See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/237011695

Measuring Attachment Dimensions in Adolescents: Development and

Validation of the Experiences in Close Relationships — Revised — General
Short Form

Article in Journal of Relationships Research · October 2011

DOI: 10.1375/jrr.2.1.53

CITATIONS READS

53 6,625

1 author:

Ross B Wilkinson
University of Newcastle Australia
74 PUBLICATIONS 2,315 CITATIONS

SEE PROFILE

All content following this page was uploaded by Ross B Wilkinson on 05 March 2014.

The user has requested enhancement of the downloaded file.

 ARTICLE AVAILABLE ONLINE
Journal of Relationships Research

Measuring Attachment Dimensions in

Adolescents: Development and Validation
of the Experiences in Close Relationships
— Revised — General Short Form
Ross B. Wilkinson
Australian National University, Australia

The assessment of attachment in adolescence remains problematic, with the most widely used measure, the
Inventory of Parent and Peer Attachment (IPPA; Armsden & Greenberg, 1987), argued to have major short-
comings. This article reports two studies examining the psychometric properties of a shortened and modified
version of the Experiences in Close Relationships — Revised (ECR-R) (Fraley, Waller, & Brennan, 2000), the
Experiences in Close Relationships — Revised — General Short Form (ECR-R-GSF), for the assessment of
general relationship attachment anxiety and avoidance in adolescents and young adults. Confirmatory factor
analyses in two independent samples of high school and university students (Total N = 1187, 11 to 22 years)
demonstrate support for the two-factor model of attachment anxiety and avoidance. The measurement model
was supported across data sets and was not significantly different with respect to either age or sex. While the
ECR-R-GSF demonstrated appropriate convergent and discriminant validity with the Relationships
Questionnaire (RQ (Bartholomew & Horowitz, 1991) it did not converge with specific, parental relationship
attachment as assessed by the IPPA. The ECR-R-GSF is argued to be a reliable and psychometrically sound
instrument for assessing general attachment dimensions across a wide age range.

Keywords: attachment, adolescent, relationships, anxiety, avoidance, assessment

Research on attachment in adult and adolescent popu- lack of convergence between the conceptualisation and
lations has flourished since the 1987 publication of two assessment of attachment for adults and adolescents.
seminal papers that introduced self-report instruments While the literature on adult attachment has now
for the assessment of this key relationship construct. evolved to focus on major dimensions of individual dif-
Hazan and Shaver (1987), drawing on Ainsworth, ferences in attachment, such as anxiety and avoidance
Blehar, Waters, and Wall’s (1978) categorical approach (Brennan, Clark, & Shaver, 1998; Mikulincer &
to the assessment of attachment aspects of infant– Shaver, 2007), the literature on adolescent attachment
mother interactions, developed a three category has been more focused on attachment security in differ-
self-report instrument for assessing individual differ- ent types of relationships (e.g., Buist, Reitz, & Dekovic,
ences in adult romantic attachment. Around the same 2008; Markiewicz, Lawford, Doyle, & Haggart, 2006;
time, Armsden and Greenberg (1987), building on Wilson & Wilkinson, 2007). The current article pres-
earlier work (Greenberg, Siegel, & Leitch, 1983), inde- ents two studies on the modification of an instrument
pendently developed a multi-scale instrument designed
to assess attachment aspects of adolescents’ attachment
relationships to their parents and peers. Both papers ADDRESS FOR CORRESPONDENCE: Dr Ross Wilkinson, Department
have had a profound influence on subsequent research of Psychology, Building 39, Australian National University, ACT
yet there has remained a division in the literature with a 0200, Australia. E-mail: [email protected]

Wilkinson, R.B. (2011). Measuring attachment dimensions in adolescents: Development and validation of the Experiences in Close
Relationships — Revised — General Short Form. Journal of Relationships Research, 2, 53–62. DOI 10.1375/jrr.2.1.53
53
 ROSS B. WILKINSON

designed to assess the two major dimensions of adult the measure’s sensitivity across the range of possible
attachment, the Experiences in Close Relationships — scores. The ECR-R has the same number of items as
Revised (ECR-R), for use in adolescent populations and the ECR, and Mikulincer and Shaver (2007) report
its subsequent psychometric evaluation. that the ECR-R and ECR scales correlate at around .95.
Following Hazan and Shaver’s (1987) development Sibley and Liu (2004) conducted a confirmatory factor
of the Attachment Style Prototype measure (ASP), a analysis of the ECR-R and their results support the
number of authors began to argue that it may be better, original two-factor solution over alternative one and
for both psychometric and conceptual reasons, to three factor solutions. They also found six-week test–
measure attachment constructs as dimensions rather retest coefficients in the low .90s for both scales. Sibley,
than as discrete categories (Bartholomew, 1994; Fischer, and Liu (2005) conducted three studies exam-
Brennan et al., 1998; Griffin & Bartholomew, 1994a). ining the factor structure and discriminant and
From quite soon after the publication of the original convergent validity of the ECR-R. They found that the
ASP, various authors had begun to modify it to generate ECR-R and Relationship Questionnaire ratings con-
continuous ratings rather than discrete categories (e.g., verged appropriately, but that the ECR-R was more
Levy & Davis, 1988; Collins & Read, 1990), and stable over a 3-week period. Using a diary methodology
various instruments based on the three attachment con- they found that the ECR-R predicted interactions with
structs of security, anxiety, and avoidance had already romantic partners, but was generally not related to inter-
begun to appear (e.g., Simpson, 1990). Bartholomew actions with family members or friends. They note that:
and Horowitz’s (1991) paper presenting a four-category ‘It remains to be seen, however, whether the ECR-R may
model of attachment style was also influential in the be suitably reworded to assess different aspects of (nonro-
push to develop dimensional, multi-item, self-report mantic) working models or whether entirely new scales
inventories. They presented a conceptual scheme in are necessary’ (p. 1534). Both the ECR and ECR-R are
which Bowlby’s putative working models of self and widely employed in the research literature examining
other formed two orthogonal dimensions that were adult attachment and may be seen as the ‘standard’ self-
argued to underlie the secure, preoccupied, dismissing, report measures of attachment dimensions.
and fearful attachment style categories (Griffin & Although there has been a long history of theory and
Bartholomew, 1994b). research examining the role of interpersonal relation-
In their 1998 review of the existing self-report meas- ships in adolescence (e.g., Coleman, 1961; Douvan, &
ures of attachment, Brennan et al. argued that Adelson, 1966; Erikson, 1968; Freud, 1958; Hall,
Ainsworth and colleagues’ (1978) original discriminant 1904), the more recent application of attachment
function analysis of their Strange Situation data had theory formulations has had a profound influence on
also suggested two dimensions were underlying attach- our understanding of relationship dynamics in this
ment categories.1 Brennan and her colleagues labelled period (Allen, 2008). As with the adult attachment lit-
these dimensions Anxiety and Avoidance and, in an erature, however, the field is divided by methodological
attempt to generate some coherence to the burgeoning approach between those adhering to interview-based
field of attachment assessment, attempted to create a assessment protocols and those accepting the viability of
new measure of these dimensions by employing 482 self-report instruments.
items from 60 available attachment construct measures. Of the self-report measures, the most frequently used
Their subsequent analyses produced the Experiences in by far has been Armsden and Greenberg’s (1987)
Close Relationships measure (ECR), a 36-item instru- Inventory of Parent and Peer Attachment (IPPA; Wilson
ment assessing the two relatively orthogonal constructs & Wilkinson, 2007). The origins of the IPPA lay in an
of attachment anxiety and avoidance. Mikulincer and earlier instrument, the Inventory of Adolescent
Shaver (2007) report ‘… (the ECR) has been used in Attachments (IAA; Greenberg et al., 1983), and neither
hundreds of studies since 1998, always with high relia- the IAA nor the later IPPA attempted to follow the
bility (the alpha coefficients are always near or above attachment styles paradigm employed in assessing infant
.90, and test–retest coefficients range between .50 and and adult attachment. Rather, these instruments focused
.75 depending on the time span and nature of the on assessing the quality of relationship as attachment.
sample)’ (p. 91). Further, the validity of the ECR has Importantly, because they recognised the developmental
been demonstrated in both observational and experi- changes in adolescent relationships, they also accepted
mental studies (Mikulincer & Shaver, 2007). that peers as well as parents could fulfil some aspects of
Fraley, Waller, and Brennan (2000), using the same an attachment relationship.
data set as Brennan et al. (1998), conducted an item The IPPA has a 28-item Parent scale (IPPA-Pa) and a
response theory analysis of the ECR items and subse- 25-item Peer scale (IPPA-Pe) each with Trust,
quently produced the Experiences in Close Alienation, and Communication subscales. Summary
Relationships — Revised (ECR-R) in order to improve Parental and Peer scores are created by summing the

54 Journal of Relationships Research

 MEASURING ATTACHMENT DIMENSIONS IN ADOLESCENTS

Trust and Communication scores and then subtracting secure base attachment functions; thus, a modified
the Alienation scores. Armsden and Greenberg (1987) measure would need to have items assessing general
argue that a self-report instrument, such as the IPPA, attachment relationship representations rather than
can access both behavioural and cognitive aspects of those specific to romantic relationships. Further, to
attachment that are part of the ‘internal working enhance the utility of the scale, it was also decided that
models’ that individuals develop. In developing the any modified measure should also be shortened to
IPPA they rejected the categorical model of attachment encourage its use in both research and clinical settings.
differences as being restricted to behavioural aspects, as Two studies are presented as part of the development
exemplified by Ainsworth and colleague’s (1978) work and evaluation process. In the first study, the items for
with infants, and were more informed by what they felt the modified scale will be selected and then the factor
were the affective/cognitive aspects of reactions to structure of the resulting measure evaluated in relation
attachment figure responsiveness. Various short-form to adherence to the widely supported two-factor struc-
versions and modifications of the IPPA are commonly ture of attachment anxiety and avoidance. The measure
used in the literature (e.g., Vignoli & Mallet, 2004; will also be evaluated in relation to attachment style
Liang, Hou, & Tian, 2006; Wilkinson, 2006). categories and the IPPA evaluation of maternal, pater-
Research has accumulated extensive evidence of the nal, and peer attachment. The second study will
relationship of the IPPA and its variants to various attempt to replicate the factor structure in an independ-
aspects of adolescent adjustment, including wellbeing, ent sample and to evaluate if the measurement model
anxiety, self-esteem, school attitudes, anti-social behav- varies depending on age or sex of participants.
ior, and empathy (Armsden, McCauley, Greenberg,
Burke, & Mitchell, 1990; Laible, Carlo, & Roesch,
2004; Marcus & Betzer, 1996; Paterson, Prior, & Field, Study 1
1995; Raja, McGee, & Stanton, 1992; Vivona, 2000; METHOD
Wilkinson, 2010; Wilkinson & Kraljevic, 2004; Participants
Wilkinson & Walford, 2001). However, the fact Participants were 479 female and 215 male high school
remains that there is doubt over the extent to which the (n = 580) and college (n = 118) students aged from 14
IPPA actually operationalises a contemporary under- to 22 years (Mean = 17.23, SD = 1.54) sampled from
standing of individual differences in attachment the Australian Capital Territory. Of this sample, 85.2%
(Mikulincer & Shaver, 2007). reported of being Australian or European background,
Apart from the fact that the IPPA has a different 7.5% of Asian background, with the remainder report-
conceptual basis to the adult self-report measures of ing ‘other’ ethnic backgrounds. Based on reports of
attachment, another major shortcoming is that it is father and mother occupations, the majority (> 80%)
debatable whether it was actually developed on a repre- reported being of upper and middle socio-economic
sentative sample of adolescents. Armsden and status.
Greenberg (1987) employed two samples of undergrad-
uate college students with a total N of only 179 and a Procedure and Materials
mean age 18.9 years (range 16 to 20 years). Certainly A questionnaire booklet was administered as part of a
this does not sample the full age range of adolescence larger study examining interpersonal relationships and
and may be said to be more representative of young psychological health in adolescents and young adults.
adults. In fact, this is a similar mean age to the samples For the high school students, the questionnaire was
used in many of the studies of adult attachment (e.g., administered in classroom settings while for college par-
Brennan, Clark, & Shaver, 1998). A further limitation ticipants the questionnaire was completed individually.
is that the original version of the IPPA targets specific
types of relationships (parents and peers) but does not The Experiences in Close Relationships — Revised
distinguish between specific relationship partners (e.g., General Short Form (ECR-R-GSF). Twenty items, 10
mother, father, or best friend, friends, clique members). from the Anxiety scale, and 10 from the Avoidance
scale, were selected from the ECR-R. A number of cri-
THE PRESENT STUDY teria were used as the basis for selection. First, the items
The aim of the present study was to evaluate if a widely needed to be amenable to alteration to wording to suit
used and well-validated self-report measure of adult a general rather than romantic partner attachment.
attachment dimensions, the ECR-R, could be modified Second, items also needed to be suitable for adolescents
successfully for use with adolescent samples. To and young adults. Items were checked for their appro-
examine this appropriately, samples representative of priate wording with a pilot sample of 20 adolescents.
the full adolescent age range will be employed. Because Table 1 presents the items selected from the ECR-R
of their developmental stage, adolescents generally do and the reworded versions employed for the ECR-R-
not have romantic relationships that have associated GSF. Participants are asked to rate each item on a

Journal of Relationships Research 55

 ROSS B. WILKINSON

TABLE1
ECR-R Items and Corresponding ECR-R-GSF Items

ECR-R item ECR-R-GSF

I prefer not to show a partner how I feel deep down* I prefer not to show others how I feel deep down
I feel comfortable sharing my private thoughts and feelings I feel comfortable sharing private thoughts and feelings with
with my partner* other people
I find it difficult to allow myself to depend on romantic partners* I find it difficult to allow myself to depend on other people
I often worry that my partner doesn’t really love me I often worry that other people close to me don’t really love me
I am very comfortable being close to romantic partners* I am very comfortable being close to other people
I often worry that romantic partners won’t care about me as I often worry that other people don’t care as much about me as
much as I care about them* I care about them
I worry a lot about my relationships* I worry a lot about my relationships
When I show my feelings for romantic partners, I’m afraid they When I show my feelings to people I care about, I’m afraid that
will not feel the same about me they will not feel the same about me
I usually discuss my problems and concerns with my partner* It is usually easy for me to discuss my problems and concerns
with other people
My romantic partner makes me doubt myself My relationships with people make me doubt myself
It helps to turn to my romantic partner in times of need* It helps to turn to others for support in times of need
I find that my partner(s) don’t want to get as close as I would like* I find that other people don’t want to be as close as I would like
Sometimes romantic partners change their feelings about me Sometimes people change their feelings about me for no
for no apparent reason apparent reason
I am nervous when partners get too close to me* I am nervous when people get too emotionally close to me
My desire to be close sometimes scares people away* My desire to be close sometimes scares people away
I feel comfortable depending on romantic partners* I feel comfortable depending on other people
I’m afraid that once a romantic partner gets to know me, I am afraid that once somebody gets to know me, he or she
he or she won’t like who I really am won’t like who I am
I find it easy to depend on romantic partners I find it easy to depend on other people
It makes me mad that I don’t get the affection and support It makes me mad that I don’t get the affection and support
I need from my partner I need from other people
It’s easy for me to be affectionate with my partner It is easy for me to be affectionate with other people
Note: * These items are also on the original ECR measure.

5-point scale from 1 (Strongly disagree) to 5 (Strongly The Relationship Questionnaire — General Version.
agree). Scale scores were created by recoding as neces- Based on the original Relationship Questionnaire
sary and taking the mean of relevant items. Based on (Bartholomew & Horowitz, 1991), a modified version
the proposed structure of the ECR-R-GSF, both the was created that asked about relationships in general
Anxiety and Avoidance items produced reliable scales rather than romantic relationships. Both ratings and
(Cronbach’s Alpha = .883 and .884 respectively) and categories were employed. Participants are asked to read
were correlated at .393. four descriptions of relationship attitudes matching
four categories of attachment style: Secure, Fearful,
The Inventory of Parent and Peer Attachment Short Dismissing, and Preoccupied. They are then asked to
Form. Based on the original IPPA (Armsden & rate each description for ‘how much it is like them’ (1 =
Greenberg, 1987) a 45-item short form of the IPPA was Not at all like me to 7 = Very much like me). Next, they
employed that included 15-item Maternal, Paternal, are asked to nominate one style that is ‘most like them’.
and Peer attachment scales. These short-forms have Bartholomew and Horowitz report appropriate stability
produced good internal consistencies (Mother of the RQ and good convergence between the RQ and
Attachment α = .91, Father Attachment α = .92, Peer an interview-based measure of attachment style.
Attachment α = .85) in previous research (Wilkinson,
2006). Participants are asked to rate each item on a 5- STUDY 1: RESULTS AND DISCUSSION
point scale from 1 (Almost never or never) to 5 (Always Examination of the univariate distributions for the
or almost always). Scale scores were created by recoding items revealed some variation from normality but no
as necessary and summing relevant items. major violations (Table 2). Based on the procedure out-

56 Journal of Relationships Research

 MEASURING ATTACHMENT DIMENSIONS IN ADOLESCENTS

TABLE 2
Descriptive Statistics for ECR-R-GSF Items

Item M SD Skew Kurt.

1. I prefer not to show others how I feel deep down. 3.223 1.052 -0.114 -0.430
2. I often worry that other people close to me don’t really love me. 2.517 1.082 0.343 -0.421
3. I find it difficult to allow myself to depend on other people. 2.992 1.060 0.072 0.841
4. I often worry that other people don’t care as much about me as I care about them. 2.944 1.134 0.029 -1.013
5. I am very comfortable being close to other people. (r) 3.743 0.896 -0.641 0.112
6. Sometimes people change their feelings about me for no apparent reason. 2.581 1.013 0.572 -0.213
7. It is usually easy for me to discuss my problems and concerns with other people. (r) 3.209 1.060 -0.435 -0.593
8. My desire to be close sometimes scares people away. 2.173 0.902 0.712 0.344
9. It helps to turn to others for support in times of need. (r) 3.810 0.882 -0.753 0.721
10. My relationships with people make me doubt myself. 2.421 1.016 0.543 -0.400
11. I am nervous when people get too emotionally close to me. 2.504 1.074 0.503 -0.362
12. When I show my feelings to people I care about, I’m afraid that they will not feel 2.750 1.100 0.031 -0.623
the same about me.
13. I find it easy to depend on other people. (r) 3.012 1.023 -0.052 -0.623
14. I am afraid that once somebody gets to know me, he or she won’t like who I am. 2.622 1.042 0.431 -0.454
15. It is easy for me to be affectionate with other people. (r) 3.457 0.994 -0.413 0-.222
16. It makes me mad that I don’t get the affection and support I need from other people. 2.462 0.973 0.372 -0.111
17. I feel comfortable sharing private thoughts and feelings with other people. (r) 3.121 1.042 -0.248 -0.422
18. I worry a lot about my relationships. 2.989 1.124 0.032 -0.541
19. I feel comfortable depending on other people. (r) 3.084 0.994 -.0233 -0.511
20. I find that other people don’t want to be as close as I would like. 2.512 0.923 0.311 -0.291
Note: Odd items are Avoidance items and even numbers are Anxiety items. Reverse coded items are indicated with (r).

lined by Tabachnick and Fidell (2007) six cases where using the criterion outlined by Byrne (2010).
identified as multivariate outliers with Mahalanobis dis- Standardised regression weights ranged from .505 to
tance less than p < .001 and where subsequently .778. The two factors were moderately correlated at
removed from further analyses. .478. Factor loadings are presented in Table 4 and
The proposed factor structure, which allowed covari- values for correlated error terms in Table 5.
ance between the two latent variables, was evaluated To evaluate construct validity the Avoidance and
using confirmatory factor analysis. Maximum likeli- Anxiety scores were correlated with the scales from the
hood estimation was employed with χ2, CFI, AGFI, RQ and the IPPA (Table 6). Anxiety was moderately
and RMSEA evaluated for model fit. The hypothesised negatively correlated with the RQ Secure ratings and
model produced a suboptimal fit, χ2 = 956.05, df = moderately to strongly, positively correlated with both
169, p < .001; AGFI = .833; CFI = .867, RMSEA = Fearful and Preoccupied ratings. It was only weakly cor-
.082 (90% CI = .077–.087). Examination of the modi- related with the Dismissing ratings. In terms of the
fication indices suggested including error covariances IPPA scales, Anxiety was moderately negatively corre-
for a number of variable pairs. A stepwise procedure lated with Peer Attachment but only weakly, negatively
was undertaken with individual changes made based on correlated with both Maternal and Paternal
the highest modification index. The criteria used was Attachment. Avoidance was moderately, negatively cor-
that the modification must be consistent with the related with the RQ Secure ratings and moderately to
overall proposed model and be explainable in terms of strongly, positively correlated with the Fearful and
shared item content, the index must be greater than 30 Dismissing ratings. It was weakly and positively corre-
and the change in χ2 must be significant. Table 3 pres- lated with Preoccupied ratings. With regard to the IPPA
ents the development of the modified CFA models. scales, Avoidance was weakly to moderately, negatively
Modifications ceased at the fifth step with a version of correlated with Maternal and Paternal Attachment and
the model that has an appropriate fit. Although χ2 was moderately to strongly, negatively correlated with Peer
still significant (665.73, df = 164, p < .001), the Attachment.
remaining fit indices, AGFI = .882; CFI = .937, Mean levels of the Anxiety and Avoidance scales were
RMSEA = .058 (90% CI = .052–.063), were acceptable also examined across the categories of the RQ. A one-

Journal of Relationships Research 57

 ROSS B. WILKINSON

TABLE 3
Modifications to the Hypothesised Model

Model Error covariance added ∆χ2 df p AGFI CFI RMSEA

1 Item 3 with Item 19 92.685 1 < .001 .852 .883 .075
2 Item 8 with Item 20 56.541 1 < .001 .862 .891 .071
3 Item 7 with Item 17 59.211 1 < .001 .871 .912 .067
4 Item 3 with Item 13 36.142 1 < .001 .876 .923 .063
5 Item 3 with Item 19 45.740 1 < .001 .882 .937 .058
Note: Items 3, 13, and 19 all refer to depending on other people.
Items 8 and 20 both refer to being ‘close’ to others.
Items 7 and 17 both have the theme of sharing or discussing things with other people.

way MANOVA was conducted with Anxiety and F = 105.84, df = 3, 682, p < .001, partial η2 = .318,
Avoidance as the dependent variables and RQ Category were significantly different across the four categories.
(Secure, Fearful, Preoccupied, Dismissing) as the inde-
Post-hoc Bonferroni tests with adjustment for multiple
pendent variable. Means and standard deviations are
comparisons revealed that the Preoccupied and Fearful
presented in Table 7. Because of the disproportionate
groups reported the highest level of Anxiety. There were
distribution of cell sizes results were also confirmed
with non-parametric tests. There was a significant mul- no significant differences between Secure and
tivariate effect across the categories, Wilk’s λ = .053. F Dismissing, and between Preoccupied and Fearful. All
= 6033.24, df = 2, 681, p < .001, partial η2 = .947. other comparisons were significant. Dismissing and
Univariate tests revealed that both Anxiety, F = 63.01, Fearful categories reported the highest levels of
df = 3, 682, p < .001, partial η2 = .217, and Avoidance, Avoidance. There was no significant difference between

TABLE 4
Factor Loadings for the Final Model Evaluated Separately for Study 1 and Study 2 Data

Item Study 1 Study 2

Avoidance Anxiety Avoidance Anxiety
1. I prefer not to show others how I feel deep down. .693 .586
2. I often worry that other people close to me don’t really love me. .720 .727
3. I find it difficult to allow myself to depend on other people. .651 .606
4. I often worry that other people don’t care as much about me as I care about them. .778 .724
5. I am very comfortable being close to other people. (r) -.697 -.609
6. Sometimes people change their feelings about me for no apparent reason. .571 .526
7. It is usually easy for me to discuss my problems and concerns with other people. (r) -.739 -.720
8. My desire to be close sometimes scares people away. .505 .450
9. It helps to turn to others for support in times of need. (r) -.615 -.576
10 My relationships with people make me doubt myself. .690 .692
11. I am nervous when people get too emotionally close to me. .538 .331
12. When I show my feelings to people I care about, I’m afraid that they will
not feel the same about me. .677 .727
13. I find it easy to depend on other people. (r) -.600 -.646
14. I am afraid that once somebody gets to know me, he or she won’t
like who I am. .663 .705
15. It is easy for me to be affectionate with other people. (r) -.517 -.556
16. It makes me mad that I don’t get the affection and support I need from
other people. .624 .554
17. I feel comfortable sharing private thoughts and feelings with other people. (r) -.691 -.691
18. I worry a lot about my relationships. .599 .563
19. I feel comfortable depending on other people. (r) -.703 -.674
20. I find that other people don’t want to be as close as I would like. .683 .675

58 Journal of Relationships Research

 MEASURING ATTACHMENT DIMENSIONS IN ADOLESCENTS

TABLE 5 model of self. Those who endorsed descriptions associ-

Correlated Error Terms for Study 1 and Study 2 ated with a negative model of other reported more
attachment avoidance than those who endorsed descrip-
Paths Study 1 Study 2
tions associated with a positive model of other.
Item 3 and Item 13 -.352 -.331 With respect to convergence with the IPPA scales,
Item 3 and Item 19 -.294 -.232 consistent with previous observations (e.g., Mikulincer
Item 7 and Item 17 .299 .144 & Shaver, 2007), attachment avoidance tended to have
Item 13 and Item 19 .446 .460 stronger relationships with the IPPA scales than did
Item 8 and Item 20 .302 .148 attachment anxiety. Further, these were quite modest
for the maternal and paternal scales and somewhat
stronger for the peer scale. The stronger relationship
the Dismissing and Fearful categories but all other com- with the IPPA-Peer scale may have occurred because it
parisons were significant. refers to peers in general rather than asking the partici-
Overall, the general measurement model for individ- pant to focus on a particular peer, like a best friend for
ual differences in attachment anxiety and avoidance was example. Thus, both the IPPA-Peer scale and the ECR-
supported by the operationalisation represented by the R-GSF are tapping more general relationship
ECR-R-GSF. Only minor modifications to the pro- representations while the IPPA-Mother and Father
posed latent variable structure, involving the inclusion scales are assessing specific relationships.
of some correlated error terms, were required to
produce a model that was an acceptable fit to the data. Study 2
A group of items (3, 13, and 19) that all included some In order to establish the stability of the measurement
reference to dependency on others were correlated inde- model and to confirm that the model would still hold
pendently of the relationship with the common for younger adolescents and across gender, a further
avoidance factor. This suggests that there is potentially study using an independent sample of adolescents was
some distinction between attachment avoidance and undertaken.
the perception of dependency in relationships that war-
rants further clarification. METHOD
In terms of convergent validity, the anxiety and Participants
avoidance scales of the ECR-R-GSF were appropriately Participants were 343 female and 150 male high school
correlated with the ratings of the four RQ descriptions. students aged from 11 to 19 years (Mean = 15.61, SD =
Consistent with expectations the avoidance scale had 2.10). Of this sample, 84% reported being
moderate to strong correlations with the RQ ratings Australian/European background, 10.6% of Asian
that involved a model of other (i.e., Dismissing and background, with the remainder reporting ‘other’
Fearful) while the anxiety scale evinced its highest cor- ethnic background. Based on reports of father and
relations with the RQ ratings involving the negative mother occupations the majority (> 80%) reported
model of self (i.e., Preoccupied and Fearful). These being of upper or middle socio-economic status.
results were essentially replicated when using the cate-
Procedure and Materials
gorical results from the RQ. Those who endorsed
descriptions associated with a negative model of self The procedure was similar to that of Study 1.
reported higher attachment anxiety scores than those The Experiences in Close Relationships — Revised
who endorsed descriptions associated with a positive General Short Form (ECR-R-GSF). The ECR-R-GSF,
developed in Study 1, was used to assess the anxiety and
avoidance dimensions of attachment.
TABLE 6
Study 1: Correlations of Anxiety and Avoidance with Alternative STUDY 2: RESULTS AND DISCUSSION
Attachment Measures Examination of the univariate distributions for the
Measures Avoidance Anxiety items revealed some variation from normality but no
major violations. Based on the procedure outlined by
RQ — Secure -.501 -.368
Tabachnick and Fidell (2007) three cases where identi-
RQ — Fearful .610 .420
fied as multivariate outliers with Mahalanobis distance
RQ — Preoccupied .266 .592
less than p < .001 and where subsequently removed
RQ — Dismissing .446 .128 from further analyses.
IPPA — Mother -.384 -.249 The model developed in Study 1 was applied and
IPPA — Father -.313 -.181 evaluated with the Study 2 data and produced an
IPPA — Peer -.596 -.440 acceptable fit, χ2 = 527.12, df = 164, p < .001; AGFI =

Journal of Relationships Research 59

 ROSS B. WILKINSON

TABLE 7
Means and Standard Deviations for Anxiety and Avoidance by Relationship Questionnaire Category

Measures Secure Fearful Preoccupied Dismissing

n = 425 n = 109 n = 116 n = 36
M (SD) M (SD) M (SD) M (SD)
Anxiety 2.382 (0.622)a 3.011 (0.683)b 3.178 (0.633)b 2.443 (0.839)a
Avoidance 2.433 (0.583) a
3.390 (0.527) b
2.953 (0.610) c
3.415 (0.678)b
Note: Within rows comparisons with different superscripts are significantly different at p < .001.

.889; CFI = .915, RMSEA = .063 (90% CI = .058– supported when assessing general, rather than romantic,
.068). Standardised regression weights for items ranged attachment in adolescent samples. The ECR-R-GSF, as
from .331 to .727 (Table 4). Correlated error terms a relatively brief measure of differences in general
ranged from .144 to .460 (Table 5). The two factors attachment dimensions, was found to produce attach-
were moderately correlated at .432. A simultaneous ment anxiety and avoidance scales that are internally
evaluation of both data sets without invariance restric- consistent and structurally similar to the full-length,
tions was then conducted to generate a baseline fit. This romantic attachment versions employed with adults.
indicated that both samples were a good fit to the Further, these scales appear to be psychometrically
model, χ2 = 1178.82, df = 326, p < .001; AGFI = .883; equivalent for the full age range of adolescence and for
CFI = .913, RMSEA = .047 (90% CI = .044–.050). A adolescent girls and boys.
model was then evaluated with the factor loadings and From a validity standpoint the ECR-R-GSF scales
factor variances constrained to be equal across the evince the expected convergent and discriminant rela-
groups, χ2 = 1207.12, df = 347, p < .001; AGFI = .887; tionships with the assessment of attachment styles as
CFI = .912, RMSEA = .046 (90% CI = .043–.049), operationalised by the Relationships Questionnaire.
and was not statistically different to the baseline fit, ∆χ2 When participants are asked to consider attachment
= 28.30, df = 21, p > .05. Finally, a model was evaluated style descriptions that are ‘most like them’, then their
with the factor loadings, factor variances, and correlated ECR-R-GSF scale scores are consistent with the
error terms (Table 7) invariant across the samples, χ2 = working models that are argued to underlie those
1214.60, df = 352, p < .001; AGFI = .887; CFI = .911, descriptions. When participants are asked to rate those
RMSEA = .045 (90% CI = .042–.048). Again, this was same descriptions, rather than selecting a single descrip-
not statistically different to the baseline fit, ∆χ 2 = tion, again the scores are consistent with the
35.78, df = 26 p > .05, indicting no differences across ECR-R-GSF scale scores. This convergence is typical of
the two samples. that found with the full-length, romantic version of the
To evaluate sex and age differences in the measure- ECR-R and the RQ (Brennan et al., 2000), and despite
ment model the data from the two studies were some debate over the extent to which these measures
combined and split into sex (male, female) and age have different underlying conceptual models
(younger: 11 to 17 years, older: 17.1 to 22.9 years) (Mikulincer & Shaver, 2007).
groups. Comparisons with the baseline model revealed The pattern of relationships the ECR-R-GSF
that neither age, ∆χ2 = 34.18, df = 26 p > .05, nor sex, Anxiety and Avoidance scales display with the IPPA
∆χ2 = 34.37, df = 26 p > .05 produced significant dif- Mother, Father, and Peer scales is consistent with the
ferences in the constrained measurement models. view that the formulation of attachment as represented
Overall, the results of the second study, using an by the IPPA is not convergent with the two dimen-
independent sample of adolescents, replicated the CFA sional model dominant in the adult attachment
results from the first study. The factor loadings, latent literature. While further evidence needs to be gathered,
variable covariances, and additional error covariances there is some support to the argument that the IPPA
were not statistically different across the samples and does not adequately assess general attachment anxiety
and is more weighted towards the assessment of attach-
the overall model was an acceptable fit. Importantly,
ment avoidance. This may be a function of the IPPA’s
when the two samples were combined, there was no
specific rather than general focus with regards to
indication that the final model was different for
Mother and Father attachment. By contrast, the Peer
younger and older adolescents or for males and females.
scale of the IPPA does not focus on a specific relation-
ship but rather a specific class of relationships. The fact
General Discussion that this scale has a somewhat higher correlation with
The results of these two studies confirm that the two- the ECR-R-GSF Anxiety scale than the IPPA Mother
factor model of attachment avoidance and anxiety is and Father scales may be because of this more general

60 Journal of Relationships Research

 MEASURING ATTACHMENT DIMENSIONS IN ADOLESCENTS

rather than specific relationship specification. As Sibley attachment in both younger and older adolescents. The
et al. (2005) have noted, it may be the case that ver- results are encouraging and indicate that this new
sions of the ECR-R that specify specific rather than instrument is superior to the IPPA when the assessment
general relationships may produce different results. of the two dimensions of attachment anxiety and avoid-
Versions of the ECR-R that specifically focus on ance is required. The brevity of the instrument and its
Mothers or Fathers may produce more convergence applicability across a wide age range suggest that it
with the IPPA Mother and Father scales when assessing could be a valuable tool to assist in further understand-
adolescent attachment. It is questionable, however, ing the role of attachment in relationships and
whether the items in such a modified measure would psychological adjustment.
adequately assess attachment with regard to parental
relationships (Wilson & Wilkinson, 2007). Further Endnote
research is required to establish whether or not other 1 This has since been replicated in both Strange Situation and
versions of the ECR-R can be produced that appropri- Adult Attachment Interview data (Fraley & Spieker, 2003;
ately assess attachment as it is related to specific Roisman, Fraley, & Belsky, 2007).
relationships.
A strength of the current study, in comparison to
many other studies in this area, is that the combined
References
Ainsworth, M.S., Blehar, M.C., Waters, E., & Wall, S. (1978).
samples employed are larger and cover more fully the
Patterns of attachment: A psychological study of the strange situa-
entire adolescent age range. Further, the issue of potential tion. Potomac, MD: Lawrence Erlbaum.
age and sex differences was specifically addressed without Allen, J.P. (2008). The attachment system in adolescence. In J.
finding any significant differences in the measurement Cassidy & P.R. Shaver (Eds.), Handbook of Attachment;
structure. It is acknowledged, however, that self-report Theory, research, and clinical applications (2nd ed., pp. 419–
was the single methodological approach taken with 435). New York: Guilford Press.
regard to the assessment of the constructs. It has been Armsden, G.C., & Greenberg, M.T. (1987). The Inventory of
argued that psychometric validity may be best established Parent and Peer Attachment: Individual differences and their
through the use of multiple methodologies (e.g., Kaplan relationship to psychological well-being in adolescence.
Journal of Youth and Adolescence, 16, 427–454.
& Saccuzzo, 2005). There are a number of significant
Armsden, G.C., McCauley, E., Greenberg, M.T., Burke, P.M., &
problems, however, in comparing the self-report assess- Mitchell, J.R. (1990). Parent and peer attachment in early
ment of attachment to major, non-self-report approaches adolescent depression. Journal of Abnormal Child Psychology,
to attachment assessment, such as the Adult Attachment 18, 683–697.
Interview (AAI) (Main & Goldwyn, 1984). Although Bartholomew, K. (1994). Assessment of individual differences in
there is good evidence to support the view that AAI adult attachment. Psychological Inquiry, 5, 23–27.
results have an underlying bidimensional structure Bartholomew, K., & Horowitz, L.M. (1991). Attachment styles
(Roisman, Fraley, & Belsky, 2007), convergence between among young adults: A test of a four-category model. Journal
that structure and that produced by self-report measures of Personality and Social Psychology, 61, 226–244.
is poor and averages on r = .09 (Roisman, Holland, Brennan, K.A., Clark, C.L., & Shaver, P.R. (1998). Self-report
measurement of adult attachment: An integrative overview. In
Fortuna, Fraley, Clausell, & Clark, 2007). There is J.A. Simpson & W.S. Rholes (Eds.), Attachment theory and
increasing evidence that self-report measures of attach- close relationships (pp. 47–76). New York: Guilford Press.
ment and the AAI are measuring distinct aspects of Buist, K.L., Reitz, E., & Dekovic, M. (2008). Attachment stabil-
relationships and that, although they share a common ity and change during adolescence: A longitudinal application
theoretical basis, the constructs they assess are distinct of the social relations model. Journal of Social and Personal
(Mikulincer & Shaver, 2007). Relationships, 25, 429–444.
Byrne, B.M. (2010). Structural equation modeling with AMOS:
Basic concepts, applications, and programming (2nd ed.). New
Conclusions York: Routledge.
Research on attachment relationships and their impacts Coleman, J. (1961). The adolescent society: The social life of a
is burgeoning, yet the measures used to assess these con- teenager and its impact on education. New York: Basic Books.
structs remain controversial. Although there has been Collins, N.L., & Read, S.J. (1990). Adult attachment, working
some consensus among groups of researchers over what models, and relationship quality in dating couples. Journal of
Personality and Social Psychology, 58, 644–663.
approach to take and what specific instruments to use
Douvan, E., & Adelson, J. (1966). The adolescent experience.
(i.e., self-report versus interview; ECR-R versus IPPA)
New York: Wiley.
this has to some extent compartmentalised progress in
Erikson, E. (1968). Identity: Youth and crisis. New York: Norton.
the area. The current study aimed to demonstrate that a
Fraley, R.C., & Spieker, S.J. (2003). Are infant attachment pat-
modified version of an instrument widely employed by terns continuously or categorically distributed? A taxometric
social psychologists researching attachment, the ECR- analysis of Strange Situation behavior. Developmental
R-GSF, could be applied to assessing general Psychology, 39, 387–404. doi: 10.1037/0012–1649.39.3.387

Journal of Relationships Research 61

 ROSS B. WILKINSON

Fraley, R.C., Waller, N.G., & Brennan, K.A. (2000). An item Paterson, J.E., Prior, J., & Field, J. (1995). Adolescent attach-
response theory analysis of self-report measures of adult ment to parents and friends in relation to aspects of
attachment. Journal of Personality and Social Psychology, 78, self-esteem. Journal of Youth & Adolescence, 24, 365–376.
350–365. Raja, N.S., McGee, R., & Stanton, W.R. (1992). Perceived
Freud, A. (1958). Adolescence. In R.S. Eissler (Ed.), attachments to parents and peers and psychological well-being
Psychoanalytic study of the child (Vol. XIII, p. 72). New York: in adolescence. Journal of Youth and Adolescence, 21, 471–485.
International Universities Press. Roisman, G.I., Fraley, R.C., & Belsky, J. (2007). A taxometric
Greenberg, M.T., Siegel, J.M., & Leitch, C.J. (1983). The nature study of the Adult Attachment Interview. Developmental
Psychology, 43, 675–686. doi: 10.1037/0012–1649.43.3.675
and importance of attachment relationships to parents and
peers during adolescence. Journal of Youth and Adolescence, 12, Roisman, G.I., Holland, A., Fortuna, K., Fraley, R.C., Clausell,
E., & Clarke, A. (2007). The Adult Attachment Interview
373–386.
and self-reports of attachment style: An empirical rapproche-
Griffin, D.W., & Bartholomew, K. (Eds.). (1994a). The meta- ment. Journal of Personality and Social Psycholog y, 92,
physics of measurement: The case of adult attachment (Vol. 5). 678–697. doi: 10.1037/0022–3514.92.4.678
London: Jessica Kingsley. Sibley, C.G., Fischer, R., & Liu, J.H. (2005). Reliability and
Griffin, D.W., & Bartholomew, K. (1994b). Models of the self validity of the Revised Experiences in Close Relationships
and other: Fundamental dimensions underlying measures of (ECR-R) self-report Measure of adult romantic attachment
adult attachment. Journal of Personality and Social Psychology, Personality and Social Psychology Bulletin, 31, 1524–1536.
67, 430–445. Sibley, C.G., & Liu, J.H. (2004). Short-term temporal stability
Hall, G.S. (1904). Adolescence: Its psychology and its relations to and factor structure of the revised experiences in close rela-
physiology, anthropology, sociology, sex, crime, religion and edu- tionships (ECR-R) measure of adult attachment. Personality
cation. New York: Appleton. and Individual Differences, 36, 969–975.
Hazan, C., & Shaver, P. (1987). Romantic love conceptualized as Simpson, J.A. (1990). Influence of attachment styles on roman-
an attachment process. Journal of Personality and Social tic relationships. Journal of Personality and Social Psychology,
Psychology, 52, 511–524. 59, 971–980.
Kaplan, R.M., & Saccuzzo, D.P. (2005). Psychological testing: Tabachnick, B.G., & Fidell, L.S. (1996). Using multivariate sta-
tistics (3rd ed.). New York: Harper Collins.
Principles, applications, and issues. Belmont, CA: Thomas
Wadsworth. Vignoli, E., & Mallet, P. (2004). Validation of a brief measure of
adolescents’ parent attachment based on Armsden and
Laible, D.J., Carlo, G., & Roesch, S.C. (2004). Pathways to self- Greenberg’s three-dimension model. European Review of
esteem in late adolescence: The role of parent and peer Applied Psychology, 54, 251–260.
attachment, empathy, and social behaviors. Journal of Vivona, J.M. (2000). Parental attachment styles of late adoles-
Adolescence, 27(703–716). cents: Qualities of attachment relationships and consequences
Liang, L.H., Hou, Z.J., & Tian, L. (2006). Adolescent attach- for adjustment. Journal of Counseling Psychology, 47, 316–329.
ment to father, mother and peers. Chinese Mental Health Wilkinson, R.B. (2006). Age and sex differences in the influence
Journal, 20, 6390642. of attachment relationships on adolescent psychological
Levy, M.B., & Davis, K.E. (1988). Lovestyles and attachment health. The Australian Educational and Developmental
styles compared: Their relations to each other and to various Psychologist, 23, 87–104.
relationship characteristics. Journal of Social and Personal Wilkinson, R.B. (2010). Best friend attachment versus peer
Relationships, 5, 439–471. doi: 10.1177/0265407588054004 attachment in the prediction of adolescent psychological
Main, M., & Goldwyn, R. (1984). Adult attachment scoring and adjustment. Journal of Adolescence, 33, 709–717.
classification system. Unpublished manuscript, University of Wilkinson, R.B., & Kraljevic, M. (2004). Adolescent psychologi-
California at Berkeley. cal health and school attitudes: The impact of attachment
Marcus, R.F., & Betzer, P.D.S. (1996). Attachment and antisocial relationships. Proceedings of the Australian Psychological
Society’s Psychology of Relationships Interest Group 4th Annual
behavior in early adolescence. Journal of Early Adolescence, 16,
Conference, 150–155.
229–248.
Wilkinson, R.B., & Walford, W.A. (2001). Attachment and per-
Markiewicz, D., Lawford, H., Doyle, A.B., & Haggart, N. sonality in the psychological health of adolescents. Personality
(2006). Developmental differences in adolescents’ and young and Individual Differences, 31, 473–484.
adults’ use of mothers, fathers, best friends, and romantic Wilson, J.M., & Wilkinson, R.B. (2007, November). Measuring
partners to fulfill attachment needs. Journal of Youth and adolescent attachment relationships: A quantitative review. Paper
Adolescence, 35, 127–140. presented at the Combined 7th Annual Conference of the
Mikulincer, M., & Shaver, P.R. (2007). Attachment in adulthood: Australian Psychological Society’s Psychology of Relationships
Structure, dynamics, and change. New York: The Guilford Interest Group and International Association for Relationship
Press. Research Mini-Conference, Melbourne, Australia.

62 Journal of Relationships Research

View publication stats