tures emphasised the disadvantages of insecurity.
Ein-Dor served as the operations manager for Mikulincer’s research group between 2002
and 2009, undertaking masters’ and doctoral study.381 When Mikulincer moved from Bar-

Ilan to become founding Dean of the School of Psychology at the Interdisciplinary Center
Herzlyia in 2007, Ein-Dor followed to help set up the new laboratory. During this time, Ein-
Dor paid close and critical attention to the data being gathered by Mikulincer’s group. Whilst
the overarching story told in the published papers was about the virtues of a secure attach-
ment style, he noticed an accumulation of research findings suggesting the benefits of inse-
cure attachment styles under specific circumstances. Some of these findings might be due
to chance. However, Ein-Dor began to see a logic to the findings, and developed hypotheses
regarding the conditions under which an insecure attachment style might be an asset. He
came to regard an anxious attachment style as a potential asset in contexts in which vigi-
lance would be rewarded, and an avoidant attachment style as a potential asset in contexts in
which instrumentalism would be rewarded. Though the co-author on many earlier papers
published by Mikulincer’s research group, Ein-Dor’s first paper as lead author was published
in 2010 entitled ‘The attachment paradox: how can so many of us (the insecure ones) have no
adaptive advantages?’, with Mikulincer and Shaver both as co-authors.382
In this paper, Ein-Dor and colleagues acknowledged that Mikulincer and Shaver had time
and again reported results indicating the benefits of a secure attachment style. Yet they also
pointed to theoretical work suggesting that insecure attachment can be adaptive, not only in
the sense of responsive to the environment but also in the sense of conferring some advan-
tages. This position, already under discussion by Ainsworth in the mid-1980s (Chapter 2),
was developed in print by researchers in the developmental tradition such as Belsky and
Crittenden.383 Belsky’s emphasis on the contribution of attachment strategies to repro-
ductive fitness was especially influential for Ein-Dor.384 As well as this theoretical tradition,
380 Mikulincer, M. & Shaver, P. (2016) Attachment in Adulthood, 2nd edn. New York: Guilford.
381 http://portal.idc.ac.il/faculty/en/Pages/resume.resume?userName=dGVpbmRvcg==&Language=1.
382 Ein-Dor, T., Mikulincer, M., Doron, G., & Shaver, P.R. (2010) The attachment paradox: how can so many of
us (the insecure ones) have no adaptive advantages? Perspectives on Psychological Science, 5(2), 123–41.
383 Belsky, J., Steinberg, L., & Draper, P. (1991) Childhood experience, interpersonal development, and repro-
ductive strategy: an evolutionary theory of socialization. Child Development, 62, 647–70; Crittenden, P.M. (1992)
Quality of attachment in the preschool years. Development & Psychopathology, 4(2), 209–241. For reflection on
Ein-Dor’s claims from critical psychology see Carr, S. & Batlle, I.C. (2015) Attachment theory, neoliberalism, and
social conscience. Journal of Theoretical and Philosophical Psychology, 35(3), 160–76. These researchers interpret
Ein-Dor’s perspective as essentially a positive framing of the fit between an avoidant attachment style and the
depleting, dehumanising late capitalist labour market. They argue that, even if there are some advantages to inse-
cure attachment styles, these advantages are vastly outweighed by their cost, to the point that claims about advan-
tages are misleading.
384 See Ein-Dor, T. & Hirschberger, G. (2016) Rethinking attachment theory: from a theory of relationships to a
theory of individual and group survival. Current Directions in Psychological Science, 25(4), 223–7, p.226.
Ein-Dor and colleagues could also point to the accumulation of studies showing some bene-
fits of the insecure attachment style. In addition to earlier findings from Mikulincer’s group,
Ein-Dor and colleagues were able to report new findings from their longitudinal follow-up
of ex-prisoners of war from the 1973 Yom Kippur war. Participants were followed up in 1991,
2003, and 2008. This research found that veterans’ avoidant attachment scores were inversely
associated with the extent to which their wives showed symptoms of trauma, such as in-
trusion or hyperarousal.385 The avoidant attachment strategy of the ex-prisoners of war ap-
peared to have kept their spouse safe from contamination by their symptoms of trauma.
However, wives’ avoidant attachment style was positively associated with veterans’ PTSD
symptoms, suggesting that, when it was not a matter of their own choosing, reduced oppor-
tunity to seek intimacy and share emotional experiences with their partner was harmful for
the veterans.

Ein-Dor and colleagues criticised the wider field of attachment research generally on
two grounds. First, following Belsky, they argued that researchers have generally designed
studies based on an implicit model of mental health as behaviour that is adaptive in indus-
trialised society, rather than developing hypotheses considering how apparently maladap-
tive behaviours might have their niche within human evolutionary history and/or particular
contemporary contexts. Second, they criticised the focus of attachment researchers on
individuals and, sometimes, dyads. Few attachment researchers besides the Minnesota
group (Chapter 4) have examined the contribution of individual differences in attachment
to group-level processes. Yet contemporary evolutionary biology, in contrast to the evolu-
tionary theory of Bowlby’s day, has come to place greater emphasis on group processes within
natural selection. For instance, the survival of the group as a whole may benefit if there are
some members who are more wary and alert to threat, some more focused on instrumental
concerns, and others more capable of coordination, negotiation, and compromise.
Ein-Dor, Mikulincer, and Shaver published a first empirical article based on this agenda
in 2011:386 46 groups of participants were observed in the laboratory room as it gradually
filled with smoke from an apparently malfunctioning computer. The results were in line with
Ein-Dor’s theory of the potential benefits of insecure attachment styles under particular cir-
cumstances. Individuals high on attachment anxiety detected the smoke more quickly and
alerted the group. This led the group as a whole to a faster response to the threat, since they
had been notified of it earlier. Effects were quite marked: a 1-point increase in attachment
anxiety was associated with an 11.5-second decrease in detection time. Individuals high
on attachment avoidance were faster at getting out the door once the danger was detected,
and contributed to a faster exit for the group as a whole. Furthermore, Ein-Dor and col-
leagues found a linear association between diversity of attachment scores in the group and
the group’s effectiveness at evacuating the room. This implied that although security might
not be associated with vigilance regarding threat or directness in avoiding the threat, it was
associated with the holistic effectiveness of the group, perhaps by facilitating coordination.
They found that the effects remained significant even with temperament measures of extra-
version and neuroticism statistically controlled. As a conceptual replication and extension
385 Ein-Dor, T., Doron, G., Solomon, Z., Mikulincer, M., & Shaver, P.R. (2010) Together in pain: attachment-
related dyadic processes and posttraumatic stress disorder. Journal of Counseling Psychology, 57(3), 317–27.
386 Ein-Dor, T., Mikulincer, M., & Shaver, P.R. (2011) Effective reaction to danger: attachment insecurities pre-
dict behavioral reactions to an experimentally induced threat above and beyond general personality traits. Social
Psychological and Personality Science, 2(5), 467–73.
Adaptation 513
of the finding that groups benefited from heterogeneity of attachment styles, Ein-Dor and
colleagues assessed the attachment styles of students at their university enrolled in courses
including a team project. The researchers found that when reported team cohesion was high,
heterogeneity in attachment anxiety and avoidance scores in the group was associated with
better grades on the group project. However, only heterogeneity in anxiety was associated
with better perceived group functioning. As such, heterogeneity of attachment avoidance
was associated with better performance evaluations, but without individuals being aware of
the increased effectiveness of their team.387
Ein-Dor and Orgad conducted another study. They led participants to believe that they
had accidently activated a computer virus that wiped the experimenter’s computer. They
were then asked to alert the department’s computer technicians. On the way to the techni-
cians’ office, they were presented with four decision-points where they could either choose

to delay the warning or continue on to their destination. Only the anxious attachment style,
and not neuroticism, was associated with continuing on to deliver the warning rather than
responding to the distractions.388 Ein-Dor and Orgad concluded that whereas anxious at-
tachment and trait anxiety might both influence threat perception, the communication of
concerns to someone who should help is a priority specific to attachment anxiety.
Such findings provided support for Ein-Dor’s claim that there can be beneficial correl-
ates of heterogeneity of attachment styles for groups, since there are specific situations in
which an insecure attachment style is a local advantage. The studies had been designed to
offer a certain analogy to human evolutionary history and the question of group survival.
However, Ein-Dor felt that he could prove his claims also on the ‘home terrain’ of attachment
research in the study of individual differences, and even in the study of contemporary pro-
fessional life. Ein-Dor, Reizer, Shaver, and Dotan proposed that individuals with avoidant
attachment styles will profit in professional fields that reward self-reliance and the ability to
work without social support.389 The researchers elegantly demonstrated this with a study of
professional singles tennis players: 58 players completed the ECR, and their progress in the
national tennis rankings was assessed over a 16-month period. Amount of training, feelings
of self-efficacy, and avoidant attachment style all predicted change in ranking movement, ac-
counting for 13.5% of variance.390 However, in a regression, only avoidant attachment style
was significant, perhaps suggesting that amount of training was in part a function of avoid-
ance. In another study, Ein-Dor and colleagues were able to demonstrate the potential bene-
fits of an anxious attachment style for individuals. Researching a card game tournament,
they found that the higher the player’s attachment anxiety score, the better their ability to
cheat without being caught and to detect others cheating.391 Research findings from other
groups have offered support for Ein-Dor’s claims regarding the potential advantages of in-
secure attachment styles under particular circumstances. Of special note are findings from
387 Lavy, S., Bareli, Y., & Ein-Dor, T. (2015) The effects of attachment heterogeneity and team cohesion on team
functioning. Small Group Research, 46(1), 27–49.

388 Ein-Dor, T. & Orgad, T. (2012) Scared saviors: evidence that people high in attachment anxiety are more ef-
fective in alerting others to threat. European Journal of Social Psychology, 42(6), 667–71.
389 Ein-Dor, T., Reizer, A., Shaver, P.R., & Dotan, E. (2012) Standoffish perhaps, but successful as well: evidence
that avoidant attachment can be beneficial in professional tennis and computer science. Journal of Personality,
80(3), 749–68.
390 Ibid. Table 2.
391 Ein-Dor, T., Perry-Paldi, A., Zohar-Cohen, K., Efrati, Y., & Hirschberger, G. (2017) It takes an insecure liar
to catch a liar: the link between attachment insecurity, deception, and detection of deception. Personality and
Individual Differences, 113, 81–7.
a randomised control trial of the Circle of Security eight-session parenting intervention.

Cassidy and colleagues reported that this intervention had no main effect on either infant
attachment security or on child mental health. However, positive effects were seen when
caregivers were high in attachment avoidance; by contrast, the intervention reduced infant–
caregiver security for caregivers who began the intervention low in attachment avoidance.392
Evaluating Ein-Dor’s position, Gillath, Karantzas, and Fraley have argued for only quali-
fied acceptance. They acknowledge that in specific circumstances, insecure attachment styles
may have local advatanges. However, they emphasise that the vast majority of studies have
found that secure attachment is associated with more positively regarded outcomes.393 Yet
Shaver and Mikulincer have acknowledged Ein-Dor’s point that this has partly been a result
of how studies have been designed and conceptualised. They have accepted that Ein-Dor’s
work has provided a helpful corrective to the tendency in their writing to treat insecure at-

tachment as, in itself, simply bad. Furthermore, Ein-Dor’s concern with the relationship be-
tween individual attachment style and group processes has been an important development
for attachment research with adults. To date, there have not been sufficient studies of small
group processes to build the critical mass for a research agenda, and research energies have
been directed elsewhere.394 However, interest in small groups has definitely been growing in
social psychology,395 and the developmental tradition has developed attachment-based par-
enting interventions that specifically make use of group dynamics such as GABI (Chapter 4).
It may be that Ein-Dor’s ongoing research efforts will contribute to small-group research as
an important new direction for attachment research.
The ECR items
For two decades, the social psychology tradition of attachment research has been under-
pinned both theoretically and methodologically by the ECR. As discussed in the section
392 Cassidy, J., Brett, B.E., Gross, J.T., et al. (2017) Circle of security-parenting: a randomized controlled trial in
Head Start. Development & Psychopathology, 29(2), 651–73. However, the assessment of outcomes was conducted
as soon as feasible after completion of the intervention, when participants low in avoidant attachment style may
have still been processing the intervention, requiring time to stabilise their caregiving. Longer before follow-up
may have provided more opportunity for positive intervention effects. It should also be noted that the researchers
appear not to have measured or controlled for the involvement of other services (e.g. child welfare involvement).
This might be another relevant moderator.

394 On the potential for self-perpetuating neglect of certain research agendas within attachment research see
Fearon, R.M.P., Bakermans-Kranenburg, M.J., & van IJzendoorn, M.H. (2010) Jealousy and attachment: the
case of twins. In S.L. Hart & M. Legerstee (eds) Handbook of Jealousy. Theory, Research, and Multidisciplinary
Approaches (pp.362–86). New York: Wiley: ‘Without an accumulation of empirical data and novel findings it may
be that we have not seen a sufficient number of new phenomena for researchers to get their teeth into, and so their
energies have, to a large extent, been directed elsewhere’ (364).
395 Marmarosh, C. & Markin, R. (2007) Group and personal attachments. Group Dynamics: Theory, Research,
and Practice, 11(3), 153–64; Yip, J., Ehrhardt, K., Black, H., & Walker, D.O. (2018) Attachment theory at work: a
review and directions for future research. Journal of Organizational Behavior, 39(2), 185–98; DeMarco, T.C. &
Newheiser, A.K. (2019) Attachment to groups: relationships with group esteem, self-esteem, and investment in
ingroups. European Journal of Social Psychology, 49(1), 63–75.
‘Creating the ECR’, Brennan and colleagues developed the ECR through a two-level factor
analysis of items from existing measures. The model of two orthogonal factors that came out
of this analysis comprised items representing attachment anxiety and attachment avoidance.
And the idea of orthogonal dimensions in a two-dimensional space, representing minimising
and maximising, became the foundation for Shaver and Mikulincer’s approach to all other be-
havioural systems. Yet there has been astonishingly little discussion of the ECR items and the
inner machinery of the measure, with researchers seeming to rest comfortably on the original
factor analysis conducted by Brennan and colleagues. An exception is Allen and colleagues,
who have called for a close study of the ECR items, but not conducted this study themselves.396
Likewise, Banai, Mikulincer, and Shaver, in a paper from 2005, urged that distinctions may be
drawn between forms of attachment anxiety, and that ‘researchers should attempt to distin-
guish among these potentially different kinds of individuals who score high on the anxiety

dimension’.397 However, these calls have not been followed up. Social psychology research on
attachment seems to have been firmly gripped in the beak of the two dimensions the ECR is
understood to embody, with the items themselves treated as of little consequence. Yet exam-
ination of the items makes this assumption all the more strange, since the items—drawn as
they are from a variety of pre-existing scales—are quite a menagerie. This was acknowledged
by Mikulincer and Shaver: ‘the items consequently range from ones concerned with relation-
ships in general to ones concerned with a particular partner. Some deal with “comfort” and
other feelings; some deal with desires and motives.’398 They expressed surprise at how well the
items have performed, ‘given their relative crudeness’. In fact, ‘it is remarkable how systematic
and cumulative our research findings have been’ on the basis of items that were ‘not designed
component-by-component with a coherent theoretical model in mind’.399
The ECR is 36 items, which is unusually long for a self-report measure. This has led to at-
tempts to produce a shortened version of the scale with the same properties. Work on trans-
lations and the ECR and attempts to produce a shortened version have represented the only
sustained conversation in the published literature about the ECR items and the latent phe-
nomena they measure, and as such the only fine-grained consideration of what participants
may be endorsing. In a 2007 paper, Wei and colleagues presented the first attempt to develop
a short version of the ECR, drawing 12 items from the 36.400 They cited a personal com-
munication from Shaver (July 2004) that the avoidant and anxious attachment styles each
have three ‘critical components’. According to Shaver, the ‘critical components’ of attachment
avoidance are:
i) concern about closeness;

ii) reluctance to depend on others;
iii) reluctance to self-disclose.
396 Allen, J.G., Stein, H., Fonagy, P., Fultz, J., & Target, M. (2005) Rethinking adult attachment: a study of expert
consensus. Bulletin of the Menninger Clinic, 69(1), 59–80: ‘We believe there is room to continue sharpening the con-
ceptual boundaries of adult attachment by further examining the content of putative attachment scale items’ (60).
397 Banai, E., Mikulincer, M., & Shaver, P.R. (2005) ‘Selfobject’ needs in Kohut’s self psychology. Psychoanalytic
Psychology, 22(2), 224–60, p.253.

(pp. 53–152). New York: Academic Press, p.141.
399 Ibid.
400 Wei, M., Russell, D.W., Mallinckrodt, B., & Vogel, D.L. (2007) The Experiences in Close Relationship Scale
(ECR)—short form: reliability, validity, and factor structure. Journal of Personality Assessment, 88(2), 187–204.
The critical components of attachment anxiety are:
i) concern about abandonment;

ii) an extensive desire for reassurance from others;401
iii) distress about the unavailability of one’s partner.
Whilst these often occur together or cause one another, the three components of each of
the forms of insecure attachment are articulated. It is possible for them to occur unalloyed.
For instance, reluctance to depend on others may often come with a concern about close-
ness, but it need not do so. Attachment figures could be regarded, for instance, as generally
untrustworthy but nonetheless attractive. Or again, distress about the availability of one’s
partner may entail concern about abandonment, but it need not do so. Attachment figures

could be regarded as distracted and inattentive, but still invested in the relationship and its
continuation.
Wei and colleagues developed the ECR-short, drawing items from the ECR with ref-
erence to the theoretical rationale of the three ‘critical components’ per form of insecure
attachment. However, in an independent attempt to make a short form of the ECR, Lo,
Mikulincer, and colleagues argued that there was a need to reconsider the Brennan study
itself. They observed that the Brennan et al. paper is usually described as creating the two
dimensions of the ECR out of the items of a factor analysis. But in fact this was a two-stage
process. The data presented by Brennan and colleagues indicated that when factored at the
item level, a higher-order factor structure was present. Twelve first-order factors were ini-
tially extracted, which Brennan and colleagues termed ‘facets’ of the two latent dimensions
found using the higher-order factor analysis.402 The 36 items did not reduce in one step to
two latent dimensions; an intermediate layer was in operation. Lo, Mikulincer, and col-
leagues conducted an exploratory factor analysis of the 36 items and found that avoidance
and anxiety each broke down into two factors. Within the avoidant attachment style, one of
the factors was discomfort with closeness. This factor accounted for 16% of variance in en-
dorsement of items on the ECR. The other factor was lack of willingness to rely on others or
willingness to disclose to others (which clustered two of the ‘critical components’ separated
by Wei and colleagues on the basis of Shaver’s personal communication). This factor ac-
counted for 14% of variance. The distinction between these two factors has been supported
in other studies.403
The anxious attachment style also broke down into two factors. The first factor repre-
sented items indicating frustration about the unavailability of the attachment figure. This
401 Wei and colleagues only had one item represent this component, Item 18 on the ECR. This item reads: ‘I
need a lot of reassurance that close relationship partners really care about me’. Wei and colleagues characterised
this item as ‘an excessive need for approval from others (Item 18)’. This is an improbable characterisation of the
item! It would seem most likely that the word ‘assurance’ was intended, rather than ‘approval’. There are items in
the ECR that represent need for approval, such as Item 34: ‘when other people disapprove of me, I feel really bad
about myself ’. However, it is Item 18, not these items about approval, that feature in the list of 12 items offered by
Wei and colleagues as the short version of the ECR. It is perhaps worth noting that a desire for approval is explicitly
one of the features specified by Shaver and Mikulincer as associated with hyperactivation of the power behavioural
system, which may account for a portion of the link between the two measures.
integrative overview. In J A. Simpson & W.S. Rholes (eds) Attachment Theory and Close Relationships (pp.46–76).
403 E.g. Olssøn, I. Sørebø, O., & Dahl, A.A. (2010) The Norwegian version of the Experiences in Close
Relationships measure of adult attachment: psychometric properties and normative data. Nordic Journal of
Psychiatry, 64(5), 340–49.
is similar to the ‘critical component’ of distress about the unavailability of the partner from
Wei and colleagues’ formulation, except for the focus on frustration. As we saw in the
section ‘Creating the ECR’, Shaver and colleagues shifted the second dimension of attach-
ment theory away from Ainsworth’s concern with inconsolability (understood as anxiety
about caregiver availability), anger, and passivity towards a focus solely on anxiety about
attachment relationships. However, the fact that one of the constitutive factors of the anx-
ious attachment style constitutes frustration suggests that the ECR has continued to tap
anger about the unavailability of the attachment figure, though with only distress/worry
about partner availability featuring focally and theoretically in accounts of what is being
measured or in the design of research studies. The ‘frustration’ factor accounted for 18%
of variance, the largest share of any of the factors. The other factor constituting the anx-
ious attachment style represented items indicating worry about the relationship. This seems

to be similar to the concern about abandonment from Wei and colleagues’ formulation.
This factor accounted for 9% of variance.404 There was no equivalent in Lo, Mikulincer,
and colleagues’ work of the wish for reassurance, the final critical component for Wei and
colleagues. However, there is only one item directly about the wish for reassurance in the
ECR: ‘I need a lot of reassurance that close relationship partners really care about me’ (Item
18).405 This item is loaded with frustration in the factor analysis, though not strongly. Here
again, there appears to be some holdover from Ainsworth’s original model, since the wish
for reassurance without active efforts to achieve it was how Ainsworth defined the C2 ‘pas-
sive’ classification.
Overall, close consideration of the ECR suggests that even if the items can be considered
to ultimately tap two latent dimensions, they do so through an intermediate layer of ‘facets’
in which relevant differences can be identified.406 Indeed, some studies do not find a two-
factor model as the optional solution, and instead identify these facets as distinct factors in
accounting for variance for endorsement of items on the ECR.407 Part of the issue is that the
items that constitute the ECR were not designed for the purpose of tapping the two latent
404 Lo, C., Walsh, A., Mikulincer, M., Gagliese, L., Zimmermann, C., & Rodin, G. (2009) Measuring attachment
security in patients with advanced cancer: psychometric properties of a modified and brief Experiences in Close
Relationships scale. Psycho-Oncology, 18(5), 490–99, p.495.
405 Reassurance is also mentioned by Item 35: ‘I turn to close relationship partners for many things,
including comfort and reassurance’. However, the formulation of the item is astonishingly unspecific—‘many
things’—so it is not clear that participants would answer based primarily on their experiences of wanting or re-
ceiving reassurance. More importantly, the item, reversed, is one that contributes to the scoring of attachment
avoidance, not attachment anxiety. A study by Lafontaine and colleagues found that Item 35 had the highest
standard error and worst discrimination of any of the avoidance items in the ECR. Lafontaine, M.F., Brassard,
A., Lussier, Y., Valois, P., Shaver, P.R., & Johnson, S.M. (2016) Selecting the best items for a short-form of
the Experiences in Close Relationships questionnaire. European Journal of Psychological Assessment, 32(2),
140–54, p.146.
406 Intermediate ‘facets’ have also been identified between the items and the latent dimensions of anxiety and
avoidance in the Attachment Style Questionnaire: confidence; relationships as secondary; discomfort with close-
ness; preoccupation with relationships; and need for approval. Karantzas, G.C., Feeney, J.A., & Wilkinson, R.
(2010) Is less more? Confirmatory factor analysis of the Attachment Style Questionnaires. Journal of Social and
Personal Relationships, 27(6), 749–80.
407 E.g. Olssøn, I., Sørebø, O., & Dahl, A.A. (2010) The Norwegian version of the Experiences in Close
Relationships measure of adult attachment: psychometric properties and normative data. Nordic Journal of
Psychiatry, 64(5), 340–49: ‘The five-factor solution. Factors 1 and 4 consist of avoidance items only. The items of
factor 1 describe avoidance of getting close or discomfort by coming close. In factor 4, the content of all items is
reluctance to self-disclosure or dependence on others ... Factors 2, 3 and 5 all consist of anxiety items. The items in
factor 2 describe worrying about abandonment or being alone. Factor 3 content statements about one’s need for
partner’s availability or reassurance, and factor 5 concerns worry that the individual wants more closeness than the
other person does’ (344).
dimensions, and item-total correlations for the anxiety and avoidance scales are frequently
low.408 Differences in the ‘facets’ that constitute the intermediate layer may contribute to dis-
crepancies between theory and measurement, reducing the precision and coherence of the
measure and the tradition of research built upon it. For instance, attachment anxiety is gen-
erally assumed by Shaver and Mikulincer to be correlated with, but distinct from, frustra-
tion in relationships. However, one of the two components of the anxiety scale is tapping
frustration whereas the other is not. Is frustration a correlate or a component of an anxious
attachment style? Wongpakaran and colleagues have argued that in fact frustration, though
related, is ultimately ‘extraneous’ to what is truly meant by attachment anxiety, and its inclu-
sion within the ECR and self-report measures based on it may be a cause of unintended noise
and imprecision.409 They also expressed concern that some ECR items may simply reflect
poor self-regard rather than an attachment-specific experience, even if this loads with at-

tachment anxiety in factor analyses. And indeed, Esbjørn and colleagues found a five-factor
solution was superior to a two-factor solution in their data, and that the ‘extra’ factor be-
yond the four identified by Lo, Mikulincer, and colleagues represented those items in the
ECR that suggest poor global self-worth (e.g. Item 34: ‘When others disapprove of me, I feel
really bad about myself ’).410 Karantzas and colleagues have argued for the ‘conceptual and
empirical importance of including both broad factors and specific facets of attachment style’,
and alleged that the constructs of attachment anxiety and avoidance are decidedly ‘blunt
instruments’. They contend that consideration of the facets will have much to offer to work
in clinical and therapeutic contexts, for instance in understanding how exactly attachment
insecurity may contribute to mental health symptoms.411
Another potential issue with the anxiety items is that a single item (Item 18) suggestive
of Ainsworth’s concept of passivity appears to have smuggled aboard the ECR, but without
a second item to stabilise measurement and without theoretical acknowledgement. Yet
passivity may also be implicated elsewhere in the measure. Item 29 is the statement ‘I feel
comfortable depending on others’; this item, reversed, was intended to represent avoid-
ance. However, it may also represent the passive desire for care Ainsworth characterised as
C2. In this light it is notable that several samples, including work by Shaver and colleagues
themselves, have found Item 29 negatively associated with avoidant attachment and posi-
tively associated with anxious attachment, rather than unrelated to anxiety as anticipated by
Brennan and colleagues.412
408 Questions have been raised about the item-total correlations of the ECR and measures based on it by sev-
eral psychometricians, e.g. Hanak, N. & Dimitrijevic, A. (2013) A Serbian Version of Modified and Revised
Experiences in Close Relationships Scale (SM–ECR–R). Journal of Personality Assessment, 95(5), 530–38.
409 Wongpakaran, T., Wongpakaran, N., & Wannarit, K. (2011) Validity and reliability of the Thai version of the
Experiences of Close Relationships–Revised questionnaire. Singapore Medical Journal, 52(2), 100–106, p.103.
410 See also Esbjørn, B.H., Breinholst, S., Niclasen, J., Skovgaard, L.F., Lange, K., & Reinholdt-Dunne, M.L.
(2015) Identifying the best-fitting factor structure of the Experience of Close Relations–Revised in a Scandinavian
example. PLoS One, 10(9), e0137218. Though this study was of the ECR-R, not the ECR, there were items falling in
the ‘extra’ factor from the original ECR, such as Items 6 and 10.
411 Karantzas, G.C., Feeney, J.A., & Wilkinson, R. (2010) Is less more? Confirmatory factor analysis of the
Attachment Style Questionnaires. Journal of Social and Personal Relationships, 27(6), 749–80, p.774; Gillath,
O., Karantzas, G.C., & Fraley, R.C. (2016) Adult Attachment: A Concise Introduction to Theory and Research.
London: Academic Press, pp.120, 245. See, for instance, work by Tasca and colleagues, which found that the associ-
ation between attachment anxiety and symptoms of depression and disordered eating was fully mediated by emo-
tion regulation strategies. Tasca, G.A., Szadkowski, L., Illing, V., et al. (2009) Adult attachment, depression, and
eating disorder symptoms: the mediating role of affect regulation strategies. Personality and Individual Differences,
47(6), 662–7.
412 See e.g. Alonso-arbiol, I., Balluerka, N., & Shaver, P.R. (2007) A Spanish version of the Experiences in Close
Relationships (ECR) adult attachment questionnaire. Personal Relationships, 14(1), 45–63.

In relation to the avoidance dimension, there seem to be distinct facets representing rela-
tive comfort with closeness on the one hand, and trust on the other hand. Again, these differ-
ences in ‘facets’ in the intermediate layer may introduce discrepancies between theory and
measurement, especially in contexts where one or the other facets may be more important.
It is easy to imagine social contexts in which trust is more salient, contexts in which comfort
with closeness is more important, as well as contexts in which both play a substantial role. In
a higher-order factor analysis, Lo, Mikulincer, and colleagues found that whereas willingness
to disclose and rely on others was entirely distinct from attachment anxiety, the items repre-
senting the facet of discomfort with closeness loaded partially on both attachment avoidance
and attachment anxiety.413 This implies that discomfort with closeness may be fed by both
attachment avoidance and attachment anxiety, albeit for different reasons.414 Consider, for
instance, Item 26 on the ECR: ‘I find that my partners don’t want to get as close as I would

like’. With what is ‘liked’ unspecified, endorsement of this item may be fed by both anxiety
and avoidance, as a factor analysis conducted by Shaver and colleagues indeed showed.415
Lafontaine, Shaver, and colleagues conducted a study to further refine the items for a short
version of the ECR. One analysis they pursued was the extent to which items could discrim-
inate particular portions of the avoidance scale. At the high end of the avoidance scale, some
items that were particularly effective were Item 29: ‘I feel comfortable depending on others’
(Reversed) and Item 9: ‘I don’t feel comfortable opening up to others’. However, ‘almost all
avoidance items performed relatively poorly on the lower portion of the avoidance scale con-
tinuum’.416 The items were quite good at picking up the avoidant attachment style, but were
poor at discriminating security. There was only one item that worked adequately, though
even then not well, in picking out security (Item 35). Yet this item had the highest standard
error and lowest discrimination of all the avoidance items. On the anxiety scale, again, the
low end was not well discriminated, perhaps because there is only one reverse item in the
anxiety scale.417 Whereas the ECR promises to capture security as the absence of anxiety
413 Lo, C., Walsh, A., Mikulincer, M., Gagliese, L., Zimmermann, C., & Rodin, G. (2009) Measuring attachment
security in patients with advanced cancer: psychometric properties of a modified and brief Experiences in Close
Relationships scale. Psycho-Oncology, 18(5), 490–99, p.495. The items representing discomfort with closeness are:
Item 7. ‘I get uncomfortable when other people want to be very close to me’
Item 13. ‘I am nervous when other people get too close to me’
Item 17. ‘I try to avoid getting too close to other people’
Item 23. ‘I prefer not to be too close to other people’
Item 9. ‘I don’t feel comfortable opening up to other people’.
414 Ibid.: ‘Conceptually, the double-loading of Discomfort on the higher-order attachment dimensions sug-
gests that both attachment anxiety and avoidance tap some discomfort with the experience of intimacy and
closeness’ (498).
415 E.g. Alonso-arbiol, I., Balluerka, N., & Shaver, P.R. (2007) A Spanish version of the Experiences in Close
Relationships (ECR) adult attachment questionnaire. Personal Relationships, 14(1), 45–63. Matters are further
complicated by findings by Mikulincer and colleagues that attachment anxiety on the ECR is associated with both
preconscious approach and avoidance goals with respect to relational closeness, which affect their motor responses
in a push–pull task. It may be that when individuals with an anxious attachment style feel overclose, as may occur
in longstanding relationships, they will endorse items representing a desire for avoidance of closeness. Mikulincer,
M., Shaver, P.R., Bar-On, N., & Ein-Dor, T. (2010) The pushes and pulls of close relationships: attachment insecur-
ities and relational ambivalence. Journal of Personality and Social Psychology, 98(3), 450–68, p.463.
416 Lafontaine, M.F., Brassard, A., Lussier, Y., Valois, P., Shaver, P.R., & Johnson, S.M. (2016) Selecting the best
items for a short-form of the Experiences in Close Relationships questionnaire. European Journal of Psychological
Assessment, 32(2), 140–54.
417 Frıas, M.T., Shaver, P.R., & Mikulincer, M. (2014) Measures of adult attachment and related constructs.
In G.J. Boyle & D.H. Saklofske (eds) Measures of Personality and Social Psychological Constructs. Philadelphia,
PA: Elsevier, pp.417–47, p.446.
and avoidance, such findings raise the prospect that there are aspects of secure attachment
that are not well captured or discriminated by the existing scales, or by theory shaped in
the image of these scales. Already in 2000, Fraley and colleagues had called on researchers
to ‘write items that tap the low ends of the Anxiety and Avoidance dimensions with better
precision’.418 However, two decades later this problem with the ECR remains little discussed.
One of the issues that this neglects is whether security really is merely the inverse of anx-
iety and avoidance, or itself makes an independent contribution to individual differences
in adult attachment style. Given trends in the developmental tradition towards Individual
Participant Meta-analysis (Chapter 6) in order to address psychometric questions, it will be
interesting to see whether the social psychological tradition will also adopt such method-
ologies for further articulating the relationship between the ECR and its latent dimensions.

The opposite of insecurity
In their landmark 1998 paper, Brennan and colleagues reported a two-factor solution as the
fundamental structure of individual differences in adult attachment. In the years after the
introduction of the ECR, the two-factor model quickly became accepted. This marvellously
elegant characterisation of individual differences in adult attachment style subsequently be-
came orthodoxy among social psychological attachment researchers, with theory and meth-
odology looping around one another to direct research questions and interpretations of
findings. Yet leading figures of the social psychological tradition of research acknowledged
that the model is a pragmatic simplification, and that in reality security might represent
more than the absence of avoidance and resistance. Judith Feeney, for instance, writing four
years after the introduction of the ECR, stated:
I would like to take issue with the implicit suggestion that in defining the two attachment
dimensions of avoidance and anxiety, researchers have settled basic questions concerning
the structure of self-reported attachment. There now seems to be considerable consensus
that avoidance and anxiety are the two primary dimensions underlying adult attachment,
and that these dimensions generally provide moderately strong prediction of relationship
outcomes (especially if both partners’ characteristics are taken into account). However, in
reducing such a complex construct as romantic attachment to two dimensions, important
information is inevitably lost.419
This had already been a concern of Brennan and Shaver themselves. In a 1995 paper, Brennan
and Shaver first offered the proposal that ‘secure adults can be characterized as the opposite
of all these insecure tendencies’.420 This was aligned with the ‘love quiz’ which had, to an ex-
tent, characterised security in terms of the absence of discomfort regarding closeness or the
418 Fraley, R.C., Waller, N.G., & Brennan, K.A. (2000) An Item Response Theory analysis of self-report measures
of adult attachment. Journal of Personality and Social Psychology, 78(2), 350–65: ‘Also notice that the two ECR-R
scales, like the original ECR scales, are not adept at assessing individuals with trait levels less than –1.00 on Anxiety
or Avoidance . . . An important next step for future research on scale development is to write items that tap the low
ends of the Anxiety and Avoidance dimensions with better precision’ (361).
419 Feeney, J.A. (2002) Attachment-related dynamics: what can we learn from self-reports of avoidance and anx-
iety? Attachment & Human Development, 4(2), 193–200, p.198.

420 Brennan, K.A. & Shaver, P.R. (1995) Dimensions of adult attachment, affect regulation, and romantic rela-
tionship functioning. Personality and Social Psychology Bulletin, 21(3), 267–83, p.280.
absence of worry about abandonment. Characterisations of security as no more than the

opposite of insecurity would remain the dominant narrative in Shaver’s work in subsequent
years. Yet already in the 1995 paper, Brennan and Shaver offered a qualification: that secure
attachment is not solely the opposite of insecurity, but also has positive characteristics of
its own such as ‘being able and willing to trust romantic partners and share ideas and feel-
ings with them in a flexible, appropriate manner that is sensitive to their partners’ needs and
concerns’.421 This idea was not abandoned after 1998 and the introduction of the ECR. For
instance, in the Shaver and Mikulincer model of the attachment system developed in the
early 2000s (discussed in the section ‘Minimising and maximising’), it would appear that
broaden-and-build cycles are facilitated by specific and vital qualities of security, and not
merely through the absence of their interruption by the insecure attachment styles. For in-
stance, broaden-and-build cycles were anticipated by Shaver and Mikulincer to contribute

to confidence and trust, which was not merely the absence of avoidance or attachment anx-
iety.422 They argued that feeling understood, feeling validated, and feeling cared about are
distinct aspects of the broaden-and-build cycle characteristic of security.423
In 2004, Shaver and Mikulincer reported their impression that by this point ‘most re-
cent adult attachment studies are based on a two-dimensional model’, theoretically and in
terms of measurement.424 They acknowledged that some studies, including their own work
using the ‘love quiz’,425 and research by Rainer Banse426 had produced evidence that security
represented a distinct construct not reducible to the absence of anxiety and avoidance.427
However, they argued that this position was compatible with their theory, since a security-
insecurity dimension represented a 45-degree turn within the two-dimensional space of
anxiety and avoidance.428 So long as the ECR could tap the low end of anxiety and avoidance
effectively, the contribution of security to individual differences could be captured by the
measure. Colleagues such as Fraley had criticised the capacity of the ECR to capture security
421 Ibid.
422 Mikulincer, M. & Shaver, P.R. (2007) Boosting attachment security to promote mental health, prosocial
values, and inter-group tolerance. Psychological Inquiry, 18, 139–56: ‘We conceptualize the sense of attachment se-
curity as an inner resource’ (139). On the contribution of distrust to both anxiety and avoidance, or as a reciprocal
pathway between the two, see McWilliams, L.A. & Fried, E.I. (2019) Reconceptualizing adult attachment relation-
ships: a network perspective. Personal Relationships, 26(1), 21–41.
423 Shaver, P.R. & Mikulincer, M. (2008) Augmenting the sense of security in romantic, leader– follower,
therapeutic, and group relations: a relational model of personality change. In J.P. Forgas & J. Fitness (eds) Social
Relationships: Cognitive, Affective, and Motivational Processes (pp.55–73). New York: Psychology Press.
425 Concurrent with the publication of the ECR was work by Mikulincer suggesting a three-factor solution,
in which security played the smallest but nonetheless material role in predicting variance. A factor analysis con-
ducted by Banai, Weller, and Mikulincer in 1998 found that 13.3% of variance in the classification of an individual’s
attachment style by themselves, their friends, and acquaintances could be accounted for by endorsement of the
secure attachment style on an adaptation of the ‘love quiz’, even taking into account avoidance and ambivalence-
resistance. Banai, E., Weller, A., & Mikulincer, M. (1998) Inter-judge agreement in evaluation of adult attachment
style: the impact of acquaintanceship. British Journal of Social Psychology, 37(1), 95–109, p.104.
426 E.g. Banse, R. (2004) Adult attachment and marital satisfaction: evidence for dyadic configuration effects.
Journal of Social and Personal Relationships, 21(2), 273–82; Asendorpf, J.B., Banse, R., Wilpers, S., & Neyer, F.J.
(1997) Relationship-specific attachment scales for adults and their validation with network and diary procedures.
Diagnostica, 43(4), 289–313.
427 See also Holmes, J.G. & Murray, S.L. (2007) Felt security as a normative resource: evidence for an elemental
risk regulation system? Psychological Inquiry, 18(3), 163–7.

Simpson (eds) Adult Attachment: Theory, Research, and Clinical Implications (pp.17–54). New York: Guilford: ‘This
45-degree rotation of the measurement axes fits well with the process model proposed by Shaver and Mikulincer
(2002)’ (51).
effectively. In 2004, Shaver and Mikulincer appear to have regarded the problem as mean-
ingful, but not of great pragmatic or theoretical significance.
Yet later work from their laboratories has continued to trouble the two-dimensional
model. In 2006, Al-Yagon and Mikulincer published a study of children’s experiences of
loneliness. Since at that time no version of the ECR had been validated for use with children,
they instead used an adaptation of the ‘love quiz’. The researchers found that secure attach-
ment made a negative contribution to loneliness over and above the positive contribution
to loneliness of the avoidant and anxious attachment styles.429 Further evidence for a three-
factor model came from work by Omri Gillath. Gillath had been a graduate student with
Mikulincer between 1998 and 2003, and then a postdoctoral fellow with Shaver from 2003 to
2006 at the University of California. Following his appointment as faculty at the University of
Kansas, Gillath worked with Joshua Hart and colleagues to develop a ‘state adult attachment

measure’ (SAAM) to assess a participant’s current feelings rather than enduring generalised
schemas about relationships (see also the section ‘Conclusion’ for a discussion of security
priming).430 The purpose of the SAAM was to allow researchers to explore the extent to
which particular stimuli or circumstances, for example changes in partner interactions over
a week, might contribute to attachment-relevant experience. The items for the SAAM were
based on the ECR. However, Gillath, Hart, and colleagues ‘also wrote additional items to re-
flect aspects of attachment styles that are underrepresented on current measures, such as the
low end of anxiety (e.g., “I feel relaxed knowing that close others are there for me right now”),
which is represented by only a single item in the ECR’.431 They conducted seven studies with
the SAAM, with a total of 2,327 participants. Across the seven studies, they repeatedly found
that a three-factor model was the best fit for the data, with independent factors for security,
avoidance, and anxiety. They concluded that security appears to be an autonomous dimen-
sion of attachment states.432
Gillath and colleagues emphasised that Shaver and Mikulincer’s theory gives a place to dy-
namics specific to security, even if these are not captured well by the ECR. However, Gillath,
Karantzas, and Fraley have urged that more needs to be done within the social psychological
tradition of attachment research to capture security-specific processes within theory and
methodology. They are unconvinced that these processes are merely the opposite of anxiety
and avoidance, and anticipate that they will have distinct correlates. They argue that greater
429 Al-Yagon, M. & Mikulincer, M. (2006) Children’s appraisal of teacher as a secure base and their socio-
emotional and academic adjustment in middle childhood. Research in Education, 75(1), 1–18.
430 Gillath, O., Hart, J., Noftle, E.E., & Stockdale, G.D. (2009) Development and validation of a state adult attach-
ment measure (SAAM). Journal of Research in Personality, 43(3), 362–73.

431 The security items in the SAAM are: I feel loved; I feel like I have someone to rely on; I feel secure and close
to other people; If something went wrong right now I feel like I could depend on someone; I feel like others care
about me; I feel relaxed knowing that close others are there for me right now; I feel I can trust the people who are
close to me.
432 Parallel findings have also been reported for the ECR-RC, the adaptation of the ECR for children and adoles-
cents to report about their relationship with their parents. Here too security has emerged as an independent latent
factor. This may be an effect of developmental stage. However, like well-established couples and unlike college
students, children and adolescents are also structurally entangled in their attachment relationships, which may
reduce the orthogonality of avoidance and resistance and contribute to the autonomy of broaden-and-build cycles.
Such a conclusion would suggest that Brennan and colleagues’ two-factor solution with orthogonal dimensions
may have been influenced by the disembedded social conditions of college students within American culture,
where neither secure-base effects are fully online, nor anxiety and avoidance are able to become especially tan-
gled. Lionetti, F., Mastrotheodoros, S., & Palladino, B.E. (2018) Experiences in Close Relationships-Revised Child
version (ECR-RC): psychometric evidence in support of a security factor. European Journal of Developmental
Psychology, 15(4), 452–63.
understanding of the secure attachment style will contribute to insights into both clinical
phenomena and experiences of human thriving. For example, it will help clarify whether
traumatic experiences ‘either increase attachment insecurity or wear away at attachment se-
curity’.433 In a review of self-report measures in 2014, Shaver and Mikulincer acknowledged
the problems with the ECR for capturing security, and praised the work of Gillath and col-
leagues. They argued that ‘the inclusion of a separate security subscale may suggest a way out
of the problem identified by Fraley et al. (2000) . . . that most previous attachment insecurity
scales discriminated poorly at their “secure” ends’.434 They noted, however, that when the
low end of avoidance and anxiety are captured, a result is that the two forms of insecurity are
no longer orthogonal, as they are both negatively associated with security: ‘expanding the
scales at their secure ends in similar ways causes the two kinds of security items to correlate
with each other, which in turn makes the scales as wholes correlate more with each other.

Whether or not this leads to weaker detection of distinct effects of anxiety and avoidance
remains unclear.’435
High anxiety/high avoidance
Yet even for the unmodified ECR, questions have been growing about its psychometric prop-
erties. Since Shaver and Mikulincer’s theory of adult attachment has been based on the two di-
mensions of the ECR, this is not simply a minor matter of methodological rigour but a concern
stretching to the very basis for their scientific project. Whereas the original factor analysis con-
ducted by Brennan and colleagues accounted for 63% of variance, subsequent studies have not
accounted for such a high proportion of variance. Factor analytic exploration of the ECR after
Brennan and colleagues has been rare among American researchers, who have generally taken
its psychometric properties for granted. However, in the wider international literature, factor
analytic studies of the ECR or its translation tend to report solutions that account for around
45% of variance.436
Furthermore, there has been growing evidence against the orthogonality of the two ECR
dimensions. In a review of several of their studies in 2005, Mikulincer, Shaver, and colleagues
reported a small association between the scales (r = .18).437 In a paper from the next year,
Condradi and colleagues conducted a further analysis of the published literature, observing
Research. London: Academic Press, p.247. See also Kanninen, K., Punamaki, R.L., & Qouta, S. (2003) Personality
and trauma: adult attachment and posttraumatic distress among former political prisoners. Peace and Conflict,
9(2), 97–126.
434 Frıas, M.T., Shaver, P.R., & Mikulincer, M. (2014) Measures of adult attachment and related constructs.
In G.J. Boyle & D.H. Saklofske (eds) Measures of Personality and Social Psychological Constructs (pp.417–47).
Philadelphia, PA: Elsevier, p.443.
435 Ibid. p.446.
436 The ECR-R has fared even less well on this front than the ECR. See e.g. Rotaru, T.Ş. & Rusu, A. (2013)
Psychometric properties of the Romanian version of Experiences in Close Relationships-Revised questionnaire

(ECR-R). Procedia-Social and Behavioral Sciences, 78, 51–5; Busonera, A., Martini, P.S., Zavattini, G.C., & Santona,
A. (2014) Psychometric properties of an Italian version of the Experiences in Close Relationships-Revised (ECR-
R) Scale. Psychological Reports, 114(3), 785–801.
437 Mikulincer, M., Shaver, P.R., Gillath, O., & Nitzberg, R.A. (2005) Attachment, caregiving, and al-
truism: boosting attachment security increases compassion and helping. Journal of Personality and Social
Psychology, 89(5), 817–39, p.821.
that ‘intercorrelations vary considerably from .04 to .30’.438 The developmental and re-
lationship tasks of students as a population may well have some differences from popula-
tions in other countries and at different lifestages.439 Condradi and colleagues observed that
these differences are evidently not sufficient to block the successful application of the ECR
to diverse research populations, as years of research has demonstrated. However, the char-
acteristics of the population used to develop the sample may have nonetheless introduced
unrecognised assumptions into the measure, denting the reliability and validity of its scales.
Specifically, Condradi and colleagues were concerned by Brennan and colleagues’ conclu-
sion that attachment anxiety and attachment avoidance were, and should be, orthogonal
dimensions. This seemed to be a holdover from Bartholomew’s model of orthogonal dimen-
sions, rather than based on a cogent theoretical justification. They expressed concern that
there was no basis in Bowlby’s theory for assuming that the two dimensions would be unre-

lated. Indeed, researchers in the developmental tradition have neither expected nor found
orthogonality between avoidance and resistance.440
Condradi and colleagues sought to empirically examine whether orthogonality might be
an effect of the student sample used by Brennan and colleagues. A two-factor model of or-
thogonal dimensions was indeed the best fit for the American and Dutch student samples.441
However, the result was less clear for the Dutch sample of adults. Condradi and colleagues
argued that anxiety and avoidance may indeed be unrelated for participants without experi-
ence in lasting relationships. In more mature relationships they predicted that anxiety and
avoidance would no longer be unrelated. Both dimensions would be negatively associated
with the security that a long-established and healthy relationship can build. And both di-
mensions would be positively associated with the anxiety and avoidance that can become
entwined within a long-established and unhealthy relationship.442
These proposals were supported by two meta-analyses conducted by Cameron and col-
leagues and by Graham and Unterschute.443 The meta-analyses found that orthogonality
between anxiety and avoidance was the exception rather than the rule in the published litera-
ture. Furthermore, older samples had larger correlations between anxiety and avoidance than
younger samples, long-term relationships had larger correlations than newly established
ones, non-student samples had larger correlations than college samples, and non-American
438 Conradi, H.- J., Gerlsma, C., van Duijn, M., & de Jonge, P. (2006) Internal and external validity of the
Experiences in Close Relationships questionnaire in an American and two Dutch samples. European Journal of
Psychiatry, 20(4), 258–69, p.268.
439 The principle is one that Shaver acknowledged in his early work, if less in his writings on attachment: Felton,
B.F. & Shaver, P.R. (1984) Cohort variation in adults’ reported feelings. In C.Z. Malatesta & C.E. Izard (eds)
Emotions in Adult Development (pp.103–123). Beverly Hills: Sage Publications: ‘Adults’ choices of coping strategies
are linked to the nature of the problems they face’ (118–19).
440 See e.g. Kroonenberg, P.M., Dam, M.V., van IJzendoorn, M., & Mooijaart, A. (1997) Dynamics of behaviour
in the strange situation: a structural equation approach. British Journal of Psychology, 88(2), 311–32.
441 Research findings from Shaver’s American and Mikulincer’s Israeli undergraduate samples have generally
aligned well in terms of the correlates of attachment styles. Cultural differences have been reported at times, but
not ones that interact with the ECR. See e.g. Mikulincer, M., Shaver, P.R., Gillath, O., & Nitzberg, R.A. (2005)
Attachment, caregiving, and altruism: boosting attachment security increases compassion and helping. Journal of
Personality and Social Psychology, 89(5), 817–39, p.836.
442 Conradi, H.-J., Gerlsma, C., van Duijn, M., & de Jonge, P. (2006) Internal and external validity of the experi-
ences in close relationships questionnaire in an American and two Dutch samples. European Journal of Psychiatry,
20(4), 258–69.
443 Cameron, J.J., Finnegan, H., & Morry, M.M. (2012) Orthogonal dreams in an oblique world: a meta-analysis
of the association between attachment anxiety and avoidance. Journal of Research in Personality, 46(5), 472–6;
Graham, J.M. & Unterschute, M.S. (2015) A reliability generalization meta-analysis of self-report measures of
adult attachment. Journal of Personality Assessment, 97(1), 31–41.
samples had larger correlations than American samples. Cameron and colleagues argued
that ‘such findings call into question the largely implicit assumption that attachment dimen-
sions are orthogonal. As such, our results may inspire researchers to revisit theoretical as-
sumptions, measurement choices, and measure creation techniques. If researchers do not
wish to seek other measurement options, they should at the very least adapt their statistical
analyses to accommodate shared variance between dimensions . . . One method of accom-
modating shared variance is to include both dimensions as predictors in the same step in a
regression, and thus control for shared variance.’444 The proposal to accommodate or explore
shared variance has not, as yet, been pursued. As a result, it remains unknown whether the
association between the anxiety and avoidance scales in the ECR lies on the basis of their
joint negative association with security or on the basis of their joint positive association with
a state of high anxiety/high avoidance.

In recent work with Birnbaum, Mikuliner and Shaver acknowledged both possibilities.445
They also offered a thought-provoking speculation that the tendency towards orthogonality
of the ECR dimensions might actually be an effect of two processes, invisible in the data
because they have suppressed one another. On the one hand, anxiety and avoidance may
function as exact opposites: Shaver and colleagues have at times argued that each is the in-
verse of the other in terms of psychological processes.446 To the extent that they represent
maximisation and minimisation of the attachment system, there can be anticipated to be a
negative association between the two dimensions. However, this negative association may be
counteracted by a positive association between the scales at their endpoints. They may have
a positive association in their mutual opposition to security. Furthermore, ‘both hyperacti-
vation and deactivation . . . represent problems in the system’s functioning, which may push
their correlation in a positive direction’.447 The result of both positive and negative associ-
ations may be an apparent tendency towards orthogonality.
Whereas Bartholomew gave fearful attachment a major place in her system, the con-
junction of high anxiety and high avoidance has not been a major concern of Shaver and
Mikulincer over their careers. In part, this is likely due to the fact that there are few high anx-
iety/high avoidance participants in Shaver and Mikulincer’s samples. One essential reason
for this, as they have acknowledged, has been their sampling: ‘The issue of “fearful avoid-
ance” is, in any case, less likely to arise in normal samples of college students and commu-
nity adults. Extremely high scores on both the anxiety and avoidance dimensions are more
common in samples of abused or clinical samples . . . In most of our studies, the results can be
444 Cameron, J.J., Finnegan, H., & Morry, M.M. (2012) Orthogonal dreams in an oblique world: a meta-
analysis of the association between attachment anxiety and avoidance. Journal of Research in Personality, 46(5),
472–6, p.475.
Journal of Personality and Social Psychology, 106(5), 822–42, p.826.
446 E.g. Fraley, R.C., Davis, K.E., & Shaver, P.R. (1998) Dismissing avoidance and the defensive organization of
emotion, cognition, and behavior. In J.A. Simpson & W.S. Rholes (eds) Attachment Theory and Close Relationships
(pp.249–79). New York: Guilford: ‘It should be noted that the processes and mechanisms we discuss with respect to
dismissing-avoidance apply to preoccupation inversely’ (275).
Journal of Personality and Social Psychology, 106(5), 822–42, p.826. An example here may be the role of distrust
in contributing to both avoidance and anxiety, or their reciprocal reinforcement. See McWilliams, L.A. & Fried,
E.I. (2019) Reconceptualizing adult attachment relationships: a network perspective. Personal Relationships,
26(1), 21–41.
adequately described in terms of either anxiety or avoidance ... with the distinction between
dismissing and fearful avoidance mattering only . . . when abuse or psychopathology are at
issue.’448 For instance, they cited as evidence a published study by Fraley and Bonanno and
unpublished data from Colin Murray Parkes showing that ‘combinations of attachment anx-
iety and avoidance produced the highest levels of anxiety, depression, grief, trauma-related
symptoms, and alcohol consumption’.449
Shaver and Mikulincer’s overriding use of student samples has been a pragmatic one.
They have readily admitted that it supplies a limitation to their work. For instance, they
have noticed that it can be difficult to recruit student participants with an avoidant attach-
ment style when adverts are frank with participants that the study is concerned with close
relationships.450 And Makariev and Shaver have acknowledged that middle-class college
populations were not ‘the kinds of people Bowlby had in mind when he began to develop

attachment theory’.451 Nonetheless, Shaver and Mikulincer have defended their use of stu-
dent samples on two grounds. In a first defence, they have argued that adversities and psy-
chopathology is dimensionally distributed, and so there is no particular problem with using
a student sample—so long as clinical variance is captured.452 This defence is relatively weak.
It is not clear that clinical variance is generally well captured by the measures Shaver and
Mikulincer tend to use; in this context, it is perhaps unsurprising that studies of adult at-
tachment styles and PTSD have tended not to use the ECR.453 Furthermore, the idea that
adversities and psychopathology can be treated as dimentional in terms of their implications
for attachment style is simply asserted, rather than adequately empirically demonstrated in
their work.
Shaver and Mikulincer have also offered a second defence of their approach in describing
themselves as pragmatists, seeing to achieve cogent simplifications whilst remaining aware
of what has been simplified:
In our work, we always try to keep in mind that there is, on the one hand, complex
everyday reality as we experience it subjectively and encounter it in the behavior of other
people. And, on the other hand, there is psychological theory, with its associated hypo-
thetical constructs, and an ever-evolving toolbox of psychological measures. The trick is
to discover and document something important and valid about real life, thereby nudging
(pp.53–152). New York: Academic Press, pp.70, 88.
449 Shaver, P.R. & Mikulincer, M. (2008) Adult attachment and cognitive and affective reactions to positive and
negative events. Social and Personality Compass, 2, 1844–65, p.1853. Citing Fraley, R. C. & Bonanno, G. A. (2004)
Attachment and loss: a test of three competing models on the association between attachment-related avoidance
and adaptation to bereavement. Personality and Social Psychology Bulletin, 30(7), 878–90.
450 Gillath, O., Bunge, S.A., Shaver, P.R., Wendelken, C., & Mikulincer, M. (2005) Attachment-style differences
in the ability to suppress negative thoughts: exploring the neural correlates. Neuroimage, 28(4), 835–47, p.945.
451 Makariev, D.W. & Shaver, P.R. (2010) Attachment, parental incarceration and possibilities for interven-
tion: an overview, Attachment & Human Development, 12(4), 311–31, p.325.

452 Brennan, K.A. & Shaver, P.R. (1998) Attachment styles and personality disorders: their connections to each
other and to parental divorce, parental death, and perceptions of parental caregiving. Journal of Personality, 66(5),
835–78: ‘One obvious limitation of our research is its use of a nonclinical sample. To some extent, this limitation
turns on whether one accepts a purely categorical (vs. dimensional) understanding of personality disorders. If, as
we do, one assumes that personality disorders can be arrayed on continua, then our results ought to generalize to
clinical populations’ (870).
453 Woodhouse, S., Ayers, S., & Field, A.P. (2015) The relationship between adult attachment style and post-
traumatic stress symptoms: a meta-analysis. Journal of Anxiety Disorders, 35, 103–117.

psychological science forward, without mistaking our tentative, overly simplified picture
for everything that is actually there.454
However, the inattention to high anxiety/high avoidance attachment styles is more than a
happenstance of sampling. Instead, the sampling strategy appears to have reflected a broader
lack of concern with adverse and clinical experiences, and states where both anxiety and
avoidance are in play.
Even before the introduction of the ECR, Shaver and Mikulincer demonstrated a marked
lack of interest in what it meant when participants endorsed both an anxious and avoidant
attachment style. In his work in the early 1990s, Mikulincer reformulated the ‘love quiz’ as
three scales to produce dimensional ratings of attachment style.455 In a validation study of
the scales with 127 undergraduates, nine participants scored differently on the scales com-

pared to the original category-based ‘love quiz’. Mikulincer dropped these participants
from the analysis. This approach set a precedent. Cooper, Shaver, and Collins used both the
category-based and the dimension-based measures in their 1998 paper ‘Attachment styles,
emotion regulation, and adjustment in adolescence’.456 In the study, 411 participants—20%
of the sample—were inconsistent across the two measures. In a footnote, the authors ac-
knowledged that this inconsistent group had some interesting properties. They were more
likely to be non-White, more likely to endorse an insecure attachment style, and more likely
to have been held back in school. Furthermore, ‘inconsistent responders reported higher
levels of phobic anxiety, paranoia, and psychoticism and were less satisfied with their body
image’.457 Though this ‘inconsistent’ group represented hundreds of participants, they were
thrown out of the analysis.458
Considering this study alone, the decision makes methodological sense. However, in
the broader context of Shaver’s work, it illustrates a trend to exclude the effects of adversity,
stigma, pharmacology, and clinical complexity from the study of adult attachment styles.
The most influential instance of this trend was in work with Brennan and Clark in devel-
oping the ECR. In their 1998 paper, the factor analysis was conducted on an undergraduate
sample from the University of Texas. However, no report was made or analysis conducted for
demographic or ethnic diversity within the sample. And no data were collected on clinical
differences. Across their published studies, Shaver and Mikulincer have rarely discussed the
implications of both high anxiety and avoidance, even in studies where they included meas-
ures of mental health.459 There seems to have been a self-perpetuating cycle between lack of
interest and lack of scientific basis for interest in high anxiety/high avoidance phenomena in
existing theory and research. This has shaped the tone and priorities of social psychological
454 Shaver, P. & Mikulincer, M. (2004) Attachment in the later years: a commentary. Attachment & Human
Development, 6(4), 451–64, p.462. Other laboratories have confirmed that high anxiety/high avoidance attach-
ment is more prevalent in psychiatric samples, e.g. Alessandri, G., Fagnani, C., Di Gennaro, G., et al. (2014)
Measurement invariance of the experiences in close relationships questionnaire across different populations.
Spanish Journal of Psychology, 17(2), E22.
455 Mikulincer, M. & Nachshon, O. (1991) Attachment styles and patterns of self- disclosure. Journal of
Personality and Social Psychology, 61(2), 321–31.
456 Cooper, M.L., Shaver, P.R., & Collins, N.L. (1998) Attachment styles, emotion regulation, and adjustment in
adolescence. Journal of Personality and Social Psychology, 74(5), 1380–97.

457 Ibid. p.1385.
458 Ibid.: ‘We adopted the procedure used by Mikulincer and Nachshon and excluded inconsistent respondents
from further analyses’ (1386).

459 E.g. Crawford, T.N., Livesley, W.J., Jang, K.L., Shaver, P.R., Cohen, P., & Ganiban, J. (2007) Insecure attach-
ment and personality disorder: a twin study of adults. European Journal of Personality, 21(2), 191–208.
research on attachment, hindering dialogue with the developmental tradition where theo-
rising and measuring the implications of trauma and fear for attachment have been a major
focus (Chapter 3).
Despite the general trend, there have been a few occasions on which Shaver and
Mikulincer have reported associations for high anxiety/high avoidance attachment.460 One
was a study by Schachner and Shaver of sexual motives, which found that high anxiety/high
avoidance—over and above the anxious attachment style alone—was associated with greater
report of having sex due to insecurity.461 Hart, Shaver, and Goldenberg found that high anx-
iety/high avoidance was distinctively associated with a desire for closeness after receiving in-
formation that was anticipated to threaten the participants’ cultural worldview, but not in the
control condition. This was in contrast to the anxious attachment style which was associated
with a desire for closeness in both conditions.462 The researchers concluded that the avoid-

ant strategy held high anxiety/high avoidance participants back from seeking support in the
control condition, but that the strategy was overwhelmed by the combination of high attach-
ment anxiety and a worldview threat, leading to behaviour shaped by the anxious strategy.463
The effects of trauma
A major qualification must be offered, however, to any allegation that inattention to attach-
ment phenomena characterised by high anxiety and high avoidance is a limitation in the
work of Shaver and Mikulincer.464 The qualification lies in Mikulincer’s two decades of re-
search on the effects of war and military trauma. However, this research has not always been
well linked-up with Mikulincer’s theoretical work on adult attachment. The relationship
between PTSD and the insecure attachment styles in Mikulincer’s work has remained un-
tamed, especially in terms of what this relationship means for the conceptualisation of high
anxiety/high avoidance states and of adult attachment in general. This issue appears to be
coming to a head in recent years, given the very sharp increase in studies of PTSD and at-
tachment styles in the past decade. However, refinements in theory and methodology appear
generally not to have caught up with this empirical concern.
460 Shaver and Mikulincer have at times reported relevant findings from other research groups, though this
has been exceptionally rare. E.g. Mikulincer, M. & Shaver, P.R. (2005) Attachment theory and emotion in close
relationships: exploring the attachment-related dynamics of emotional reactions to relational events. Personal
Relationships, 12, 149–68, p.160.
461 Schachner, D.A. & Shaver, P.R. (2004) Attachment dimensions and sexual motives. Personal Relationships,
11(2), 179–95, p.192.

462 Hart, J.J., Shaver, P.R., & Goldenberg, J.L. (2005) Attachment, self-esteem, worldviews, and terror manage-
ment: evidence for a tripartite security system. Journal of Personality and Social Psychology, 88, 999–1013, p.1008.
463 For another example of specific correlates of high anxiety/ high avoidance attachment see Gillath, O.,
Giesbrecht, B., & Shaver, P.R. (2009) Attachment, attention, and cognitive control: attachment style and perform-
ance on general attention tasks. Journal of Experimental Social Psychology, 45(4), 647–54, p.651.
464 It might be thought that an additional qualification could be offered on the basis of the major efforts by
Shaver and Mikulincer to review empirical studies of psychopathology, e.g. Mikulincer, M. & Shaver, P. (2016)
Attachment in Adulthood, 2nd edn. New York: Guilford, Chapter 13. However, these reviews are lacking precisely
in consideration of high anxiety/high avoidance attachment styles except, on a handful of occasions, when re-
porting findings from Bartholomew’s measures. The question of whether studies using the ECR have found effects
for high anxiety/high avoidance, and more generally the question of whether the high ends of the scales are indi-
vidually or in interaction effective at capturing clinical phenomena, is not in view. The focus is instead on anxiety
and avoidance as, separately, dimensionally associated with forms of mental pathology. Throughout the reviews, a
linear and dimensional association between the ECR scales and psychopathology is assumed rather than demon-
strated or explored.
In the late 1980s, Solomon, Mikulincer, and colleagues examined Israeli army veterans
who had experienced symptoms of post-traumatic combat stress during the 1982 Lebanon
war. On the one hand, the researchers observed that one of the symptoms associated with
this condition was ‘psychic numbing’, which was associated with low emotional expressive-
ness in the veterans’ adult relationships with family members. The researchers accepted that
‘there is little way of knowing whether it is the family that did not permit the veteran to
express himself or the veteran who did not make use of the avenues of expression open to
him’.465 However, they expressed their suspicion that ‘many PTSD casualties who report a
low level of expressiveness may themselves have avoided discussion’.466 On the other hand,
Mikulincer, Solomon, and colleagues also found that, among veterans who had experienced
symptoms of post-traumatic combat stress, ‘the anxiety feelings aroused during battle by
the fear of death are crystalized one year later in anxiety that is not specifically related to

death, but is generalized to every potential threat’.467 This would, presumably, include rela-
tional threats. These early observations suggest, though this point was not drawn out by the
researchers, that general forms of avoidance and anxiety stemming from trauma may feed
more specific forms of avoidance and anxiety within adult close relationships.
Mikulincer, Solomon, and Benbenishty also conducted clinical interviews with a sample
of 104 veterans with histories of post-traumatic combat stress symptoms a year after the end
of the Lebanon war. From these interviews, they identified 26 different manifestations of
post-traumatic combat stress and nine battle events that seemed relevant to the emergence or
maintenance of symptoms. Analysis of the manifestations resulted in six factors, accounting
for 62% of variance. The six factors were: psychic numbing, anxiety reactions, guilt, loneli-
ness, loss of bodily control (such as uncontrollable crying or vomiting), and disorientation.
It is interesting, if perhaps not surprising, how well these agree with Bowlby’s own clinical
observations of war veterans during his time as an army psychiatrist (Chapter 1). In the work
of Mikulincer and colleagues, psychic numbing and anxiety reactions appeared to be the
most important, between them accounting for 31% of variance in symptoms reported in the
clinical interview. They explicitly argued that psychic numbing should be conceptualised as
a kind of ‘avoidance’ and considered as quite an all-purpose response, since it was not asso-
ciated with any particular kind of battle event. By contrast, anxiety reactions were predicted
best by poor unit functioning. Mikulicer and colleagues argued that psychic numbing and
anxiety reactions seemed to correspond well, respectively, to ‘avoidance and numbing’ and
‘hyperarousal’ as two of the main clusters of general PTSD symptoms. They therefore sug-
gested that post-traumatic combat stress symptoms should not be regarded solely as local-
ised responses to particular war experiences, but as reflecting the basic forms of the human
trauma response.468
Pursuing these questions further, in the mid-1990s Mukulincer and colleagues asked
40 Israeli Jewish settlers living within Palestinian Authority territory in the Gaza Strip to
465 Solomon, Z., Mikulincer, M., Freid, B., & Wosner, Y. (1987) Family characteristics and posttraumatic stress
disorder: a follow-up of Israeli combat stress reaction casualties. Family Process, 26(3), 383–94, p.390.
466 Ibid.
467 Mikulincer, M., Solomon, Z., & Benbenishty, R. (1988) Battle events, acute combat stress reaction and long-
term psychological sequelae of war. Journal of Anxiety Disorders, 2(2), 121–33, p.131.
468 Solomon, Z., Mikulincer, M., & Benbenishty, R. (1989) Combat stress reaction—clinical manifestations and
correlates. Military Psychology, 1(1), 35–47, p.44. Later work by the researchers documented powerful longitudinal
effects of these traumatic symptoms. See Solomon, Z. & Mikulincer, M. (2007) Post traumatic intrusion, avoid-
ance, and social functioning: a 20-year longitudinal study. Journal of Consulting and Clinical Psychology, 75(2),
316–24.
fill out the ‘love quiz’ and measures of mental distress and PTSD. The settlers were physic-
ally cut-off from the rest of Israel. The settlement had been repeatedly attacked, including
a suicide-bombing in 1994 which killed six Jewish members of the settlement and six
Palestinians.469 The settlement was also at the time facing the threat of eviction by the
Israeli government (indeed, this threat would later be realised in August 2005, when the
government sent in the army to forcibly remove the settlers). Though the identities and
lives of the settlers would be potentially quite different from the combat veterans he had
studied, Mukulincer nonetheless hoped that he could treat this group as representing a
‘high threat’ condition. The settler participants were compared to 40 matched controls
living in villages of equivalent size but within the borders of Israel. They also examined the
correlates of different symptom clusters of PTSD: intrusive symptoms (e.g. flashbacks),
avoidant symptoms (e.g. emotional numbing), and hyperarousal (e.g. sudden irritability

or startle responses).470 Mukulincer and colleagues found that overall attachment styles
were a much better predictor of mental health than whether participants were settlers or
in the control group, accounting for five to six times more variance in general distress and
mental health and PSTD-specific symptoms.471
However, the threat condition was a powerful moderator of attachment style. The anxious-
ambivalent attachment style was associated with general mental health symptoms and
avoidant, hyperarousal, and intrusive PTSD symptoms for both the settlers and the matched
controls; this association was not stronger for the settlers. By contrast, the avoidant attach-
ment style was associated with general distress and mental health symptoms only among the
settler group. It was also associated with PTSD symptoms from the avoidance cluster. An as-
sociation with PTSD symptoms from the hyperarousal/intrusion cluster was marked but fell
just short of significance, which was interpreted as an effect of the small sample. Mikulincer
and colleagues interpreted their findings as suggesting that the high threat condition was
putting strain on the avoidant attachment strategy, and that this strain was most clearly seen
in the form of PTSD symptoms that reflected this attachment strategy.472 However, there
were indications that the high threat condition additionally made a contribution to distress
and mental illness among participants with an avoidant attachment style.
469 See also Haberman, C. (1994) Palestinians arrest 100 Islamic militants after bicycle bombing. New York
Times, 13 November 1994.

470 The distinction between these symptom clusters had been introduced in DSM-III-R. Brett, E., Spitzer, R., &
Williams, J.B. (1988) DSM-III-R criteria for posttraumatic stress disorder. American Journal of Psychiatry, 145(10),
1232. The distinction was embedded in psychological tools such as the Post-traumatic Stress Disorder Inventory,
developed by Solomon, Weisenberg, Schwarzwald, and Mikulincer in the late 1980s, and used in the 1999 study of
the Jewish settlers.
471 Mikulincer, M., Horesh, N., Eilati, I., & Kotler, M. (1999) The association between adult attachment style and
mental health in extreme life-endangering conditions. Personality and Individual Differences, 27(5), 831–42, p.837.
472 Ibid.: ‘Interestingly, this distress was mainly manifested in avoidance rather than intrusive responses,
which seem to be avoidant persons’ habitual affect regulation strategy’ (839). The proposal that PTSD avoid-
ance and hyperarousal symptoms parallel the Ainsworth avoidant and ambivalent/resistant classifications
had first been made a few years earlier by Crittenden, P.M. (1997) Toward an integrative theory of trauma: a
dynamic–maturational approach. In D. Cicchetti & S. Toth (eds) The Rochester Symposium on Developmental
Psychopathology, Vol. 10. Risk, Trauma, and Mental Processes (pp.34–84). New York: University of Rochester Press.
In a later study, Besser and colleagues used the ECR with civilians exposed to terrorist attacks in southern Israel.
Unlike Mikulincer, they found few or weak associations from attachment avoidance. They found stronger associ-
ations from attachment anxiety with all three clusters of PTSD symptoms (intrusive, avoidance, and hyperarousal).
The reasons for these different results may stem from the different sample. It may be that an avoidant attachment
style was still a viable strategy for participants subject to long-term missile threat, in contrast to the settlers studied
by Mikulincer. Besser, A., Neria, Y., & Haynes, M. (2009) Adult attachment, perceived stress, and PTSD among ci-
vilians exposed to ongoing terrorist attacks in southern Israel. Personality and Individual Differences, 47(8), 851–7.
The avoidant PTSD symptoms reported by the settlers could be interpreted as a conse-
quence of an avoidant strategy under severe strain.473 It might also be that as the avoidant
strategy became overwhelmed and unviable, individuals would experience other forms of
distress and anxiety. Conversely, Shaver and Mikulincer considered that the anxious attach-
ment strategy might become unviable in some situations and start to fail. The comparison
of the settlers and matched controls did not reveal differences in the anxious attachment
strategy or its correlates. However, Shaver and Mikulincer acknowledged that, in other
conditions, it might be ‘possible to speak about the failure of hyperactivating strategies’.474
Indeed, Mikulincer and colleagues’ work with the combat veterans from the Lebanon war
suggested that war trauma could perhaps contribute to forms of PTSD resembling or ex-
pressing both deactivating and hyperactivating strategies.475
In the early 2000s, Mukulincer and Shaver identified the need for further attention to the

diverse conditions that might promote the activation of the avoidant attachment style in dif-
ferent forms.476 They proposed that these conditions might include (i) consistent rejection
from attachment figures; (ii) threats of punishment by attachment figures for the display of
attachment behaviours; (iii) traumatic or abusive experiences in the context of the desire
for comfort; and (iv) contexts that encourage self-reliance. The conditions that would pro-
mote an anxious attachment style might include: (i) care unrelated to signals about need; (ii)
care that punishes or prevents the development of self-regulation skills or autonomy; (iii)
messages from the attachment figure that emphasise the individual’s helplessness; and (iv)
traumatic or abusive experiences in the context of separation from attachment figures. So,
for instance, trauma in both the context of the wish for comfort and the context of separ-
ations could be expected to promote both anxious and avoidant strategies. The most potent
conditions shaping the selection of attachment strategy, Shaver and Mikulincer suggested,
473 Mikulincer, M., Dolev, T., & Shaver, P.R. (2004) Attachment- related strategies during thought suppres-
sion: ironic rebounds and vulnerable self-representations. Journal of Personality and Social Psychology, 87(6), 940–
56: ‘Deactivating strategies can be inadequate and overwhelmed, resulting in a marked decline in functioning
and what Horowitz (1982) called “avoidance-related” posttraumatic symptoms (e.g., psychic numbing, behavioral
inhibition)’ (941).
474 Shaver, P.R. & Mikulincer, M. (2002) Dialogue on adult attachment: diversity and integration. Attachment
& Human Development, 4(2), 243–57, p.247. Shaver and Mikulincer did not flesh out the conditions under which
failure of hyperactivating strategies might be expected. It is also not fully clear whether failure would entail the
flooding of anxiety in some internal sense, or the breakdown of the capacity to strive for closeness with attach-
ment figures, or both. The former appears to be described in a characterisation of the hyperactivating strategy
early in Shaver and Mikulincer’s work together: Mikulincer, M. & Shaver, P.R. (2003) The attachment behavioral
system in adulthood: activation, psychodynamics, and interpersonal processes. In M.P. Zanna (ed.) Advances in
Experimental Social Psychology, Vol. 35 (pp.53–152). New York: Academic Press: ‘Intrapsychically, amplification
of threat appraisals heightens the chronic accessibility of negative thoughts and makes it likely that new sources
of distress will mingle and become confounded with old accessible ones . . . the person experiences an endless and
uncontrollable flow of negative thoughts and moods, which in turn may lead to cognitive disorganization and, in
certain cases, culminate in psychopathology’ (82–3).
475 These concerns may also be placed in other discussions at the time in social psychology regarding the poten-
tial for different forms of personality disorder to reflect extreme forms of attachment system activation or deacti-
vation, especially in the context of loss, trauma, or chronic adversities. See Bartholomew, K., Kwong, M.J., & Hart,
S.D. (2001) Attachment. In W.J. Livesley (ed.) Handbook of Personality Disorders: Theory, Research, and Treatment
476 E.g. Mikulincer, M., Gillath, O., & Shaver, P.R. (2002) Activation of the attachment system in adult-
hood: threat-related primes increase the accessibility of mental representations of attachment figures. Journal of
Personality and Social Psychology, 83(4), 881–95: ‘It would be interesting to explore further how avoidant persons’
inhibitory processes work, what they are designed to accomplish (e.g., protection from a potentially angry, puni-
tive attachment figure; reduction of the attachment figure’s tendency to threaten abandonment or decrease support
if a particular separation is resisted), and when they arise—either in the course of development or in the course of
a particular long-term relationship’ (893–4).
would be those that related directly to the use of the caregiver as secure base or safe haven.477
However, as discussed in the section ‘Minimising and maximising’, over time, Shaver and
Mikulincer came to explicitly acknowledge that other behavioural systems could feed into
and alter schemas about relationships. This would, presumably, include the fear behavioural
system: threat experiences are anticipated to inform Module 1 in Shaver and Mikulincer’s
model of the attachment system. The researchers argued that, overall, more needed to be
done to understand the conditions that would allow the avoidant and anxious attachment
strategies to mitigate, intensify, or express mental illness.478
Yet in the 2000s, the lines of causality remained tangled. It could be that traumatic ex-
periences were relevant to the selection and intensification of maximising and minimising
strategies, including in the domain of attachment. This was suggested by Mikulincer’s ori-
ginal work with combat veterans from the Lebanon war. However, it could be that attach-

ment strategies from early life were predisposing and shaping traumatic experiences. This
was suggested by Mikulincer’s study of the Israeli settlers, and would be congruent with
Shaver and Mikulincer’s emphasis in the early 2000s on attachment styles as shaped pri-
marily by childhood experiences and then predisposing later mental illness. Of course, it was
also possible that both processes were taking place simultaneously and reciprocally. To pick
apart the temporal relationship between attachment styles and trauma, Solomon, Dekel, and
Mikulincer reported findings in 2008 from a longitudinal study of Israeli former prisoners of
war from the 1973 Yom Kippur War.479 Participants were followed up in 1991 and 2003, and
asked at both times to complete the scaled version of the ‘love quiz’ and a self-report measure
of PTSD symptoms. The researchers found that both attachment anxiety and attachment
avoidance increased over time among the former prisoners of war, whereas they remained
at least stable among the matched control veterans. Both anxious and avoidant attachment
styles were positively associated with PTSD symptoms for the former prisoners of war and
the matched controls. PTSD avoidance was a potent predictor of later attachment avoidance
(r = .68), but it was also predicted by PTSD hyperarousal (r = .59). PTSD avoidance and
hyperarousal both also predicted later attachment anxiety (r = .58, .50).480
The longitudinal nature of the study allowed Solomon and colleagues to analyse the re-
spective contribution of these factors to one another over time. Contrary to expectations,
the results showed that early PTSD symptoms predicted later attachment styles much better
than early attachment styles predicted later PTSD symptoms. The researchers accepted that
‘this finding cannot be easily explained by adult attachment theory (Mikulincer & Shaver).
According to this theory, attachment insecurities are a risk factor for the emergence and
increase of PTSD symptoms and not the reverse. As a result, this finding is a major nov-
elty of the current prospective study. It seems that traumatic events and post-traumatic re-
sponses cause changes in people’s resources and resiliency and then deteriorate their sense
477 Mikulincer, M., Shaver, P.R., & Pereg, D. (2003) Attachment theory and affect regulation: the dynamics, de-
velopment, and cognitive consequences of attachment-related strategies. Motivation and Emotion, 27, 77–102,
pp.97–8.
478 Ibid.: ‘Researchers should examine the conditions under which secondary attachment strategies seem to
work sufficiently well to avoid severe psychopathology. We still do not know why some insecurely attached indi-
viduals function within the normal range whereas others require clinical intervention’ (100).
479 Solomon, Z., Dekel, R., & Mikulincer, M. (2008) Complex trauma of war captivity: a prospective study of at-
tachment and post-traumatic stress disorder. Psychological Medicine 38(10), 1427–34.

480 Ein-Dor, T., Doron, G., Solomon, Z., Mikulincer, M., & Shaver, P.R. (2010) Together in pain: attachment-
related dyadic processes and posttraumatic stress disorder. Journal of Counseling Psychology, 57(3), 317–27, p.321,
Table 1.
Conclusion 533
of attachment security.’481 By the late 2000s, Shaver and Mikulincer had come to acknow-
ledge that attachment style would reflect not only early experiences with attachment fig-
ures, but also information relevant to close relationships from other behavioural systems
and from adolescent and adult experiences. Nonetheless, Mikulincer was still taken aback
by the fact that PTSD symptoms seemed to be better predictors of attachment style in the
sample than vice versa. It also remained unclear whether the additional attachment anxiety
and avoidance predicted by the PTSD was the same kind of anxiety and avoidance as before
the trauma. It could be that the attachment measure was actually picking up PTSD avoidance
and hyperarousal as it figured within the attachment relationship.
Conclusion

Writing in the New York Times in 1974, Bowlby acknowledged that it might often be per-
plexing why psychological researchers fell into agreement or disagreement when os-
tensibly discussing the same phenomena, and even sometimes the same theory or data.
Acknowledging that there could be a variety of factors involved, he underlined one as cen-
tral. When psychological knowledge is in dissensus, ultimately ‘we differ, I believe, on how
we picture the raw human nature than emerges into the world when a baby is born’.482 The
difference between the developmental and social psychological traditions must be recog-
nised as only partially a divergence in method. Differences in method have at times reflected
and at times stabilised important differences in how the two traditions conceptualise attach-
ment and its role within the human condition. As Chapter 3 described, Mary Main’s account
of conditional strategies was not merely a lens on the Strange Situation, but a model of the
three basic ways that humans can direct attention in responding to distress and the activa-
tion of the attachment system. It was this general model that led to the development of the
AAI coding system, since infant behaviour and adult autobiographical discourse could be re-
garded as forms affected by the same processes. In turn, Main’s methodological innovations
and their predictive value have helped stabilise the general theoretical idea of minimising
and maximising strategies as the best way to capture individual differences in the Strange
Situation and AAI. In a similar manner, the methodological decisions of the social psycho-
logical tradition have had a bidirectional relationship with their model of human nature.483
481 Solomon, Z., Dekel, R., & Mikulincer, M. (2008) Complex trauma of war captivity: a prospective study of at-
tachment and post-traumatic stress disorder. Psychological Medicine 38(10), 1427–34, p.1431. See also Mikulincer,
M., Ein-Dor, T., Solomon, Z., & Shaver, P.R. (2011) Trajectories of attachment insecurities over a 17-year period: a
latent growth curve analysis of the impact of war captivity and posttraumatic stress disorder. Journal of Social and
Clinical Psychology, 30(9), 960–84. A related finding was reported by Ghafoori and colleagues in their study of US
military veterians. Ghafoori, B., Hierholzer, R.W., Howsepian, B., & Boardman, A. (2008) The role of adult attach-
ment, parental bonding, and spiritual love in the adjustment to military trauma. Journal of Trauma & Dissociation,
9(1), 85–106.
482 Bowlby, J. (1974) A guide to the perplexed parent. New York Times, 2 March 1974.
483 Granqvist, P. (2020) Attachment, Religion, and Spirituality: A Wider View. New York: Guilford: ‘The principal
ideas about attachment-related individual differences largely derive from, and are intertwined with, the methods
used to measure such individual differences.’ See also Verbeke, W., Belschack, F., Bagozzi, R.P., Pozharliev, R., &
Ein-Dor, T. (2017) Why some people just ‘can’t get no satisfaction’: secure versus insecure attachment styles affect
one’s ‘style of being in the social world’. International Journal of Marketing Studies, 9(2), 36–55: ‘In using general
attachment scales to study why attachment styles are so pervasive in affecting people’s relationships, we take an
ontological stance (philosophical study of the nature of being) in order to look at how individual differences in at-
tachment styles reflect people’s sense of being in the social world’ (37).
The developmental tradition has more often, though not exclusively, adopted a narrower
definition of attachment, focused on the properties of a behavioural system shared with other
primates and not requiring elaborate symbolic capacities. By contrast, the social psycho-
logical tradition has tended to adopt a much broader image of attachment as manifest in all
human relational dynamics, including how humans ‘co-explore their social world, share their
enjoyment in the pleasures of their world, and cope with stress in that world’.484 This has fol-
lowed a thread in Bowlby’s writings, where sometimes he used the idea of attachment to refer
to this broader sense, including the symbolisation of safe haven and secure base phenomena
in relationships with ideas and institutions. However, more structurally, this broader notion
of attachment reflected and continues to express the concerns of social psychology as a sub-
discipline. In very general terms, and especially in its American incarnation, since the 1960s,
developmental psychology has tended to focus on the study of the emergence of the capacities

of individuals, especially as these occur in the context of their family relationships.485 The
work of Ainsworth and second-generation attachment researchers contributed to this dis-
cipline formation and was shaped by it. From the 1980s, again especially in America, experi-
mental social psychology has tended to examine the role of personality and social processes
within the diversity of adult individual perceptions, attitudes, and practices, including but not
limited to intimate relationships.486 (Exceptions can readily be found to these limited char-
acterisations, of course, which are intended to convey general and predominant differences.
Nonetheless, researchers who seem the prime exceptions can often be found expressing con-
cern about the limitations of predominant disciplinary trends.)487
The subdisciplines can be regarded as aligned by a basic commitment to imagining hu-
mans as interdependent with others, and therefore to the study of interactions. And they
find many points of cross-over, for instance in the study of the development of moral rea-
soning. However, in general terms, in developmental psychology humans are attended to
as interdependent in their formation. Main’s theory is an account of repetoires of response
to this interdependence as it is expressed by the demands of the attachment system. This is
sometimes confused by the assumption that the AAI measures a ‘thing’ called ‘attachment’
and by the individualising language of ‘secure’ and ‘insecure’ infants. Nonetheless, following
Ainsworth, the official line of the developmental tradition has been that attachment is, or
at least starts as, a dyadic property. This is a central reason why developmental attachment
484 Ibid.
485 Developmental cognitive psychology can here be identified as an exception, since the focus here is often
less relational. There are certainly exceptions, though, such as the study of babies’ recognition of human faces,
e.g. Farroni, T., Menon, E., Rigato, S., & Johnson, M.H. (2007) The perception of facial expressions in newborns.
European Journal of Developmental Psychology, 4(1), 2–13.
486 On the history of social psychology see Danziger, K. (2000) Making social psychology experimental: a
conceptual history, 1920–1970. Journal of the History of the Behavioral Sciences, 36(4), 329–47; Greenwood, J.D.
(2004) What happened to the ‘social’ in social psychology? Journal for the Theory of Social Behaviour, 34(1), 19–
34; Jahoda, G. (2007) A History of Social Psychology: From the Eighteenth-century Enlightenment to the Second
World War. Cambridge: Cambridge University Press; Kruglanksi, A.W. & Stroebe, W. (eds) (2012) Handbook for
the History of Social Psychology. Bristol: Psychology Press; Pettigrew, T.F. & Cherry, F. (2012) The intertwined his-
tories of personality and social psychology. In M.R. Leary & R.H. Hoyle (eds) Handbook of Individual Differences
in Social Behavior (pp.13–32). New York: Guilford. On distinctions between trends in American and European
social psychology see Schruijer, S.G. (2012) Whatever happened to the ‘European’in European social psychology?
A study of the ambitions in founding the European Association of Experimental Social Psychology. History of the
Human Sciences, 25(3), 88–107.
487 See e.g. Jahoda, G. (2016) Seventy years of social psychology: a cultural and personal critique. Journal of
Social and Political Psychology, 4(1), 364–80. On the role of ongoing boundary-work in the construction of legit-
imate and illegitimate exceptions to predominant trends in a discipline see Good, J.M. (2000) Disciplining social
psychology: a case study of boundary relations in the history of the human sciences. Journal of the History of the
Behavioral Sciences, 36(4), 383–403.
Conclusion 535
researchers have been reluctant to support the ‘attachment disorder’ diagnosis of individual
children, leaving this diagnosis an odd and poorly integrated appendage to the develop-
mental research tradition (Chapter 1).
By contrast, in experimental social psychology the focus tends to be on the study of already
largely formed adult humans as influenced by social and relational factors in their percep-
tions, attitudes, and practices. As Stainton Rogers and colleagues have argued, there is a lib-
eral humanist quality to social psychology: humans are registered at the point that they can
act independently, though the concern is then with how they ultimately live together.488 This
has helped make use of college student samples a backbone of the subdiscipline. It has also
helped make self-report methodology more acceptable than for developmental psycholo-
gists. Shaver and Mikulincer are more committed than most social psychologists to the idea
that, at a fundamental level, relationships are primary, individuals are secondary: ‘Rather

than conceptualizing human beings as separate entities whose interactions with each other
need to be understood, it makes more sense to consider social relatedness and its mental
correlates as the normal “baseline” condition’.489 Nonetheless, aligned with the focus on in-
dividuals and their interactions in wider American social psychology, their predominant
methodological orientation has been towards the study of the schemas of individuals about
close relationships as reflected in their self-reports. Using self-report measures of attach-
ment styles, Shaver and Mikulincer’s work has explored how an individual’s schemas about
close relationships shape a diverse range of their perceptions, attitudes, and practices, ran-
ging from how new romantic relationships are established to how recklessly they drive a car.
One important by-product of the self-report methodology used by social psychological
attachment researchers was that it lowered an important barrier to experimental research.
This contributed to the identification of the effects of security priming by social psycholo-
gists, which is discussed in the next chapter. It also lowered a barrier to public engagement.
As documented in previous chapters, Bowlby’s work had contributed to a widespread public
for attachment theory and research. However, the developmental tradition of attachment re-
search had faced obstacles in continuing this legacy. Concepts such as the attachment be-
havioural system were complex and subtle, and the observational methodology favoured by
Ainsworth was complex and required extensive training to fully understand. Furthermore,
488 Stainton Rogers, R., Stenner, P., Gleeson, K., & Stainton Rogers, W. (1995) Social Psychology: A Critical
Agenda. Cambridge: Polity Press; Danziger, K. (1992) The project of an experimental social psychology: historical
pespectives. Science in Context, 5, 309–328. See also Spini, D., Elcheroth, G., & Figini, D. (2009) Is there space for
time in social psychology publications? A content analysis across five journals. Journal of Community & Applied
Social Psychology, 19(3), 165–81. Allport’s ‘official’ definition of social psychology as the study of how ‘the thought,
feeling, and behavior of individuals are influenced by the actual, imagined, or implied presence of others’ is illus-
trative: the individual is influenced by others, but is registered already as an individual. Allport, G.W. (1954) The
historical background of modern social psychology. In G.L. Lindzey (ed.) Handbook of Social Psychology, Vol. 1
(pp.3–45). Reading, MA: Addison-Wesley, p.5. Nonetheless, continuities in social psychology over time should
not be overemphasised. See Lubek, I. & Apfelbaum, E. (2000) A critical gaze and wistful glance at Handbook
histories of social psychology: did the successive accounts by Gordon Allport and successors historiographically
succeed? Journal of the History of the Behavioral Sciences, 36(4), 405–428. In particular, the rise of relationship
research in social psychology since the 1980s has roots in earlier work, but should be recognised as a distinct devel-
opment. See Reis, H.T. (2012) A brief history of relationship research in social psychology. In A.W. Kruglanski &
W. Stroebe (eds) Handbook Online Dating (pp.363–82). Bristol: Psychology Press.
489 Mikulincer, M. & Shaver, P.R. (2012) An attachment perspective on psychopathology. World Psychiatry,
11(1), 11–15, p.14. See also Shaver, P.R. & Mikulincer, M. (2012) Attachment processes in relationships: reply
to commentaries. Journal of Family Theory and Review, 4, 311–17: ‘If interpersonal relations are somewhat like
dances or doubles tennis games, it is reasonable to expect that some of what is going on in the dance or in a match
is “in” the individual dancers or players (e.g., their skills, their muscle development, their history of training and
performance), some of it is in the interpersonal dynamics of a particular dyad, and some of it is in the context of a
particular performance (e.g., the other players, a particular piece of music, the audience)’ (313).
Ainsworth and her students were focused on creating a differentiated space within which
empirical attachment research could thrive, giving little time to communicating in popular
forums. There may also have been some reticence to take on Bowlby’s mantle as a public in-
tellectual. For the social psychological tradition, the barrier to public understanding was sub-
stantially lowered. The Hazan and Shaver ‘love quiz’ has its subtlties in the formulation of the
statements, but the measure’s presumption that there are distinct adult attachment styles that
can be reported by any individual about themselves contributed to a fundamental accessiblity.
An interesting illustration is the recommendation by Fonagy and colleagues of the use of
the Hazan and Shaver ‘love-quiz’ at the start of brief therapeutic work with patients rather
than the AAI or the ECR. Though they acknowledge that the ‘love-quiz’ is generally too crude
to pick up any changes in attachment style associated with therapy, they argued that it offers
an excellent initial basis for a therapeutic conversation in which both therapist and patient

are treated as knowledgeable but capable of change in their perspectives.490 There is no need
for the tallying of responses indexing latent constructs of anxiety and avoidance required for
the ECR, let alone the labour-intensive process of having an AAI coded. Part of the appeal of
the ‘love-quiz’ for Fonagy and colleagues in a therapeutic context was that, despite being a
self-report measure, the ECR is a deceptively subtle tool, oriented towards mental schemas
expressed in feelings, beliefs, expectations, and behaviours, but which may not be consciously
known. Fonagy and colleagues at the Anna Freud Centre have also shown increasing interest
in the use of self-report measures of attachment-related constructs in research contexts, in
part because of the greater potential viability of these measures for use at scale.491
Over the past 20 years or more, the ECR has successfully served as the dominant meth-
odological tool in the social psychological tradition, and additionally the basis for the model
of anxiety and avoidance as the dominant theory. Further discussion was largely excluded by
the fact that the ECR seemed to work well in predicting various expected correlates, and the
factor analysis by Brennan and colleagues was taken as proof of the psychometric standing of
the measure. Yet, for the third-generation researchers most versed in the social psychological
tradition, questions have been raised about the theory and method that have organised the
approach for the past 20 years. For example, Gillath, Fraley, and Birnbaum have wondered
whether enough variance in individual difference is captured in theory and method by the
constructs of anxiety and avoidance. Of course, they acknowledge, pragmatic compromises
are always a part of scientific measurement.492 Yet it is also possible for these compromises to
be renegotiated if these compromises are recognised as having incrementally caused enough
hinderance over time to warrant the effort. On the one hand, psychometric inquiries have
suggested that security may not be reducible to the absence of anxiety and avoidance, but may
represent its own dimension. On the other hand, it is not clear that the ECR as a measure or
existing theory in the tradition is adequate to capturing the effects of trauma on experiences
in close relationships. These are questions that Shaver and Mikulincer have acknowledged as
limitations to their work. Their former students and younger colleagues will need to appraise
whether these limitations are significant enough to warrant methodological or theoretical
change, or, if measures are left unchanged, how to respond to the ensuing limitations.
490 Lemma, A., Target, M., & Fonagy, P. (2011) Brief Dynamic Interpersonal Therapy: A Clinician’s Guide.

491 Fonagy, P., Luyten, P., Moulton-Perkins, A., et al. (2016) Development and validation of a self-report measure
of mentalizing: the Reflective Functioning Questionnaire. PLoS One, 11(7), e0158678.

492 See also Pickering, A. (1995) The Mangle of Practice: Time, Agency, and Science. Chicago: University of
Chicago Press.
6
Conclusion
Introduction
In Becoming Attached, Karen described the origins of Bowlby and Ainsworth’s ideas, and
considered emergent lines of research by the end of the 1980s, including work by Main and
colleagues at Berkeley, Sroufe, Egeland, and colleagues at Minnesota, and Shaver and col-

leagues at Buffalo/Davis. These have been some of the cornerstone research groups for the
development of contemporary attachment research. At the time Karen was writing, lead-
ership of attachment research was passing from Bowlby and Ainsworth to a second gen-
eration. Furthermore, attachment as an area of academic study was rapidly expanding as
perceptions of the field’s credibility and relevance opened the door to research funding. The
rapid growth of attachment research since the 1990s means that no book today about the
history of the field could even hope to be comprehensive. Work is already underway on an-
other book to address developments from the 1990s onwards. Yet even with attention to
these later developments still to come, it is possible to draw together some reflections on the
ideas, methods, and priorities of the early groups of attachment research considered in the
present book. These have set the scene for subsequent attachment researchers in important
ways. This concluding chapter will begin by reviewing three structural dynamics that have
contributed to the present state of attachment research as a paradigm: the apparent accessi-
bility of attachment concepts; the subtlety and complexity of its theory and methods; and the
differentiation of a field of cumulative attachment research. The chapter will then examine
the problems the intersection of these structural dynamics have caused for communication
between researchers and practitioners, using the case of child welfare practitioners as illus-
tration. Threats to the credibility and apparent relevance of attachment research from such
breakdowns has led to claims that the paradigm is reaching exhaustion. The chapter will
evaluate this claim and discuss three lively areas of research that build precisely on the bene-
fits of the field’s long development and history. It will close by considering what comes next
for attachment research.
Structural issues
Over the course of this book, we have seen how the circulation and the shape of attach-
ment theory and research have been influenced by the remarkable and at times mislead-
ing accessibility of some of Bowlby’s basic ideas. This can be regarded as the first of three
structural dynamics that in interaction have formed the basis for the strengths and weak-
nesses of attachment as a paradigm today. In part, the intimacy of Bowlby’s ideas stemmed
from their emergence out of psychoanalysis. Bowlby was intent on developing a theory
that was integrated with the scientific developments of his day in other disciplines and that
could offer testable hypotheses, but that also retained a portion of the capacity of psycho-
analysis to address intimate life, ugly feelings, and inner conflict. Attachment theory speaks
to themes that are constitutive of human life and that readily absorb our attention: being a
538 Conclusion
child, becoming an adult; closeness and distance; the difficulties posed both by loss and by
belonging. However, a further contribution to the power of Bowlby’s attachment theory was
his appeal to ordinary language. Bowlby’s language in the 1950s was built to be able to both
circulate to widespread publics and to persuade both academic and clinic audiences. The
originating language of attachment theory used terms—such as ‘mother’, ‘attachment’, ‘sep-
aration’, ‘love’, ‘anxiety’—that had a strong and emotive set of connotations in ordinary lan-
guage. Bowlby overlaid these connotations with a set of technical meaning in his scholarly
work. ‘Love’ proved unworkable in this regard and was abandoned. And the second gener-
ation mostly stopped using ‘mother’ to mean attachment figure. Other terms have, however,
been retained over the decades. This grounding in ordinary language has contributed to the
flexibility, urgency, and reach of ‘attachment’ discourses, allowing it to seem plausible and at
times to catch at the heart. These qualities, together with the credibility provided by attach-

ment research as an empirical paradigm, have contributed to the appeal of attachment to
diverse audiences.1
A second structural dynamic that has shaped attachment research has been the way that
theory has been enshrined in complex measures. This has been a mixed blessing. Over time,
the Ainsworth Strange Situation has replaced the interactions in the home that it had been
intended to capture and preserve. It likewise embodied but also supplanted discussions of the
attachment behavioural system. The development of the Attachment Q-Sort by Waters and
colleagues sought to preserve Ainsworth’s concern with naturalistic observation and with the
capacity of caregivers to offer their child a secure base. However, this measure too stimulated
no further discussion of the attachment behavioural system, which has increasingly func-
tioned as a memento in the developmental tradition of attachment research rather than as an
active object of further theoretical work. Its place has been taken by the ideas of ‘minimising’
and ‘maximising’ strategies, though without clarity about what exactly is being minimised
or maximised. Main’s theory was that it was attention to attachment-relevant information.
This position was the basis of her interpretation of the Ainsworth classifications and her
group’s introduction of the disorganised attachment classification and the Adult Attachment
Interview (AAI). The innovations by Main and her group gave the Ainsworth categories an
apparently universal resonance, across cultures and across lifespan development. However,
with Main’s attentional theory largely unrecognised, there has been little precision in discus-
sions of how the four-category coding system can be applied cross-culturally.
Similarly, without clarity on what exactly the AAI measured, controversies were primed
when a social psychological tradition of attachment research emerged, which also made
appeal to the Ainsworth Strange Situation categories but as part of a self-report measure.
Unlike the developmental attachment tradition, Shaver and Milkulincer have further in-
terrogated the idea of behavioural systems and attempted to articulate the concept of at-
tachment. The developmental and social psychological traditions of attachment research
have come to a better coexistence over recent years, both contributing to the Handbook of
1 Ziv and Hotam have stated that ‘the “attachment language” is undeniably unique, coherent, rich, and complex
enough to interest professional academic audiences, to be clinically meaningful to practitioners, and to be straight-
forward enough to attract laypersons’. Despite their reference to coherence, Ziv and Hotam would likely agree
that these audiences do not encounter a unitary discourse, but at least in part encounter different discourses that
have overlapping vocabularies. Terms such as ‘attachment’ and ‘separation’ (Bowlby) or ‘security’ and ‘sensitivity’
(Ainsworth) may be used by academics, clinicians, and laypersons alike—but with little overlapping meaning. Ziv,
Y. & Hotam, Y. (2015) Theory and measure in the psychological field: the case of attachment theory and the Strange
Situation procedure. Theory & Psychology, 25(3), 274–91, p.279.
Structural issues 539
Attachment and the journal Attachment & Human Development. However, dialogue between
the two traditions has been profoundly hindered by their use of the same central vocabulary
to mean quite different things.
An influential study by Brennan, Clark, and Shaver in 1998 found that existing self-report
measures could be integrated on the basis of two latent dimensions: avoidance and anxiety.
The Experiences in Close Relationships scale was developed using items that reflected these
two latent constructs. However, the exact relationship remains uncertain between the anx-
iety construct and Ainsworth’s ambivalent/resistant category, given that the latter incorp-
orates anxiety, anger, and passivity. And it is not clear how well the Experiences in Close
Relationships scale handles experiences of security and trauma. Like the Strange Situation
and the AAI, the Experiences in Close Relationships scale embodies a particular theory of
individual differences relevant to attachment, but has also served to close down certain ques-

tions about that theory. Nonetheless, these questions have generated increasing discussion
in recent years from some of Shaver and Milkulincer’s closest students.
A third structural dynamic made salient by taking a historical perspective on early attach-
ment research groups has been the work of ‘field building’: the construction of attachment as
a differentiated paradigm with its own characteristic dispositions, methods, theories, values,
allies, enemies, and nodal institutions.2 The Strange Situation provided the initial basis for
a cumulative research programme, capable of attracting a sustained flow of research funds
and empirical findings. Whereas Bowlby was writing primarily for clinical audiences and the
general public, the central audience for Ainsworth and the second generation of attachment
researchers was the American community of academic psychologists. With the construction
of attachment research as a differentiated empirical paradigm following Ainsworth and her
immediate colleagues, attachment research became a ‘non-formative’ activity, in which re-
shaping parenting practices was not a focal concern.3
Certainly there are some second-generation attachment researchers for whom clinical
relevance has been the guiding principle of their work. Peter Fonagy, Patricia Crittenden,
Alicia Lieberman, and Karlen Lyons-Ruth are salient examples.4 And social historians and
sociologists have sometimes extrapolated from Bowlby’s early writings and the reception of
attachment ideas in welfare practice to assume that contemporary attachment researchers
are oriented towards the pathologizing and disciplining of families.5 Yet, overall, what is
striking on close examination is how little developments in the second generation of at-
tachment research, especially in America, seem to have been responsive to the concerns of
2 On field differentiation see Bourdieu, P. (1989) The conquest of autonomy: the critical phase in the emergence
of the field. In The Rules of Art, trans. S. Emanuel (pp.47–112). Cambridge: Polity Press.
3 On the concept of non-formativeness see Wood, M. (2009) The nonformative elements of religious life: ques-
tioning the ‘sociology of spirituality’ paradigm. Social Compass, 56(2), 237–48.

4 E.g. Lyons-Ruth, K. & Spielman, E. (2004) Disorganized infant attachment strategies and helpless-fearful pro-
files of parenting: integrating attachment research with clinical intervention. Infant Mental Health Journal, 25(4),
318–35. Crittenden, P.M. (2015) Raising Parents, 2nd edn. London: Routledge; Fonagy, P., Luyten, P., Allison, E.,
& Campbell, C. (2017) What we have changed our minds about: Part 2. Borderline personality disorder, epistemic
trust and the developmental significance of social communication. Borderline Personality Disorder and Emotion
Dysregulation, 4, 9. Van IJzendoorn and Bakermans-Kranenburg are also partial exceptions, albeit primarily in the
past decade, for instance co-authoring a popular book aimed at child welfare professionals: van IJzendoorn, M.H.
& Bakermans-Kranenburg, M.J. (2010) Gehechtheid en Trauma. Diagnostiek en Behandeling voor de Professional.
Amsterdam: Hogrefe.
5 E.g. Kanieski, M.A. (2010) Securing attachment: the shifting medicalisation of attachment and attachment
disorders. Health, Risk & Society, 12(4), 335–44; Garrett, M.P. (2017) Wired: early intervention and the ‘neuromo-
lecular gaze’. British Journal of Social Work, 48(3), 656–74.
540 Conclusion
clinical and social welfare practice, and how much by the demands of institutional academic
psychology. Indeed, with some exceptions, the structural articulation between research and
clinical attachment discourses has frequently been what Foucault termed ‘feeble’ and ‘slack’.6
For instance, the developmental research community have generally looked askance at both
the ‘attachment disorder’ diagnosis and the use of attachment measures in assessment of
risk in a context in which clinical and welfare services and training are largely structured
by diagnostic pathways and risk assessment. Main has never written about what parents
should do, and she has emphasised in print that practitioners should not regard her ten-
tative suggestions as authoritative for what they should do.7 The Minnesota Longitudinal
Study of Risk and Adaptation did contribute to the development of STEEP. However, this
was a sideline for the Sroufe and Egeland group at Minnesota, whose primary focus was the
cohort study. Shaver and Mikulincer have focused their attention predominantly, though

certainly not exclusively, on samples of university students. Their remarks about the implica-
tions of attachment anxiety and avoidance for clinical and child welfare practice are rare and
underdeveloped.8
Meeting the demands of institutional academic psychology provided advantages. Whilst
the labour-intensive measures of the developmental tradition generally constrained sample
size,9 a cumulative programme could be pursued over decades by the second generation of
attachment researchers. Based on a massive number of studies, a series of meta-analyses in
the early 2010s confirmed the capacity of the Strange Situation to predict later social and
mental health outcomes. The extent of this prediction is either moderate or very substantial—
depending on how effect sizes are interpreted.10 In any case, Groh and colleagues observed
that they are comparable or stronger than other psychological assessments of socioemotional
development conducted in infancy.11 The dedication of attachment researchers to commen-
surate measures is a marker of important continuities over time. Attachment has appeared
as a strangely stable theoretical paradigm, in a context in which psychological theories were
6 The sociological literature discussing attachment research has tended towards the assumption that the psy-
disciplines always tend to work in the same way. However, this prefabricated ‘critical’ account of psychological
discourse misses important heterogeneity. Foucault urged that it is important to register and to study not only
effective medical and professional power/knowledge relations, but also ‘a domination that grows feeble, poisons
itself, grows slack’, and not to mistake one for the other. Foucault, M. (1971, 1994) Nietzsche, genealogy, history.
In J.D. Faubion (ed.) Aesthetics, Method, and Epistemology, Essential Works of Foucault 1954–1984. New York: The
New Press, p.381. In turn, it may be gently acknowledged that the prefabricated discourse of ‘critical’ social sci-
ence is not itself free from imbrication with forms of domination, for instance as a capital-accrual strategy within
academia.
7 The only sustained consideration appears in Main, M., Hesse, E., & Hesse, S. (2011) Attachment theory and
research: overview with suggested applications to child custody. Family Court Review, 49(3), 426–63. Here Main
and Hesse firmly situated themselves as non-experts on professional practice, offering reflections as best they could
in response to a direct request rather than because custody decisions were central to their work as researchers.
8 What remarks are offered seem to generally stem from work led by Shaver’s wife Gail Goodman, rather than
reflecting the priorities of Shaver and Mikulincer’s research agenda, e.g. Lawler, M.J., Shaver, P.R., & Goodman,
G.S. (2011) Toward relationship-based child welfare services. Children and Youth Services Review, 33(3), 473–80.
9 The central early exception here was NICHD Early Child Care Research Network (1997) The effects of infant
child care on infant–mother attachment security: results of the NICHD Study of Early Child Care NICHD Early
Child Care Research Network. Child Development, 68(5), 860–79. A larger exception was Jaddoe, V.W., van Duijn,
C.M., Franco, O.H., et al. (2012) The Generation R Study: design and cohort update 2012. European Journal of
Epidemiology, 27(9), 739–56.
10 Funder, D.C. & Ozer, D.J. (2019) Evaluating effect size in psychological research: sense and nonsense.
Advances in Methods and Practices in Psychological Science, 2(2) 156–68.

11 Groh, A.M., Fearon, R.P., van IJzendoorn, M.H., Bakermans- Kranenburg, M.J., & Roisman, G.I. (2017)
Development Perspectives, 11(1), 70–76.
Attachment in child welfare practice 541
in continual churn and psychological theory in general was in decline. Highlighting this
continuity, Waters and colleagues could claim that ‘Bowlby’s attachment theory is the only
current theoretical framework in developmental psychology that is cast in the grand theory
model, widely accepted, and empirically productive’.12
To anthropologists, this continuity has appeared as an imperviousness to criticism among
attachment researchers, contributing to a sense of frustration and of being ignored by critics
of the paradigm (Chapter 2). Yet, as Latour has urged, social studies of science must see
the stability of social institutions and practices not as merely the default state of things, but
as ‘exactly what has to be explained by appealing to costly and demanding means’.13 In the
case of attachment as a research paradigm, these costly and demanding means have included
the labour-intensive Strange Situation and astonishing amounts of convergent effort. Yet the
stability of attachment research as a unified paradigm has also been achieved through main-

tenance of a common terminology, the cost of which has been confusion and miscommu-
nication. One site of such problems has been in communication between research groups.
Likewise, the underpinnings of the AAI were obscured by Main’s choice to frame this meth-
odological innovation in terms of Bowlby’s concept of internal working models, miscuing
decades of subsequent discussion of the measure.
Attachment in child welfare practice
The apparent accessibility of attachment concepts, the subtlety and complexity of its theory
and methods, and the differentiation of a field of cumulative attachment research have to-
gether contributed to many of the strengths and attraction of the paradigm today. However,
their interaction has also contributed to serious and pervasive misunderstandings regarding
the concepts that organise attachment theory and research. The language of psychological
research often grows out of ordinary language, since it is from ordinary problems that psych-
ology frequently takes its starting point. Yet attachment research has been especially vulner-
able to problems of mistaken identity between technical and ordinary language.
Summarising themes from across the previous chapters, four factors may be identified as
of special importance. First, Bowlby’s early vocabulary was drawn deliberately and strategic-
ally from ordinary language in order to support the popular appeal of his ideas. Second, the
coding systems of the developmental tradition are complicated and information about how
constructs were actually operationalised has been unduly limited in circulation to an oral
culture. Third, the appearance of attachment as a unified paradigm has been maintained
in part through the stretching of terminology, with concepts invested with some overlap-
ping and some non-overlapping senses. The capacity for concepts to be used flexibly has
allowed attachment to seem relevant in diverse areas and diverse ways. However, this has
come at a price. Areas of practice that appreciate conceptual precision and clearly operation-
alised concepts have tended to reject the attachment paradigm.14 And communication by
and among attachment researchers has been hindered by the way that its concepts serve as
things go together. Human Development, 48(1–2), 80–84, p.81.

13 Latour, B. (2005) Reassembling the Social. Oxford: Oxford University Press, p.35.
14 E.g. Bolen, R. (2000) Validity of attachment theory. Trauma, Violence & Abuse, 1, 128– 53; Bosmans, G.
(2016) Cognitive behaviour therapy for children and adolescents: can attachment theory contribute to its efficacy?
Clinical Child and Family Psychology Review, 19(4), 310–28.
542 Conclusion
magnets attracting quite heterogenous investments. Fourth, then, it can be observed that at-
tachment research has been distinctively ill-equipped in terms of infrastructures for pruning
how concepts are used in technical discussions. It is to be hoped that the journal Attachment
& Human Development, future editions of the Handbook of Attachment, and organisations
like the Society for Emotion and Attachment Research and the International Association for
the Study of Attachment might give greater attention to this issue.15
Yet, if there has been trouble in communication between researchers, the situation has
been even worse for dialogue between researchers and practitioners. In 1999, Rutter and
O’Connor described extensive ‘conceptual confusion’ in appeals to attachment theory and
research by clinicians and social workers.16 As the previous chapters have shown, there have
been several ways in which the attachment research community have made effective dissem-
ination of their ideas harder than many other fields. If Bowlby demonstrated the dangers of

populism, his successors have demonstrated the dangers of turning away from public en-
gagement. Misunderstandings and simplifications based on Bowlby’s early and popularising
texts have abounded, and continue to shape public perceptions of the paradigm today. The
gap between researchers and their public, where it is not left empty, has often been occu-
pied by professional trainers, whose distance from empirical research on attachment and
the field’s oral culture opens varied opportunities for quick-fix misapplications. A recent ex-
ample was commercial training provided for thousands of UK social workers in using dis-
organised attachment as seen in naturalistic settings as an indicator of child maltreatment.17
Such misunderstandings and simplifications should be recognised as predisposed
by obstacles to clear understanding of the disorganised category. These included the re-
stricted circulation of Main’s texts, her use of ordinary language terms in highly technical
ways without definitions, and by the repetition of a superficial account of Main’s ideas and
findings by other attachment researchers. The lack of psychometric scrutiny of the disor-
ganised classification has helped naturalise its status as a category, contributing to its quasi-
diagnostic appearance, especially when combined with Carlson’s finding—intended as
exploratory—showing that the category was associated prospectively with a general index
of mental illness. More generally, the reification of attachment and attachment classifica-
tions no doubt facilitated the assumption of the commercial trainers that behaviour seen at
home by child welfare practitioners would have the same meaning as behaviour seen in the
research laboratory.18
Recently, problems in the role of attachment theory within child welfare practice have
been the topic of a dedicated book by White, Gibson, Wastell, and Walsh.19 White and
15 One set of definitions has been offered by Schuengel, C., de Schipper, J.C., Sterkenburg, P.S., & Kef, S. (2013)
Attachment, intellectual disabilities and mental health: research, assessment and intervention. Journal of Applied
Research in Intellectual Disabilities, 26(1), 34–46.
16 Rutter, M. & O’Connor, T. (1999) Implications of attachment theory for child care policies. In J. Cassidy & P.R.
Shaver (eds) Handbook of Attachment: Theory, Research and Clinical Applications (pp.823–44). New York: Guilford
Press, p.823.
17 Granqvist, P. (2016) Observations of disorganized behaviour yield no magic wand: response to Shemmings.
Attachment & Human Development, 18(6), 529– 33; Granqvist, P., Sroufe, L.A., Dozier, M., et al. (2017)
Disorganized attachment in infancy: a review of the phenomenon and its implications for clinicians and policy-
makers. Attachment & Human Development, 19(6), 534–58.

colleagues observe that Bowlby was active in promoting his theory to social workers.20 And
in turn, social workers found in attachment a knowledge base to help them claim profes-
sional status, in which cases could be interpreted in terms of a credible theory. Attachment
offers welfare professionals a framework that appears to predict later risk to a child’s health
and development from the child or parents’ observable behaviour.
Over the years, both social work academics and policy documents have encouraged wel-
fare professionals to use the image of secure attachment as the point of comparison when
making assessments of parenting capacity. For instance, in the Department of Health prac-
tice guidance Assessing Children in Need and their Families, published in 2000, practitioners
were told that ‘secure attachments are so important in the early years. Where these attach-
ments are absent or broken, decisions to provide children with new attachment figures must
be taken as quickly as possible to avoid developmental damage.’21 Taken at its word, in the

absence of a secure attachment the guidance seems to suggest that children need to be con-
sidered for separation from their parents. White and colleagues identify that this position
seems underpinned by the idea that ‘attachment patterns, once formed, are stable and set
forever’.22
In recent years, with cuts to supportive services for families, the balance of child welfare in
the UK has tilted firmly towards a focus on statutory investigation of families.23 White and
colleagues regard attachment discourses as complicit in this shift. They suggest that, at least
in UK child welfare practice, attachment provides ‘a handy vocabulary, a diagnostic gaze,
learned-sounding re-descriptions of messy relationships and often a foil for moral judge-
ments’.24 They report from an extensive ethnographic study of social workers in a child pro-
tection service. They found that attachment was never mentioned by social workers in their
discussions or personal reflections. Nonetheless, appeals to attachment appeared repeatedly
in official reports when judgements were made about parenting capacity. This expressed a
contradiction facing social work, stemming structurally from its dominated position com-
pared to both legal and medical professionals. Social workers require access to a category-
based and prognostic knowledge system in order to operate and make qualitative judgements
as professionals, but are not regarded as qualified to make use of reserved clinical diagnoses.
A quasi-diagnostic use of attachment language appeals as a workable, if imperfect, solution
to this structural conflict.
Uses of attachment discourse in the reports observed by White and colleagues were gen-
erally vague, essentially functioning as synonyms for a poor child–parent relationship.
Similar observations have been made by other researchers. Potter’s ethnographic observa-
tions and North’s qualitative interviews suggest that vagueness in the use of attachment lan-
guage assessments may reflect the climate of child welfare services and the family courts,
20 E.g. Association for Psychiatric Social Workers (1955) Presentation at the Annual General Meeting 1955:
Dr John Bowlby on preventative activities. Modern Records Centre Warwick University, MSS.378/APSW/P/
16/6/19-20.
21 E.g. Department of Health (2000) Assessing Children in Need and their Families: Practice Guidance.
London: TSO.

23 Bilson, A. & Munro, E.H. (2019) Adoption and child protection trends for children aged under five in
England: increasing investigations and hidden separation of children from their parents. Children and Youth
Services Review, 96, 204–11.

544 Conclusion
and ambiguities around the professional status of social workers. This would imply that
social workers draw on attachment concepts in their reasoning, but are vague about this in
their reports, so as to avoid claiming expert levels of knowledge where this might be chal-
lenged by higher-status professionals.25 One of North’s participants, Bryony, described this
predicament:
Whereas although we can say we have concerns about the attachment, I don’t feel we’re
qualified enough to say, you know [softly], ‘They’ve got an attachment issue, you know,
they’ve got a dis . . . organised [almost inaudible] . . . attachment or whatever,’ because
I don’t feel we’re qualified enough. I don’t feel qualified enough to say that.26
North observed that across the interview, Bryony ‘spoke softly when she referred to attach-

ment, seemingly to emphasise her shame at not feeling competent to make theoretically in-
formed judgments. This was a response echoed by many interview participants.’ In North’s
analysis, ‘social workers often want to be more proficient in their application of attachment
theory and in how they describe their utilisation of it in assessments. They also experience
frustration at lacking the skills to explain effectively how they have used attachment theory
to indicate the potentially harmful outcomes for a child of experiencing emotional abuse.’27
By contrast, White and colleagues adopt a more sceptical conclusion. They claim that at-
tachment language is being drawn upon, post-hoc, in reports as authority to justify practi-
tioners’ intuitions about adequate and inadequate care: ‘the social workers referred to “good”
or “positive” attachments, rather than the scientific community concepts of secure and inse-
cure, to describe a good quality relationship between a child and their parents’.28 However,
White and colleagues felt that the attachment frame of reference, even if post-hoc, had nega-
tive effects. It helped legitimate ‘the narrow focus on the mother–child relationship, and the
responsibility of the mother for this relationship’, directing attention away from the family
socioeconomic context, the availability of social support, and potential neurodevelopmental
issues experienced by the child.29 In this, White and colleagues suggest that practitioners’
use of attachment theory was shaped by the priorities of child protection practice in the UK,
25 Potter, A. (2019) Judging social work expertise in care proceedings. In D.S. Caudill, S.N. Conley, M.E. Gorman,
& M. Weinel (eds) The Third Wave in Science and Technology Studies (pp.71–85). London: Palgrave. Further find-
ings from Potter’s doctoral research are forthcoming. See also Shemmings, D. (2018) Why social workers shouldn’t
use ‘attachment’ in their records and reports. Community Care, 28 June 2018. https://www.communitycare.co.uk/
2018/06/28/social-workers-shouldnt-use-attachment-records-reports/.
26 North, G. (2019) Assessing for bruises on the soul: identifying and evidencing childhood emotional abuse.
Journal of Social Welfare and Family Law, 41(3), 302–320, p.313.

27 Ibid.
28 Similar observations have been made by Wilkins, D., Shemmings, D., & Shemmings, Y. (2015) A–Z of
Attachment. London: Palgrave: ‘We have also seen what can only be described as quite crude and, in all likeli-
hood, mistaken applications of attachment theory in practice. For example, we have read reports by contact
supervisors observing that a child stays very physically close to their attachment figure throughout the session,
with this then interpreted as a sign of a “positive attachment”; we see conclusions being drawn about a child’s at-
tachment relationships based on one, short observation; descriptions of parents being “attached” to their babies;
of a child’s attachment relationship with their father being almost completely overlooked and of social workers
describing young children as always happy and content without reflecting on whether this is a “good thing” or not.
Additionally, we often hear workers speak of “strong attachments” . . . We have become increasingly worried that
attachment theory and research are often used “against” families: to highlight “problems” and “gaps” in parenting
or caregiving relationships—particularly when writing reports for the courts or in relation to child protection pro-
cedures’ (xv–xvi).

which focus on procedure and risk assessment at the expense of a rounded attention to the
context of families over time.
The primary target of White and colleagues is not attachment research, but the way that
in the UK child welfare practice ideas from attachment theory may at times be ‘used with a
mixture of excessive credulity and zealotry, a cavalier heavy-handedness and unsophisti-
cated reductionism’. In fact, White and colleagues claim that attachment theory and research
has a lot to contribute to child welfare practice: ‘we agree with the proponents of the theory
that policy and practice is not sufficiently attachment minded. The complexities and nuances
of the concepts are hidden by simplification . . . at its best, attachment research has produced
ideas that practitioners can use to understand the quality of child–carer relationships when
the child is anxious, scared, or upset, and to guide them in their work to improve familial
relationships.’30

White and colleagues give the Love Barrow Families project as an exemplary use of at-
tachment theory and measures by helping professionals. Love Barrow Families is an innova-
tive service-delivery model for families with multiple complex needs based in Cumbria, UK.
The model was codesigned with local families, and local families sit on the steering group.
Key principles include working with the whole family and support for families to integrate
with their local community. Rather than multiple services working with families in crisis at
once, one member of the Love Barrow Families team acts as a key-worker for each family.
Decisions about how to direct and prioritise work with the family are supported through
use of attachment assessments with family members, which feed into a whole-family for-
mulation and plan. An evaluation undertaken by Vincent documented that, compared to a
matched comparison sample, the Love Barrow Families project reduced the number of chil-
dren taken into care and the number of children on child protection plans.31 White and col-
leagues express appreciation that Love Barrow Families does not use attachment measures or
theory to diagnose or label, but to sensitise practitioners to both the strengths and the needs
of the different members of family systems.
Love Barrow Families draws especially on the ideas and assessments developed by
Patricia Crittenden, which have been influential in the UK.32 Love Barrow Families regards
Crittenden’s work as offering a range of clinically relevant archetypes, and proposals for
distinct forms of therapeutic approach to clients displaying different forms of behaviour.33
There have been recent debates among attachment researchers about whether the assess-
ments developed by Crittenden have sufficient sensitivity and specificity for use in court
assessments of parenting capacity.34 White and colleagues are generally hostile to the use
30 Ibid.
31 Vincent, S. (2017) ‘The Magic Is in the Co-Production’: Summary Report from the Evaluation of the Love Barrow
Families Project. Newcastle: Northumbria University.
32 Crittenden, P.M. (2015) Raising Parents, 2nd edn. London: Routledge. See also Baim, C. & Morrison, T. (2014)
Attachment-Based Practice with Adults. Brighton: Pavilion: ‘There is a general truism about attachment theory,
which is that the more you learn about attachment theory the more cautious you become about attaching labels to
people. This includes labels such as “securely attached”, “reactive attachment disorder”, “disorganised attachment”
and so on, which tend to be over-used and under-defined’ (23).
33 See also Baim, C. (2019) DMM vs ABC+D—a controversial discussion. DMM News, 32. https://www.iasa-
dmm.org/images/uploads/DMM%20News%20%2332%20May%2019%20English.pdf: ‘The DMM community

attempts to find a way of assessing and informing treatment that is more accurate and useful because it is focused
on the function of behaviour rather than labelling symptoms—which puts people into boxes.’
34 Spieker, S.J. & Crittenden, P.M. (2018) Can attachment inform decision-making in child protection and fo-
rensic settings? Infant Mental Health Journal, 39(6), 625–41; Van IJzendoorn, M.H., Bakermans, J.J., Steele, M., &
Granqvist, P. (2018) Diagnostic use of Crittenden’s attachment measures in Family Court is not beyond a reason-
able doubt. Infant Mental Health Journal, 39(6), 642–6.
546 Conclusion
of attachment measures in court-mandated parenting assessments, in contrast to the use

of these assessments to identify how best to proceed in supporting a family. This seems to
parallel recent work of Madigan and colleagues in their creation of a brief assessment, us-
able by practitioners in real-time, for identifying forms of frightening or disrupted parenting
(AMBIANCE-Brief).35 Madigan and her collaborators anticipate that this assessment will
be able to help professionals identify forms of parenting associated with adverse child soci-
oemotional development and to target supportive interventions. However, the assessment
has been designed in such a way as to minimise its likelihood of being used in court assess-
ments of parenting capacity, for instance in avoiding the implication that certain forms of
caregiving are in themselves pathological. Madigan and colleagues are now in the process of
validating the AMBIANCE-Brief in terms of its capacity to predict Strange Situation classi-
fications, and evaluating the measure’s reliability when used in clinical and welfare practice.

One contributing factor supporting Madigan’s work has been the availability of funding in
Canada for research–practitioner collaborative inquiry. There also seems to be a broader
shift towards greater engagement with practitioners and publics associated with the third
generation of attachment researchers.36
In their book, White and colleagues at times depict practitioners as enthralled to sim-
plified and haughty guidance about attachment theory. Presumably this is intended as a
counterweight to what the authors regard as the untroubled circulation of lazy assumptions
about attachment. One qualification that can be made regarding their account, though,
is that there is more variety among guidance for practitioners than White and colleagues
suggest. The Circle of Security graphic of the caregiver as a secure base and safe haven has
gone into wide circulation, and represents an admirably effective visual representation of
Ainsworth’s concepts; the Circle of Security Intervention itself is used in many countries
and there is initial evidence that it contributes to less judgemental forms of supportive
work with families.37 There are several books that do an excellent job in representing the
available research evidence and theory, whilst also offering reflections relevant to helping
professionals and parents. The best available in English is perhaps either Understanding
Attachment and Attachment Disorders by Prior and Glaser or Attachment in Therapeutic
Practice by Slade and Holmes.38 However, it should not be thought that effective mediation
35 Madigan, S. (2019) Beyond the academic silo: collaboration and community partnerships in attachment re-
search. Paper presented at International Attachment Conference, Vancouver, 20 July 2019; Haltigan, J.D., Madigan,
S., Bronfman, E., et al. (2019) Refining the assessment of disrupted maternal communication: using item response
models to identify central indicators of disrupted behavior. Development & Psychopathology, 31(1), 261–77.
36 This transition may be placed in the broader context of the rise of discourses, since the late 1990s, questioning
whether or how academic research meaningfully contributes to addressing the challenges faced by practitioners
and/or the public good. Irwin, A. (2001) Constructing the scientific citizen: science and democracy in the bio-
sciences. Public Understanding of Science, 10(1), 1–18; Gunn, A. & Mintrom, M. (2016) Higher education policy
change in Europe: academic research funding and the impact agenda. European Education, 48(4), 241–57.
37 Powell, B., Cooper, G., Hoffman, K., & Marvin, B. (2016) The Circle of Security Intervention. New York: Guilford;
McMahon, C., Huber, A., Kohlhoff, J., & Camberis, A.L. (2017) Does training in the Circle of Security framework
increase relational understanding in infant/child and family workers? Infant Mental Health Journal, 38(5), 658–68.
38 Prior, V. & Glaser, D. (2006) Understanding Attachment and Attachment Disorders: Theory, Evidence and
Practice. London: Jessica Kingsley Press; Slade, A. & Holmes, J. (2017) Attachment in Therapeutic Practice.
London: SAGE. The best available in Dutch is likely van IJzendoorn, M.H. & Bakermans-Kranenburg, M.J.
(2010) Gehechtheid en Trauma: Diagnostiek en Behandeling voor de Professional. Amsterdam: Hogrefe. Another
thoughtful work is Page, T. (2017) Attachment theory and social work treatment. In F. Turner (ed.) Social Work
Treatment: Interlocking Theoretical Approaches, 6th edn (pp.1–22). Oxford: Oxford University Press. Beyond the
published literature, some trainings offered to clinicians and foster carers are also exemplary in effective charac-
terisation of the technical aspects of attachment theory and research, and its relevance. Those by John Sands and
Lydia Fransham can be mentioned, though as yet neither has published their training materials.
of research and practice always entails the implementation of the former in the latter. An ex-
ample of a book that offers thoughtful integration of theory and practice, in an area where
there is little empirical attachment research, is Blood and Guthrie’s Supporting Older People
Using Attachment-Informed and Strengths-Based Approaches.39 There are many works that
take care to distinguish contemporary scholarly consensus from ‘allodoxia’—a cut-price,
simplified account offered up as if it had the same meaning as the technical conclusions of
researchers. For example, in Golding’s Nurturing Attachments Training Resource for deliver-
ing parenting groups, a flipchart of potentially ambiguous terms is used to help participants
keep in mind these distinctions.40
Nonetheless, White and colleagues are unquestionably right that a great deal of guid-
ance for clinicians and social workers is overconfident and poorly informed.41 Moreover,
this guidance is disproportionately likely to be available for free or cheaply, in contrast

to works that offer greater access to attachment research in its complexity, which are fre-
quently more expensive or behind journal paywalls. Three features can be identified that
especially characterise allodoxic guidance about attachment for clinicians and social
workers and related policy discourses. First, these texts offer little or no explanation of
Bowlby’s behavioural systems model, Ainsworth’s operationalisation of sensitivity, or how
disorganised attachment is actually coded. Instead they conjure with the broad ordinary
language connotations of attachment, security, sensitivity, and disorganisation.42 Second,
empirical attachment research over recent decades is ignored in favour of statements made
by Bowlby in popularising writings, as well as selected other statements from early attach-
ment theory implying a massive and stable causal influence of early care on later social
behaviour.43 Attachment researchers are treated as generally all much the same, which
allows for changes over time to be ignored. For instance, the concept of ‘felt security’ is
routinely attributed to Bowlby, rather than Sroufe and Waters’ critique of Bowlby. Third,
no reference is made to the qualified findings reported in meta-analyses or the findings
from these meta-analyses regarding moderators of the effect of caregiving on attachment,
or attachment on later development. It should be acknowledged that allodoxic attachment
discourse contributes, often helpfully, to recognition that child–parent relationships are
very important to children and their socioemotional development. However, the impres-
sion of a basis in empirical attachment research is spurious or heavily overstated. And
some topics, such as disorganised attachment and attachment disorders, are pervasively
mischaracterised.
39 Blood, I. & Guthrie, L. (2018) Supporting Older People Using Attachment- Informed and Strengths-Based
Approaches. London: Jessica Kingsley Press.
40 Golding, K. (2017) Nurturing Attachments Training Resource: Running Parenting Groups for Adoptive Parents
and Foster or Kinship Carers. London: Jessica Kingsley Press.

41 One example is Pearce’s A Short Introduction to Attachment and Attachment Disorder, which is admirably
accessible but contains many outright errors alongside confused oversimplifications. Pearce, J. (2009) A Short
Introduction to Attachment and Attachment Disorder. London: Jessica Kingsley Press. Even in otherwise good
works there are major errors, such as the conflation of controlling-punitive/controlling-caregiving with Reactive
Attachment Disorder in Howe, D., Brandon, M., Hinings, D., & Schofield, G. (1999) Attachment Theory, Child
Maltreatment and Family Support: A Practice and Assessment Model. London: Palgrave, pp.135–6. A decade
later, these problems were corrected in Howe, D. (2011) Attachment Across the Lifecourse: A Brief Introduction.
London: Palgrave.
42 E.g. Perry, B. (2001) Attachment: the first core strength. Early Childhood Today, 16(2): ‘Attachment is the cap-
acity to form and maintain healthy emotional relationships’ (28).

43 E.g. Marshall, N. (2014) The Teacher’s Introduction to Attachment: Practical Essentials for Teachers, Carers and
School Support Staff. London: Jessica Kingsley Publishers.

548 Conclusion
At first sight it seems curious that many hostile academic discussions of attachment re-
search, such as those of Vicedo (Chapter 1) and anthropologist critics (Chapter 2), possess
the same three qualities: they mistake technical for ordinary language (e.g. regarding the
meaning of ‘sensitivity’); recent attachment research is ignored in favour of classic state-
ments by Bowlby; and the findings of meta-analyses are neglected. However, these qualities
are less mysterious in light of Keller’s reflection (Chapter 2) that such critiques of attach-
ment research are, at least in part, a proxy for criticism of the uses of attachment discourse
in child welfare contexts. In this regard, a limitation of the work of White and colleagues is
that they lean on statements by Vicedo that homogenise and caricature attachment research
and its applications. Though White and colleagues refer to ethnographic observations and
scrutinise several court reports, the empirical basis for their claims about the uses of attach-
ment remains relatively limited.

As mentioned in the Introduction to this book, colleagues and I have several studies
underway exploring how attachment theory and measures are used by child welfare prac-
titioners and clinicians working with children. Findings are just in, and results will be re-
ported over the coming couple of years; this presentation of initial impressions should not
be taken to pre-empt the formal analysis. However, already these initial impressions indicate
more heterogeneity than suggested by Vicedo or White and colleagues. In a project led by
Sarah Foster, we conducted interviews with 24 children’s services social workers, and asked
them to offer responses to two vignettes of fictionalised cases drawn partly from Serious
Case Reviews. This was part of a wider project comparing the responses of different profes-
sional groups: social workers, general practitioners, and clinical psychologists. What was
especially striking from the interviews with the social workers was the significant individual
variation in knowledge of and views on the potential value and actual use of attachment
theory concepts and discourse in practice. This was despite the fact that all the social workers
were employed by just two Local Authorities, both in the same region of the UK. For some
of the social workers interviewed, they regarded attachment theory and discourses as having
little role in their work, regardless of guidance suggestions that they take attachment into
account in assessments of children. This stance was most frequently underpinned by a lack
of confidence in any theories relevant to social work. Theory was seen as ‘high’ knowledge,
beyond the reach of ordinary practitioners like themselves. However, some social workers
drew on other theories but were specifically critical of attachment theory. For example, one
social worker was concerned that use of attachment theory could risk drawing attention
away from fully acknowledging different cultural family contexts.
However, many of the interviewed social workers felt attachment was directly important
to their work with children and families. Practitioners reported using attachment to inform
their thinking about a variety of matters, including in thinking about what a child’s repeated
pattern of behaviour in the family context might mean, and how a parent’s own childhood
experiences might be affecting how they are currently responding to their child. However,
among practitioners who appreciated attachment theory and its relevance to their work, here
again there was variation. Some practitioners were enthusiastic about attachment theory and
research because it offered a lens on the relationships, emotional life, and socioemotional de-
velopment of everyone. Other practitioners were enthusiastic but viewed attachment as ‘one
aspect of everything that you take into consideration’ and only ever helpful alongside other
theories and frameworks.
In direct contrast to the finding in White and colleagues’ research, there were multiple
interviewees who said in interview that they used attachment theory and research extensively
to inform their practice, whilst intentionally avoiding the use of attachment terminology in
their reports. Explaining this stance, some of the social workers stated that identification of
individual differences in attachment, such as disorganised attachment, required specialist
assessment training which they did not have. Many expressed concern over being challenged
on their expertise in court if they used such terms. Another set of concerns raised by inter-
viewees was that many felt that using attachment terminology was inappropriate, as it may
be inaccessible to families and unhelpful or misleading for other practitioners:
I try not to put labels on, this child is suffering disorganised or ambivalent attachment,
I think in terms of your theory . . . you know what you’re looking for in a secure attachment
and you know when it’s not good! But it’s focusing more on the behaviours and what the
behaviours will be telling you as opposed to sticking a label, an attachment theory label on
it. Because it’s just words, we need to understand what that means for these children and

what we’re actually seeing.
Whilst much rarer amongst those interviewed, a few social workers stated that they did use
attachment classifications in their written reports despite not having trained in conducting
attachment assessments or carried out formal attachment assessments in these cases.
Practitioners drew on attachment language in responding to the two vignettes. There
was a good deal of confused use of attachment terminology in these responses. Examples
included: talking about the ‘strength’ of attachment, reasoning about ‘poor’ attachment,
conflating attachment with love, using the terms ‘attachment difficulties’ and ‘attachment dis-
orders’ interchangeably, and confusion about whether a child can be attached at all to a mal-
treating caregiver. This does support the concerns White and colleagues have raised about the
use and understanding of attachment. However, what was also clear from the interviews was
that many social workers were reflective about the limitations of their knowledge and of the
version of attachment theory in circulation within child welfare contexts. We were told:
I think it [attachment theory] can be vague and I think it can be overused by people, or
not always used appropriately. So I sometimes read reports or I hear professionals talking
about bad attachment and good attachment, and I understand what they’re talking about
but I don’t think it really explains anything.
One of the reasons I don’t like using that word [attachment] too much is because I think
it’s widely kind of used in, if not, it might be harsh to say it’s misused but it’s used in lots of
different ways and so you can’t really be confident that you’re talking about the same thing
when you talk to people.
In another study, led by Barry Coughlan, we asked practitioners from various clinical
services (including child and adolescent mental health, neurodevelopmental teams, and
primary care) to discuss cases they perceived as ‘attachment-related’ versus ‘neurodevelop-
mental’. This was part of a broader study of how clinicians distinguish autism, ADHD, and
attachment-related difficulties, drawing on interviews and examination of clinical records
from children’s mental health services. Regarding neurodevelopmental conditions, clin-
icians tended to lean on psychiatric nosology and standardised assessments to buttress their
formulations. Yet these frames of reference did not seem available or desirable to clinicians
when thinking about attachment-related cases. Instead, practitioners tended to place greater
weight on unstructured observations and the non-standardised taking of a developmental
history. In interview, few clinicians used standardised attachment assessments. In fact, ref-
erences to attachment literature were relatively sparse in the interviews, though several
550 Conclusion
clinicians described keeping attachment ‘in mind’ alongside the array of other frameworks
when making decisions about case conceptualisation and intervention planning. However,
they tended to be vague about what behaviours would prompt them to think about attach-
ment. Instead, attachment theory seemed to be drawn in when clinicians had other evidence
of insufficient care or maltreatment, or where there was a discrete precipitating event such as
a major separation. Most practitioners indicated a preference for the general phrase ‘attach-
ment difficulties’ over the technical clinical term ‘attachment disorder’. But practitioners did
not seem confident in articulating the difference between these terms. Practitioners were
also unsure what interventions or support should be offered for families where attachment
was identified as a relevant issue.44
Finally, in a third study, led by Helen Beckwith, we used Q-methodology to examine the
perspectives held about attachment concepts and research findings among two groups: (i)

30 child mental health clinicians and (ii) 30 established attachment researchers. A variety of
other self-report data were also collected. We have been fascinated to see both the conver-
gences and divergences in the perspectives of these groups. Clinicians and researchers alike
saw value in attachment theory as a framework for facilitating personalised care for chil-
dren and for making decisions about fostering and adoption placements. Further, all agreed
that ‘attachment theory could be used more precisely within mental health practice’, and that
attachment research and measures were largely inaccessible to clinicians. Both researchers
and practitioners were mindful of the difficulty of maintaining the integrity of attachment
concepts and methods when applying them to practice.45 Interestingly, both researchers
and clinicians were unsure whether callous and unemotional traits in children originate
from their early attachment experiences, or what the role is of attention within attachment
processes.
One line of difference was that attachment researchers held strongly that good quality
care throughout childhood is a better predictor of future mental health than a child’s early at-
tachment pattern. Clinicians were unsure. Researchers also felt that the most effective inter-
ventions to improve child–caregiver attachment are those that target security, presumably
following the meta-analysis from Bakermans-Kranenburg and colleagues (Chapter 4). By
contrast, clinicians appeared not to regard this as a practicable area of focus, and held that
they could achieve improvements in child–caregiver attachment through other means. We
were also interested that clinicians endorsed strongly the idea that early attachment experi-
ences determine how the brain develops, whereas the research community offered neither
agreement nor disagreement, presumably waiting on further evidence. This may reflect the
role of allodoxia, perhaps intersecting with the wider discourse that if something influences
the brain then it is more real.46
Clinicians were enthusiastic about the relevance of attachment language to helping them
make sense of cases, especially in understanding dyadic processes. However, they were less
44 See also Alexander, S.L., Frederico, M., & Long, M. (2018) Attachment and children with disabilities: know-
ledge and views of early intervention professionals. Children Australia, 43(4), 245–54; Morison, A., Taylor, E., &
Gervais, M. (2019) How a sample of residential childcare staff conceptualize and use attachment theory in practice.
Child & Youth Services, 1(25).
45 See also Oppenheim, D. & Goldsmith, D.F. (2007) Attachment theory in clinical work with children: bridging
the gap between research and practice. Journal of Canadian Academic Child and Adolescent Psychiatry,
16(4), 186–7.
46 See also Wastell, D. & White, S. (2017) Blinded by Science: The Social Implications of Epigenetics and
Neuroscience. Cambridge: Policy Press.

The exhaustion of attachment research? 551
positive about the relevance of existing attachment assessments. Researchers held that at
root, attachment is a dyadic and relational phenomenon, for all that it can stabilise with de-
velopment. Clinicians agreed in principle that attachment is dyadic in nature, but regarded
this as rather immaterial in clinical services structured by individual diagnoses and labels.
Though these are just indications of some initial findings, already they suggest some provi-
sional conclusions. They indicate that researchers and practitioners have many important
points of convergence; that divergences in perspective often express the specific and different
demands of the research and clinical contexts; and that there is appetite from both sides to
improve the precision with which attachment theory is used in clinical practice with chil-
dren. One line of future collaborative work may be efforts to validate the use of attachment
assessments in applied contexts, to see if they can demonstrate benefits compared to assess-
ment as usual. Another may be efforts to make the findings of attachment research more

readily available to practitioners and publics.
The exhaustion of attachment research?
As we saw in the above section, White and colleagues advocate for a better understanding
of attachment research among practitioners, to help counteract the dangers of the distorted
and deadened version currently in circulation, and especially the dependence on assump-
tions from Bowlby’s early writings. Yet, in their book, they offer little discussion of changes
in attachment research over time or diversity among attachment researchers.47 Relevant to
their concerns are discussions that have been taking place among attachment researchers
themselves about what within attachment research remains lively and what is fading from
view. These discussions have a long history, even if they have become especially salient in the
past few years.
In the late 1980s, van IJzendoorn sought to characterise the contemporary state of at-
tachment research in terms from the philosophy of science. He drew distinctions between
research paradigms as characterised, by degrees, in terms of ‘formation’, ‘construction’, ‘sat-
uration’, and ‘exhaustion’.48 In ‘formation’, research is starting up, exploring the possibility of
a programme of work. In the ‘construction’ stage, research is focused on the verification of
bold hypotheses, and social communication in a growing community is facilitated by pa-
pers and symposia. There is a dancing, exhilarating sense of excitement and discovery to the
work, alongside endless hard work. In ‘saturation’, research is focused on developing consist-
encies and responding to inconsistencies in the theory, journals are established to formalise
the field, and ideas from research get applied in practice in ways that do not feed back into
further scientific developments. In this stage, the big discoveries seem to have been made,
and empirical work becomes increasingly about filling in gaps and qualifying effect sizes.
47 White and colleagues do praise the work of Fonagy and colleagues for advocating a shift ‘away from instinct
and into interaction, and the social and psychological circumstances in which the mother and infant may find
themselves’. White, S., Gibson, M., Wastell, D., & Walsh, P. (2019) Reassessing Attachment Theory in Child Welfare.
Bristol: Psychology Press. However, the characterisation of attachment research as focused on instincts is half a
century out of date (Chapter 1). And Fonagy gives no more attention to interaction and social and psychological
circumstances than many attachment researchers.
48 Van IJzendoorn, M.H. & Tavecchio, L.W.C. (1987) The development of attachment theory as a Lakatosian re-
search program. In L.W.C. Tavecchio & van IJzendoorn, M.H. (eds) Attachment in Social Networks: Contributions
to the Bowlby–Ainsworth Attachment Theory (pp.3–31). New York: Elsevier Science, Table 1.
552 Conclusion
In ‘exhaustion’, a research paradigm becomes primarily concerned with defending estab-

lished orthodoxies, and communication becomes rigidified in handbooks and training insti-
tutes. Alternative perspectives gain ground in achieving funding, recruiting new researchers,
and securing institutional recognition. Van IJzendoorn characterised Bowlby’s development
of attachment theory as the ‘formation’ stage for the paradigm and Ainsworth’s introduction
of the Strange Situation procedure as the ‘construction’ stage, which lay the basis for a cumu-
lative empirical research programme in which new findings continually inform the develop-
ment of theory.49 However, looking about in the late 1980s, van IJzendoorn also identified
symptoms of the saturation stage. There was a sense that the major discoveries had been
made, and evidence that applications of attachment theory in psychotherapy, parent educa-
tion, and policy were not feeding back into the priorities of research practice.
It is interesting to reflect today on van IJzendoorn’s characterisation of the field of at-

tachment research in the 1980s. There seem to be aspects of the research paradigm that
resemble construction and saturation—and exhaustion. Since van IJzendoorn’s analysis, at-
tachment research has seen major developments, not least the extension of the paradigm
to study attachment phenomena in adulthood. Today, some developments are still in the
stage of ‘construction’. For instance, research on the relationship between individual differ-
ences in attachment and sexuality remains an area in which early hypotheses are still being
validated (Chapters 3 and 5). By contrast, some developments seem to have achieved satur-
ation. For instance, the relationship between caregiver sensitivity and infant attachment is
now well established, and current discussions have been mostly concerned with qualifying
effect-sizes and communicating with practitioners. However, there remains lively interest in
other factors contributing to intergenerational attachment processes.50 And the proposal by
Woodhouse, Cassidy, and colleagues (Chapter 2) to refine sensitivity to secure base provi-
sion may also reignite discussions in this area.
Over recent years, a few commentators have characterised attachment research as reach-
ing exhaustion. Unquestionably, criticisms of the paradigm and of individual attachment
researchers have gained ground in academic circles. Harkness has observed that ‘earlier
critics of attachment theory have recently been joined by others; if not exactly a chorus
of critics, there are now enough to form the basis for an organized, multi-referenced, and
multi-faceted counter-offensive’.51 While popular and clinical interest in attachment remains
high, it has become increasingly difficult to get funding for attachment research. Discussing
the UK context, Fonagy recalls that ‘attachment theory had ten good years in the research
community from about 1985–1995. We had excellent research grants from the Wellcome,
the Economic Social Research Council and the European Union; we were invited speakers
at British Psychological Society congresses, the British Association and all that. Attachment
theory is still good, but research interest has moved to neuroscience.’52
49 Ibid. p.16.
50 E.g. Bailey, H.N., Tarabulsy, G.M., Moran, G., Pederson, D.R., & Bento, S. (2017) New insight on intergenera-
tional attachment from a relationship-based analysis. Development & Psychopathology, 29(2), 433–48.
51 Harkness, S. (2015) The strange situation of attachment research: a review of three books. Reviews in
Anthropology, 44(3), 178–97, p.179.

52 White, K. & Schwartz, J. (2007) Attachment here and now: an interview with Peter Fonagy. Attachment, 1(1),
57–61, p.57. Thompson likewise has marked the standing of attachment research in this period, commenting in
2000 that attachment has been ‘the dominant approach to understanding early socioemotional and personality
development during the past quarter-century of research’. Thompson, R.A. (2000) The legacy of early attachments.
Child Development, 71(1), 145–52, p.145.
In a watershed development in 2016, the National Institute of Mental Health in the USA
removed the Ainsworth Strange Situation from its list of recommended procedures for pub-
licly funded mental health research, citing the debt of attachment theory to psychoanalysis
and its tendency to ‘reify . . . theoretical claims’ as essential problems with the paradigm.53
The framing of this decision was striking: in contrast to other judgements in the document,
the exclusion of the Strange Situation was made without any justification in terms of any
of the scientific criteria identified by the National Institute of Mental Health as relevant to
recommendations, such as psychometric properties, longitudinal stability, standardised
administration, and cultural specificity. Instead, it was simply asserted that the Strange
Situation was an invalid measure. And a letter of query led by Lyons-Ruth was ignored by
the National Institute of Mental Health.54 From the information available, it would appear
that the Strange Situation has been rejected for public funding on the basis of the disfavour

of attachment research among psychological researchers, rather than on the grounds the
National Institute of Mental Health itself had established as criteria. Though raked by the
claws of this development, attachment research in the USA has continued. The central pre-
occupations of contemporary developmental science—brains, genes, big samples, and clin-
ical application—are all viable for attachment research.55 However, work in the USA using
the Strange Situation has had to stow aboard grant applications that justify themselves in
other terms. Whilst the situation is less severe in other countries, the National Institute of
Mental Health announcement illustrates the challenge of credibility felt by attachment re-
search more broadly.
Fonagy and Campbell have offered the view that, essentially, the party is over for devel-
opmental attachment research pursued in the manner of the second-generation researchers
such as Main and Sroufe.56 Above all, Fonagy and Campbell claim that even if the research
paradigm was built upon the categories from the Strange Situation, it is now coming to grief
upon them. They regard categories as an outdated way of representing human differences.
This presentation of this criticism seems to reflect both the psychometric criticisms discussed
in Chapter 4 and wider contemporary discourses that treat categories for human beings as
a set of oppressive and misleading expectations, inappropriate for societies characterised
53 National Institute of Mental Health (2016) Behavioral assessment methods for RDoC constructs. https://
www.nimh.nih.gov/about/advisory-boards-and-groups/namhc/reports/behavioral-assessment-methods-for-
rdoc-constructs.shtml.
54 Lyons-Ruth, K., Belsky, J., Booth-LaForce, C., et al. (2016) Letter to Joshua Gordon—Director of NIMH and
Sarah Morris—Acting Director of the RDoC Unit, responding to the behavioral assessment methods for RDoC
constructs. Unpublished letter shared by Karlen Lyons-Ruth.
55 The work of Cicchetti, Rogosch, Toth, and colleagues (Chapter 4) provides one illustration. The work of
van IJzendoorn, Bakermans-Kranenburg, and colleagues offers another. See Kok, R., Thijssen, S., Bakermans-
Kranenburg, M.J., et al. (2015) Normal variation in early parental sensitivity predicts child structural brain
development. Journal of the American Academy of Child & Adolescent Psychiatry, 54(10), 824–31; Bakermans-
Kranenburg, M.J. & van IJzendoorn, M.H. (2015) The hidden efficacy of interventions: gene × environment ex-
periments from a differential susceptibility perspective. Annual Review of Psychology, 66, 381–409.
56 Fonagy, P. & Campbell, C. (2015) Bad blood revisited: attachment and psychoanalysis, 2015. British Journal
of Psychotherapy, 31(2), 229–50. This paper represents thinking with a wider group of colleagues also reported
in Fonagy, P., Luyten, P., Allison, E., & Campbell, C. (2017) What we have changed our minds about: Part 2.
Borderline personality disorder, epistemic trust and the developmental significance of social communication.
Borderline Personality Disorder and Emotion Dysregulation, 4(1), 9. In the USA, there have likewise been claims
that the field of attachment research is moving from saturation to exhaustion. Waters and colleagues have reported
a growing perception of attachment research among some American colleagues that ‘There is some great work
there. But it is a mature field now and I think all the big studies have been done’. Waters, E., Petters, D., & Facompre,
D. (2013) Epilogue: reflections on a special issue of attachment & human development in Mary Ainsworth’s 100th
year. Attachment & Human Development, 15(5–6), 673–81.
554 Conclusion
by fluidity and precarity.57 Fonagy and Campbell also argue that whilst correlations be-
tween caregiving, attachment, and mental health have been in the expected direction across
thousands of studies, the associations have been less strong than Bowlby appeared to pre-
dict. Furthermore, the sciences on which attachment theory was erected have moved on.58
Fonagy and Campbell acknowledge that attachment research has brought valuable atten-
tion to the question of the evolutionary function of mental health and illness. However, they
feel that since Main’s concept of conditional strategies, the dialogue between attachment re-
search and evolutionary biology has generally petrified (though Belsky and Crittenden are
cited favourably as exceptions).59 Fonagy and Campbell argue that developments in evo-
lutionary biology in thinking about cultural or gene–culture co-evolution models seem to
have passed by attachment research, which is looking rather old-fashioned.60
Ultimately, Fonagy and Campbell predict not a replacement of the developmental trad-

ition of attachment research, but its supersession by an approach that incorporates its
strengths.61 In the process, they anticipate that familiar aspects of attachment research,
above all the labour-intensive Strange Situation and its lumbering categories, will—and
should—be put aside.62 One tradition that may fill part of the remaining space is the social
psychological approach. However, given that this approach remains without an adequately
elaborated account of trauma, its clinical relevance remains limited. Fonagy and Campbell
instead recommend their own assessments of mentalising. Mentalising is the capacity to im-
agine, perceive, and interpret human behaviour in terms of intentional mental states (e.g.
needs, desires, feelings, beliefs, goals, and reasons).63 Fonagy and Campbell regard deficits
in mentalising as associated with attachment-related processes since they are predisposed by
problems within early family relationships. But they argue that it is mentalising and the cap-
acity to learn from interactions with others rather than attachment that is important for the
development of adult mental health and caregiving behaviour.
Fonagy and Campbell’s intent in their paper is clearly in part advocacy of their own theory
and measures. The attendant predictions of the demise of attachment research also have
57 See e.g. Bauman, Z. (2000) Liquid Modernity. Cambridge: Polity Press; Weeks, J. (2015) Beyond the categories.
Archives of Sexual Behavior, 44(5), 1091–7.

58 See also Fonagy, P. & Target, M. (2007) The rooting of the mind in the body: new links between attachment
theory and psychoanalytic thought. Journal of the American Psychoanalytic Association, 55(2), 411–56: ‘Advances
in the sciences to which Bowlby’s ideas are coupled dictate a reconsideration’ (420).
59 Fonagy, P. (2016) The role of attachment, epistemic trust and resilience in personality disorder: a trans-
theoretical reformulation. DMM News, 26. http://www.iasa-dmm.org/images/uploads/DMM%20%2322%20

Sept%2016%20English.pdf.
60 An exception is Simpson, J.A. & Belsky, J. (2016) Attachment theory within a modern evolutionary frame-
work. In J. Cassidy & P.R. Shaver (2018) Handbook of Attachment: Theory, Research, and Clinical Applications, 3rd
edn (pp.91–116). New York: Guilford. However, Simpson and Belsky themselves readily acknowledge that attach-
ment research has been slow to absorb developments in modern evolutionary theory. This issue is also discussed
in Granqvist, P. (2020) Attachment, culture, and gene-culture co-evolution: expanding the evolutionary toolbox.
Attachment & Human Development, 2 January, 1–24.
61 Fonagy, P. & Campbell, C. (2015) Bad blood revisited: attachment and psychoanalysis, 2015. British Journal of
Psychotherapy, 31(2), 229–50, p.230.

62 See also Fonagy, P. (1999) Points of contact and divergence between psychoanalytic and attachment the-
ories: is psychoanalytic theory truly different. Psychoanalytic Inquiry, 19(4), 448–80: ‘Attachment theory, far closer
to empirical psychology with its positivist heritage, has been in some ways method-bound over the past 15 years.
Its scope was determined less by what fell within the domain defined by relationship phenomena involving a
caretaking-dependent dyad and more by the range of groups and behaviors to which the preferred mode of ob-
servation, the strange situation, the adult attachment interview, and so forth, could be productively applied. This
sheltered the theory from a range of ideas’ (472).
63 Fonagy, P., Gergely, G., & Target, M. (2007) The parent–infant dyad and the construction of the subjective self.
Journal of Child Psychology and Psychiatry, 48(3–4), 288–328, p.288.

something of the air of a challenge, more than simply a statement of beliefs about the future.
Some similar points about the limitations of attachment research have been made by Pehr
Granqvist, who sees them as areas for renewal and revision of the paradigm, rather than
as evidence of exhaustion.64 Though not explicitly mentioned by Fonagy and Campbell or
by Granqvist, a factor in the wider context of their appraisal of the state of attachment as a
paradigm is the retirement of the second generation of attachment researchers. The infra-
structure built by this generation has been inherited in recent years by a new generation
of research leaders.65 Some examples, among many, in the developmental tradition include
Sheri Madigan, Pasco Fearon, Carlo Schuengel, Chantal Cyr, and Glenn Roisman; in the
social psychological tradition Chris Fraley, Omri Gillath, and Gurit Birnbaum. Granqvist
pursues work in both developmental and social psychology, making use of each set of at-
tachment measures. Patrick Luyten and others represent a younger cohort of researchers on

mentalisation who likewise draw from both traditions.
If a new generation of attachment researchers can be identified, this is not defined by their
chronological age. Nor, besides Roisman’s work at times,66 is there a sense of insurrection
against an older generation. Rather the point is that a bundle of problems that could previ-
ously be ignored or postponed are coming due, at a point where long-time research leaders
have retired, and the field is seeing a changing of the guard. As such, a third generation of
attachment researchers have been constituted by the playing out of problems in theory and
method at a time when a generation of researchers who had direct contact with Ainsworth
are putting down their tools. This third generation has also been shaped by active efforts to
guide its development: in the developmental tradition, many of the third generation have
received mentorship from van IJzendoorn, who has encouraged cross-group, international
collaboration in meta-analytic research and the adoption of new methodologies from medi-
cine and biostatistics.67 Shaver and Mikulincer have likewise played a concerted role in men-
toring a younger generation of rather ingenious and highly skilled social psychologists.
The first question the new generation of laboratory heads face is what must be preserved,
altered, or rejected from the legacy of their teachers in responding to the field’s challenges
and opportunities. This is the challenge pointed to by Fonagy and Campbell’s predictions of
the supersession of attachment theory and methods. Inheritance is never a given; it is always
64 Granqvist, P. (2020) Attachment, Religion, and Spirituality: A Wider View. New York: Guilford. Granqvist ad-
mits that, compared to the past, ‘I am less optimistic about the prospect of attachment theory . . . This is because
of attachment theory’s conceptual boundaries, its rudimentary defense mechanisms, and the attachment research
habit of “cross-tabulating” people into types (secure versus insecure and organized versus disorganized), despite
no individual being reducible to a type. Although these features have indisputably contributed to attachment
theory’s prosperity as an empirical research program, the attachment framework remains somewhat schematic
and impoverished.’ Unlike Fonagy and colleagues, Granqvist is firmly of the view, however, that these matters
can be resolved through innovations within attachment theory and research, and especially from renewed cross-
fertilisation with developments in evolutionary theory.
65 On the concept of infrastructures see Star, S.L. (1999) The ethnography of infrastructure. American Behavioral
Scientist, 43, 377–91; Berlant, L. (2016) The commons: infrastructures for troubling times. Environment and
Planning D: Society and Space, 34(3), 393–419.
66 E.g. Haltigan, J.D. & Roisman, G.I. (2015) Infant attachment insecurity and dissociative symptom-
atology: findings from the NICHD Study of Early Child Care and Youth Development. Infant Mental Health
Journal, 36(1), 30–41.
67 Van IJzendoorn is perhaps best regarded as generation 2.5. He learnt the Strange Situation from Brian Vaughn
and Mary Main and first met Ainsworth only at an AAI training with Main in the late 1980s. As well as his ten-
dency to be an early adopter of new technologies and quantitative methodologies, some ‘third-generation’ char-
acteristics to van IJzendoorn’s work may stem from the closeness of his collaboration with his former students
Marian Bakermans-Kranenburg and Femmie Juffer. He has also mentored and in turn learnt from collaborations
with Madigan and Fearon.
556 Conclusion
a task of filtering and sorting, of deciding between alternative pasts and the different futures
they might make possible. Inheriting effectively is hard and takes courage. In a sense, the
third generation of leaders of attachment research need to figure out together what genre of
story they are in. Is it a swashbuckler? Fonagy and Campbell seem willing to battle on deck.
Is it a dystopian story of complicity with oppressive state power, as critics such as White and
colleagues suggest? Is it a mystery, with new questions, methods, or interdisciplinary collab-
orations drawing the paradigm back towards the stages of construction and formation?
Will the story be an elegy, in which the present is oriented only towards the achievements
of the past? It looks not. So far, the early chapters are of a coming-of-age story. Ainsworth’s
immediate students remained loyal to the four-category system. Yet as we saw in Chapter 4,
renegotiation of this position seems to have a symbolic role, as a banner marking the pres-
ence of some new leaders, who now hold the datasets and the research agenda.68 There are

other changes too, coincident with the retirement of the second generation of attachment
researchers. One is particular attention to moderators of stability or intergenerational trans-
mission, including predictable forms of movement or cross-transmission between patterns
of attachment. The third generation also seems more able to cross-pollinate methods and
ideas between the developmental and social psychology traditions of attachment research.
In particular, the psychometric rigour and priorities of the social psychological tradition
have been adopted by younger researchers in the developmental tradition. However, there
have also been a variety of other migrations. One of special interest is that the model of min-
imising and maximising strategies as elaborated by Shaver and Mikulincer has frequently
been re-imported back into developmental psychology—though, at least to date, the wider
behavioural systems theory of Shaver and Mukulincer, and their measures of minimising
and maximising strategies in the domains of sexuality, caregiving, and dominance, have not
been taken up by developmentalists.
In an attempt at stock-taking the current state of attachment research on children,
Schuengel and colleagues reported from a bibliometric study. The results suggest that the past
few years represent a period of transition. For instance, citations of Patterns of Attachment
have been in decline since 2015, whereas citations of meta-analyses have gone from strength
to strength.69 Among meta-analytic research on child attachment Schuengel found a com-
parative decline in focus on the concerns of Ainsworth and Main. He found a comparative
increasing focus on mental health symptoms and supportive interventions with families, a
growing concern with fathers, and rising attention to anxiety and avoidance conceptualised
as the latent dimensions of individual differences in attachment.
Even if there are definitely some pulls towards ‘exhaustion’ of the developmental tradition,
Schuengel’s findings indicate that, in general, attachment research is changing, not fading.
There are significant strengths that mark its ongoing vitality. Three developments can be
used as illustrations, though by no means are intended as exhaustive. These developments
are: Individual Participant Data meta-analysis; security priming; and intervention research.
68 See e.g. Raby, L., Verhage, M.L., Fearon, R.P.M., et al. (2019) The latent structure of the Adult Attachment
Interview: large sample evidence from the Collaboration on Attachment Transmission Synthesis. Paper presented
at Biennial Meeting of the Society for Research in Child Development, Baltimore, 21–23 March. On the symbolic
work demanded in the constitution of a generation see Eyerman, R. & Turner, B.S. (1998) Outline of a theory of
generations. European Journal of Social Theory, 1, 91–106; Purhonen, S. (2016) Generations on paper: Bourdieu
and the critique of ‘generationalism’. Social Science Information, 55(1), 94–114.
69 Schuengel, C., Verhage, M., & Duschinsky, R. (2019) Representing attachment through meta-analyses: a
move to the level of collaboration, manuscript under review. https://psyarxiv.com/anf6t/.

Individual Participant Data meta-analysis 557
Many other lively areas of research have been discussed in previous chapters. Two of par-
ticular note are the growth of attention to secure base scripts and observations of secure
base/safe haven provision—with a cadre of talented younger researchers currently pursuing
a programme of work very much in the ‘construction’ stage (Chapter 2). Other lively areas of
research build from later research groups, to be considered in a later volume, and so would
not make sense to discuss here. This includes extensive research using biomedical measures,
in acknowledgement of the embodied aspects and underpinnings of attachment relation-
ships. This was touched on in Chapter 4, discussing the work of Dante Cicchetti. In making
a selection, Individual Participant Data meta-analysis, security priming, and intervention
research have been chosen as illustrations because they capitalise precisely on the maturity
of the field’s theory and methods discussed in this book, whilst also reflecting the changing
priorities and methodological infrastructure of attachment research as an area of science

inquiry.
Individual Participant Data meta-analysis
A first area of particular vitality in contemporary attachment research has made use pre-
cisely of the advantages of ‘saturation’. Over the decades, attachment research on young
children has used the labour-intensive Strange Situation or measures adapted from it like
the MacArthur preschool system70 or the Preschool Assessment of Attachment.71 This has
permitted the accumulation of thousands of relatively small-scale studies that provide little
pieces of the puzzle. Each on its own is unable to offer confident guidance for interventions
with families, justify commissioning decisions, or inform policy to protect and support
young children. Attachment researchers, particularly van IJzendoorn and colleagues, were
early adopters of meta-analysis as a methodology for studying the overall effect sizes iden-
tified by these numerous small studies and for examining factors accounting for variance in
effect sizes between studies.72 There were also attempts to pool data from different studies
more directly in order to increase statistical power and explore potential moderating fac-
tors.73 This offered a richer array of variables and greater depth than reliance on the pub-
lished record as in traditional meta-analysis. Data pooling permits investigations that are
wider than single studies and deeper than traditional meta-analyses. However, data pooling
efforts in attachment research foundered in the 1990s, in part because the statistical tech-
niques for multilevel modelling were not available to developmental psychologists. Instead,
the use of traditional meta-analyses flourished.74
70 Cassidy, J. & Marvin, R., with the MacArthur Network on Attachment in the Preschool Years (1992)
Attachment organisation in pre-school children: procedures and coding manual. Unpublished manuscript.
71 Crittenden, P.M. (1981) The Pre-School Assessment of Attachment Coding Manual. Miami: Family Relations
Institute.
72 Van IJzendoorn, M.H., Goldberg, S., Kroonenberg, P.M., & Frenkel, O.J. (1992) The relative effects of ma-
ternal and child problems on the quality of attachment: a meta-analysis of attachment in clinical samples. Child
Development, 63(4), 840–58. See also van IJzendoorn, M.H., Bakermans-Kranenburg, M.J., & Alink, L.R.A. (2011)
Meta-analysis in developmental science. In B. Laursen, T.D. Little, & N.A. Card (eds) Handbook of Developmental
Research Methods (pp.667–86). New York: Guilford.
73 Lamb, M.E., Sternberg, K.J., & Prodromidis, M. (1992) Nonmaternal care and the security of infant mother
attachment: a reanalysis of the data. Infant Behavior & Development, 15(1), 71–83.
74 Another factor may have been that attitudes among researchers were not yet favourable to data sharing—a
situation that has, by degrees, shifted in recent years. Tenopir, C., Dalton, E.D., Allard, S., et al. (2015) Changes in
data sharing and data reuse practices and perceptions among scientists worldwide. PLoS One, 10(8), e0134826.
558 Conclusion
One recent such meta-analysis was completed by Verhage and the Collaboration
on Attachment Transmission Synthesis in 2016.75 The researchers examined the asso-
ciation between AAI for parents and Strange Situation classifications for dyads from
95 samples, with 4,819 participants in total. They found that the association between
secure-autonomous state of mind on the AAI and secure attachment in the Strange
Situation was r = .31 and for unresolved state of mind to disorganised infant attach-
ment r = .21. This was very substantially down from the meta-analysis conducted by van
IJzendoorn in 1995 (r = .47 and r = .31, respectively).76 Caregivers with dismissing states
of mind were more likely to have avoidant attachment relationships and less likely to
have secure attachment relationships, but they were no less likely than other caregivers
to be part of dyads receiving an ambivalent/resistant classification. Similarly, preoccu-
pied states of mind were associated with more ambivalent/resistant and fewer secure,

but not with fewer avoidant attachment relationships. This suggests that classical and
widely repeated assumptions since Main and colleagues regarding the mechanisms that
lead to avoidant and resistant attachment might need additional thought, and especially
the underarticulated concept of ‘inconsistent care’ inherited from Ainsworth. Such a
conclusion aligns with proposals by diverse researchers, proposing ways that dismiss-
ing states of mind may also contribute to resistance, and preoccupied states of mind to
avoidant attachment relationships.77
Another curious finding was that speakers with unresolved/disorganised states of mind
regarding attachment were less likely to have secure and avoidant attachment relationships
with their infant, but the dyads were no less likely to receive an ambivalent/resistant classi-
fication. Verhage and colleagues observed that this finding has pertinence to debates about
how hard or how fluid the distinction is between resistance and disorganised attachment
(Chapter 4). Around 25% of the relationship between AAI and Strange Situation classifica-
tions could be accounted for on the basis of assessments of caregiver sensitivity; after correc-
tion for test–retest reliability, this left a little under half of variance explained by sensitivity.78
Examining moderators, Verhage and colleagues reported that the association between the
AAI and the Strange Situation was stronger in low-risk samples and weaker in at-risk sam-
ples. At first sight, the decline in the association since van IJzendoorn’s meta-analysis in 1995
could be thought to reflect the fact that early samples were generally with low-risk sam-
ples and subsequent samples have more frequently been with high-risk samples. However,
Verhage and colleagues found that, in fact, risk status did not account for the effect of publi-
cation year on the reported effect size.
75 Verhage, M., Schuengel, C., Madigan, S., et al. (2016) Narrowing the transmission gap. Psychological Bulletin,
142(4), 337–66.
76 Van IJzendoorn, M.H. (1995) Adult attachment representations, parental responsiveness, and infant
attachment—a meta-analysis on the predictive validity of the Adult Attachment Interview. Psychological Bulletin,
117(3), 387–403.
77 Crittenden, P.M., Partridge, M.F., & Claussen, A.H. (1991) Family patterns of relationship in normative and
dysfunctional families. Development & Psychopathology, 3(4), 491–512; Shah, P.E., Fonagy, P., & Strathearn, L.
(2010) Is attachment transmitted across generations? The plot thickens. Clinical Child Psychology and Psychiatry,
15(3), 329–45; Kondo-Ikemura, K., Behrens, K., Umemura, T., & Nakano, S. (2018) Japanese mothers’ prebirth
Adult Attachment Interview predicts their infants’ response to the Strange Situation procedure. Developmental
Psychology, 54(11), 2007–2015.
78 See also Bernier, A., Matte-Gagné, C., Bélanger, M.-È., & Whipple, N. (2014) Taking stock of two decades of
attachment transmission gap: broadening the assessment of maternal behavior. Child Development, 85(5), 1852–
65; van IJzendoorn, M.H. & Bakermans-Kranenburg, M.J. (2019) Bridges across the intergenerational transmis-
sion of attachment gap. Current Opinion in Psychology, 25, 31–6.
Individual Participant Data meta-analysis 559
Subsequently, the Collaboration on Attachment Transmission Synthesis has pursued

Individual Participant Data meta-analysis, in which data are pooled from studies using the
same measures. The Collaboration on Attachment Transmission Synthesis received data
from 58 research laboratories from around the world that had used the Strange Situation
and the AAI (4,396 parent–child dyads); 67% of eligible studies contributed their data.
The first analyses from this collaboration are just appearing, and already are offering im-
portant qualifications to existing theory and new insights, for instance for understanding
how poverty and social adversity impact parenting. Adding to their previous traditional
meta-analysis, Verhage and colleagues found that risk factors lowered the likelihood of a
secure child–caregiver attachment if the caregiver has a secure-autonomous state of mind
regarding attachment. The finding suggests that secure-autonomous states of mind do not
offer as much resilience as previously expected by attachment researchers against contextual

risks.79 However, low education and single parenthood had no effect on the strength of the
association between the AAI and the Strange Situation. Neither did the gender of the care-
giver. In a further study, Verhage found that contextual risk factors weaken the association
between caregiver sensitivity and attachment, again illustrating the importance of the chal-
lenges faced by families in the development of attachment relationships.80
Another study by the Collaboration on Attachment Transmission Synthesis was led by
Lee Raby and sought to examine the psychometric structure of the AAI. As we have seen,
Main, Goldwyn, and Hesse initially offered a four-category model (Chapter 3). However,
Roisman, Fraley, and colleagues have subsequently proposed instead a two-dimensional
model modelled on the ECR, with avoidance and anxiety as latent dimensions (Chapter 4).
Raby and colleagues found that the distribution of the data-points suggested that individual
differences in attachment states of mind reflect differences in degree, not kind. However,
the findings were not clear regarding what dimensions should be used. A two-factor model
based on dismissing and preoccupation was found to be an adequate fit. But a three-factor
model based on dismissing, preoccupation, and lack of resolution was also found to be an
adequate fit. Raby and colleagues argued in favour of the two-factor model on the basis of
parsimony.81 However, they did not examine which model might best serve to predict vari-
ables of interest such as Strange Situation classifications or other indices of child mental
health. It is likely that predictive validity will be a key criterion for any potential overhaul of
the dominant approach to reporting results from attachment research.
Over the coming years, it is possible that the social psychological tradition will adopt
Individual Participant Data meta-analysis as a methodology. Pooled data would permit ex-
ploration of some of the issues faced by this tradition, such as the status of security and the
implications of trauma. Individual Participant Data meta-analysis, much like the present
book, is oriented by a sense that the past can open anew in response to curation and the
questions of the present. It may be that many social psychology laboratories will have thrown
79 Verhage, M.L., Fearon, R.P., Schuengel, C., et al. (2018) Examining ecological constraints on the intergen-
erational transmission of attachment via Individual Participant Data meta-analysis. Child Development, 89(6),
2023–37.
80 Verhage, M.L., Fearon, R.P., Schuengel, C., et al. (2019) Does risk background affect intergenerational trans-
mission of attachment? Testing a moderated mediation model with IPD. Paper presented at Biennial Meeting of
the Society for Research in Child Development, Baltimore.
81 Raby, L., Verhage, M.L., Fearon, R.P.M., et al. (2019) The latent structure of the Adult Attachment
Interview: large sample evidence from the Collaboration on Attachment Transmission Synthesis. Paper presented
at Biennial Meeting of the Society for Research in Child Development, Baltimore, 21–23 March.
560 Conclusion
away their scale scores. That has certainly been the case for some research groups in the
developmental attachment tradition: a third of eligible laboratories did not contribute data
to the Collaboration on Attachment Transmission Synthesis. But enough may have been
preserved to make application of Individual Participant Data meta-analysis a feasible and
fruitful possibility for the social psychological tradition and for answering other questions
in the developmental tradition. Work is currently underway, for example, to assemble an
Individual Participant Dataset for randomised control trials of attachment-based interven-
tions. The ambition is to ask questions with greater power and precision than have been
possible before, about what works for whom, and how, in supportive interventions with
families.82
Security priming

A second area in which attachment research seems to be thriving and breaking new ground
is in studies of priming and attachment-relevant states—as opposed to longer-term attach-
ment patterns or styles. The concept of ‘attachment states’ originated in the developmental
tradition of attachment research, introduced by Fonagy, Steele, and Steele in 1991. Fonagy
and colleagues argued that Bowlby’s ‘internal working model’ construct may well not be a
unitary whole. Instead it likely encompasses some components that are relatively unrespon-
sive to environmental cues and other components that are relatively responsive to specific
cues.83 This distinction was accepted by other researchers in the developmental tradition of
attachment research.84 However, as discussed in Chapter 2, researchers in the developmental
tradition have tended often, in practice if not in theory, to treat attachment as a trait. In line
with this, state attachment fluctuations have not been much studied in childhood by devel-
opmentalists, since attachment has generally been treated as relatively stable.85
The differentiation between global adult attachment style and more particular states was
developed strongly by the social psychology tradition, especially by Mark Baldwin.86 An im-
portant early study was reported by Baldwin and colleagues in 1996. Participants were asked
to characterise their most significant relationships and then to indicate which attachment
style best captured their feelings about each relationship. Baldwin and colleagues found that
most participants reported having experienced secure, avoidant, and anxious experiences
82 Schuengel, C., Verhage, M., & Duschinsky, R. (2019) Representing attachment through meta-analyses: a
move to the level of collaboration. Manuscript under review. https://psyarxiv.com/anf6t/.

83 Fonagy, P., Steele, H., & Steele, M. (1991) Maternal representations of attachment during pregnancy predict
the organization of infant–mother attachment at one year of age. Child Development, 62(5), 891–905, p.902.
84 E.g. Thompson, R.A. & Raikes, H.A. (2003) Toward the next quarter-century: conceptual and methodological
challenges for attachment theory. Development & Psychopathology, 15, 691–718.

85 Attention to state attachment variation in childhood has been a concern of Guy Bosmans, whose work
has increasingly spanned the divide between developmental and social traditions of attachment research, e.g.
Bosmans, G., Van de Walle, M., Goossens, L., & Ceulemans, E. (2014) (In)variability of attachment in middle
childhood: secure base script evidence in diary data. Behavior Change, 31, 225–42; Vandevivere, E., Bosmans, G.,
Roels, S., Dujardin, A., & Braet, C. (2018) State trust in middle childhood: an experimental manipulation of ma-
ternal support. Journal of Child and Family Studies, 27(4), 1252–63.
86 E.g. Baldwin, M.W., Carrell, S.E., & Lopez, D.F. (1990) Priming relationship schemas: my advisor and the Pope
are watching me from the back of my mind. Journal of Experimental Social Psychology, 26(5), 435–54. Attachment
styles have been conceptualised by researchers in the social psychological tradition as having trait-like and state-
like qualities, e.g. Fraley, R.C. & Roberts, B.W. (2005) Patterns of continuity: a dynamic model for conceptual-
izing the stability of individual differences in psychological constructs across the life course. Psychological Review,
112(1), 60–74.
Security priming 561
across their relationships. Despite this diversity, participants reported more relationships
that fitted with their global attachment style. Baldwin and colleagues also found that this
global attachment style was influential for shaping self-report of dating preferences. Priming
secure, avoidant, and anxious experiences did not influence self-report global attachment
style. But the researchers did find that priming altered reported dating preferences among
their participants, for instance with a prime for avoidance decreasing the attractiveness of a
partner offering an intimate and committed relationship.87
The findings suggested to Baldwin and colleagues that a person’s global attachment style
is generally more accessible for informing semantic appraisals. However, other information
about attachment experiences, which could inform alternative attachment styles, is poten-
tially available in the hierarchical network, and specific cues in the environment may in-
fluence this availability. The qualities of an avoidant relationship, for example, may make

available insecure generalised schemas even for someone with a secure attachment style.
Following up on the work of Baldwin and colleagues, Mikulincer and Arad published a
vignette-based study in 1999 of the effects of attachment style and security priming on ac-
cessibility for participants of incongruent information about their partner.88 They found that
a secure attachment style was associated with greater openness to and better recall of in-
congruent information when it suggested that their partner was caring and available. And
asking participants to think about a time when others were available to them for support and
comfort—a security prime—made an independent additional contribution to openness to
and recall of incongruent positive information. Attachment style did not moderate the ef-
fect of the prime. Mikulincer and Arad concluded that a security prime led to behaviour that
resembled a secure attachment style by activating relevant parts of the hierarchical network
of associations. However, curiously, this activation of the ‘attachment state’ appeared to op-
erate through an independent process to the role of global attachment style in influencing
behaviour, as there was no interaction between attachment style and the security prime in
predicting behaviour.89
Over the subsequent decades, Mikulincer, Shaver, and colleagues in the social psychology
tradition have regularly found that similar effects can be obtained on the basis of a secure
attachment style or on the basis of a prime for security.90 Furthermore, they have demon-
strated that the effect of the secure base prime cannot be accounted for in terms of variations
in mood or self-esteem.91 Part of what has been electrifying about the new line of research
87 Baldwin, M.W., Keelan, J.P.R., Fehr, B., Enns, V., & Koh-Rangarajoo, E. (1996) Social-cognitive conceptu-
alization of attachment working models: availability and accessibility effects. Journal of Personality and Social
Psychology, 71(1), 94–109.
88 Mikulincer, M. & Arad, D. (1999) Attachment working models and cognitive openness in close relation-
ships: a test of chronic and temporary accessibility effects. Journal of Personality and Social Psychology, 77, 710–25.
89 In a recent study, Hudson and Fraley reported intriguing findings that repeatedly priming attachment anx-
iety over time contributed to reduced anxiety on the ECR over time, no less than repeated priming attachment
security. The researchers concluded that repeated priming of attachment anxiety or attachment security offered
participants the opportunity to reflect on their overall network of associations, contributing to change in the global
attachment style. Hudson, N.W. & Fraley, R.C. (2018) Moving toward greater security: the effects of repeatedly
priming attachment security and anxiety. Journal of Research in Personality, 74, 147–57.
90 Mikulincer, M., Gillath, O., Halevy, V., Avihou, N., Avidan, S., & Eshkoli, N. (2001) Attachment theory and
reactions to others’ needs: evidence that activation of the sense of attachment security promotes empathic re-
sponses. Journal of Personality and Social Psychology, 81(6), 1205–224; Mikulincer, M., Shaver, P.R., Gillath, O., &
Nitzberg, R.A. (2005) Attachment, caregiving, and altruism: boosting attachment security increases compassion
and helping. Journal of Personality and Social Psychology, 89(5), 817–39.
91 Gillath, O., Hart, J., Noftle, E.E., & Stockdale, G.D. (2009) Development and validation of a state adult at-
tachment measure (SAAM). Journal of Research in Personality, 43(3), 362–73; Gillath, O. & Karantzas, G. (2019)
Attachment security priming: a systematic review. Current Opinion in Psychology, 25, 86–95.
562 Conclusion
on priming has been that, in many domains, a prime seems able to produce prosocial ef-
fects for all participants that would otherwise only be limited to those with a secure attach-
ment style. For instance, Mikulincer and Shaver found that a secure attachment style and a
security prime contributed to greater tolerance towards out-groups.92 Mikulincer, Shaver,
Gillath, and Nitzberg reported that secure attachment style and a security prime contributed
to greater willingness to help someone in need.93 And Shaver, Mikulincer, Lavy, and Cassidy
found that a secure attachment style and a subliminal security prime could contribute to less
defensive and more constructive responses to feeling hurt by a romantic partner.94
However, some research has documented an interaction between attachment style and
a security prime. One suggestive finding reported by Mikulincer, Hirschberger, Nachmias,
and Gillath was that attachment insecurity can moderate the effect of a security prime on
positive mood. If a security prime is followed by visualisation of an experience of separ-

ation from attachment figures, only participants with a secure attachment style reported an
increase in positive mood from the security prime. The effect disappeared for participants
with an avoidant or anxious attachment style.95 Another relevant finding by Mikulincer,
Shaver, Sahdra, and Bar-On was that a secure base prime could overcome mental depletion,
allowing a participant to remain responsive to their romantic partner during a laboratory-
based task.96 The prime also removed the negative effect of an avoidant attachment style on
the extent of help offered. By contrast, there was no interaction between attachment style and
the security prime in intention to help strangers.
Mikulincer and colleagues speculated that with any longstanding romantic partner
there will be a rich network of associations, more so than for a stranger. The secure base
prime prior to the couple interaction may have activated the ‘reservoir of positive feelings
and memories’ within the hierarchical network of associations, knocking out the mini-
mising strategy that might otherwise be enacted.97 Later research demonstrated, however,
that this effect was specific to help with practical rather than emotional tasks. Mikulincer,
Shaver, and colleagues observed participants responding to their romantic partner in two
conditions: discussing a distressing problem and exploring personal goals. An avoidant at-
tachment style was generally associated with lower supportiveness in both conditions. A se-
curity prime eradicated this effect in relation to the partner’s exploration of personal goals.
However, the security prime had no effect on the condition where the partner was discussing
a personal problem and needed emotional support.98
92 Mikulincer, M. & Shaver, P.R. (2001) Attachment theory and intergroup bias: evidence that priming the
secure base schema attenuates negative reactions to out-groups. Journal of Personality and Social Psychology, 81(1),
97–115.
93 Mikulincer, M., Shaver, P.R., Gillath, O., & Nitzberg, R.A. (2005) Attachment, caregiving, and al-
truism: boosting attachment security increases compassion and helping. Journal of Personality and Social
Psychology, 89(5), 817–39.
94 Shaver, P.R., Mikulincer, M., Lavy, S., & Cassidy, J. (2009) Understanding and altering hurt feelings: an
attachment-theoretical perspective on the generation and regulation of emotions. In A.L. Vangelisti (ed.) Feeling
Hurt in Close Relationships (pp.92–119). Cambridge: Cambridge University Press.
95 Mikulincer, M., Hirschberger, G., Nachmias, O., & Gillath, O. (2001) The affective component of the secure
base schema. Journal of Personality and Social Psychology, 81(2), 305–321, Study 7.
96 Mikulincer, M., Shaver, P.R., Sahdra, B.K., & Bar-On, N. (2013) Can security-enhancing interventions over-
come psychological barriers to responsiveness in couple relationships? Attachment & Human Development, 15(3),
246–60.
97 Ibid. p.257.
98 Mikulincer, M., Shaver, P.R., Bar-On, N., & Sahdra, B.K. (2014) Security enhancement, self-esteem threat,
and mental depletion affect provision of a safe haven and secure base to a romantic partner. Journal of Social and
Personal Relationships, 31(5), 630–50.
Security priming 563
Yet perhaps the most curious set of interactions between attachment style and priming
have been in relation to anxious attachment. Though, in general, security primes tend to
foster comfort, happiness, and more positivelyregarded behaviour, the interaction between
security priming and an anxious attachment style in several cases has had the opposite effect.
For example, Taubman-Ben-Ari and Mikulincer examined the effect of attachment style
and security priming on driving using a simulator. They found that an anxious attachment
style was associated with higher willingness to drive recklessly. Furthermore, whereas for
other participants a security prime reduced reckless driving, among participants with an
anxious attachment style the security prime increased recklessness.99 The researchers won-
dered whether the security prime was being interpreted by their participants as signalling
the availability of a secure base, contributing to overconfidence, without the feeling of being
cherished that would otherwise reduce willingness to endanger oneself. Mikulincer and col-

leagues suspected that rumination may be another factor involved in the paradoxical effects
of security priming on participants with an anxious attachment style. The security prime
may cause participants to think about the unavailability of their attachment figures. Support
for this conclusion came from a study by Mikulincer, Shaver, and Rom.100 The researchers
found that an explicit security prime (recalling experiences of being cared about), but not
a subliminal presentation of the names of attachment figures, was associated with less cre-
ativity and effectiveness in solving laboratory-based tasks for participants with an anxious
attachment style. It was concluded that an explicit security prime provides a prompt for ru-
mination, in contrast to a subliminal prime.
In opening priming effects as a domain of inquiry, the social psychology tradition has
allowed the attachment paradigm as a whole to put out new branches. It has taken some
time, but researchers in the developmental tradition have begun to explore priming tech-
niques, such as in the collaboration between Bosmans, van IJzendoorn, and Bakermans-
Kranenburg.101 One obstacle is that the AAI is not especially well adapted for repeated
administration to assess attachment states or validate the effects of a prime. But neither is
the ECR. The State Adult Attachment Measure has been developed by Gillath and colleagues
(Chapter 5) to circumvent this issue in the social psychological tradition. There is likewise
no intrinsic obstacle to the use of priming for researchers in the developmental tradition.
For instance, repeated observation of caregiver–child interaction in free-play situations (not
the Strange Situation) could be pursued by developmental researchers to examine the effects
of a prime. The recent development of briefer assessments for the study of caregiver–child
interaction, such as the brief Attachment Q-Sort or the AMBIANCE-Brief, may facilitate
this by reducing the labour required to code observational data.102 There remain some con-
cerns about research on attachment priming, since, in general, studies of priming in social
99 Taubman-Ben-Ari, O. & Mikulincer, M. (2007) The effects of dispositional attachment orientations and con-
textual priming of attachment security on reckless driving. Transportation Research Part F, 10, 123–38.
100 Mikulincer, M., Shaver, P.R., & Rom, E. (2011) The effects of implicit and explicit security priming on creative
problem solving. Cognition and Emotion, 25(3), 519–31.

101 E.g. Verhees, M.W., Ceulemans, E., Bakermans-Kranenburg, M.J., van IJzendoorn, M.H., De Winter, S., &
Bosmans, G. (2017) The effects of cognitive bias modification training and oxytocin administration on trust in ma-
ternal support: study protocol for a randomized controlled trial. Trials, 18(1), 326. Granqvist has made extensive
use of priming in his research, though in studies using measures from the social psychological tradition, and not in
his research using measures from the developmental tradition.
102 E.g. Haltigan, J.D., Madigan, S., Bronfman, E., et al. (2019) Refining the assessment of disrupted maternal
communication: using item response models to identify central indicators of disrupted behavior. Development &
Psychopathology, 31(1), 261–77; Cadman, T., Belsky, J., & Fearon, R.M.P. (2018) The Brief Attachment Scale (BAS-
16): a short measure of infant attachment. Child: Care, Health and Development, 44(5), 766–75.
564 Conclusion
psychology have an uneven record of replication.103 Nevertheless, so far, results regarding

attachment priming specifically have been encouraging.
Intervention research
A third development has been the growing prominence of intervention research both for the
sake of clinical relevance and in order to articulate potential causal mechanisms. Description
of a variety of attachment-based interventions for families with children of different ages is
offered in the recent Handbook of Attachment-Based Interventions, edited by Howard and
Miriam Steele.104 A concern with interventions among attachment researchers is, of course,
far from new. Attachment theory was initially developed in a clinical context by Bowlby, and

many of Ainsworth’s graduate students pursued training as clinicians. The development of
supportive interventions with parents based directly on attachment principles and evaluated
using the Strange Situation began in the 1980s. The first such interventions were STEEP and,
though perhaps more grounded in psychoanalytic than attachment principles, Lieberman’s
child–parent psychotherapy (Chapter 4). The Circle of Security intervention, discussed in
previous chapters, emerged out of a determinate attempt to translate the central principles of
Ainsworth’s concept of sensitivity into a supportive intervention with families.
Over time, the evidence base has grown for interventions designed with attachment prin-
ciples in mind and evaluated with attachment measures. A systematic review of this literature
by Mohamed and colleagues identified 32 empirical studies of attachment-based interven-
tions. The most frequent elements of the interventions were psychoeducation, increasing
parents’ awareness of the functioning of the attachment system, and supporting the parent
relationship. Other elements, more specific to some interventions, included the intervener
‘subtitling’ the child’s behaviours by indicating what the behaviours seem to be saying, video-
feedback, and supporting the development of caregiver insight into their own behaviour.105
Among this growing literature, two attachment-based interventions for families with young
children have generated both considerable research evidence and clinical interest: Video-
feedback Intervention to promote Positive Parenting (VIPP) and the Attachment and
Biobehavioural Catchup (ABC). Both have had significant, and growing, penetration into
child health and welfare services.
Video-feedback Intervention to promote Positive Parenting
VIPP was developed by Femmie Juffer, Marian Bakermans-Kranenburg, and Marinus van
IJzendoorn. It was first used in a randomised control trial in a sample of 130 families with
infants who had been internationally adopted in their first weeks of life. When the children
were 9–12 months, home visits were undertaken, during which sessions of video-feedback
were used to support parenting sensitivity. Whereas video-based instruction to parents had
103 Locke, E.A. (2015) Theory building, replication, and behavioral priming: where do we need to go from here?
Perspectives on Psychological Science, 10(3), 408–414.

104 Steele, H. & Steele, M. (eds) (2018) Handbook of Attachment-Based Interventions. New York: Guilford.
105 Mohamed, A.R, Sterkenburg, P., van Rensburg, E., & Schuengel, C. (2019) The development of a coding
system to examine the effective elements of attachment-based interventions. Paper presented at International
Intervention research 565
proven ineffective or even counterproductive,106 it was anticipated that drawing on 10-to

30-minute clips of the parent’s own behaviour with the child in ordinary situations would be
more relatable and empowering.
In the intervention, the intervener discusses these film clips with the caregiver, helping
the caregiver to consider the baby’s signals, their meaning, and how the baby responds to
the caregiver’s behaviours, especially when these show sensitivity. In a strategy inherited
from STEEP, as well as Fraiberg’s approach to therapeutic work,107 the intervener interprets
the baby’s behaviours as if giving them subtitles, translating the infant signals into a verbal
form that the caregiver may find easier to reflect upon and consider. Positive film fragments
are emphasised; the intervention builds precisely on parents’ own expertise, with caregivers
serving as a reinforcing role model for themselves.108 An illustration is offered by Juffer and
Bakermans-Kranenburg from a detailed case study:

Using the recordings, she pointed out how happy and proud Ava reacted to Noah’s compli-
ments and how they were peacefully playing together afterward. The intervener explained
that research has shown that it works much better to praise your child and give compli-
ments when she does things well than to punish her when things go wrong. Sometimes
it is even better to ignore naughty or difficult behavior because that way your child is not
receiving attention for that type of behavior. The intervener summarized this message by
saying: ‘Because you are so important for Ava, she really loves to have your attention and
compliments, and that is why it works so well to praise Ava when you want her to listen to
you or be compliant.’109
Juffer, Bakermans-Kranenburg, and van IJzendoorn anticipated that, compared to STEEP, a

shorter intervention focused on specific behaviours would prove more manageable for par-
ents and more targeted to supporting change in attachment-relevant processes.110 VIPP is
conventionally six to eight sessions, which makes it short enough even for use in assessing
caregiving capacity, as well as directly for parenting support.111 The first session of VIPP
focuses on exploration and attachment behaviour; the second session helps the caregiver
consider the meaning of the infant’s signals; the third session addresses the role of prompt
and adequate response to these signals; and the fourth session emphasises shared emotions
and affective attunement with the child. Two additional sessions are often used to review
106 Lambermon, M.W. & van IJzendoorn, M.H. (1989) Influencing mother–infant interaction through video-
taped or written instruction: evaluation of a parent education program. Early Childhood Research Quarterly, 4(4),
449–58.
107 Carter, S.L., Osofsky, J.D., & Hann, D.M. (1991) Speaking for the baby: a therapeutic intervention with ado-
lescent mothers and their infants. Infant Mental Health Journal, 12(4), 291–301.
108 Juffer, F. & Steele, M. (2014) What words cannot say: the telling story of video in attachment-based interven-
tions. Attachment & Human Development, 16(4), 307–314, p.311.

109 Juffer, F. & Bakermans- Kranenburg, M.J. (2018) Working with Video-feedback Intervention to pro-
mote Positive Parenting and Sensitive Discipline (VIPP-SD): a case study. Journal of Clinical Psychology, 74(8),
1346–57, p.7.
110 Van IJzendoorn, M.H., Juffer, F., & Duyvesteyn, M.G. (1995) Breaking the intergenerational cycle of insecure
attachment: a review of the effects of attachment-based interventions on maternal sensitivity and infant security.
Journal of Child Psychology and Psychiatry, 36(2), 225–48; Bakermans-Kranenburg, M.J., van IJzendoorn, M.H.,
& Juffer, F. (2003) Less is more: meta-analyses of sensitivity and attachment interventions in early childhood.
Psychological Bulletin, 129(2), 195–215.
111 Bakermans-Kranenburg, M.J., Juffer, F., & van IJzendoorn, M.H. (2019) Reflections on the mirror: on video-
feedback to promote positive parenting and infant mental health. In C. Zeanah (ed.) Handbook of Infant Mental
Health, 4th edn (pp.527–42). New York: Guilford.
566 Conclusion
the information and feedback from the first four.112 Issues around sensitive discipline of the
child are also addressed, helping the caregiver provide non-coercive discipline and positive
reinforcement, to de-escalate tantrums, and achieve empathy and consistent limit-setting
for the child.
In a first evaluation of VIPP with 130 families, the researchers found that compared to the
control group receiving treatment as usual, the intervention enhanced sensitivity as meas-
ured by Ainsworth’s scale. The dyads were also less likely to receive a disorganised attach-
ment classification on the basis of the child’s behaviour in the Strange Situation procedure.113
In a later trial with 237 families with toddlers with externalising behaviour problems, the
mothers who completed VIPP showed more sensitive discipline, and their children showed
fewer conduct problems at a later follow-up.114 A process evaluation revealed that neither
mothers nor home visitors were able to anticipate the relative effectiveness of the interven-

tion based on their first impressions of one another.115 Whereas advocates for STEEP had
argued that behavioural-focused targeted interventions would be less effective in families fa-
cing multiple adversities and daily hassles (Chapter 4), in fact VIPP was found to be more ef-
fective in such circumstances.116 A meta-analysis of the first twelve randomised control trials
of VIPP found an effect size of d = .47 for sensitivity, d = .36 for infant–caregiver attachment
classifications, and d = .26 for reduced child behaviour problems, with no decrease in effect
size resulting from length of follow-up.117 Across multiple samples, VIPP has been found to
decrease disorganised attachment as assessed in the Strange Situation. However, why this is
112 Juffer, F., Bakermans- Kranenburg, M.J., & van IJzendoorn, M.H. (eds) (2008) Promoting Positive
Parenting: An Attachment-Based Intervention. New York: Psychology Press.
113 Juffer, F., Bakermans-Kranenburg, M.J., & van IJzendoorn, M.H. (2005) The importance of parenting in the
development of disorganized attachment: evidence from a preventive intervention study in adoptive families.
Journal of Child Psychology and Psychiatry, 46(3), 263–74.
114 Klein Velderman, M., Bakermans-Kranenburg, M.J., Juffer, F., Van Ijzendoorn, M.H., Mangelsdorf, S.C., &
Zevalkink, J. (2006) Preventing preschool externalizing behavior problems through video-feedback intervention
in infancy. Infant Mental Health Journal, 27(5), 466–93. However, both maternal discipline practices and child be-
havioural problems seem to have been independently affected by the intervention. Analysis did not suggest effects
of maternal discipline on child behaviour. See Mesman, J., Stoel, R., Bakermans-Kranenburg, M.J., et al. (2009)
Predicting growth curves of early childhood externalizing problems: differential susceptibility of children with
difficult temperament. Journal of Abnormal Child Psychology, 37(5), 625–36, p.633.
115 Stolk, M.N., Mesman, J., van Zeijl, J., et al. (2008) Early parenting intervention aimed at maternal sensitivity
and discipline: a process evaluation. Journal of Community Psychology, 36(6), 780–97.

116 Van Zeijl, J., Mesman, J., Van IJzendoorn, M.H., et al. (2006) Attachment-based intervention for enhancing
sensitive discipline in mothers of 1 to 3 year-old children at risk for externalizing behavior problems: a randomized
controlled trial. Journal of Consulting and Clinical Psychology, 74(6), 994–1005. Subsequent research has compli-
cated this picture: Euser, S., Alink, L.R., Stoltenborgh, M., Bakermans-Kranenburg, M.J., & van IJzendoorn, M.H.
(2015) A gloomy picture: a meta-analysis of randomized controlled trials reveals disappointing effectiveness of
programs aiming at preventing child maltreatment. BMC Public Health, 15(1), 1068. ‘The “less is more” effect in
attachment-based interventions found by Bakermans-Kranenburg and colleagues seems only partly applicable to
programs aimed at reducing or preventing child maltreatment. We found a curvilinear association with program
duration and number of program sessions. Programs with a moderate duration (6–12 months) or a moderate
number of sessions (16–30) yielded significantly higher effect sizes compared to shorter or longer programs and
programs with fewer or more sessions’ (11). Additionally, initial research suggests that parent learning disabilities
and child physical disabilities may moderate the capacity of VIPP to increase caregiver sensitivity or sensitive dis-
cipline, though other benefits were identified from the intervention. Platje, E., Sterkenburg, P., Overbeek, M., Kef,
S., & Schuengel, C. (2018) The efficacy of VIPP-V parenting training for parents of young children with a visual or
visual-and-intellectual disability: a randomized controlled trial. Attachment & Human Development, 20(5), 455–
72; Hodes, M.W., Meppelder, M., de Moor, M., Kef, S., & Schuengel, C. (2018) Effects of video-feedback inter-
vention on harmonious parent–child interaction and sensitive discipline of parents with intellectual disabilities: a
randomized controlled trial. Child: Care, Health and Development, 44(2), 304–311.
117 Bakermans-Kranenburg, M.J., Juffer, F., & Van IJzendoorn, M.H. (2019) Reflections on the mirror: on video-
feedback to promote positive parenting and infant mental health. In C. Zeanah (ed.) Handbook of Infant Mental
Health, 4th edn (pp.527–42). New York: Guilford.
the case remains an outstanding question. VIPP focuses on increasing caregiver sensitivity,
and caregiver sensitivity has only a very weak association with disorganised attachment.
One proposal put forward by Out, Bakermans-Kranenburg, and van IJzendoorn, building
on Main and Hesse’s thinking, is that the intervention directs the caregiver’s attention to the
child, which reduces absorption and the intrusion of unresolved memories and affects.118
This proposal remains to be tested. However, some supportive evidence is available: an inter-
vention focused on supporting caregiver attention to the child’s experience—Minding the
Baby—reduced rates of disorganised attachment without impacting PTSD symptoms.119
Such findings suggest that the behavioural and attentional focus on the here-and-now of
the child supported by VIPP may reduce disorganised attachment because it helps care-
givers avoid absorption or mental states that may prompt the intrusion of segregated sys-
tems. An alternative explanation, implied by the researchers working on Minding the Baby,

is that Bakermans-Kranenburg and colleagues have underestimated the capacity of parent-
ing interventions to reduce disorganised attachment by increasing the caregiver’s reflective
functioning about attachment relationships.120
Attachment Biobehavioral Catch-up
Another widely studied attachment- based parenting intervention is Attachment

Biobehavioral Catch-up (ABC), developed by Dozier and colleagues. ABC is a ten-session,
in-home intervention to help families meet the needs of infants and toddlers.121 Holding the
intervention in the home is anticipated to help the integration of learning into daily life, and
to allow others in the home to see and gain benefits from the intervention. Part of what is
intriguing about ABC is its parsimony. The focus of the intervention is extremely strict: the
comments made by the coaches only seek to help parents achieve three things: (i) to be nurt-
uring when the child is distressed; (ii) to follow the child’s lead, where possible taking cues
from the child rather than dictating activities to them; and (iii) to avoid displaying intru-
sive, harsh, or frightening behaviours.122 Dozier and colleagues take these to be respect-
ively the central proposals of Bowlby, Ainsworth, and Main and Hesse. Bowlby emphasised
the importance of a child’s confidence in their caregiver when distressed, which Dozier
and colleagues regard as facilitated by nurturing behaviour and tone. Ainsworth empha-
sised heeding and responding to the child’s signals, which she described as sensitivity; she
118 Out, D., Bakermans-Kranenburg, M.J., & van IJzendoorn, M.H. (2009) The role of disconnected and ex-
tremely insensitive parenting in the development of disorganized attachment: validation of a new measure.
119 Slade, A., Holland, M.L., Ordway, M.R., et al. (2019) Minding the Baby®: enhancing parental reflective func-
tioning and infant attachment in an attachment-based, interdisciplinary home visiting program. Development &
120 See also Tereno, S., Madigan, S., Lyons-Ruth, K., et al. (2017) Assessing a change mechanism in a randomized
home-visiting trial: reducing disrupted maternal communication decreases infant disorganization. Development
121 Dozier, M. & Bernard, K. (2017) Attachment and Biobehavioral Catch-up: addressing the needs of infants
and toddlers exposed to inadequate or problematic caregiving. Current Opinion in Psychology, 15, 111–17.
122 Dozier, M. & Infant Caregiver Project (2016) Attachment and Biobehavioral Catch-up for Infants Who Have
Experienced Early Adversity (ABC-1). Intervention Manual. Unpublished manuscript; Dozier, M. & Bernard,
K. (2019) Coaching Parents of Vulnerable Infants: The Attachment and Biobehavioral Catch- up Approach.
New York: Guilford. For the adaptation of ABC for toddlers the focus on avoidance of intrusive, harsh, or frighten-
ing behaviours is partially supplanted by, partly incorporated within, a focus on encouraging parents to collaborate
with their children in regulating difficult emotions.
568 Conclusion
anticipated that this would contribute to a child’s capacity for self-regulation.123 And Main
and Hesse argued that alarming caregiver behaviour disrupts the capacity of the child to co-
ordinate attention and behaviour in making coherent use of the caregiver, either directly or
conditionally, as a secure base and safe haven.
As the first attachment-based intervention, STEEP had a broad set of goals and a broad rep-
ertoire of intervention strategies, as the relationship between attachment theory and interven-
tion science was just getting started (Chapter 4). By contrast, the focus of ABC on only three
concerns represents a fierce, hard-earned, and specific theoretical security. One consequence is
that training in delivering the intervention can be completed in two days. Another consequence
is that the large majority of those interested to train as parent coaches can do so and gain certi-
fication following a fidelity test: potential coaches are screened out only if they find it difficult to
make in-the-moment comments to parents, or if they show a dismissing state of mind regarding

attachment on a cut-down version of the AAI. Following Main’s theory (Chapter 3), Dozier and
colleagues assume that a dismissing state of mind is incompatible with cognitive openness and
comfort with nurturance.124
The sessions of ABC have a few components. A component early in the intervention is psy-
choeducation: parents are informed about some key findings from attachment research, such
as Ainsworth’s findings regarding the effects of prompt response to infant crying (Chapter 2).
These findings are used as an opportunity for discussion of parents’ expectations of themselves
and their child, for instance beliefs about the potential for care to spoil an infant or for harsh
care to help an infant by promoting toughness. Another component of ABC is the use of films
of caregivers responding to infants in desirable and undesirable ways, which likewise serve as
a basis of discussion with the parent. Parents are also asked to complete tasks with their child,
such as playing with blocks or making pudding together. These are filmed, and clips may be
brought by the caregiving coach to review with the parent in a subsequent session. During the
intervention sessions, caregiving coaches provide frequent and specific comments to parents ‘in
the moment’, around 60 times per hour. Dozier and colleagues have found that the frequency
of coaches’ comments, as well as their quality, could predict the magnitude of change in parent
sensitivity.125
Comments address exclusively the three areas of focus of the intervention, drawing par-
ents’ attention to specific behaviours, their purpose, and what ensues from them. This in-
tensity of focused commenting is one of the most distinctive aspects of ABC—especially
when compared to more psychoanalytically inspired forms of parenting intervention, such
as GABI (Chapter 4), where clinicians are advised to address a wider range of topics and
practice reticence in making comments. Caron, Bernard, and Dozier have documented that
27% of caregivers leave ABC before the end of the intervention: families who dropped out
tended to receive fewer comments in sessions and more comments on issues away from the
123 Bernard, K., Meade, E.B., & Dozier, M. (2013) Parental synchrony and nurturance as targets in an attach-
ment based intervention: building upon Mary Ainsworth’s insights about mother–infant interaction. Attachment
& Human Development, 15(5–6), 507–523. Another important influence for Dozier and colleagues was Raver, C.C.
(1996) Relations between social contingency in mother–child interaction and 2-year-olds’ social competence.
124 Caron, E.B., Roben, C.K., Yarger, H.A., & Dozier, M. (2018) Novel methods for screening: contributions
from Attachment and Biobehavioral Catch-up. Prevention Science, 19(7), 894–903.

125 Caron, E.B. Bernard, K., & Dozier, M. (2018) In vivo feedback predicts parent behavior change in the
Attachment and Biobehavioral Catch- up intervention. Journal of Clinical Child & Adolescent Psychology,
47(1), 35–46.
three areas of focus of the intervention.126 A related aspect of ABC is the meticulous use of
video-based supervision to ensure the fidelity of coaches to the goals of the intervention,
ensuring that coaches do not miss opportunities to offer relevant comments.127 In the early
sessions of ABC, comments focus on identifying positives in the parent’s behaviour. As the
sessions progress, the coach may identify areas in which the parent may improve on being
nurturing, following the child’s lead or avoiding intrusive, harsh, or frightening behaviours.
Though the focus of ABC is on behaviour, in the later sessions the coach also works with the
parent to recognise aspects of their past experience that may be proving an obstacle to pro-
viding sensitive and non-frightening care. These may not be the aspects of past experience
that are most salient for the parent in their day-to-day life; the focus is rather on cognitive
and procedural obstacles to the effective functioning of the caregiving behavioural system.
At the end of the tenth session, the coach presents the parent with a montage of film clips

from earlier sessions, celebrating occasions when their behaviour was nurturing or when
they followed the child’s lead.
Bernard, Dozier, and colleagues conducted a randomised control trial of ABC with par-
ents identified as at risk for neglecting their young children. The children were assessed in the
Strange Situation. Classifications of disorganised attachment were less frequent in the ABC
group (32%) than in the control intervention (57%), and classifications of secure attachment
were more frequent (52%) relative to the control (33%).128 Mothers in the ABC group showed
greater increases in sensitivity and decreases in intrusiveness than participants from the con-
trol intervention.129 Mothers in the intervention group also showed higher secure base script
knowledge (Chapter 2) than parents assigned to the control group.130 In a recent study, Dozier’s
team have reported that, contrary to expectations, no improvements were seen in frightened,
frightening, disoriented, or role-confused parenting behaviours.131 However, base rates of the
behaviours were low, given that the observations were exceptionally brief, which may have con-
tributed to false negatives. Nonetheless, the researchers found that parents in the ABC interven-
tion group were less likely to display withdrawing behaviour towards their child’s attachment
behaviours; this difference accounted for around 19% of the effect of ABC on the proportion of
child–caregiver dyads receiving a disorganised attachment classification.
In a follow-up, children from the ABC intervention group showed less externalising be-
haviours than the control group in response to the frustrating problem-solving task devel-
oped for the Minnesota study (Chapter 4).132 Such intervention effects have been replicated
126 Ibid.
127 Meade, E.B., Dozier, M., & Bernard, K. (2014) Using video feedback as a tool in training parent
coaches: promising results from a single-subject design. Attachment & Human Development, 16(4), 356–70.
128 Bernard, K., Dozier, M., Bick, J., Lewis-Morrarty, E., Lindhiem, O., & Carlson, E. (2012) Enhancing at-
tachment organization among maltreated children: results of a randomized clinical trial. Child Development, 83,
623–36.
129 Yarger, H.A., Hoye, J.R., & Dozier, M. (2016) Trajectories of change in Attachment and Biobehavioral Catch-
up among high-risk mothers: a randomized clinical trial. Infant Mental Health Journal, 37(5), 525–36. An important
finding has been that the effects of ABC on caregiver sensitivity appear to be just as strong when implemented by
welfare organisations as when applied as part of a randomised clinical trial. This is quite unusual. Effect sizes gen-
erally drop when interventions are disseminated in the community. Dozier, M. & Bernard, K. (2019) Coaching
Parents of Vulnerable Infants: The Attachment and Biobehavioral Catch-up Approach. New York: Guilford, p.177.
130 Ibid. p.143.
131 Yarger, H.A. (2018) Investigating longitudinal pathways to dysregulation: The role of anomalous parenting
behaviour. Unpublished doctoral dissertation, University of Delaware.

132 Lind, T., Bernard, K., Ross, E., & Dozier, M. (2014) Intervention effects on negative affect of CPS-referred
children: results of a randomized clinical trial. Child Abuse & Neglect, 38(9), 1459–67. See also Lind, T., Raby, K.L.,
570 Conclusion
at multiple sites, with diverse population groups.133 A large-scale implementation of ABC by

the State of New York has recently reported successful impacts on caregiver sensitivity and
reduced frightening behaviours.134 Further trials are underway in Australia, Germany, South
Africa, and Russia. There is emergent endocrinal, physiological, and neurological evidence
that converges with behavioural findings to suggest that the ABC group display greater skills
at emotion regulation in the face of challenges, even several years after the intervention.135
Dozier and colleagues regard ABC as a targeted intervention for parenting, ideally to be de-
livered in the community alongside other targeted support for families depending on their
needs, whether caregiver depression, substance use, housing, or other difficulties. It remains
to be adequately explored, however, whether contextual factors moderate the success of the
ABC intervention. One finding recently reported by Berlin was that the impact of ABC on
caregiving behaviour is moderated by self-reported attachment style. The intervention was

more successful for participants self-reporting a less avoidant attachment style, but actually
reduced caregiver sensitivity among participants endorsing an avoidant attachment style.
Berlin and colleagues suggest that ‘it may be that more avoidant mothers experienced the
intervention model’s emphasis on nurturance, along with the parent coaches’ frequent com-
ments, as dissonant or even aversive, which, in turn, led to iatrogenic effects’.136
There remains further work to be done to maximise the benefits of attachment-based
interventions. Attachment-based interventions are seeing increasing take-up by child wel-
fare and clinical services. However, there remains much we do not yet know about their
cost-effectiveness and, perhaps more to the point, their capacity to survive in cash-strapped
and continually restructuring services. Further research is also needed to understand how
attachment-based interventions can best be delivered with different families and commu-
nities, with sensitivity to signals regarding their needs and to the ways of working they find
acceptable and satisfying.137 Attachment-based interventions offer a fruitful possible point
Caron, E.B., Roben, C.K., & Dozier, M. (2017) Enhancing executive functioning among toddlers in foster care with
an attachment-based intervention. Development & Psychopathology, 29(2), 575–86.
133 Grube, W. & Liming, K. (2018) Attachment and biobehavioral catchup: a systematic review. Infant Mental

134 Chapin Hall at the University of Chicago (2019) Strong families New York City. Final evaluation report, June
2019. https://www1.nyc.gov/assets/acs/pdf/initiatives/2019/CHFinalReport.pdf.
135 Bernard, K., Hostinar, C.E., & Dozier, M. (2015) Intervention effects on diurnal cortisol rhythms of child
protective services-referred infants in early childhood: preschool follow-up results of a randomized clinical trial.
JAMA Pediatrics, 169(2), 112–19; Tabachnick, A.R., Raby, K.L., Goldstein, A., Zajac, L., & Dozier, M. (2019) Effects
of an attachment-based intervention in infancy on children’s autonomic regulation during middle childhood.
Biological Psychology, 143, 22–31; Dozier, M. & Bernard, K. (2019) Coaching Parents of Vulnerable Infants: The
Attachment and Biobehavioral Catch-up Approach. New York: Guilford.
136 Such findings also stand in thought-provoking contrast to a trial of the Circle of Security intervention, which
led to change in children’s attachment classification only for mothers higher in attachment avoidance (Chapter 5).
Berlin and colleagues speculate that Circle of Security is a gentler intervention than ABC, and so less likely to
be aversive to caregivers with strong assumptions that devalue nurturance and emotional needs. Berlin, L.J.,
Martoccio, T.L., & Jones Harden, B. (2018) Improving early head start’s impacts on parenting through attachment-
based intervention: a randomized controlled trial. Developmental Psychology, 54(12), 2316–27. See also Cassibba
and colleagues who found iatrogenic effects of discussions with caregivers about their attachment representations,
if these began secure/autonomous. Cassibba, R., van IJzendoorn, M.H., Coppola, G., et al. (2008) Supporting
families with preterm children and children suffering from dermatitis. In F. Juffer, M.J. Bakermans-Kranenburg,
& M.H. van IJzendoorn (eds) Promoting Positive Parenting: An Attachment-Based Intervention (pp.91–110).
New York: Psychology Press.
137 For work to date see Klein Velderman, M., Juffer, F., Bakermans-Kranenburg, M.J., & van IJzendoorn, M.H.
(2008) A case study and process evaluation of video feedback to promote positive parenting alone and with rep-
resentational attachment discussions. In F. Juffer, M.J. Bakermans-Kranenburg, & M.H. van IJzendoorn (eds)
Promoting Positive Parenting: An Attachment-Based Intervention (pp.23–36). New York: Lawrence Erlbaum; Hodes,
M.W., Meppelder, H.M., Schuengel, C., & Kef, S. (2014) Tailoring a video-feedback intervention for sensitive
Concluding remarks 571
of connection between researchers and practitioners. They do so in the context of serious

obstacles to mutually satisfying and informative communication. Nonetheless, they form
one of several valuable bases for such a conversation, especially in the context of the turn to
greater engagement and collaboration with practitioners among the third generation of at-
tachment researchers. There is certainly a new wind blowing through attachment research
at the moment, with the windows held wider than before to the field’s different audiences.
Given the mutual benefit of mutual learning between research and practice, it is heartening
to see attachment researchers taking steps to facilitate this two-way dialogue.
Concluding remarks

The need for attachment research to be open to different audiences and to the future was a
central theme in the keynote address to the International Attachment Conference in 2019
by Carlo Schuengel. In reflecting on the state of the field, Schuengel described a number of
questions that colleagues in attachment research have been raising with one another. These
include pinning down further: the origin of individual differences in child attachment, the
meaning of attachment disorganisation, the attachment-specific elements of interventions
and how they work, and the cognitive or symbolic aspects of attachment.138 For instance,
complacent and vague appeals to the internal working model concept need to be superseded
by a more detailed account of the cognitive or symbolic components of attachment, their
parameters and features, their interrelations, and links with perceptions and actuators.139
Schuengel also urged further scrutiny of the developmental mechanisms studied by attach-
ment research, the psychometric properties of its measures, the scalability of its interven-
tions, and its broader relevance to child welfare and clinical practice.
Thompson, Simpson, and Berlin have likewise taken the opportunity of 50 years since
the publication of Attachment, Volume 1 to consider the fundamental outstanding questions
for attachment research.140 To Schuengel’s list they have added several further questions.
Among these, they feel that more work is needed to understand what kinds of relationships
qualify as attachment relationships. They urge concern with how attachment processes are
manifested in different cultures, and how culture manifests itself in attachment processes.
They feel attachment research needs a better conceptual model of what domains of later be-
haviour early security of attachment should predict and what domains it would not be an-
ticipated to predict. And they emphasise that there is much more to be done to understand
how attachment theory and research might best inform services for children and families,
including divorce and custody proceedings, home visiting, education, and foster care.
discipline to parents with intellectual disabilities: a process evaluation. Attachment & Human Development, 16(4),
387–401; Scourfield, J., Allely, C., Coffey, A., & Yates, P. (2016) Working with fathers of at-risk children: insights
from a qualitative process evaluation of an intensive group-based intervention. Children and Youth Services
Review, 69, 259–67.
138 Schuengel, C. (2019) Representing attachment: the future is open. Paper presented at International

139 Schuengel, C. & Tharner, A. (2020) Patterns of parenting: revisiting mechanistic models. Attachment &
Human Development, 22. See also Petters, D.D. (2019) The attachment control system and computational mod-
eling: origins and prospects. Developmental Psychology, 55(2), 227–39.
140 Thompson, R.A., Simpson, J.A., & Berlin, L. (2020) Introduction: synthesizing the fundamental questions
and issues in attachment theory. In R.A. Thompson, J.A. Simpson, & L. Berlin (eds) Attachment: The Fundamental
Questions. New York: Guildford.
572 Conclusion
If such questions as those posed by Schuengel and by Thompson and colleagues form
some of the present horizon of empirical attachment research, they do so knowingly on the
basis, in part, of the work of research groups considered in this book. The critical historical
consideration of these groups, with their different perspectives and ambitions, has aimed to
put more firmly at the disposal of the present a portion of the resources of the past in their
liveliness and diversity, and in their potential relevance for attachment research during and
after this period of transition. If attachment research has faced pressures that contort and ab-
breviate its ideas in their circulation, it is hoped that a historical perspective has the potential
to exert some contrary force, whilst at the same time offering observations relevant to the
wider history of science. In this regard, Cornerstones has sought to add to existing historical
studies, which to date have not considered developments after Ainsworth. Those interested
to understand, use or change attachment research must walk with and negotiate with the

past. In addition to Bowlby and Ainsworth, this entails the important legacy from a second
generation of attachment researchers.
Index
For the benefit of digital users, indexed terms that span two pages (e.g., 52–53) may, on occasion, appear
on only one of those pages.
AAI see Adult Attachment Interview (AAI) coherence, 288–93

Abraham, K., 78 determining what is measured, 538–39

abuse of child, 237 development by Main, 17, 211–12, 216,
Bowlby, views of, 12, 32–34, 57 232–33, 428–29
emotional, 305–6, 359–60, 384–85, early experiences and continuity of
397–98, 544 care, 203
factors breaking abuse cycle, 392 earned security, 293–97
long-term effects, 96, 339–40, 353–54 four-category mode, 274–75, 559
and Minnesota study, 338, 339–40, linked trajectories, 293–94
358–60, 392 and Minnesota study, 410, 411–12
physical, 33, 65, 88, 196, 237, 303–4, 318–19, and priming, 214–15
359–60, 376–78, 384–85, 389, 393 and rejection, 280, 281–82
sexual, 33, 304, 305, 318–19, 355–56, 359–60, representations, 286–88
398, 495 secure base script, assessing, 180
traumatic, 303–9 and self-report measures, 462–64, 465–67,
verbal, 33–34 468–69, 482–85
see also intrusive intimacy; neglect semantic and episodic
action schemes, 346–47 information, 279–88
actual experience, psychoanalytic theory, 9–13, and six-year systems, 259
96, 98–99 speech patterns, 282
adaptation, 17–18, 155–56 states of mind, 296, 301–2, 303, 558
global adaptive functioning, 386 and Strange Situation classification system,
and Minnesota study, 354 267, 285–86, 435–36, 437, 558
and pathology, 380, 381 traditions of attachment research, 461–62
self-report measures, 506–14 unresolved loss, 297–303
ADHD see Attention Deficit Hyperactivity unresolved traumatic abuse, 303–9
Disorder (ADHD) see also Main, M. (and Berkeley group)
adolescence, Minnesota study group, 365–69 affect theory, 344–57
Adult Attachment (Gillath, Karantzas and emotional development, 344–49
Fraley), xiii felt security, xxv, 229–30, 349–52, 547
Adult Attachment Interview (AAI), x, 193, intrusive intimacy, 355–57
214–15, 216, 232–33, 235–36, 279–309, social affects, 352–55
312, 320–21, 399 see also emotion
‘Cannot Classify’ category, 307–9 affectionate bonds, 16–17, 195–96
clinical application, 325–26 aggression, xxiii–xxiv, 14–15, 74–75, 82, 140
coding manuals and systems, 265–66, 267, in adolescence, 388–89
285–86, 291, 294–95, 299–300, 301–3, Ainsworth and Strange Situation, 126, 140,
304, 305–6, 308–9, 315–16 142, 177–78
574 Index
aggression (cont.) oral histories by, 37–38

and behavioural systems, 36–37, 43–44, President of Society for Research in Child
47, 49–50 Development, 109
see also aggression behaviour system qualitative research by, 112, 184–85, 193, 194
Minnesota study group, 359, 362–63 on security, xxiv, 113–16, 179–81
self-report measures, 459, 495–97, 499–500 critics, 167–68
see also anger; emotion/emotional secure attachment, x, 150–56, 193–94
development and self-report measures, 442, 443, 447,
aggression behaviour system, 14, 479–80, 534
65–68, 109–10 adaptation, 507–8, 509, 511–12
see also aggression; anger avoidance and anxiety, 448, 461–62, 464
Aguilar, B., 388 Experiences in Close Relationships (ECR)

Ainsworth, L. (husband of Mary), 109, scale, 516–17
116–17 religion and mindfulness, 502–3
Ainsworth, M., viii–ix, xviii, 45–46, 95–96, on separation, 37–39, 40–41, 109–10, 228
111–12, 124, 241, 427, 428–29, 437, and Strange Situation see Strange Situation
537, 572 classification system; Strange Situation
association with Bowlby, 25–26, 28, 36–37, 80, procedure
99–100, 109–11, 113, 151–52, 165–66, students, xii, 49, 100, 119, 134–36, 144–45,
169–70, 182, 187 150–51, 166–67, 169, 183
correspondence, viii–ix, 121, 128, 147–48, see also Bell, S.; Cassidy, J.; Crittenden, P.;
158, 218–19 Main, M.; Marvin, B.; Waters, E.
feeding behaviour, 127–29 in therapy, 111–12
association with Hinde, 155–56, 163–64 Uganda ethnography see Uganda ethnography
association with Main and Berkeley group, (Ainsworth)
144–45, 148, 159–60, 165–67, 169–70, writings
187, 198–200, 214–15, 232–33, 282, Infancy in Uganda, 117, 118–19
301–2, 433–34 Measuring Security in Personal
audiences, 539 Adjustment, 116
Baltimore longitudinal study, xxiv, 120–21, ‘Object relations, dependency and
148, 152–53, 196, 229 attachment,’ 112
biography, xxiv, 109 Patterns of Attachment (co-written with
on caregiving, 116–95 Waters et al), 109, 122, 134–37, 138–39,
clinical postdoctoral researcher, as, 1 140–42, 144–45, 152, 162, 164–66, 167,
coding system see Strange Situation 191, 196, 205–6, 212, 357, 449–50
classification system see also Blatz, E.; Bowlby, J.; Waters, E.
drawing, use of, 275–76 alarm, 2–3, 16, 23, 32
on exploratory behavioural system, 62–63 attachment-relevant perceptions and
on feeding (compared with Bowlby), 127–29 memories, 311, 312–13
hypothesis generation, 103, 113 behavioural systems, 47–49, 51–52, 63, 64
at Johns Hopkins University, Baltimore, 109, caregivers, alarming behaviour in, 249, 251–53,
112, 120, 166–67 255–56, 272, 311, 567–68
Mary Ainsworth Archive, xiv proximity-seeking, 118–19
and Minnesota study, x, 337–38, 341, 350–51, Alexander, C., 29–30
395, 423–24, 425 Allen, J.P., 216
observational research, 109, 120–21, 122, 128, allodoxia, 102
153, 463–64 Allyn, G., 120
Index 575
Al-Yagon, M., 522 intrapsychic, 14

AMBIANCE (Atypical Maternal Behavior and proximity-seeking, 118–19
Instrument for Assessment and Separation Anxiety Test, 214–15, 261, 263–64
Classification), 253, 254–55, 545–46 see also anxious attachment style;
AMBIANCE-Brief, 545–46, 563–64 neuroticism
ambivalence, 127, 448–49 anxious attachment style
‘ambivalent’ anxiety and ‘withdrawn’ adaptation, 510, 513–14
avoidance, 38–39, 40–41 affect theory, 354
psychological conflict, causes, 13, 14–15 anxious-ambivalent, 138, 439–42, 446–47,
ambivalent/resistant attachment, see resistant 448, 449–50, 530
attachment anxious-resistant, 138, 190–91
Ammaniti, M., 303–4 avoidance and anxiety, 459–61

anger, 14, 82, 289–90, 539 behavioural systems theory, 487, 488
Ainsworth and Strange Situation, 111–12, broad and narrow attachment, 476–77,
137–38, 140, 160, 161, 191–92 478–80, 481
angry parent, 219–20, 249 Experiences in Close Relationships (ECR)
behavioural systems, 49–50, 56, 65–67 scale, 515, 516–18
case study (Mrs Q), 82, 83, 86 factors, 516–17
of despair, 66–68, 188, 467–68 high anxiety and high avoidance, 527, 528
of hope, 67–68, 188 and insecurity, 522
and Berkeley study, 214–15, 218–19, 224, and neuroticism, 460–61
225–27, 233, 268–69, 270–71 and religion, 502–3, 505–6
and Minnesota study, 354–55, 398, 408–9 security priming, 562, 563
self-report measures, 431–32, 448–50, sexuality, 492
497–98, 516–17 trauma, effects of, 531–32
self-reporting attachment, 439, 440 Appleyard, K., 388
symptoms, 74–76 Arad, D., 561
see also aggression behaviour system archives, xiv, 6
animal behaviour, study of see ethology (study of Arsenian, J., 130–31
animal behaviour) Assessing Children in Need and their Families
Anna Freud Centre, 12, 536 (DoH), 543
antecedents of attachment, 357–61 Association for the Study of Animal
anthropology, 181–82, 183, 187, 188, 206, 541 Behaviour, 20–21
anxiety, 448–69 atheoretical research, 103
abandonment, fear of, 450 Attachment & Human Development (journal),
attachment anxiety, 448–49 xxiii, 467–68, 538–39, 541–42
attachment style see anxious attachment style attachment behavioural system, 47–53,
and avoidance 54–55, 113
‘ambivalent’ anxiety and ‘withdrawn’ availability of caregiver, 132
avoidance, 38–39, 40–41 defensive exclusion, 159
Experiences in Close Relationships (ECR) and Berkeley study, 219, 222–23
scale, 446–47 multiple stresses, 132
high-anxiety/high-avoidance, 523–28 set-goal, 351–52
case study (Mrs Q), 82 Strange Situation procedure, 138, 144
and conflict in relation to attachment termination of, 115–16, 117
figures, 79 Attachment Biobehavioural Catch-up
defining, 448–49 (ABC), 567–71
576 Index
attachment disorders, vii–viii, 87–94 convergent validity, 177–78

attachment figures, xvii–xviii, 5–6, 76–77, development, 176
99–100, 175–76, 270, 276, 286, 310–11, methodology, 176, 177
347–48, 364, 442–43, 444–45, 461, 478, attachment research/theory, vii–viii, xxiii, 111
480–81, 482, 488–89, 516 antecedents of attachment, 357–61
abusive or threatening, 100, 153–54, 249, attachment to God, 501–3
255–56, 273, 303–4, 305, 306, 312–13, autobiographical interviews, x, 111–12
317, 381 correlates of attachment, 357–78
availability, 12–13, 27, 28–30, 43, 49, 56–57, division between traditions, 462–63
63, 131, 144, 222, 270–71, 448–49, 450, ethological see ethology (study of animal
451–52, 454–58, 459, 475–76, 487–88 behaviour)
avoidance, 221–23 exhaustion of, 551–57

behavioural systems, 49–50, 61–63, and historical time, xv–xvi
64–65, 113 Individual Participant Data Meta-Analysis,
and caregiving behavioural system, 54–56 208, 289–90, 556–60
child welfare issues, 543 intervention research, 564–72
childhood experiences, 482–83, 532–33 narrow and broad attachment
conditional strategies, 221–23 Bowlby on, 25–26, 27, 51–52
conflict-inducing, 78, 79, 448–49 self-report measures, 469–85
deceased, 297–98 observational, 34, 109, 120–21, 122,
discrimination, 21–24, 63 128, 463–64
fathers as, 212–13 and oral culture, 207
greeting behaviours, 117–18, 347–48 origins, 9–10
hierarchy of, 59 self-reporting attachment, 435–47
proximity-seeking, 453, 475–76 significance of in psychological theory, ix
see also proximity and/proximity-seeking site attachment, 23–24
behaviour terminology challenges, xxii–xxiii, xxiv,
psychoanalytic theory, 8–9, 11–12 29–30, 53–54
rejection by, 281–82, 451, 531–32 traditions of (developmental and social
see also rejection psychological), 181–88, 250–51,
representation of, 477–78 461–69, 538–39
as a safe haven, 488–89, 504 see also developmental psychology
separation from, 212, 562 tradition; social psychology tradition
see also separation attachment styles, xxv, 427–28
social referencing of, 62–63 dimensional ratings, 527
unavailability or blocked access to, 64, 78, factors shaping, 488–89
179–80, 453, 459, 486–87, 516–17, 563 fearful, 445–46, 451
see also caregivers/caregiving; parent–child gender factors, 491–92
relationship; parenting and security priming, 562, 563
Attachment in Adulthood (Mikulincer and self-report measures, 437, 440, 441,
Shaver), 468 445–46, 527
attachment language, xxiii, 537–38, 543–45, see also anxious attachment style; avoidant
549, 550–51 attachment; disorganised attachment
attachment parenting, 29–30 classification; insecure attachment;
Attachment Q-Sort (AQS) measure, 176–79, resistant attachment; secure attachment
410, 538, 563–64 Attachment: The Fundamental Questions
Brief Attachment Scale (BAS-16), 178–79 (Thompson, Simpson and Berlin), xiii
Index 577
Attachment Transmission Synthesis, 303, 412 unavailability, 64, 78, 179–80, 453, 459,
Collaboration on Attachment Transmission 486–87, 516–17, 563
Synthesis, 559–60 vigilance, 43–44, 224–25, 226–27, 431–32,
attention, 71, 218, 233, 533, 538 448–49, 450, 472
‘approach/withdrawal’ model, 217–18 see also attachment figures; caregivers/
attending to the infant’s signals, 124–25 caregiving; parent–child relationship;
and cognitive schemas, 482–85 parenting
conditional strategies, 219–20, 222–23, avoidance, 448–69
224, 226 and anxiety, 139–40, 523–28
conditional strategy of, 217–24 ‘ambivalent’ anxiety and ‘withdrawn’
direction of, 215–16, 226–27, 259, 260 avoidance, 38–39, 40–41
and dissociation, 310–11 Experiences in Close Relationships (ECR)

and ethology, 218 scale, 446–47
intensification of, 306 high-anxiety/high-avoidance, 523–28
levels of, 285–86, 311 and attention, 217
looping of, 311–13 conditional strategies, 219–21, 223–24
Main’s intended meaning, 226–27 conflict behaviours, 213–14
manipulation of, 226–27, 229 gaze aversion, 219–20, 222–23, 224
minimising and maximising, 227–28, and Berkeley study, 217
235–36, 494–95 protective function, 159
narrowing of, 74–75 redirecting attention away from caregiver,
objects of, 147, 219–20 139–40, 217–18, 222–23, 224, 226–27,
organisation of, 222–23, 260, 287–88 249, 268–69, 281–82
processes of see attentional processes Strange Situation procedure, 140–42,
redirection of, 192, 229, 236–37, 300 146, 221–22
away from caregiver, 139–40, 217–18, 222–23, avoidant attachment, 569–70
224, 226–27, 249, 268–69, 281–82 Ainsworth and Strange Situation, 121, 156–60,
and security, 217–18 199–200, 292
see also attentional processes conflict, 163–64
Attention Deficit Hyperactivity Disorder cross-cultural aspects, 189–90, 192, 193
(ADHD), 387–88, 484–85 early experiences and continuity of care,
attentional processes, 43–44, 54–55, 218, 223–24, 202, 203
230–31, 235–36, 259, 260–61, 285–86, interpretation of Strange Situation, 149–50
300–1, 326–27, 472, 482–83, 484 origins of Strange Situation
see also attention procedure, 133
Atypical Maternal Behavior Instrument reasons for categories, 142–43
for Assessment and Classification scales, 140–41
(AMBIANCE) see AMBIANCE and Waters, 167–68, 177, 181
(Atypical Maternal Behavior Instrument versus resistant attachment, 225–26
for Assessment and Classification) clinic settings (Bowlby), 36–37, 38–39,
autobiographical interviews, x, 111–12 40–41, 42–43
availability of attachment figure or caregiver, conflict, 79
28–29, 48–49, 131, 132, 455–56 defensive exclusion, 74–75
self-report measures, 446, 453, 455, 459, exploratory behavioural system, 62–63
486–87, 516–17 and Berkeley study, 212–14, 217, 322–23
Strange Situation procedure, 131, 132 Adult Attachment Interview (AAI), 280–82,
Uganda ethnography, 117, 118 283–84, 292, 301–2, 303
578 Index
avoidant attachment (cont.) Banai, E., 440–41, 488–89

conditional strategies, 219–23, 224–26, Banse, R., 521–22
229–30, 233, 234 Barnett, D., 250, 406
conflict, 236–37 Barrier Box task, 396–97
dissociation, 247–48 Barry, R.A., 153
fear and causality, 249, 250–51 Bartholomew, K., xii, 444–46, 448, 450,
six-year data, 259, 260–61, 266–67, 277–78 451, 461–63
meta-analysis, 558, 562 positive and negative working models, 447
and Minnesota study, 342–43, 359, 362–65, ‘Relationship Questionnaire,’ 444, 461
369–71, 373, 376–78, 381, 389–90, 391, Barton, M.L., 260
403–4, 406, 407, 423–24 Bateson, P., 220–21
prosociality, 213–14 Baumrind, D., 361–62

self-report measures, 432–33, 438, 439–40, Beck, A., 79
441–42, 522 Beckwith, H., 550
adaptation, 508, 509, 510, 511–12, 513–14 behavioural systems, 42–68, 230–31, 427–28,
avoidance and anxiety, 526, 527, 528 547
behavioural systems, 487, 488–89 activating and terminating conditions, 45–46,
beliefs about self and others, 444, 445–47 47–49, 50–51, 54–55, 65
dominance, 500–1 aggression, 65–68, 109–10
Experiences in Close Relationships (ECR) attachment see attachment
scale, 448–50, 451–52, 516, 518, 519–20 behavioural system
minimising and maximising, 453, Bowlby distinguished from
456–57, 459–61 Ainsworth, 109–10
pivoting concepts, 472, 474, 476–77 caregiving, 53–61
and religion, 502–3, 505 cognitive aspects, 44, 46, 344, 429
secure base/safe haven, 479–80, 481 concept, 42–44
sexuality, 493, 494 conflict, displays of see conflict
traditions of attachment research, and control systems, 46, 230–31
465–66, 468 desire, 45–47
trauma, 530, 531–33 dominance, 495–501
Strange Situation procedure, 156–60, 181 emotional aspects, 44, 97–98
see also avoidance environment, 46–47, 48
awareness, 122 ethology, developed within, 43, 44
exploratory system, 61–63, 86, 345, 488–89
Bailey, H.N., 71, 254–55, 434–35 fear system, 42, 43, 63–65, 77
Bakermans-Kranenburg, M., xii, 394–95, 550 and innate responses, 45–46
Adult Attachment Interview (AAI), 203–4, and learning, 45–46
305–6, 309 limitations of behaviourism, 1
behavioural systems, 53–54, 65–66 metaphor, 43, 87
dissociation, 318, 319–20 motivation theory, 43, 45–46
infant disorganised classification, 254, power system, 497–98, 499–501
255–56, 257–58 responsiveness, 47, 61–62
intervention research, 565–66, 567 self-report measures, 485–501
and Minnesota study, 348–49, 394–95, 409, set-goal, 43–44, 46, 47–49, 243, 351–52
411–12, 418–20 sexuality, 50–51, 489–95
self-report measures, 474–75 theory, 485–89
Baldwin, M.W., 169–70, 560–61 types of system, 47–68
Index 579
see also Bowlby, J.; ethology (study of animal association with Main and Berkeley
behaviour) group, xii–xiii, 23–24, 28–29, 40–41,
Behrens, K.Y., 138–39, 190–91, 261–63 71–72, 97–98, 100, 101, 217, 221,
Bell, S., 134–36, 138, 166–67, 195–97 233–36, 322–23
Belsky, J., xii, 464–65, 468–69, 511–12 association with Mead, 181–82
Benjafield, J.G., xxii association with Shaver, 17, 37, 65–66
Bennett, S., ix association with Sroufe, xvii, 3, 10–11,
Bento, S., 152–53 95–97, 102
Berant, E., 456, 510 attention, alleged neglect of, 218
Berkeley group see Main, M. (and Berkeley audiences, 539
group) and behavioural systems, 42–68
Berkeley–Leiden Adult Attachment biography, 1

Questionnaire Unresolved States of at Cambridge, 1, 31–32, 39–40
Mind scale, 324, 433 and category formation, 134
Berlin, I., 569–70, 571 clinic settings, 80–94
Attachment, xiii case files, 80–81
Bernard, K., 313, 569 and conditional strategies, 233–36
Bernier, A., 466–67, 479–80 and continuity, 94–96
Bertalanffy, L von, 6–7 critics, x, 27, 28, 39–40, 45–46, 53–54, 71–72,
Berzenski, S.R., 384–85 97–98, 218, 291–92, 547
Bielefeld Centre for Interdisciplinary Research, poor relationship with, 4, 22
West Germany, 189–90, 220–22 critique/revision of psychoanalytical theory,
bimodal distribution, 141–42 x, 1, 8–17, 68–69
Birnbaum, G., xii, 492–93, 494–95, 525, 536, 554–55 following and attachment, 24–25, 26–27
Blatz, E., 109, 113–16, 130–31, 133, 151–52, and separation, 32–33
215–16, 229 on crying behaviour, 195–96
see also Ainsworth, M.; Main, M. (and on Darwin, xv, 7, 19, 52–53
Berkeley group) discrepancies in writings, 4–5, 22
Blehar, M., 166–67 early childhood challenges, 1–2, 30–31
Block, J. and J., 241 early claims, viii–ix, x
Blome, W.W., ix on emotion, 96–99
Blood, J., 546–47 ethology, appeal of, 18, 19–21, 102, 111
Bosmans, G., 369–70 on evolutionary theory, 99–101
Boston, M., 35 fatherhood, 16, 19–20
Boston Change Process Study Group, 21 on feeding
boundary-dissolving behaviours/violations, compared with Ainsworth, 127–29
355–57, 365–66 compared with Klein, 14–15
Bourdieu, P., xix and self-preservation, 16
Bowlby, J., xxiii–xxiv, 537, 572 and historical enquiry, xv–xvii
association with Ainsworth, 25–26, 28, 36–37, internal working model theory, 10–13,
80, 99–100, 110–11, 113, 151–52, 165–66, 27, 179–80
169–70, 182, 187 interpretation of work, xviii
correspondence, viii–ix, 121, 128, 147–48, John Bowlby Archive, Wellcome
158, 218–19 Collection, xiv, 6
feeding behaviour, 127–29 key concepts in writings, 104
association with Hinde, 18–19, 25–26, 27, 28, and language, xxi–xxii, 21
30–31, 102, 115–16, 117 and Berkeley study, 258–59, 260, 261
580 Index
Bowlby, J. (cont.) Loss, 1, 43–44, 51–52, 69–70, 81, 234,

on maternal role, 2–3, 5–6, 8, 22, 47–48, 235–36, 270, 279–80, 431
54–56, 59–60 The Making and Breaking of Affectional
maternal deprivation concept, xxii, 4, 33–34, Bonds, 56–57
52, 109–10, 382–83 ‘Notes on child attachment and
on media use, 208–9 monotropy,’ 22–23
and Minnesota study, x, 337–38, 341, Personality and Mental Illness, 4–5, 88
370, 382–83 Protest, Despair and Detachment, 36–37,
on monotropy, 21–23 40–41, 42–43
narrow focus on parental attachment, 23–24 The Roots of Parenthood, 58
populist writings, 3, 7, 8, 37, 55–56, 542 A Secure Base, 57
on predation, 99–101 Separation, 1, 23–24, 34, 63, 64, 65, 66–67,

scholarly, scientific and clinical works, 4–5, 7, 81, 88–89, 222, 225, 442–43, 448
8, 37, 47, 103 ‘Separation Anxiety,’ 240–41
and self-report measures, 435–36, 442, 448–49, ‘Wariness,’ 64
470–71, 495–96, 503, 533, 535–36 see also key concepts in writings above;
avoidance and anxiety, 462–63, 464 populist writings above; scholarly,
beliefs about self and others, 442–43 scientific and clinical works above;
on separation, 1, 23–24, 30–42, 63, 64, 65, unpublished work above
66–67, 81, 88–89, 222, 225, 228, Bowlby, U. (wife of John), 16, 74–75
442–43, 448 Bowlby-Ainsworth-Sroufe tradition, 337–38
on sexuality, 489–90, 495 Bradshaw, D., 480–81
and symptoms, 68–79 Brassard, A., 492
and Tavistock Clinic, London, 1, 4, 8 Brazelton, B., 217–18, 222–23
founding of Separation Research Unit by breast-feeding see feeding
Bowlby, 35 Brennan, K.A., 449–50, 465, 514–15, 516, 520,
as a therapist, 83–87 523–24, 527–28
unpublished work, xiv, 22–23, 25, 26–27, Brennan, K.B., xx–xxi
31–32, 44, 64 Bretherton, I., 72–73, 166–67, 232–33, 259, 261
clinic settings, 81, 86 British Psychoanalytic Institute, 12
psychoanalysis, 11–13 British Psychoanalytic Society, 15
separation, 32, 34 British Psychological Society, 91–92
symptoms, 68–69, 73 broad and narrow attachment see narrow and
writings broad attachment
Attachment. Volume I, 1, 4–6, 21–22, Bronfman, E., 254
27, 42–43, 44, 46, 48–49, 54, 58–59, Bronson, G., 63
61–62, 65–66, 68, 90–91, 99–100, Buddhism, 502, 504, 505–6
128–29, 199–200, 217, 240–41, 349–50, Burlingham, D., 8
489–90, 495–96 Byng-Hall, J., 179–80
Charles Darwin: A Life, 52–53
Child Care and the Growth of Love, Cadman, T., 177–78
2–3, 25 Caldwell, B., 130
Defences that Follow Loss, 68–69, 70–71, 73, Call, I., 91
74, 81, 83, 375–76 Cameron, J.J., 524–25
‘Developmental psychiatry comes of age,’ 88 Campbell, C., 101, 285, 553–56
‘Forty-Four Juvenile Thieves,’ 31, 32, caregivers/caregiving
35, 275–76 African-American population, 120
Index 581
Ainsworth on, 116–95 Carlson, E., 172–73, 203, 424–25, 542

alarming behaviour in caregivers, 249, 251–53, and Berkeley study, 245, 256–57, 277–78
255–56, 272, 311, 567–68 and Minnesota study, 360–61, 369–70, 382–83,
versus attachment, 53–54 387–88, 425
availability see availability of caregiver Carlson and Putnam Dissociative Experiences
Baltimore study (Ainsworth), xxiv, 120–21, Scale, 376
148, 152–53, 196, 229 Cassidy, J., xii, 96–97, 204, 261, 264, 475–76
caregiver-infant touch, 218–19 conditional strategies, 225–26, 231, 472
caregiving behavioural system see caregiving Handbook of Attachment, xiii, 413, 464–65,
behavioural system 468–69, 538–39, 541–42
childcare culture, 117–18 reunion system, 263–64, 268–75
compulsive caregiving, 56–57 self-report measures, 453, 464–65, 468–69,

controlling behaviour, 271–72, 307 479–80, 482
and emotion, 96–97 and Strange Situation procedure,
insensitive, 122 152–53, 166–67
multiple caregivers, 117–18, 182 causes of psychological conflict, 13–15
paid professionals, 47 Charles, G., 29–30
primary see primary caregivers Chaudhary, N., 60
proximity-seeking behaviour of infant to Child Attachment Interview, 261–63
see proximity and proximity-seeking Child Guidance Clinic, London, 9–10, 30–31,
behaviour 32–33, 35, 80–81
redirecting attention away from, 139–40, child welfare practice, attachment in, 541–51
217–18, 222–23, 226–27, 249, 268–69, attachment language, 543–45, 549, 550–51
281–82 Department of Health guidance, 543
retrieval response, 54–55 ethnographical research, 543–44, 548
safe haven/secure base, caregiver seen as, Love Barrow Families project, 545–46
131, 132 social work practice, 548–49
semi-standardisation of caregiving UK practice, 543, 545
behaviour, 179 see also foster care
and sensitivity, 122–27 childcare culture, 117–18
sensitivity see sensitivity of caregiver child–caregiver relationship see infant/child–
states of mind, 315–16 caregiver relationship
Uganda ethnography see Uganda ethnography Chisholm, J., 186–87
(Ainsworth) Cicchetti, D., xxv, 126, 250–51, 399–408, 419–20
see also attachment figures; child–caregiver Circle of Security framework, xiii–xiv,
relationship; fathers; mothers; parent– 208–9, 513–14
child relationship; parenting; proximity Clark, C.L., 449–50, 527–28
and proximity-seeking behaviour clinging behaviour, 115, 381, 448–49, 489–90
caregiving behavioural system, 53–61 Bowlby, views of, 25–26, 48, 100
and attachment behavioural Main, M (and Berkeley group), 219–20,
system, 54–56 222, 225
and consoling system, 54–55 separation, 39, 40
motivational and behavioural Strange Situation procedure, 133, 139
aspects, 56 clinic settings (Bowlby)
problem with Bowlby’s characterisation attachment disorders, 87–94
of, 57–58 Bowlby as therapist, 83–87
CARE-Index, 152–53 case study (Mrs Q), xxiii–xxiv, 80–85, 86, 298
582 Index
clinic settings (Bowlby) (cont.) resolution of, 78, 346

diagnosis-focused, 92 Strange Situation procedure, 163–66
and exploratory behavioural system, 86 Connelly, K., 495–96
familial context, attention to, 85 continuity
guidance, role of, 85–86 and Bowlby, 94–96
past and present experiences, 86 early experiences and continuity of
phases of therapy, 86 care, 200–4
transference, 83–85 and Minnesota study, 340–41
Clown experiment and crying behaviour, 213–14, Cooper, M.L., 440
237, 252, 281–82 coping strategies, 69, 377–78, 432, 441–42
coding manuals and systems, xiv, 252, 261–63, adaptation, 509
324, 329, 434 avoidant and distancing, 510

Adult Attachment Interview (AAI), 265–66, correlates of attachment, 357–78
267, 285–86, 291, 294–95, 299–300, Coughlan, B., 549–50
301–3, 304, 305–6, 308–9, 315–16 Cowan, C. and P., 294
see also Adult Attachment Interview (AAI); Cox, M., xii
Strange Situation classification system Cranley, M.S., 53–54
Coffino, B., 382–83 Crawford, T.N., 451–52, 535
coherence, xxiii Crittenden, P., xii, 72–73, 208–9, 232–33, 406,
Adult Attachment Interview (AAI), 288–93 499–500, 539–40
Cohn, D., 166–67 CARE-Index, 152–53
Colin, D., 166–67 child welfare practice, 545–46
Collaboration on Attachment Transmission self-report measures, 475, 490–91
Synthesis, 559–60 and Strange Situation procedure, 152–53,
Collins, N.I., 440 165–67, 198–99, 250
Collins, W.A., 368, 373 Cronbach, L., 175
conditional strategies, 217–36, 472, 474 cross-cultural research, Strange Situation
attention, 217–24, 533 procedure, 181–95
avoidance, 219–21, 223–24 anthropology, 181–82, 183, 187, 188, 206
and Bowlby, 233–36 applicability, 181–95
and direct approach, 221–22 classification system, 183
evolutionary theory, 155–56 crying behaviour, 191–92
game theory, 221, 223–24 Germany, 189–90, 192
gaze aversion, 219–20, 222–23, 224 Japan, 193
proximity-seeking behaviour, 221–22 overseas distribution of classifications,
and reproductive strategies, 221 185–86, 189–95
resistance and universality, 224–33 Sapporo study, 191–93
responsiveness, 227, 229, 234 traditions, 181–88
sexuality, 494–95 validity, 193–94
terminology, 221, 227–28 Crowell, J.A., 179, 461, 464, 465
Uppsala Longitudinal Study, 220–21 crying behaviour, 19, 43, 449–50
Condradi, H.-J., 523–24 Clown experiment, 213–14, 237,
conflict 252, 281–82
and avoidance, 213–14 cross-cultural research, 191–92
conditional strategies, 220–21 low-level fussing, 197
displays of, 75–79, 213–14 responsiveness to, 197–98
psychological, causes of, 13–15 Strange Situation procedure, 191–92, 195–97
Index 583
Cummings, M., 166–67, 245–46, 350–51 self-report measures, 427–28, 437,

curiosity, 61–62 446, 534–35
Cyr, C., 554–55 avoidance and anxiety, 460–61,
462–64, 467–68
Dagan, Or, 60–61 see also social psychology tradition
Dalai Lama, 504 developmental psychopathology, xxv, 379–86
Dallaire, D.H., 261–63 conference (1979–1980), 379–80
Darwin, C. defining, 380–81
Bowlby on, xv, 7, 19, 52–53 developmental pathways, 386–87
sea lizard study, 219–93 disease model, 387
The Voyage of the Beagle, 219 dual risk model, 380
see also evolutionary theory origins, 379

Davidowitz, R., 509–10 studies modelling, 422
Davies, P.T., 126 Diagnostic and Statistical Manual of Mental
Dawkins, R., 155–56, 220–21 Disorders (DSM)
Dazzi, N., 303–4 Adult Attachment Interview (AAI), 284–85
defence concept, 68–70, 113 attachment disorders, 90–91
and coping strategies, 509 discriminant function analysis, 140–41,
defensive exclusion, 72–75, 159, 226 289–90, 449–50
dissociation, 311 discrimination of attachment figures, 21–24, 63
and Berkeley study, 260 disorganised attachment classification, 328, 406,
and segregation, 70–71 407–8, 412–13, 425, 538, 542, 547, 548–
Deinard, A., 338 49, 558, 566–67, 569
delinquency, 31 adolescence, 368, 369–71
DeMoss, A., 298 antecedents of attachment, 360–61
depression, 16–17 attachment disorders, 93–94
and conflict in relation to attachment conflict behaviours, 236–38
figures, 79 criticism by Ainsworth, 270–71
Main, M (and Berkeley group), 320–23 disorganised/disoriented attachment, 28–29,
separation, 39–40 192, 236–58
desire, 45–47 dissociation, 376–78
despair, anger of, 66–68, 188, 467–68 early social competence, 362–63
detachment, 36–37 fear and causality, 248–58
Development and Psychopathology ‘FR coding system,’ 252, 253, 254–55
(journal), 386–87 holism, 418–20
developmental anomalies, 155 indices of disorganised attachment,
developmental pathways, 88–89, 95, 192, 242–43
386–90, 414 Direct Indices of Apprehension, 244–45
probabilistic, 400–1 and later aggressive behaviour, 247–48
Developmental Psychology (journal), 146, 285 Leiden study, 252–53
developmental psychology tradition, x, 4–5, Berkeley study, 211–12, 215–16,
99–100, 540–41, 556 235–36, 325–27
and Ainsworth/Strange Situation, 122, 146, coherence, 292–93
167, 183, 186, 197, 204–5 depression, 321–23
American, x, 197–98, 427–28, 528 dissociation, 310, 314, 315–17, 320
and Berkeley study, 246, 249 fear and causality, 250–51, 252–58
and Minnesota study, 344–45, 364–65, 379 reunion system, 270–71
584 Index
disorganised attachment classification (cont.) regulating, 134–36, 459

six-year data, 273–75, 277–78 self-report measures, 444, 445–46, 449–50
unresolved loss or trauma, 301–4, self-reporting attachment, 441
305–7, 309 and separation, 23–24, 34, 36, 40, 43–44, 132,
and maternal depression, 322 133, 437–38, 480–81
meaning of disorganisation, 242–48 signals, 126–27, 160, 218–19
compared with dictionary six-year data, 259, 270–71
definition, 243–45 soothing, 128
mental and physical health, 374–75 Strange Situation procedure, 130–31, 132,
and Minnesota study, 203, 376–77, 408–9 133, 136, 141, 149–50, 156, 158, 159,
self-report measures, 445–46, 452 160–62, 163–64
Strange Situation procedure, 203, 207 subjective, 444

terminology, 240–41 symptoms, 73–74, 76–78, 88, 115–16, 180,
test-retest stability, 247 449–50, 529–31
and trauma, 377–78 domestic violence, 376–77, 385–86
‘unclassifiable’ cases in Strange Situation dominance, 495–501, 502
procedure, 238–39 Dozier, M., xii, 281–82, 313, 318–19
disorientation on separation, 40–42, 75 Attachment Biobehavioural Catch-up (ABC),
disinhibited social engagement, 42 568, 569
freezing, 42 DRD4 (dopamine-related)
and reunion, 41 polymorphisms, 407–8
and strangers, 41, 42 dreams, 9
dissociation drive, concept of, 18–19
Bowlby on, 7–8, 75 dual risk model, psychopathology, 380
as a defence, 311 Duggal, S., 371–72, 398, 399
Dissociative Experiences Scale, 316–17 Duschinsky, R., Trauma and Loss, 6,
information retrieval, 310–11 68–69, 91–92
and Berkeley study, 310–20
and Minnesota study, 370–71, 375–78 Eagle, M., 11–12, 200
Post-Traumatic Stress Disorder (PTSD), earned security, AAI, 293–97
317–18, 319–20 Edelstein, R., xii
recovered memories, 310 Egeland, B., x
symptoms, 318–19 and Minnesota study, 337–426
distress, 215–16, 282–83, 346, 347, 381, 388–89, and Strange Situation procedure, 123,
403–4, 432–33, 434–35 147, 202
Adult Attachment Interview (AAI), 293–94, Eichberg, C., 166–67, 232–33, 301–3, 307
300–1, 307–8 Ein-Dor, T., 452, 510–14
Attachment Q-Sort (AQS) measure, 175–76 Elicker, J., 365–66
behavioural systems, 44, 167–68, 498 Elkin, E., 338
clinic settings (Bowlby), 188, 189, Emde, R., 220–21
191–92, 195–96 emotional abuse, 305–6, 359–60, 384–85,
conditional strategies, 218–19, 220–21, 229, 397–98, 544
233, 235–36 emotional neglect, 359–60
disorganised attachment classification, emotion/emotional development, 97–98
241, 249 and behavioural systems, 44, 97–98
distress ‘language,’ 165–66 and Bowlby, 96–99
hidden or masked, 167–68, 229–30, 234, 282 and caregiving, 96–97
Index 585
cognitive aspects, 346, 347 conditional strategies, 221, 222

intensity, 79, 102 game theory, 221, 223–24
in Minnesota study, xxv, 344–49 sea lizard study (Darwin), 219–93
regulation of emotions, 226, 441 see also behavioural systems; Bowlby, J.;
see also alarm; anger; anxiety; depression; Darwin, C.; ethology (study of animal
despair, anger of; distress; fear behaviour)
behavioural system; grief; hope, exclusion
anger of; states of mind defensive, 72–75, 159, 226
empathy, 31, 54–55, 565–66 levels of, 73–75
Englund, B., 373 selective, 71, 73
episodic and semantic information exhaustion of attachment research, 551–57
(AAI), 279–88 and Strange Situation procedure, 553–54

Erickson, M.F., 414, 418 experience see actual experience,
Erikson, E., 6–7 psychoanalytic theory
Esposito, A., 286–87 Experiences in Close Relationships (ECR) scale,
Estes, D., 170–72 xxv, 427, 456–57, 539, 559
ethnographical research and adaptation, 513–14
child welfare issues, 543–44, 548 aim, 450–51
and cross-cultural research, 185–86 anxious attachment style, 515, 516–18
qualitative research, 184–85, 193, 194 avoidance and anxiety, 446–47, 463, 466–67
quality, 186–87 broad and narrow attachment,
quantitative research, 186 479–80, 481–85
social affects, 352–53 creating, 448–52
Strange Situation procedure, 183–84, critical components, 515–16
185–86 dominance, 499
Uganda (Ainsworth) see Uganda ethnography factor analysis, 450–51
(Ainsworth) items, 514–20
ethology (study of animal behaviour), xxiii–xxiv, length of, 515
17–30, 80, 155–56, 472 orthogonal dimensions, 514–15
appeal to Bowlby, 18, 19–21, 102, 111 and priming, 214–15
behavioural systems, 43, 44 and Roschach inkblot test, 456
bird-watching and Bowlby, 18 scale, 450–51
and defence concept, 69–70 SSFS modelled on, 492–93
desire for proximity versus traditions of attachment research, 461–62
nutrition, 19–20 exploratory behavioural system, 61–63, 86,
dominance, 500–1 345, 488–89
ethological attachment theory, 24–25
geese, 19–20 Fairbairn, W.R.D., 11–12, 69
imprinting, 19–20 family drawing system (Kaplan and Main),
motivation theory, 18–19, 20 249, 275–79
primates/monkeys, 39–40, 48, 58–59 coding system, 277–78
questions asked by, 17–18, 221 fantasy, 9, 12, 96
and security, 114, 115–16 versus actual experience, 96, 98–99
see also Bowlby, J. Farber, E.A., 359
evolutionary theory, 20–21, 99–101, 155–56 fathers
and adaptation, 512 as attachment figures, 212–13
behavioural systems, 230–31 inclusion in Berkeley study, 212–13
586 Index
fathers (cont.) self-report measures, 466–67, 470, 489,

as primary caregiver, 422–23 522–23, 525–26
Strange Situation procedure, 149 Strange Situation procedure, 145–47, 177
see also mothers; parent–child relationship; freezing response, 42, 64
parenting Freud, A., 8, 11–12
fear Freud, S., xiv, 9, 13, 16–17, 18–19, 50–51, 76,
attachment response and flight response, 481 98–99, 470
behavioural system see fear Introductory Lectures on Psychoanalysis, 1, 15
behavioural system see also psychoanalytic theory
and causality, 248–58 Frisch, K. von, 17–18
fearful attachment style, 445–46, 451 Furnivall, J., ix
freezing response, 42, 64 Fury, G., 277–78

of rejection, 444, 445–46, 450
fear behavioural system, 42, 43, 63–65, 77 GABI (Group Attachment-Based Intervention),
Fearon, P., xii, 178, 554–55 420–22, 514, 568–69
feeding game theory, 221, 223–24
Ainsworth on, 127–29 Garmezy, N., 379–81, 390–91
Bowlby on, 14–15, 16 Gaskin, S., 245–46
development of feeding response, 128–29 gaze aversion, 219–20, 222–23, 224
Feeney, J., 520 Geertz, C., xxi
Fehr, B., 169–70 gender factors
felt power, 497 and attachment, 395–96, 492, 559
felt security, xxv, 229–30, 349–52, 547 attachment styles, 491–92
Firth, R., 182 boundary violations, 365–66
5-HTTLPR (serotonin-transporter-linked cross-gender relationships, 356,
polymorphic region), 403–4, 407–8 365–66, 385–86
Fleeson, J. (later Sroufe), 353–55 and Minnesota study, 394–96
Florian, V., 441, 509, 510 self-concept, 395
following and attachment, 24–30 socialisation, 394–95, 396
Fonagy, P., xii, 539–40, 560 see also sexuality
and Ainsworth/Strange Situation, generation of attachment researchers concept,
153–54, 208–9 xii–xiii, 10–11, 208–9, 534
and Bowlby, views of, 11–12, 23–24, and Minnesota study, 400, 411–12, 423
26–27, 101 structural issues facing attachment research,
and exhaustion of attachment 537–38, 539, 540–41
research, 553–56 George, C., 57–59, 214–15, 237, 238–39, 261,
and Berkeley study, 285, 291, 307–8, 313 272–73, 307
and self-report measures, 475, 479–80, Germany, 160, 189–90
490–91, 536 Group C dyad, 192
Fortuna, K., 465–66 Giesbrecht, B., 510
Foster, S., 548 Gillath, O., 477–78, 491–92, 510, 522,
foster care, viii–ix, 60, 133, 393, 397, 407, 536, 554–55
411–12, 571 Adult Attachment, xiii
Foucault, M., xxiii, 539–40 Glaser, D., xiii, 546–47
Fraiberg, S., 407 Goal-Corrected Partnership in Adolescence
Fraley, C., xii, 17, 536, 554–55, 559 Coding System, 199–200
and Minnesota study, 409–11, 423–25 Goldberg, S., 208–9
Index 587
Goldenberg, J., 487–88, 497, 502 on romantic relationships, 431–32, 433–34,

Golding, K., 546–47 443–44, 446, 502–3, 535–36
Goldsmith, H.H., 451–52 see also Mikulincer, M.; self-report measures;
Goldwyn, E., 264–65 Shaver, P.
Goldwyn, R. health, mental and physical, 369–75
and Adult Attachment Interview (AAI), 281– anxiety and depression, 371–72
83, 288, 293–94, 296–98, 559 secure base/safe haven, 374
and Minnesota study, 411–12 trauma, effects of, 372
and six-year systems, 264–65, 266–67, 268 Heckendorf, E., 468–69
Goodman, G.S., 508–9 hermeneutic philosophy, 266
Gorer, G., 182 Hesse, E., xxiv–xxv, 144, 567–68
Graham, J.J., 524–25 and Berkeley study, 211–12, 216, 235–36, 328

Granqvist, P., 23–24, 204–5, 503, 554–55 Adult Attachment Interview (AAI), 283–
Greenberg, M., 166–67 84, 288, 289–91, 292–93, 297–98, 300–1,
Grice, P., 288–89 303–4, 305–6
grief, 16–17, 297–98, 430 disorganised attachment classification,
Groh, A.M., 149–50, 248, 322–23, 358, 540–41 249–50, 251–53, 254–56, 257–58
Grossman, K. and K., 3 dissociation, 310, 311, 312–16, 318
Bielefeld study, 189–90 and Minnesota study, 408–9, 411–12
and Berkeley study, 220–21, 222–23, personal archive, xiv
235–36, 237 and self-report measures, 433, 445–46, 468,
and Strange Situation procedure, 160, 188, 483, 485
190, 191–92, 198, 200–1, 205–6 see also Main, M. (and Berkeley group)
Group Attachment-Based Intervention see Hinde, R., xxiii–xxiv, 24–25, 39–40,
GABI (Group Attachment-Based 61–62, 75–76
Intervention) association with Ainsworth, 155–56, 163–64
Gruneau Brulin, J., 23–24 association with Bowlby, 18–19, 25–26, 27, 28,
guess and uncover memory game, 30–31, 102, 115–16, 117
267–68, 282–83 association with Main and Berkeley group,
Guthrie, L., 546–47 211–12, 224–25, 239, 286–87
see also Bowlby, J.; ethology (study of animal
Haltigan, J.D., 378, 389–90 behaviour)
Handbook of Attachment (Cassidy and Shaver), history and attachment, xv–xvi
xiii, 413, 464–65, 538–39, 541–42 concept of ‘development,’ xvii
Handbook of Attachment-Based Interventions developmental psychology, xvii
(Howard and Steele), 564 nature of psychological theory, xix–xx
Hansburg, H., 56 textual records, xviii
Hargreaves, R., 6–7 Hofer, M., 100–1
Harkness, S., 45, 181–82 holism
Harlow, H., 19–20, 115–16, 220–21 dissociation, 375
Hart, J., 487–88, 497, 502, 522 interventions, 414–22
Harvard Child Mistreatment Project, 406 in theory, 412–14
haven of safety see safe haven Holmes, J., 216, 337–38
Hazan, C., 17, 37, 320–21, 427, 432–33, 461, home, attachment to, 23–24
476, 480–81 home observation see observational research
first attempt at self-reporting hope, anger of, 67–68, 188
attachment, 435–42 Horowitz, I.M., 444, 451
588 Index
hospitalisations, 37–40, 41, 43–44, 76, conditional strategies, 235–36, 316–17

109–10, 134 attention, 216, 223–24
conflict, displays of, 75, 76 resistance and universality, 230–31,
see also separation 232–33
Hotam, Y., 327 in early attachment, 17, 51–52, 103, 144–46,
Howland, M., vii–viii 147, 150, 153–54, 310–11
Hrdy, S., 58–60 Experiences in Close Relationships (ECR)
Mother Nature, 58–59 scale, 539
Hubbard, F., 196–97 fixed traits, seen as, 199–200
Hughes, D., 475 formation, 326–27
Huston, L., 339–40 Main, M (and Berkeley group), 216, 310–11
Hutt, M., 439–40 minimising and maximising, 231

Huxley, J., 17–18 and Minnesota study, 411
hyperactivity, 448–49, 454–56, 457–58, 493, 531 psychoanalytic theory, 112
and deactivation, 455–56, 457–58, 492–93, self-report measures, 444–45, 484–85
499, 500–1, 525 Strange Situation procedure, 144–45, 153–54,
power system, 497–99, 500–1 169–70, 200, 259
and sexuality, 493–94, 495 category formation, 134, 136–37
see also Attention Deficit Hyperactivity cross-cultural context, 183–84,
Disorder (ADHD); sexuality 185–86, 193
hyperarousal, 511–12, 529–30, 532–33 Main, M (and Berkeley group), 224, 259
hypervigilance, 318–19, 371, 431–32 between transcripts, 288
Individual Participant Data Meta-Analysis, 208,
IJzendoorn, M. van, xii, 565–66 289–90, 556–60
and Adult Attachment Interview (AAI), infant disorganised classification see
287–88, 302–3 disorganised attachment classification
and Bowlby, 53–54, 65–66 infant/child–caregiver relationship, 14–15
and depression, 321–22 emotional development, 347–48
and dissociation, 312, 318–20 emotional neglect, 359–60
emotional development, 348–49 and feeding, 127–29
and exhaustion of attachment historical differences, 133
research, 551–52 infant–grandparent attachment, 60–61
and infant disorganised classification, 251, proximity-seeking behaviour see proximity
254–56, 257–58 and proximity-seeking behaviour
and Minnesota study, 411–12, 418–19 separation see separation
and self-report measures, 433, 474–75 and Strange Situation procedure see Strange
and Strange Situation procedure, 144, 147, Situation procedure
149–50, 152–53, 181–82, 196–97, see also Ainsworth, M.; caregivers/caregiving;
203–4, 208 mother; parent–child relationship;
Immelmann, K., 220–21 Strange Situation procedure
immersion parenting, 29–30 inhibition, 68
imprinting, 19–20 insecure attachment, 1–2, 12, 59–60, 222
individual differences, 14, 19, 493, 552, 559 Ainsworth and Strange Situation, 173–74,
adult attachment, 408–9, 484, 520, 524–25 201, 203
Adult Attachment Interview (AAI), 274–75, insecure-avoidant, 163–64, 169, 189–90, 268
279, 285–88, 484–85 and Minnesota study, 403–4
avoidance and anxiety, 453 antecedents of attachment, 357
Index 589
early social competence, 362–63 intrusive intimacy, 355–57

mental and physical health, 371–72, 374–75 IQ scores, 37–38, 361
self-report measures, 520–21, 527, 528
adaptation, 506–7, 508, 510–11, 512–14 Jacobovitz, D., 252–53, 261–63, 393, 445–46
behavioural systems, 488 James, W., 21–22
Experiences in Close Relationships (ECR) Johnson, S.M., 96–97, 475
scale, 516 Juffer, E., 565–66
and religion, 502–3
see also insecurity; secure attachment; Kagan, J., 340–41, 348–49, 366–67
security Kaplan, N., 214–15, 258–59, 275–80
insecurity, 113, 115 drawing system, 249, 275–79
insecure attachment see insecure attachment Karantzas, G.C., 522–23

Main, M (and Berkeley group), 214–15, 222 Adult Attachment, xiii
opposite of, 520–23 Karen, R., x–xi, xii, 110–11
see also security Becoming Attached, x, 537
insensitivity, 122, 124–25, 159–60, Kaufman, J., 384–85
254–55, 434–35 Keller, H., 60, 124–25, 186, 187–88, 548
maternal, 203, 253, 463 Kermoian, R., 185
see also sensitivity of caregiver Kirkpatrick, L., 501–3
Institute of Psychoanalysis, 1–2 Klagsbrun, M., 258–59
institutional care, 47, 59–60, 90, 92–93 Klein, M., 5–6, 31, 66, 69, 127
internal working model concept (Bowlby), 27, on actual experiences of child, 9–10, 11–12
179–80, 223–24, 472–74 on causes of psychological conflict, 14–15
actual experience, psychoanalytic Contributions to Psychoanalysis, 14
theory, 10–13 Kobak, R., xii, 166–67, 453
and emotion, 344–45 and Berkeley study, 225–26, 231, 235–36,
and Berkeley study, 260, 261, 276, 281–82, 286–87
286–87 and Minnesota study, 351–52, 369–70
International Classification of Diseases (ICD), Kochanska, G., 153
attachment disorders, 92 Kondo-Ikemura, K., 193
International Conference on Infant Studies, Korfmacher, J., 414, 418
201, 507 Kroonenberg, P.M., 144
International Journal of Psychoanalysis, 1 Kuhn, T.S., The Structure of Scientific
intervention research, 564–72 Revolutions, 426
Attachment Biobehavioural Catch-up, 567–71
video-feedback intervention, 564–67 Lafontaine, M.F., 519–20
interviews Lamb, M., 144–45, 166–67, 170–72, 196, 200–1,
autobiographical, x, 111–12 203, 222
Child Attachment Interview, 261–63 language, 21, 537–39
child welfare practice, attachment in, 548–49 Ainsworth and Strange Situation, 123, 206
‘friendship,’ 343 attachment, xxiii, 537–38, 543–45, 549, 550–51
longitudinal research, 121 ordinary and scientific, xx–xxiii
mental and physical health, 369–70 Laplanche, J., 68–69
qualitative research, 339–40, 543–44 Lapsley, D.K., 462–63
self-report measures, 439–40 Latour, B., xix
Uganda ethnography, 119 laugh response, 344–45
see also Adult Attachment Interview (AAI) Lavi, N., 457–58
590 Index
Lawler, M.J., 508–9 Main, A.N. (husband of Mary), 211–12, 219

learned helplessness, 432, 451 Main, M. (and Berkeley group), xiii–xiv,
Leerkes, F.M., 126–27 xxiv–xxv, 138, 211–329, 437, 567–68
Leiderman, P.H., 185 Adult Attachment Interview (AAI) see Adult
Leipzig Research Centre for Early Child Attachment Interview (AAI)
Development, 187–88 association with Ainsworth, 144–45, 148,
Levine, S. and R., 182, 183 159–60, 165–67, 169–70, 187, 198–200,
Levinson, A., 307–8 214–15, 232–33, 282, 433–34
Lieberman, A., 150–51, 166–67, 257–58, association with Bowlby, xii–xiii, 23–24,
407, 539–40 28–29, 40–41, 62, 65–66, 71–72, 97–98,
life stress, 173, 359, 373, 386, 388, 394, 100, 101, 217, 221, 233–36
416–17, 465–66 Bielefeld study, 220–22

life-history model of attachment, 227–28 biography, 211–12
Liotti, G., 377–78 characteristics of study sample, 212–13
Lo, C., 516–18, 519 Clown experiment and crying behaviour,
longitudinal research, x 213–14, 237, 252, 281–82
Baltimore study (Ainsworth), xxiv, 120–21, conditional strategies, 217–36, 472,
148, 152–53, 196, 229 494–95, 533
early experiences and continuity depression, 320–23
of care, 207 disorganised attachment classification, 192,
mental health, 432 236–58, 360–61, 408–9
and Minnesota study, 202–3, 341, 413 meaning of disorganisation, 242–48
multigeneration, 341 dissemination of work, restricted, 324,
post-traumatic stress disorder 325, 329
(PTSD), 532–33 dissociation, 310–20
separations, 35 early experiences and continuity of
Uppsala Longitudinal Study, 220–21 care, 207
see also Minnesota Longitudinal Study of Risk family drawing system of Kaplan and Main,
and Adaptation and study group; Main, M. 249, 275–79
(and Berkeley group) graduate student of Ainsworth, Main
Lorenz, K., 6–7, 17–20, 239 as, 165–67, 211–12, 214–15, 217–19,
loss, experiences of 236, 241
Klein on, 31 interpretation of work, 44
in psychoanalytic theory, 14–15, 17 Johns Hopkins University,
unresolved, in AAI, 297–303 Main at, 211–12
love, 25 methodology, xxiv–xxv
Lussier, Y., 492 and Minnesota study, 248, 250, 294–95,
Luyten, P., xii 370, 408–9
Lyons-Ruth, K., xii, 21, 92–93, 198–99, personal archive, xiv
539–40, 553 reunion system, 268–75
and Adult Attachment Interview (AAI), and self-report measures, 437, 443, 445–46
261–63, 307–8 avoidance and anxiety, 446–47, 451–52,
and infant disorganised classification, 252–53, 453, 468
254, 256 broad and narrow attachment, 470–71,
477, 479–80
MacArthur preschool system, 557 six-year data, 211–12, 214–15,
Madigan, S., xii, 152–53, 254, 545–46, 258–79, 320–21
554–55 doll-play, 272–73
Index 591
family drawing system (Kaplan and maximising and minimising see minimising and
Main), 275–79 maximising
guess and uncover memory game, Maynard-Smith, J., 155–56, 220–21
267–68, 282–83 Mead, M., 6–8, 181–82
move to next level of Meehl, P., 175
representation, 258–67 Mehr, D.G., 506–7
reunion system (Cassidy and Meins, E., ix–x, 153–54
Main), 268–75 Mental Health Inventory, 481–82
St. John’s College, Main at, 211–12, 219 Mesman, J., 181–82
and Strange Situation procedure, 144–45, meta-analysis, 208, 213, 247–48, 254, 321–23,
165–67, 169–70, 187, 192, 198–200, 209, 342–43, 348–49, 358, 360–61, 419–20,
211–12, 213 463–64, 524–25, 550

avoidant attachment, 157, 159–60, Individual Participant Data Meta-Analysis,
225–26 208, 289–90, 556–60
interpretation of, 148 Strange Situation procedure, 149–50, 152–54,
six-year data, 214–16 162–63, 173–74, 177–78, 193–94
on trauma, xxiv–xxv Miga, E.M., 216
writings of Main Mikulincer, M., xii–xiii, xxv, 65–66, 432–33, 525,
‘Avoidance in the service of proximity: a 539, 556
working paper,’ 222 Attachment in Adulthood, 468, 500–1, 510–11
‘Avoidance of the attachment figure under attachment training, 428, 432
stress,’ 221–22 biography, 427
Behaviour and the Development of critics, 474
Representational Models of Attachment, security priming, 561–63
263–64, 324 self-report measures
‘Cross-cultural studies of attachment adaptation, 506–7, 509–13
organization,’ 228, 233 avoidance and anxiety, 448–49,
Four Patterns of Attachment Seen 451–52, 453–58, 459–60, 462–63,
in Behaviour, Discourse and 523–24, 525–28
Narrative, 263–64, 266, 324 behavioural systems, 485–87, 488
‘The “ultimate” causation of some infant broad and narrow attachment, 469–71, 476,
attachment phenomena,’ 222 477–78, 479–81
see also Ainsworth, M.; Hesse, E.; Main, M. dominance, 497–501
Makariev, D.W., 526 Experiences in Close Relationships (ECR)
Manchester Child Attachment Story, 261–63 scale, 514–15, 517–18, 519
Mandelbaum, D., 182 mindfulness, 504
Manning, I.G., 126 opposite of insecurity, 520–23
Marvin, B., xii, 28–29, 49, 120, 183 religion, 503
and reunion system, 273–75 sexuality, 489–93, 494–95
Marvin, R., 166–67 trauma, effects of, 528–30, 531–32
Masten, A., 390–91 see also Hazan, C.; self-report measures;
Masters, J., 168–69 Shaver, P.
Maternal Behaviour Q-sort, 152–53 mindfulness, 504–6
maternal deprivation concept (Bowlby), xxii, 4, minimising and maximising
33–34, 52, 109–10, 382–83 attention, 227–28, 235–36, 494–95
maternal warmth, 123 coding systems, 261
Maternal–Fetal Attachment Scale, 53–54 individual differences, 231
Matte-Gagmé, C., 466–67 self-report measures, 453–61, 531–32
592 Index
Minnesota Longitudinal Study of Risk and mothers

Adaptation and study group, x, 337–426 Bowlby on role of, 2–3, 5–6, 8, 22,
abuse, 338, 339–40, 358–60 54–56, 59–60
adolescence, 365–69 as primary caregiver, 47–48
and Adult Attachment Interview (AAI), 411–12 depression, 322
affect theory, 344–57 Maternal Behaviour Q-sort, 152–53
antecedents of attachment, 357–61 maternal deprivation concept (Bowlby), xxii,
broad-band competencies, focus on, 342–43 4, 33–34, 52, 109–10, 382–83
case studies, 396–400 maternal warmth, 123
Cicchetti, 400–8 Maternal–Fetal Attachment Scale, 53–54
correlates of attachment, 357–78 non-recognition following separation, 41
developmental pathways, 386–90 as primary caregiver, Bowlby’s view

developmental psychopathology, 379–86 of, 59–60
disorganised attachment classification, 203 responsiveness of, 161
dissociation, 375–78 stereotypes, 29–30
duration of study, 343–44 Strange Situation, focus on infant–mother
early experiences and continuity of care, 207 interaction in, 149, 190–91, 212–13, 215
early social competence, 361–65 see also fathers; parent–child relationship;
emotion, xxv parenting
empirical research, 352–53 motivation theory, 18–19, 20, 158
gender factors, 394–96 behavioural systems, 43, 45–46
holism, 412–22 mourning, 16–17, 297–98, 430
hypothesis generation, 363 see also grief; loss, experiences of
importance, 337–38 Murphy, A., 420–21
laboratory settings, 343
longitudinal research, 202–3 narrow and broad attachment, 469–85
and Berkeley study, 248, 250, 294–95 Bowlby on, 27, 51–52
mental and physical health, 369–75 pivoting concepts, 469–79
origins of study, 337, 338 WHOTO security measure, 477–79
preschool behaviour, 363–64 National Academy of Sciences, 401–2
resilience, 390–94 National Center on Child Abuse and Neglect,
STEEP intervention see STEEP (Steps Toward US, 401
Effective, Enjoyable Parenting) National Institute for Health and Clinical
and Strange Situation procedure, 141, 173, Excellence (NICE), UK, 256
338–39, 348–49, 370, 371–72 National Institute of Child Health and
structural issues facing attachment Human Development (NICHD), 120,
research, 539–40 141–42, 178–79
three-generational processes, 374–75 dataset, 202–3
see also Egeland, B.; Sroufe, A. and Minnesota study, 378, 389–90, 409
Miyake, K., 193 National Institute of Mental Health, US,
Mizrahi, M., 494 401–2, 553
monotropy, 21–23 natural selection, 28
see also discrimination of attachment figures naturalistic fallacy, 2–3
Moran, G., 152–53 neglect, 33–34, 93–94, 134, 141, 369–70, 374
Morison, A., ix emotional, 359–60
Moss, E., 274–75 see also abuse of child; sexual abuse
Moss, H., 340–41 neuroticism, 222, 433, 498, 505–6
Index 593
adaptation, 512–13 Piaget, J., 6–7

avoidance and anxiety, 459–61, 468 Pinquart, M., 173–74
broad and narrow attachment, 478–79 Place of Attachment in Human Behaviour, The
trait, 451–52, 481–82 (Parkes and Stevenson-Hinde), 222–23
see also anxiety; anxious attachment style; Planta, R.C., 395
fear pleasure principle, 11–12
NICHD see National Institute of Child Health Pontalis, J.B., 68–69
and Human Development (NICHD) Popper, K., 5–6, 103
North, G., 543–44 Posada, G., 185–86, 187
Nurturing Attachments Training Resource positive attachment, 544–45
(Golding), 546–47 post-traumatic stress disorder (PTSD), 248
combat-related, 432–33, 529–30, 532

object relations theory, 11–12 criteria, 359–60
observational research, 109, 122, 128, 153, 217, dissociation link, 317–18, 319–20
463–64, 498 effects, 528–33
Baltimore study (Ainsworth), xxiv, 120–21, hyperarousal, 529–30, 532–33
148, 196, 229 and physical health, 372
early social competence, 363 symptoms, 529, 530, 532–33
O’Connor, C., 502, 508–9 Potter, A., 543–44
Oedipus complex, 13, 14, 68–69 power behavioural system, 497–98, 499–501
Ogawa, J.R., 376–77 Power Behavioural System Scale, 498
Olds, D., 166–67 predation, 99–101
Orgad, T., 513 Preschool Assessment of Attachment, 557
Out, D., 567 primary caregivers, 14, 40, 58–59, 119, 133, 200
fathers as, 422–23
Padrón, E., 256–57, 425 mothers perceived as, 5–6, 47, 59–60, 259
Paetzold , R.L., 256 priming, 186–87, 214–15, 259, 479, 498, 505–6
Palgi, P., 182 sexual, 491
parenting, vii–viii, xi–xii see also security priming
attachment parenting, 29–30 Prior, V., xiii, 546–47
co-parenting, 60–61 Prosada, G., 185–86
dimensions, 147 prosociality, 213–14
‘ideal,’ xix, 123 proximity and proximity-seeking behaviour,
instinct, 55–57 26–27, 28–29, 115–16
physical abuse by parents, 304 and anxiety, 118–19, 453–54
positive, 128–29, 570–71 attachment behavioural system, 47–53, 219
video-feedback to promote, 564–67 and availability of care-giver, 28–29
sensitive, 203 conditional strategies, 221–22, 224–25
see also attachment figures; caregivers/ desire for proximity versus nutrition, 19–20
caregiving; fathers; mothers; parent– emotional development, 351–52
child relationship; STEEP (Steps Toward literal/physical proximity, 351–52, 453–55
Effective, Enjoyable Parenting) minimising and maximising, 453–54
Parkes, C., 222–23, 525–26 and predation, 101
Pederson, D., 152–53 self-report measures, 489
Philibert, R.A., 153 Strange Situation procedure, 138–39,
physical abuse, 33, 65, 88, 196, 237, 303–4, 318– 141, 146
19, 359–60, 376–78, 384–85, 389, 393 Uganda ethnography, 118–19
594 Index
psychoanalytic theory of Ainsworth, 112, 117, 183, 184–85, 186,

actual experience, 9–13, 96, 98–99 193, 194
assumptions regarding infant behaviour, 115 attachment theory, ix
backlash against attachment theory, 12 descriptions/impressions, 36–38, 219–20,
causes of psychological conflict, 13–15 281–82, 301–2
conflict, displays of, 78 ethnographic, 184–85, 193, 194
continuity, 94–95 interviews, 339–40, 543–44
critique/revision by Bowlby, x, 1, records, 417–18
8–17, 68–69
following and attachment, 24–25, 26–27 Raby, L., 396, 412, 559
and separation, 32–33 rejection, 127, 488–89
defence see defence concept Adult Attachment Interview (AAI), 280, 281–82

and emotion, 344–45 by attachment figure/caregiver, 39–40, 67–68,
individual differences, 112 95, 111–12, 118–19, 155–56, 219–20,
internal working model concept (Bowlby) 362–63, 445–46, 451, 531–32
see internal working model concept see also maternal below
(Bowlby) avoidance, 142–43, 492–93
limitations, 1, 12–13 and Bowlby, 1–2, 4, 7–8
loss, 14–15, 17 by child, 37, 41, 269
mourning, 16–17 disorganised/disoriented attachment, 445–46
Oedipus complex, 13, 14, 68–69 expectations of, 1, 482
and other forms of knowledge, 12 fear of, 444, 445–46, 450
pleasure principle, 11–12 maternal, 111–12, 118–19, 155–56, 159–60,
representational model, 12–13 161–62, 191–92, 218–19, 222, 224–25,
repression, 59–60, 83 234, 249, 272–73, 281–82
self-preservation, 15–17 peer, 372–73
on sexuality, 17, 20 psychological conflict, causes, 14, 15
Wolf Man case, 14 religion, 501–6
see also Bowlby, J.; Freud, S.; Klein, M. and anxious attachment style, 502–3, 505–6
‘Psychobiology of the Child’ study group, WHO, attachment to God, 501–3
6–7, 23 Buddhism, 502, 504, 505–6
psychology repression
developmental see developmental psychology Adult Attachment Interview (AAI), 291–92
tradition of attachment response, 41
personality psychology, 428–29, psychoanalytic theory, 59–60, 83
459–60, 461–63 and segregation, 70–71
psychological conflict, causes, 13–15 resilience
social see social psychology tradition conditional strategy of, 224–33
see also personality traits factors breaking abuse cycle, 392
psychometric analysis, 246, 410, 523, 559 and Minnesota study, 379–80, 390–94
PTSD see post-traumatic stress disorder (PTSD) resistant attachment, 217–18, 224–33, 403–4,
Puig, J., 373 424–25, 539
adaptation, 508
Q, Mrs (Bowlby case study), xxiii–xxiv, 80–85, avoidance and anxiety, 448–50, 453
86, 298 behavioural systems, 149, 160–63
Q-methodology, 550 broad and narrow attachment, 472, 474
qualitative research, 112, 237 continuity, 202
Index 595
and Berkeley study, 213 pointing to a toy on, 226–27

Adult Attachment Interview reunion system (Cassidy and Main),
(AAI), 282–83, 301–2, 303 263–64, 268–75
conditional strategies, 217–18, 222, Strange Situation procedure, 140–41, 164
223–27, 229–30, 233 Rheingold, H.L., 130–31
six-year data, 261, 270–71, 277 Richards, A., 117
meta-analysis, 558 Richardson, E., 214–15
and Minnesota study Richters, J.E., 138–39, 171–72
adolescence, 365–66, 368 Rieu, D.C.H., 21
affect theory, 348–49, 350, 351–52 Rifkin, A., 261–63
antecedents of attachment, 358, 359 Robertson, J., 1, 35, 92, 113, 134, 214–15, 375–76
case studies, 396–97 Protest, Despair and Detachment, 36–37,

early social competence, 362 40–41, 42–43
Main, M (and Berkeley group), 226–27, 250, on separation, 36–38, 40–43, 75, 76
268–69, 84, 277, 283–303 Robins, R.W., 496–97
mental and physical health, 371, 373 Roisman, G., xii, 17, 208
meta-analysis, 558 and Adult Attachment Interview (AAI), 294–95
and Minnesota study, 371, 406, 423–24 on avoidance and anxiety, 465–67, 468
self-report measures, 439, 444, 494–95 broad and narrow attachment, 470
separation, 138, 142–43 and exhaustion of attachment
Strange Situation classification system, 140–41, research, 554–55
160–63, 181, 223–24 Individual Participant Data
category formation, 138 Meta-Analysis, 559
cross-cultural aspects, 193–94 and Minnesota study, 378, 396, 400,
reasons for categories, 141, 146 408–12, 426
symptoms, 181 romantic relationships, 427, 431–32, 433–34, 443
see also ambivalence ‘love quiz,’ 437–41, 442, 446, 502–3, 536
Strange Situation procedure, 142–43, and peer relationships in childhood, 368–69
149–50, 162 ‘Relationship Questionnaire,’ 444, 461
responsiveness, 161, 169–70, 268–69 Roschach inkblot test, 456
affective, 212–13 Rosenblatt, P.C., 182
attachment response and flight Rosenbluth, D., 35
response, 481 Rosmalen, L. van, 130–31
behavioural systems, 47, 61–62 Rothbaum, F., 193, 254
child’s lack of confidence in, 118–19 Röttger-Rössler, B., 194–95
conditional strategies, 227, 229, 234 Rubinstein, C., 436–37
confidence in, 118, 150–51 Rutter, M., 4–5, 25, 29, 34, 326–27, 363, 394–95, 413
to contingencies, 243–44 developmental psychopathology, 379–83
to crying, 197–98
to following by the safe haven, 23–24, 48, 219, 345, 374
infant, 117–18 Ainsworth and Strange Situation procedure,
security needs, 477, 482 115–16, 119, 145
social, 90 attachment figure as, 488–89, 504
retrieval response, 54–55 caregiver seen as, 132
reunion and Minnesota study, 345, 374
following separation from caregiver, 36–37, and self-report measures, 471–72, 477–82
39–41, 133 Sagev, M., 497–98, 499
596 Index
Sagi-Schwartz, A., xii, 60–61, 181–82 mental and physical health, 374
Sahdra, B.K., 505–6 and self-report measures, 471–72
Salter, M. (later Ainsworth) see Ainsworth, M. Strange Situation procedure, 145, 175
Sameroff, A., 338–39, 380 secure base script, 179–81, 363–64
Sandler, J., 69 security
Schachner, D.A., 528 and attention, 217–18
Schaffer, H.R., 133 Blatz on, 215–16
Scheper-Hughes, N., 186–87 Circle of Security framework, xiii–xiv, 208–9,
Schieche, M., 243–44 513–14, 546–47
Schore, A., 475 earned (AAI), 293–97
Schuengel, C., xii, 252–53, 312, 316–17, 474–75, ethological concept, 114
554–55, 556, 572 felt see felt security

Scott, J.P., 19–20 and independence, 113–16
secure attachment, ix, 67–68, 543 and insecurity, 113, 115, 520–23
Ainsworth and Strange Situation, 121, 134, responsiveness, 477, 482
150–56, 173–74 security system, 488
classification system, 145 Strange Situation procedure, x, 113–16, 150–
cross-cultural aspects, 183, 184–85, 187–88, 56, 167–68, 179–81
190, 193–94 traditions of attachment research, 461
‘broaden and build cycle of attachment see also safe haven; secure attachment; secure
security,’ 453–54 base; security metasystem
child welfare issues, 549 security metasystem, 497
ideal, seen as, 505–6 security priming, 522, 535–36, 556–57, 560–64
intervention research, 569 and attachment style, 562, 563
Main, M (and Berkeley group), 213–14, social psychology tradition, 563–64
293–94, 301–2 segregation, 70–72
and Minnesota study, 403–4, 405–6, 407 conflict, displays of, 76–78
affect theory, 350–51 selective exclusion, 71, 73
antecedents of attachment, 359 self and others, beliefs about, 442–47
case studies, 396–97 self-preservation, 15–17
developmental psychopathology, 382 self-reliance, 115–16, 160
early social competence, 361, 362, 364 self-report measures, x, 359, 427–536
mental and physical health, 371–72, 374–75 adaptation, 506–14
security priming, 561–62 and Adult Attachment Interview (AAI), 462–
self-report measures, 437–38, 519–20, 536 64, 465–67, 468–69, 482–85
adaptation, 506–7, 508–9, 510–12, 514 avoidance and anxiety, 448–69, 523–28
avoidance and anxiety, 448, 460–61, 468–69 behavioural systems, 485–501
behavioural systems, 488 beliefs about self and others, 442–47
beliefs about self and others, 444, 447 creating the ECR, 448–52
broad and narrow attachment, 476–77, 482 first attempt by Shaver and Hazan, 435–42
and insecurity, 520–21, 522–23 ‘love quiz,’ 437–41, 442, 446, 502–3, 536
and religion, 502–3, 505–6 mindfulness, 504–6
self-reporting attachment, 437–38, minimising and maximising, 453–61, 531–32
439–40, 441–42 narrow and broad attachment, 469–85
see also insecure attachment; safe haven; ‘Relationship Questionnaire,’ 444, 461
secure base; security religion, 501–6
secure base, 23–24, 115–16, 119, 477–82 romantic relationships, 427, 431–32,
caregiver as, 131 433–34, 443
Index 597
self-reporting attachment, 435–47 in psychoanalytic theory, 17, 20

and Strange Situation classification system, see also gender factors
435–36, 468–69 Shaffer, A., 339–40
traditions of attachment research, 461–69 Shaver, P., xx–xxi, xxv, 114, 320–21, 413, 435–37,
trauma, effects of, 528–33 439, 449–50, 537, 539, 556, 561–62
see also Experiences in Close Relationships attachment training, 428–29
(ECR) scale; Mikulincer, M.; Shaver, P. biography, 427
semantic and episodic information (AAI), 279–88 and Bowlby, 17, 37, 65–66
sensitivity of caregiver, 122–27, 547 critics, 474
attending to the infant’s signals, 124–25 self-report measures
attention, 217 adaptation, 506–7, 509–13
awareness, 122 avoidance and anxiety, 449–52, 453–57,

and insensitivity see insensitivity 459–60, 461–63, 464, 468–69, 523–24,
maternal and paternal, 160 526–28
measures, 197 behavioural systems, 485–87, 488
ordinary language, 206 broad and narrow attachment, 469–71, 476,
Strange Situation procedure, 122, 123, 124, 477, 480–81
125–26, 152–54, 155, 172–73 dominance, 496–501
terminology challenges, 122–23 Experiences in Close Relationships (ECR)
separation, 30–42 scale, 514–16, 517–18, 519–20
Ainsworth on, 37–39, 40–41, 109–10 first attempt at self-reporting
classes of responses to, 448 attachment, 435–42
depression, 39–40 mindfulness, 504, 505–6
disorientation, 40–42 opposite of insecurity, 520–23
early, 37 religion, 501–3
and evacuated children, 31–32 romantic relationships, 427, 431–32,
freezing, 42 433–34, 443–44, 446, 502–3,
hospitalisations, 37–40, 41, 43–44, 75, 535–36
109–10, 134 sexuality, 489–92
and impact of affection, 16–17 trauma, effects of, 528, 531–32
long-term, 16–17, 30–31, 32, 37, 441–42 in therapy, 428–59
minor infant–mother separations, US, 190–91 writings
prospective studies, 35–40 Attachment in Adulthood, 468,
reunion see reunion 500–1, 510–11
Separation Anxiety Test, 214–15, 261, 263–64 Handbook of Attachment, xiii, 413, 453,
as a threat, 129–30 464–65, 468–69, 538–39, 541–42
Uganda ethnography, 118 see also Hazan, C.; Mikulincer, M.; self-report
sexual abuse, 33, 304, 305, 495 measures
boundary-dissolving behaviours, 355–57 Shemmings, D., 325–26
inter-generational, 356 Siegel, D., 325–26
intra-familial, 355–56 Simpson, J.A., vii–viii, xii, 468, 571
and Minnesota study, 355–56, 359–60, 398 Attachment, xiii
trauma, effects of, 318–19 and Minnesota study, 368–69, 373, 426
Sexual System Functioning Scale (SSFS), Sinason, V., 256
492–93, 494 site attachment, 23–24
sexuality six-year systems (Berkeley developmental
behavioural systems, 50–51, 489–95 study) see under Main, M.
infantile, 20 (and Berkeley group)
598 Index
Slade, A., 337–38, 369–70 Sroufe, A., x, 537

Smeeton, J., xiii–xiv ‘Attachment as an organizational
smile reflex, 346–47 construct,’ 350
Smith, J.M., 155–56, 220–21 and Bowlby, xvii, 3, 10–11, 95–97, 102, 547
Smith, P.K., 22, 495–96 and Berkeley study, 241
Smith, R.S., 390–91 Adult Attachment Interview
social affects, 352–55 (AAI), 294–96
social competence, early, 361–65 disorganised attachment classification, 245,
social learning, vii–viii, 43, 61–62, 94–95, 248, 250–51, 257–58
145, 507–8 six-year data, 277–78
Strange Situation procedure, 174, 195–96 and Minnesota study, 337–426
social psychology tradition, x, 25, 144, 556, leader, 337–38

559–61, 563–64 and Strange Situation procedure, 147, 152,
American, 535 167–68, 172–73, 176, 190–91, 200, 202
experimental, 440–41, 534, 535 Stainton Rogers, R., 535
and Berkeley study, 246, 285–86 state adult attachment measure (SAAM), 522
and Minnesota study, 352, 424–25 states of mind
self-report measures, 427–28, 431, Adult Attachment Interview (AAI), 296,
436–37, 534 301–2, 303, 558
avoidance and anxiety, 450–51, 459–61, Berkeley–Leiden Adult Attachment
462–63, 464–65, 468–69 Questionnaire Unresolved States of
behavioural systems, 489 Mind scale, 324, 433
beliefs about self and others, 444–45, 446 caregivers/caregiving, 315–16
broad and narrow attachment, 475–76, unresolved, 411
477, 479–81 Stayton, D., 195–96
Experiences in Close Relationships (ECR) Steele, H., 291, 420–21, 467–68, 560, 564
scale, 514–15 Steele, M., 291, 420–21, 560, 564
see also developmental psychology tradition STEEP (Steps Toward Effective, Enjoyable
Society for Research in Child Development, Parenting), 208–9, 539–40
Toronto, 259 ABC contrasted, 568
Society for Research on Adolescence, intervention research, 564, 565–67, 568
Wisconsin, 365 and Minnesota study, 404–5, 414–16, 417,
Solomon, J., 192, 214–15, 445–46 418–19, 420–21
and Bowlby, views of, 28–29, 57–59 VIPP contrasted, 565–66
and Berkeley study Steps Toward Effective, Enjoyable Parenting
Adult Attachment Interview (AAI), see STEEP (Steps Toward Effective,
300, 307 Enjoyable Parenting)
disorganised attachment classification, Stevenson-Hinde, J., 85, 222–23, 476
238–39, 240–42, 243–45, 248–49, 250– Strage, A., 258–59, 263–64
51, 253, 255–56, 406 Strange Situation classification system, 124–25,
dissociation, 314 133–38, 175, 197–200
six-year data, 261, 270, 272–74, 275 and AAI classification system, 267,
and Minnesota study, 406, 425 285–86, 558
Solomon, Z., 432, 529 acceptance within developmental science
Spangler, G., xii, 92–93, 243–44 community, 167
Spencer-Booth, Y., 39–40 additional information provided by, 144
Spieker, S., xii, 145–47, 177 category versus dimension, 147
and Minnesota study, 410, 423–25 coding individuals, 197–200
Index 599
coding protocols, xxiv, 141, 143, 171–72, 191, Strange Situation procedure, x, 62, 129–66
300–1, 307–8, 358–59, 423, 438 resistant attachment, 138, 160–63, 181,
disorganised attachment classification, 193–94, 217–18, 223–24, 225
238–39, 250, 256–57 attachment behavioural system, 138, 144
cross-cultural research and anthropology, availability of caregiver, 131, 132, 446
181–82, 183, 187, 188, 206 avoidance, 140–42, 146, 221–22
defence of categories, 141–43 avoidant attachment, 156–60, 181
distribution of classifications, 189 Bielefeld study, 189–90, 220–22
Group A dyad, 134, 137–38, 140–41, 142–43, categories/classification system see Strange
155, 156–59, 249 Situation classification system
Group A1, 136–37 conflict behaviours, 163–66, 213–14, 236
Group A2, 136–37 cross-cultural applicability, 181–95

and Berkeley study, 217, 219–20, 227, 229, classification system, 183
230–31, 233, 300 overseas distribution of classifications,
Group B dyad, 134–36, 137–38, 140–41, 185–86, 189–95
142, 145, 150–52, 158, 175–76, 227, Sapporo study, 191–93
249, 358–59 traditions, 181–88
Group B1, 136, 142–43, 157–58, validity, 193–94
167–68, 437–38 crying behaviour, 195–97
Group B2, 134–36 depression, 320–21
Group B3, 134–36, 154–56, 167–68 description of, 109, 131
Group B4, 164–65, 167–68 early experiences and continuity of
Group C dyad, 136–38, 140–41, 142, 144, care, 200–4
155, 160–61, 162–63, 192, 230–31, and exhaustion of attachment
233, 249 research, 553–54
C1, 137–38, 141, 160–61, 163 individual differences, 144–45, 153–54,
C2, 109, 137–38, 141, 160–61, 169–70, 200, 259
162–63, 517–18 category formation, 134, 136–37
Group D dyad, 142, 162–63, 164–65, 239, cross-cultural context, 183–84, 185–86, 193
241–42, 243, 245–46, 248–49 Main, M (and Berkeley group), 224, 259
initial importance of categories, 145 infant-mother interactions, focus by
Main’s use of categories, 211–12, 213, 215–16 Ainsworth on, 149, 190–91, 212–13, 215
and Minnesota study, 367–68, 370, 371–72 interpretation of, 147–50
new category requirement, 238–42 limitations of Ainsworth’s method, xxiv,
normativity, 154–55 170–71, 183–85, 194, 208–9
patterns of Ainsworth’s method, 215–16, 232, meta-analysis, 149–50, 152–54, 162–63,
239, 243, 359 173–74, 177–78, 193–94
see also under Ainsworth, M. Minnesota Strange Situation training
reasons for categories, 141–47 institute, 245
romantic relationships scale modelled and Minnesota study, 141, 173,
on, 427 338–39, 348–49
and self-report measures, 435–36, 468–69 origins, 129–33
social psychological tradition, 538–39 passive–aggressive behaviour, 137–38
testing of stability of categories, 169–75 prominence given to by Ainsworth, 194, 223–24
‘unclassifiable’ cases, 238–39 proximity-seeking behaviour, 138–39,
see also Ainsworth, M.; Main, M. (and 141, 146
Berkeley group); Strange Situation resistance, 138, 140–41, 160–63
procedure reunion, 140–41, 164
600 Index
Strange Situation procedure (cont.) anxious attachment style, 531–32

scales, 138–43, 144 and disorganised attachment, 377–78
avoidance, 139–40 effects, 528–33
proximity-seeking, 139 unresolved traumatic abuse, 303–9
resistance, 140–41 see also post-traumatic stress disorder (PTSD)
sensitivity, 122, 123, 124, 125–26, 172–73 Trivers, R., 224–25, 494–95
secure attachment, 150–56, 193–94 True, M., 193–94, 252–53
semi-standardisation of caregiving Tulving, E., 279–80
behaviour, 179
sensitivity see sensitivity of caregiver Uganda ethnography (Ainsworth), xxiv, 112,
and traditions of attachment 113, 116–19, 123, 127–28, 131
research, 463–64 availability of caregiver, 117, 118

video recordings, 256–57 interviews, 118–19
see also Ainsworth, M. proximity-seeking behaviour, 118–19
Stroop task, 479 separation, 118
structural issues facing attachment Understanding Attachment and Attachment
research, 537–41 Disorders (Prior and Glaser), xiii, 546–47
Suomi, S., 220–21 unresolved loss, 297–303
Supporting Older People Using Attachment- Unterschute, M.S., 524–25
Informed and Strengths-Based Uppsala Longitudinal Study, 220–21
Approaches (Blood and Guthrie), 546–47
symptoms, mental health van IJzendoorn, M. see IJzendoorn, M. van
conflict, displays of, 75–79 Vaughn, B.E.
defence concept, 68–71 and Ainsworth/Strange Situation,
meaning and structures, 68–79 138–39, 170–72
post-traumatic stress disorder (PTSD), 529, and Minnesota study, 337, 338–39
530, 532–33 Verhage, M., 558
segregation, 70–72, 76–78 Vicedo, M., 55–56, 548
Szepsenwol, O., 494 video-feedback, xix, 128–29, 256–57, 341–42
Video-feedback Intervention to promote
Takahashi, K., 190–91, 192, 193 Positive Parenting (VIPP), 564–67
Target, M., 12, 26–27, 310 vigilance, 217, 229, 492–93, 511
Taubman-Ben-Ari, O., 563 attentional, 226–27, 494–95
Tavistock Clinic, London, 1, 4, 8, 40–41, 133, 430 caregiver availability, 43–44, 224–25, 226–27,
Separation Research Unit, 35, 42–43 431–32, 448–49, 450, 472
see also Bowlby, J. hypervigilance, 318–19, 371, 431–32
taxometric analysis, 145–46, 181, 409, 411–12 perceptual, 43–44, 494–95
Tellegan, A., 390–91 threat-related, 454–56, 479–80, 497–98
Thompson, R.A., xiii–xiv, 29–30, 170–72, 187–88, Vincent, S., 545
201, 571, 572
Attachment, xiii Waddington, C.H., 88–89
Tinbergen, N., 17–18, 23–24, 239, 491 Wall, S., 166–67
Toth, S.L., 216, 404–6, 407, 408–9 Wanaza, A., 244
Tracey, R.I., 157 Ward, M.J., 355–56
Tracy, J.L., 496–97 Wartner, U., 166–67
transference, 83–85 Waters, E., 461–62, 540–41
trauma, xxiv–xxv ‘Attachment as an organizational
and abuse, 307–8 construct,’ 350
Index 601
Attachment Q-Sort (AQS) measure, 176–79, Winnicott, D., 3–4, 8–9, 99–100
410, 538 Wittig, B., 120, 121
and Bowlby, 67–68, 96–97, 547 Wolff, M.S. de, 149–50
and language, xxii–xxiii, 123 working memory, 226–27, 312–13
and Berkeley study, 219, 222–23, 232–33, 241, World Health Organisation (WHO),
250–51, 259 7–8, 181–82
and Minnesota study, 337, 339 , –62, 409 ‘Infancy, Childhood and Adolescent
emotional development, 349–50, 351–52 Disorders’ committee, 90–91
Patterns of Attachment see under International Classification of Diseases
Ainsworth, M. (ICD), 92
secure base scripts, 179–81 ‘Psychobiology of the Child’ study group,
and self-report measures, 470, 472, 474, 506 6–7, 23

and Strange Situation procedure, 109–10, 123, World Trade Organisation, Study Group of
140–41, 167–81, 201, 203–4 the Psychological Development of the
Waters, H., 179–81, 203–4 Child, 50–51
Wei, M., 515, 516–17
Weigand, R., 172–73 Yates, T.M., 385–86, 391, 392
Weingarten, S., 464–65 Young Children in War Time (Burlingham and
Weiss, R., 430 Freud), 8
Weller, A., 440–41
Wellman, H., 168–69 Zeanah, C., 92–93, 260
Werner, E.E., 390–91 Zeegers, M.A., 153–54
Weston, D.R., 213–15, 238–39, 348–49 Zigler, E., 384–85
White, K., Trauma and Loss, 6, 68–69, 91–92 Zilcha-Mano, S., 481–82
White, S., 542–45, 546–47, 548–49 Zimmerman, R., 115–16
William T. Grant Foundation, 109, 113, 212, 401 Ziv, Y., 327

Cornerstones of Attachment Research by ROBBIE DUSCHINSKY

Uploaded by

Copyright:

Available Formats

Cornerstones of Attachment Research by ROBBIE DUSCHINSKY

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cornerstones of Attachment Research by ROBBIE DUSCHINSKY

Uploaded by

Copyright:

Available Formats

Downloaded from https://academic.oup.

com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

This is an open access publication, available online and distributed under

1. John Bowlby and the Tavistock Separation Research Unit 1

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Review, 4(4), 306–​10.

Family Theory & Review, 4(4), 282–​9, p.282.

recognition of the importance of stable, trusting relationships for children’s socioemotional

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Partnership UK. https://​www.pipuk.org.uk/​sites/​default/​files/​PIPUK%20Rare%20Jewels%20FINAL.pdf. In the

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

inistvoices.com/​assets/​Women-​Past/​Ainsworth/​Mary-​Ainsworth-​CPA-​Oral-​History.pdf: ‘I think it was just the

attachment theory in practice. Child & Youth Services, DOI: 10.1080/​0145935X.2019.1583100.

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

New York: Guildford.

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Fraley, Klaus and Karin Grossmann, and Alan Sroufe.

attachment research. Attention to a number of research groups offers an ‘arsenal of exem-

Attachment and historical time

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

personality symptoms. Development & Psychopathology, 21(4), 1311–​34, p.1315.

One of the primary forms of attachment-​based intervention with families is video-​

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

as complementary developmental science. Science in Context, 32(3), 309–​26.

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Ordinary and scientific language

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

49(3), 474–​82, p.474.

fied. Psychoanalytic Inquiry, 35(1), 5–​23.

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

and facilitators to the use of historical sociology as complementary developmental science.

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Klein, 10 January 1984. PP/​Bow/​A.5/​7.

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Explorations (pp.423–​32). London: Karnac, p.427.

1976, 37–​41, p.38.

preschool children. American Journal of Orthopsychiatry, 51(1), 4–​28.

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

red pen for emphasis.

Capacity to Love. Oxford: Oxford University Press, p.29.

Downloaded from https://academic.oup.com/book/28760 by guest on 01 February 2024

perplexed parent. New York Times, 2 March 1974.

Bowlby and psychoanalysis

Review, 4(4), 306–10.

Family Theory & Review, 4(4), 282–9, p.282.

Partnership UK. https://www.pipuk.org.uk/sites/default/files/PIPUK%20Rare%20Jewels%20FINAL.pdf. In the

inistvoices.com/assets/Women-Past/Ainsworth/Mary-Ainsworth-CPA-Oral-History.pdf: ‘I think it was just the

attachment theory in practice. Child & Youth Services, DOI: 10.1080/0145935X.2019.1583100.

personality symptoms. Development & Psychopathology, 21(4), 1311–34, p.1315.

One of the primary forms of attachment-based intervention with families is video-

as complementary developmental science. Science in Context, 32(3), 309–26.

49(3), 474–82, p.474.

fied. Psychoanalytic Inquiry, 35(1), 5–23.

Klein, 10 January 1984. PP/Bow/A.5/7.

Explorations (pp.423–32). London: Karnac, p.427.

1976, 37–41, p.38.

preschool children. American Journal of Orthopsychiatry, 51(1), 4–28.

of Bowlby’s ideas by second-generation attachment researchers was criticism and revision

Affectional Bonds (pp.150–88). London: Routledge.

of Ed Tronick to developmental and psychotherapeutic thinking. Psychoanalytic Inquiry, 35(4), 355–69.

Psychological Works of Sigmund Freud, Vol. 17 (pp.1–124). London: Vintage, p.116.

Klein, 10 January 1984. PP/Bow/A.5/7.

87 Bowlby, J. (1933–36) Research notes for MD thesis. PP/Bow/D.2/44/3.

Washington 1986. PP/BOW/A.5/1.

Psychoanalytic Study of the Child, 25(1), 360–400.

Association, 11(3), 500–41, p.521.

text. Ethology, 120(3), 215–23.

Psychology, 30(4), 230–40.

Psychology, 30(4), 230–40.

125 Bowlby, J. (c.1955) Notes on child attachment and monotropy. PP/Bow/H146.

Bowlby, J. (1955) Letter to C.F. Rycroft, 26 April 1955. PP/Bow/H.116.

Psycho-Analysis, 25, 107–28, p.121.

‘disorganised/disoriented attachment’ had already been established by Main and Solomon as

168 Bowlby, J. (1986) Interview with the BBC. PP/Bow/F.5/8.

included ‘tempers, sullenness, disobedience, stealing, sleeplessness, bed-wetting, tim-

order. Child: Care, Health and Development, 36(5), 639–45.

the file Maternal Behaviour: Humans. PP/Bow/H.136.

Controversies 1941–45, 2nd edn. London: Routledge, p.369.

Child, 7, 82–94, p.85.

six days. Journal of Child Psychology & Psychiatry, 7, 179–97, p.187.