PLM 59

VOLUME FIFTY NINE
The Psychology of
LEARNING AND
MOTIVATION
Series Editor
BRIAN H. ROSS
Beckman Institute and Department of Psychology
University of Illinois, Urbana, Illinois
VOLUME FIFTY NINE
The Psychology of
LEARNING AND
MOTIVATION
Edited by
BRIAN H. ROSS
Beckman Institute and Department of Psychology
University of Illinois, Urbana, Illinois
AMSTERDAM • BOSTON • HEIDELBERG • LONDON

NEW YORK • OXFORD • PARIS • SAN DIEGO
SAN FRANCISCO • SINGAPORE • SYDNEY • TOKYO
Academic Press is an imprint of Elsevier
Academic Press is an imprint of Elsevier
225 Wyman Street, Waltham, MA 02451, USA
525 B Street, Suite 1800, San Diego, CA 92101-4495, USA
Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands
The Boulevard, Langford Lane, Kidlington, Oxford, OX5 1GB, UK
32 Jamestown Road, London NW1 7BY, UK
Copyright © 2013 Elsevier Inc. All rights reserved.
No part of this publication may be reproduced, stored in a retrieval system or transmitted

in any form or by any means electronic, mechanical, photocopying, recording or otherwise
without the prior written permission of the publisher
Permissions may be sought directly from Elsevier’s Science & Technology Rights
Department in Oxford, UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333;
email: [email protected]. Alternatively you can submit your request online by
visiting the Elsevier web site at http: //elsevier.com/locate/permissions, and selecting
Obtaining permission to use Elsevier material
Notice
No responsibility is assumed by the publisher for any injury and/or damage to persons
or property as a matter of products liability, negligence or otherwise, or from any use or
operation of any methods, products, instructions or ideas contained in the material herein.
Because of rapid advances in the medical sciences, in particular, independent verification of
diagnoses and drug dosages should be made
ISBN: 978-0-12-407187-2
ISSN: 0079-7421
For information on all Academic Press publications

visit our website at store.elsevier.com
Printed and bound in USA

13 14 15 10 9 8 7 6 5 4 3 2 1
CONTENTS
Contributors ix
1. Toward a Unified Theory of Reasoning 1

P. N. Johnson-Laird, Sangeet S. Khemlani
1. Introduction 2
2. What Is Reasoning? 4
3. Models of Possibilities 6
4. Icons and Symbols 9
5. The Principle of Truth 11
6. Models as Counterexamples 13
7. Modulation and the Use of Knowledge 16
8. Induction and Abduction 20
9. Probabilities: Extensional and Intensional 23
10. Mental Simulations and Informal Programs 27
11. Toward a Unified Theory 33
12. Conclusions 37
Acknowledgments37
References 38
2. The Self-Organization of Human Interaction 43

Rick Dale, Riccardo Fusaroli, Nicholas D. Duran, Daniel C. Richardson
1
. Introduction: The “Centipede’s Dilemma” of Interaction Research 44
2 . An Example Theoretical Debate and the Need for Integration 46
3 . Self-Organization and Human Interaction 49
4 . Cognitive Dynamics under Social Constraints 57
5 . Coordination, Complementarity, and Interactive Performance 68
6 . Conclusion: Time for More Models 77
Acknowledgments84
References 84
3. Conceptual Composition: The Role of Relational Competition in

the Comprehension of Modifier-Noun Phrases and Noun–Noun
Compounds 97
Christina L. Gagné, Thomas L. Spalding
1.
Introduction 98
2.
Modifier-Noun Phrases and Compounds as Expressions of Combined Concepts 100
v
vi Contents
3. Theoretical Framework: A Three-Stage Theory of Conceptual Combination 101

4. Evidence of the Modifier’s Role in Relation Suggestion 104
5. The Nature of Relations and the Nature of Relational Competition 108
6. The Role of Relation Competition in the Processing of Compounds that
Lack an Underlying Relation 115
7. Evaluation of Relational Interpretations 119
8. Elaboration of Combined Concepts Following Relation Selection 121
9. Summary 124
10. Concluding Remarks 126
References 127
4. List-Method Directed Forgetting in Cognitive and Clinical

Research: A Theoretical and Methodological Review 131
Lili Sahakyan, Peter F. Delaney, Nathaniel L. Foster, Branden Abushanab
1
. Introduction 133
2 . List-Method DF: Design and Measurement 134
3 . Our Framework of List-Method DF 138
4 . Forgetting is a Strategic Decision 140
5 . Context Change as an Explanation for DF Impairment 148
6 . Areas of Disagreement Across Studies 165
7 . Strategy Change Explains DF Benefits 167
8 . Implications for Clinical Populations 172
9 . Concluding Thoughts 181
Acknowledgments182
References 182
5. Recollection is Fast and Easy: Pupillometric Studies of Face

Memory 191
Stephen D. Goldinger, Megan H. Papesh
1
. Introduction 192
2 . Recognition Memory 192
3 . Models of Memory 193
4 . Estimating Recollection and Familiarity 198
5 . Pupillometry 203
6 . Psychophysiological Correlates of Memory for Faces 210
7 . General Conclusions 215
References 215
Contents vii
6. A Mechanistic Approach to Individual Differences in Spatial

Learning, Memory, and Navigation 223
Amy L. Shelton, Steven A. Marchette, Andrew J. Furman
1
. Introduction 224
2 . What Does It Mean to Measure Spatial Learning and Navigational Ability? 225
3 . Dual Systems for Spatial Learning in Rodents 229
4 . Place and Response Learning in Humans 232
5 . The Place/Response Framework for Individual Differences 239
6 . Connections to Other Sources of Variability 246
7 . Competition or Interaction of Systems 250
8 . Conclusions 252
References 255
7. When Do the Effects of Distractors Provide a Measure of

Distractibility? 261
Alejandro Lleras, Simona Buetti, J. Toby Mordkoff
1
. Introduction 262
2 . When Do “Distractors” Cause Distraction? 264
3 . A Brief Case Study on Distraction 291
4 . A Theory of Attention and Distractibility 300
5 . Conclusions 307
References 310
Index 317
Contents of Previous Volumes 331
This page intentionally left blank

CONTRIBUTORS
Branden Abushanab
Department of Psychology, University of North Carolina at Greensboro, Greensboro,
NC, USA
Simona Buetti
Department of Psychology, University of Illinois at Urbana-Champaign, Champaign, IL, USA
Rick Dale
Cognitive and Information Sciences, University of California Merced, Merced, CA, USA
Peter F. Delaney
NC, USA
Nicholas D. Duran
Cognitive and Information Sciences, University of California Merced, Merced, CA, USA;
Division of Psychology and Language Sciences, University College London, London, UK
Nathaniel L. Foster
NC, USA
Andrew J. Furman
Department of Psychological and Brain Sciences, Johns Hopkins University, Baltimore,
MD, USA
Riccardo Fusaroli
Cognitive and Information Sciences, University of California Merced, Merced, CA, USA;
Interacting Minds Center and Center for Semiotics, Aarhus University, Aarhus, Denmark
Christina L. Gagné
Department of Psychology, University of Alberta, Edmonton, AB, Canada
Stephen D. Goldinger
Department of Psychology, Arizona State University, Tempe, AZ, USA
P. N. Johnson-Laird
Department of Psychology, Princeton University, Princeton, NJ, USA;
Department of Psychology, New York University, New York, NY, USA
Sangeet S. Khemlani
Navy Center for Applied Research in Artificial Intelligence, Naval Research Laboratory,
Washington, DC, USA
Alejandro Lleras
Department of Psychology, University of Illinois at Urbana-Champaign, Champaign,
IL, USA
ix
x Contributors
Steven A. Marchette
MD, USA
J. Toby Mordkoff
Department of Psychology, University of Iowa, Iowa City, IA, USA
Megan H. Papesh
Department of Psychology, Louisiana State University, Baton Rouge, LA, USA
Daniel C. Richardson
Division of Psychology and Language Sciences, University College London, London, UK
Lili Sahakyan
NC, USA
Amy L. Shelton
MD, USA
Thomas L. Spalding
CHAPTER ONE
Toward a Unif ied Theory of

Reasoning
P. N. Johnson-Laird*,†,1, Sangeet S. Khemlani‡
*Department of Psychology, Princeton University, Princeton, NJ, USA
†Department of Psychology, New York University, New York, NY, USA
‡Navy Center for Applied Research in Artificial Intelligence, Naval Research Laboratory, Washington, DC, USA
1Corresponding author: E-mail: [email protected]
Contents
1. Introduction2
2. What Is Reasoning? 4
3. Models of Possibilities 6
4. Icons and Symbols 9
5. The Principle of Truth 11
6. Models as Counterexamples 13
7. Modulation and the Use of Knowledge 16
8. Induction and Abduction 20
9. Probabilities: Extensional and Intensional 23
10. Mental Simulations and Informal Programs 27
11. Toward a Unified Theory 33
12. Conclusions37
Acknowledgments37
References38
Abstract
This article describes a theory that uses mental models to integrate deductive, induc-
tive, and probabilistic reasoning. It spells out the main principles of the theory and
illustrates them with examples from various domains. It shows how models underlie
inductions, explanations, estimates of probabilities, and informal algorithms. In all
these cases, a central principle is that the mind represents each sort of possibility in a
separate mental model and infers whatever holds in the resulting set of models. Finally,
the article reviews what has been accomplished in implementing the theory in a single
large-scale computer program, mReasoner.
Psychology of Learning and Motivation, Volume 59 © 2013 Elsevier Inc.

ISSN 0079-7421, http://dx.doi.org/10.1016/B978-0-12-407187-2.00001-0 All rights reserved. 1
2 P. N. Johnson-Laird and Sangeet S. Khemlani
1. INTRODUCTION
The capacity to reason underlies mathematics, science, and technol-
ogy. It is essential for coping with everyday problems—without it, social life
would be almost unimaginable. The challenge to psychologists is to explain
its underlying mental mechanisms. Since Störring’s (1908) pioneering study,
they have discovered several robust phenomena. Perhaps the most important
is that naive reasoners—those with no training in logic—can make valid
deductions, that is, inferences in which the conclusion is true in all the cases
in which the premises are true (cf. Jeffrey, 1981, p. 1). And they are happy
to do so about abstract matters with no ecological validity, as in Sudoku
puzzles (Lee, Goodwin, & Johnson-Laird, 2008).
Fifty years ago, psychologists took for granted that human reasoning
was rational. Individuals developed deductive competence during child-
hood, and the psychologists’ task was to pin down the nature of the formal
logic underlying this ability. As Inhelder and Piaget (1958, p. 305) wrote,
“Reasoning is nothing more than the propositional calculus itself ”. There
may be vagaries in performance, but faulty reasoning does not occur (Henle,
1962) or is attributable to local malfunctions in the system—a spanner in
the works rather than an intrinsic flaw (Cohen, 1981). Indeed, theories of
deduction in cognitive psychology began with accounts based on formal
logic (e.g. Braine, 1978; Johnson-Laird, 1975; Osherson, 1974–1976). These
views, of course, echo those of Enlightenment philosophers. But, another
robust phenomenon that psychologists discovered is that individuals differ
in their ability to reason. A few are very good, a few are very bad, and most
are somewhere in the middle. Differences in ability are vast, and correlate
with the tests of academic achievement, as proxies for measures of intel-
ligence (Stanovich, 1999). For everyone, however, failures are inevitable:
complex inferences are computationally intractable.
Nowadays, a consensus exists that that the psychology of reasoning has
undergone a deep change—even, some say, a paradigm shift. The acces-
sibility of digital computers was a license for theorizing, and psychological
theories of reasoning have multiplied at a startling rate. Quite what the new
foundations of reasoning should be is controversial. One view is that humans
are rational but in terms of the probability calculus rather than logic (e.g.
Oaksford & Chater, 2007; Tenenbaum & Griffiths, 2001). One view is that
natural selection has equipped the mind with modules for reasoning about
special topics, such as social exchange (Cosmides & Tooby, 2005). One view
is that rationality presupposes a normative system, and psychologists should
Toward a Unif ied Theory of Reasoning 3
abandon norms in favor of descriptions (Evans, 2012). One revenant is that

logic, or logics, provides the inferential machinery (Rips, 1994; Stenning &
van Lambalgen, 2008). We will not try to assess these views, but for those
who espouse such a theory, we recommend the answers to two questions as
a recipe for resipiscence: Has the theory been implemented in a computer
program, and does it predict most of the 60 or more experimental results
reported here? The goal of the article, however, is not polemical, but to
describe a different theory.
Craik (1943) postulated that thinking was based on making mental sim-
ulations of the world to anticipate events. This idea in turn has historical
antecedents, although Craik was unlikely to have known them (see Johnson-
Laird, 2004, for the history of mental models). Oddly, however, Craik did
not consider reasoning, other than to make a casual remark that it was based
on “verbal rules” (Craik, 1943, p. 81). In the spirit of Craik, we argue that
the mind is neither a logical nor a probabilistic device, but instead a device
that makes mental simulations. Insofar as humans reason logically or infer
probabilities they rely on their ability to simulate the world in mental mod-
els. The application of simulation to reasoning is based on mental models
of the possibilities to which the premises refer, and a valid deduction has
a conclusion that holds in all these models. This idea was first proposed a
generation ago ( Johnson-Laird, 1975). Since then, its proponents and critics
have revised and extended it in hundreds of publications.
The theory of mental models—the model theory for short—is contro-
versial, as are all current theories of reasoning, and the only way to put it
beyond controversy calls for two crucial steps. The first step follows Leibniz
(1685, 1952), who dreamt of replacing argument with calculation. It is to
implement a unified theory of reasoning in a computer program that, for any
inferential task, outputs the responses that human reasoners should make, the
respective likelihoods and latencies of these responses, the processes underly-
ing them, and, where relevant, valid or ideal responses. The second step is
to show in stringent experiments that the program’s predictions are correct.
We are a long way from the two steps. Researchers have applied the model
theory to many sorts of inferential task, implemented computational models of
these applications, and tested the theory experimentally. But, until recently,
the work has been piecemeal rather than unified.
What are the essentials of the model theory, and what counts as a mental
model? This article is going to answer these questions step by step, and it
aims above all to enable readers to understand the model theory without
having to read anything else. It illustrates the theory’s application to most
sorts of reasoning. Its plan mirrors these aims. It begins with an outline of the
main sorts of reasoning. It follows with sections that elucidate each of the
theory’s main principles. It then considers the role of models in inductive
reasoning, explanatory reasoning, reasoning about probabilities, and reason-
ing that yields informal algorithms. The final section of the article reviews
what has been accomplished in unifying the theory, and in implementing it
in a single computer program to achieve Leibniz’s (and our) dream.
2. WHAT IS REASONING?
Suppose you infer:
If the ink cartridge is empty then the printer won’t work.
The ink cartridge is empty.
So, the printer won’t work.
You are making a deduction: your inference is valid because your conclusion
holds in any case in which the premises hold. Suppose instead you infer:
The printer won’t work.
So, the ink cartridge is empty.
You are making an induction. Your inference isn’t valid because there
may be another reason that the printer won’t work. Yet, your conclusion
may be true, especially if the printer is producing blank pages. For many
theorists—Aristotle for one, all inferences fall into one of these two catego-
ries: deduction and induction.
Aristotle defined induction as an inference from a particular assertion to
a universal one (Topics, 105a13). But, inductions are often from the particu-
lar to the particular, as is your induction about the printer. Hence, a better
way to distinguish between the two sorts of inference is in terms of seman-
tic information ( Johnson-Laird, 1983, chap. 2). The more possibilities that
an assertion rules out, the more information it conveys (Bar-Hillel, 1964).
An inference to a conclusion that refers to all the same possibilities as the
premises do, or at least includes them all in what it refers to, is a deduction.
Consider again your earlier deduction:
The ink cartridge is empty.
So, the printer won’t work.
The premises refer to just one possibility:
The ink cartridge is empty and the printer won’t work.
Hence, your inference is valid because its conclusion holds in the one pos-
sibility to which the premises refer. The conclusion therefore does not
increase information. But semantics should not be confused with epistemol-

ogy: a conclusion may be news to the person who draws it, bringing to mind
a novel proposition. An inference to a conclusion that refers to only some of
the possibilities to which the premises refer, although it may add some new
possibilities too, is an induction. Consider again your earlier induction:
So, the ink cartridge is empty.
The premises refer to two possibilities:
The ink cartridge is empty. The printer won’t work.
The ink cartridge isn’t empty. The printer won’t work.
Your conclusion, however, refers to only one of these two possibilities. It goes
beyond the information in the premises, and it is consistent with them, that is,
it is possible that the ink cartridge is empty. But, the conclusion does not follow
validly. A special case of induction is one that also introduces new ideas to
explain something, and this sort of reasoning is known as abduction, for example:
Hence, there’s a fault in the connection between the computer and the
printer.
Inferences either maintain or throw information away (deductions) or they
increase information (inductions). One other relation between the premises
and conclusion remains. If they refer to disjoint possibilities, they contradict one
another. In logic, any conclusion whatsoever follows validly from a contradic-
tion: the premises don’t refer to any possibility in which the conclusion fails to
hold because the premises don’t refer to any possibility. Naive reasoners, how-
ever, reject inferences from self-contradictions. For them, the definition of valid-
ity has a rider: a valid inference is one in which the conclusion holds in every
possibility to which the premises refer, and there is at least one such possibility.
Reasoners aim to draw conclusions that are true, or at least plausible.
But, they also aim to draw novel and parsimonious conclusions, and so they
would feel silly just to form a conjunction of all the premises even though
such an inference is valid. They know more when they know:
It’s raining
than when they know:
It’s raining or it’s cold, or both.
Yet, the disjunctive conclusion follows validly from the categorical premise.
Hence, not all valid deductions are sensible, and it would be silly to make this
particular inference because it throws away information by adding a disjunctive
alternative to a premise. So, a theory of deductive competence—of what the

inferential system computes—assumes that individuals have the potential to
be rational and an awareness of this potential. They abide by the foregoing
constraints. In sum: “To deduce is to maintain semantic information, to sim-
plify, and to reach a new conclusion” (Johnson-Laird & Byrne, 1991, p. 22).
These constraints are not easy to embody in a theory based on formal logic,
and this difficulty explains why such theories, like automated theorem provers
in artificial intelligence, focus on the evaluation of given conclusions. Induc-
tive competence also aims for parsimony and novelty, but it goes beyond the
information given and ultimately aims to explain phenomena.
3. MODELS OF POSSIBILITIES
The fundamental assumption of the model theory is that each mental
model represents what is common to a distinct set of possibilities. Hence, an
assertion such as:
A triangle is on the right of a circle
has a single mental model, which we depict in this diagram:
The left-to-right axis of the model corresponds to the left-to-right axis

of a scene, and the disposition of the triangle and circle in the model cor-
responds to their disposition in a scene for which the assertion is true. The
model represents an indefinite number of possibilities that have in common
only that a triangle is on the right of a circle. Of course, the relative sizes
of the figures in the model, their distance apart, and so on, play no role in
reasoning from the model, but we defer an explanation of how their irrel-
evance is represented until Section 11.
Everyone prefers to think about just one possibility at a time. Intuitions
work in this way. And the theory postulates two separate systems for reason-
ing, one for intuitions and one for deliberations—a familiar distinction in
“dual process” theories of reasoning (see, e.g. Evans, 2008; Kahneman, 2011;
Reitman, 1965; Sloman, 1996; Stanovich, 1999; Verschueren, Schaeken, &
d’Ydewalle, 2005). The model theory distinguishes between the two sys-
tems in computational power, and we have implemented both of them
in computer programs (Khemlani & Johnson-Laird, 2012a; Khemlani,
Lotstein, & Johnson-Laird, 2012). The intuitive system, which is sometimes
known as “system 1”, has no access to working memory, and so it can repre-
sent only one mental model at a time (Johnson-Laird, 1983, chap. 6), and it
cannot carry out recursive processes, including arithmetical operations such

as counting. It lacks even the computational power of a finite-state automa-
ton (Hopcroft & Ullman, 1979) because it can carry out a loop of opera-
tions for only a small finite number of times—a restriction that is built into
its computer implementation. In contrast, the deliberative system, which is
sometimes known as “system 2”, has access to working memory, and so it
can search for alternative mental models, and carry out recursive processes,
such as counting and arithmetical operations, until they overwhelm its pro-
cessing capacity.
One way in which to overwhelm the deliberative system is to force it to
reason about disjunctions. An inclusive disjunction, such as:
There’s a triangle or there’s a circle, or both
includes the joint possibility of both the triangle and the circle, and so it
refers to three sorts of possibility. It therefore calls for three mental models,
which we show on separate rows in this diagram:
In this case, spatial relations play no role in the use of the models. An exclu-
sive disjunction, such as:
Either there’s a triangle or there’s a circle, but not both
exclude the joint possibility, and so it calls for only two mental models:
Models preoccupy system 2, and so more models mean more work. The
theory therefore predicts that deductions from exclusive disjunctions should
be easier than those from inclusive disjunctions, as when either of the dis-
junctions above occurs with the categorical assertion:
There isn’t a circle.
This assertion eliminates any model in which there is a circle, and so it
follows validly in both cases that:
There is a triangle.
Evidence corroborates the prediction (e.g. Johnson-Laird, Byrne, &
Schaeken, 1992), and it also shows that inferences from conjunctions, which
have just one model, are easier than those based on disjunctions (García-
Madruga, Moreno, Carriedo, Gutiérrez, & Johnson-Laird, 2001).
The sorts of inference that can overwhelm the deliberative system are
“double disjunctions” ( Johnson-Laird et al., 1992), which are from pairs of
disjunctive premises, such as:
June is in Wales, or Charles is in Scotland, but not both.
Charles is in Scotland, or Kate is in Ireland, but not both.
What follows?
The two possibilities compatible with the first premise are relatively easy
to envisage, but it is difficult to update them with those from the second
premise, although the result is just two possibilities:
June in Wales Kate in Ireland
Charles in Scotland
Of course, real mental models represent these spatial relations, and are not
phrases, which we use here for convenience. The two models yield the
conclusion:
Either June is in Wales and Kate is in Ireland or Charles is in Scotland.
Inferences become even harder when disjunctions are inclusive. In one
experiment, 25% of the participants, who were from the general public,
drew valid conclusions from exclusive disjunctions, but this figure fell to
below 10% for inclusive disjunctions. The result is hardly surprising, but
what was striking was the nature of the modal errors: for all the inferences,
the participants drew conclusions corresponding to a model of a single
possibility. Just under a third of all the participants’ responses were conclu-
sions of this sort. The result suggests that when the task was too much for
them, they fell back on their intuitions and envisaged just a single model
of the premises. So, their conclusions were consistent with the premises
but did not follow from them. The performance of undergraduates showed
the same pattern. But, when the disjunctions were presented in equivalent
electrical circuit diagrams or analogs of them, they performed better and
faster (Bauer & Johnson-Laird, 1993). Their conclusions, however, still bore
out a failure to consider all the possibilities, and so most errors were at least
consistent with the premises.
Mental models represent possibilities, and so the more models that are
necessary to make an inference, the harder that inference is to make. Indi-
viduals are in danger of overlooking a model. When the deliberative system
is vastly overburdened, reasoners may even fall back on the intuitive system
and draw a conclusion that is consistent with only a single model. One
side effect of the use of models is that reasoners are most unlikely to draw
conclusions that throw semantic information away by adding disjunctive

alternatives.
4. ICONS AND SYMBOLS

Mental models are iconic insofar as possible. What “iconic” means is
that their structure corresponds to the structure of what they represent (see
Peirce, 1931–1958,Vol. 4, paragraph 447). One example is the mental model
of the assertion, the triangle is on the right of the circle, which we diagrammed
in the previous section. Another example is an electrical circuit diagram
with the same structure as the circuit it denotes. The great advantage of an
icon, as Peirce realized, is that its inspection yields new information. Given
the premises:
The triangle is on the right of the circle.
The square is on the right of the triangle.
The intuitive system can build the model:
It yields a new relation, namely, the square is on the right of the circle, and so
this transitive inference emerges from scanning the model. To establish its
validity, reasoners need to call on the deliberative system to check that no
alternative model of the premises refutes the conclusion.
When premises are consistent with more than one spatial layout, infer-
ences are more difficult than the preceding example (e.g. Byrne & Johnson-
Laird, 1989; Carreiras & Santamaría, 1997; Vandierendonck, Dierckx, & De
Vooght, 2004). Likewise, reasoners try to construct initial models that do
not call for a rearrangement of entities (e.g. Jahn, Knauff, & Johnson-Laird,
2007; Knauff & Ragni, 2011), and inferences that call for such rearrange-
ments are more difficult than those that do not (e.g. Krumnack, Bucher,
Nejasmic, Nebel, & Knauff, 2011). Analogous results bear out the use of
iconic representations in temporal reasoning, whether it depends on rela-
tions such as “before” and “after” (Schaeken, Johnson-Laird, & d’Ydewalle,
1996a) or on the tense and aspect of verbs, as in:
John has cleaned the house.
John is taking a shower.
John is going to read the paper.
Mary always does the dishes when John cleans the house.
Mary always drinks her coffee when John reads the paper.
What is the relation between Mary drinking coffee and doing the dishes?
Participants inferred that Mary drinks her coffee after doing the dishes, in
an experiment that controlled such factors as order of mention (Schaeken,
Johnson-Laird, & d’Ydewalle, 1996b).
The earlier example of a transitive inference is child’s play (see, e.g. Bryant &
Trabasso, 1971). But, many inferences based on iconicity are more complex,
such as those that combine both spatial and temporal relations in kinematic
simulations (see Section 10). The intuitive system can also mislead adult
reasoners. It constructs only a simple model of a typical situation. Given this
sort of problem:
Ann is a blood relative of Beth.
Beth is a blood relative of Cal.
Is Ann a blood relative of Cal?
Many adult reasoners respond, “Yes”. The relation holds in their model,
which represents lineal descendants or siblings. They fail to search assidu-
ously for an alternative model—it takes work to engage the deliberative sys-
tem, or a clue to a possible alternative model, such as a reminder that people
can be related by marriage. Indeed, Ann and Cal could be Beth’s parents,
not blood relatives of one another (Goodwin & Johnson-Laird, 2005, 2008).
Visual images are iconic, and so some theorists suppose that they play a
key role in reasoning (e.g. Kosslyn, 1994, p. 404). They do play a role: they
impede reasoning.To see why, it is crucial to distinguish among relations that
elicit visual images, such as “dirtier than”, relations that elicit spatial relations,
such as “on the right of ”, and relations that are abstract, such as “better than”.
Individuals are slowest in reasoning from visual relations (Knauff & Johnson-
Laird, 2002), but do not differ reliably in reasoning from the other sorts of
relation. As an fMRI study showed, only visual relations elicited additional
activity in visual cortex (Knauff, Fangmeier, Ruff, & Johnson-Laird, 2003).
Knauff (2013) tells the whole story: visual imagery is not necessary for rea-
soning, which is just as well because some relations, such as those between
sets, have iconic representations that may not be visualizable.
Not everything can be represented in an icon. A crucial example is a
negation, such as:
The triangle is not on the right of the circle.
Reasoners could try to list all the alternative affirmative possibilities—the
triangle is on the left of the circle, behind it, and so on—but it would be criti-
cal, not only to include all the possibilities but also to make explicit that the
list is exhaustive. Alas, neither these conditions nor the meaning of negation
itself can be represented in an icon.The model theory accordingly introduces
a symbol for negation, which is linked to its meaning: a negative assertion
or clause is true if, and only if, its corresponding affirmative is false. This
meaning goes back to Aristotle’s De Interpretatione, and with some exceptions,
it holds for English usage (Khemlani, Orenes, & Johnson-Laird, 2012). The
mental model of the preceding assertion is denoted in the following diagram:
where “¬” denotes the symbol for negation, and the brackets symbolize the
scope of the negation, that is, what it applies to. Hence, a comparison of this
model with an actual scene would yield the value “true” if, and only if, the
relevant circle and triangle were not in the spatial relation represented in
the embedded model.
Few people grasp the concept of “negation”, and so prudent experi-
menters ask them about the “denial” of assertions. But, even so, most reason-
ers err in enumerating the possibilities referred to by compound assertions,
such as:
He denied that John was watching TV and smoking, or else Ann was
writing a letter.
Once again, however, number of models is the key variable (Khemlani et al.,
2012). It is harder to enumerate the possibilities for the denial of a conjunc-
tion, A and B, which has three models:
¬A ¬B
¬A B
A ¬B
than to enumerate the possibilities for the denial of an inclusive disjunction,
A or B, which has one model:
¬A ¬B
There are plenty of other abstract concepts, such as “possibility”, “truth”,
and “obligation” that transcend iconicity.
5. THE PRINCIPLE OF TRUTH

The model theory postulates a principle of truth: mental models rep-
resent what is true, not what is false unless assertions refer to falsity. Here is
an example of an exclusive disjunction with a negative clause:
Either there isn’t a triangle or there’s a circle.
It has two mental models:
They represent the possibilities in which the disjunction is true, not the
possibilities in which it is false. But, the principle of truth applies at a lower
level. Both of the preceding models represent clauses in the disjunction only
when they are true. In contrast, fully explicit models also represent clauses
that are false. In the first model above, it is false that there is a circle; and
in the second model, it is false that there isn’t a triangle, that is, there is a
triangle. Hence, the fully explicit models of the exclusive disjunction are:
where we use negation to represent falsity. The fully explicit models show
that the disjunction is equivalent to the biconditional assertion:
There isn’t a triangle if, and only if, there isn’t a circle.
Reasoners don’t immediately grasp this equivalence—a failure that shows
that they rely on mental models, not fully explicit models.
When participants are given a compound assertion, such as a disjunction,
and are asked to list what is possible, the principle of truth constrains them, and
so they list the possibilities corresponding to mental models (see, e.g. Barres &
Johnson-Laird, 2003; Johnson-Laird & Savary, 1995). The advantage of the prin-
ciple is that it reduces the processing load of reasoning. But, when we implemented
the principle in a computer program, we discovered an unexpected downside.
Could both of these disjunctions be true at the same time?
Either the pie is on the table or the cake is on the table, but not both.
Either the pie isn’t on the table or the cake is on the table, but not both.
Most people say,“Yes” ( Johnson-Laird, Lotstein, & Byrne, 2012).The mental
models of what’s on the table according to the first disjunction are:
Pie
Cake
And the mental models of what’s on the table according to the second dis-
junction are:
¬ Pie
Cake
The presence of the cake is common to both sets of models, and so it seems
that the two assertions can both be true at the same time, that is, when the
cake is on the table. In contrast, the fully explicit models of the two disjunc-
tions are as follows:
Pie ¬ Cake
¬ Pie Cake
and:
¬ Pie ¬ Cake
Pie Cake
As readers can see, no model is common to both assertions, and so they can-
not both be true at the same time.
The program implementing the model theory predicts these fallacies,
and others too. Their occurrence has been corroborated in many sorts of
deductions, including those based on:
• Disjunctions and biconditionals of conditionals ( Johnson-Laird & Savary,
1999);
• Disjunctions of conjunctions (Walsh & Johnson-Laird, 2004);
• Disjunctions of disjunctions (Khemlani & Johnson-Laird, 2009);
• Disjunctions of quantified assertions (Yang & Johnson-Laird, 2000).
The fallacies tend to be compelling and to elicit judgments of high con-
fidence in their conclusions, and so they have the character of cogni-
tive illusions. Other illusions led to conclusions about what is probable
(Johnson-Laird & Savary, 1995), possible (Goldvarg & Johnson-Laird, 2000),
and permissible (Bucciarelli & Johnson-Laird, 2005). And still others con-
cerned the evaluation of the consistency of assertions (Legrenzi, Girotto, &
Johnson-Laird, 2003). Each study examined several sorts of illusion and
matched control problems. Why so many studies of illusions? Because only
the model theory predicts them, and so they are a litmus test for the use of
mental models.
6. MODELS AS COUNTEREXAMPLES
In reasoning, a counterexample is a possibility that is consistent with a
set of premises, but not with a putative conclusion, and so it shows that the
conclusion does not follow validly from the premises. The intuitive system
can generate at most a single model at a time. To establish the validity of a
conclusion, the deliberative system has to search for alternative models and
to show either that no other mental model can be formed from the premises
or that the conclusion holds in the alternatives. If the deliberative system
creates a model that is a counterexample then it can search for an alterna-
tive conclusion that holds in all the models or, if this search fails, declare that
no valid conclusion follows from the premises. As we pointed out earlier, a
conclusion such as a conjunction of the premises follows validly from any
set of premises, and so an alternative conclusion needs to be parsimonious
and to establish a new relation not explicitly asserted among the premises.
In short, counterexamples are crucial for rationality. Without the ability to
create them, individuals can infer conclusions, but they have no ready way to
establish their invalidity. So, to what extent do individuals make use of them?
On the one hand, reasoners often fail to use counterexamples when they
are drawing conclusions from premises—to the degree that one model-
based theory makes no use of them (Polk & Newell, 1995). On the other
hand, all the participants in one study spontaneously used them to revise
their responses (Bucciarelli & Johnson-Laird, 1999).
There are two sorts of invalid conclusion. One sort contradicts the
premises—their respective sets of possibilities are disjoint. The other sort is
consistent with the premises, but does not follow from them, that is, there
are possibilities to which the premises, but not the conclusion, refer. The
model theory predicts that the invalidity of contradictions should be easier
to detect than the invalidity of consistent premises: the former don’t have
a mental model in common with the premises whereas the latter do. The
theory also predicts that when individuals are asked to explain why a con-
clusion does not follow from the premises, they should tend to point out
the contradiction in the first case but to exhibit a counterexample in the
second case. A study corroborated both of these predictions (Johnson-Laird &
Hasson, 2003). The participants were more accurate in identifying invalid
inferences in which the conclusion contradicted the premises (92% cor-
rect) than those in which the conclusion was consistent with the premises
(74% correct). To justify their judgments, they used counterexamples more
often for conclusions consistent with the premises (51% of cases) than for
conclusions inconsistent with them (21% of cases). Of course, they used
other strategies too. One participant, for instance, pointed out that a piece of
necessary information was missing from the premises. But, the use of coun-
terexamples correlated with accuracy in the evaluation of the inferences.
An fMRI study contrasted reasoning and mental arithmetic from
the same premises (Kroger, Nystrom, Cohen, & Johnson-Laird, 2008).
The participants read a statement of the problem, then three premises,

and finally either a conclusion or an arithmetical formula, which they
had to evaluate. The experiment included easy inferences that followed
immediately from a single premise and difficult inferences that should lead
individuals to search for counterexamples, as in this case:
There are five students in a room.
Three or more of these students are joggers.
Three or more of these students are writers.
Three or more of these students are dancers.
Does it follow that at least one of the students in the room is all three: a
jogger, a writer, and a dancer?
Most people think first of a possibility in which the conclusion holds. But,
those who search for a counterexample may find one, such as this model
in which each of the five individuals shown in separate horizontal rows is
a student:
Jogger Writer
Jogger Writer
Jogger Dancer
Writer Dancer
Dancer
Hence, it doesn’t follow that a student is all three. While the participants
were reading the premises, the language areas of their brains were active
(Broca’s and Wernicke’s areas), but then other areas carried out the solu-
tion to the problems. Right prefrontal cortex and inferior parietal lobe
were more active for reasoning than for calculation, whereas regions in left
prefrontal cortex and superior parietal lobe were more active for calcula-
tion than for reasoning. Right prefrontal cortex—a region known as the
right frontal pole—was active only during the difficult inferences calling
for a search for counterexamples. Other studies have shown that difficult
inferences activate right frontal cortex (Kroger et al., 2002; Waltz et al.,
1999). The anterior frontal lobes evolved most recently, they take longest to
mature, and their maturation relates to measured intelligence (Shaw et al.,
2006). Whether they are activated merely by problems calling for delibera-
tion remains unclear.
7. MODULATION AND THE USE OF KNOWLEDGE

In logic, the interpretation of logical terms is constant, as for its ideal-
ized connectives akin to “if ”, “and”, and “or”. But, the model theory rec-
ognizes that their interpretation in everyday language can be modified by
the meanings of the clauses that they connect, the entities referred to in
these clauses, and general knowledge. We refer to the process as modulation,
and we illustrate it with the most notorious case, conditional assertions (see
Johnson-Laird & Byrne, 2002).
Conditionals of the grammatical form, if A then B, receive a logical
interpretation by default. When individuals have to list the fully explicit
possibilities to which a conditional refers, they tend to list:
A B
¬A B
¬A ¬B
where A and B have as values actual propositions (see, e.g. Johnson-Laird &
Savary, 1995). Barrouillet and his colleagues have shown that children
around the age of 8 years list only one possibility, A and B, a conjunctive
interpretation; around the age of 11 years, they include another possibility,
¬ A and ¬ B, a biconditional interpretation; and around the age of 15 years,
they list the three possibilities above (Barrouillet & Lecas, 1998). The pro-
cessing capacity of working memory is a better predictor than chronologi-
cal age for the number of possibilities that children list (Barrouillet, Grosset, &
Lecas, 2000; Barrouillet & Lecas, 1999).
In reasoning, individuals rely on the mental models of conditionals, which
consist of an explicit model of the salient case in which both clauses hold, and
a content-less placeholder for other possibilities in which the if-clause is false:
A B
. . .
One corollary concerns inferences of the form:
If A then B.
A.
What follows?
Individuals easily infer the conclusion, B. It follows at once from the mental
models. A contrasting inference is:
If A then B.
Not-B.
What follows?
The second premise eliminates the one explicit mental model, and so it seems
that nothing follows—a common response. Only if reasoners flesh out their
mental models, or adopt some analogous strategy, can they make the valid
inference: Not-A.The difference between the two sorts of inference is highly
robust. A more striking corroboration of the model theory is that the presen-
tation of the premises in the opposite order improves performance with the
difficult inference—it renders unnecessary the need to construct the explicit
mental model of the conditional, thus making room for models of the pos-
sibilities in which not-A holds (see Girotto, Mazzocco, & Tasso, 1997).
Modulation has several effects, and one of them is to block the construction
of models of possibilities. A conditional, such as:
If she played a musical instrument then it wasn’t the flute
refers to only two possibilities because knowledge that a flute is a musical
instrument blocks the construction of the possibility that she didn’t play a
musical instrument but did play the flute. Hence, the conditional alone yields
the conclusion that she didn’t play the flute.The principal possibility to which
almost all conditionals refer is the one in which both the if-clause and the
then-clause hold. Hence, the theory postulates that if a conditional refers to
more than one possibility, then this possibility must be one of them. Modula-
tion can accordingly yield the preceding interpretation or the biconditional
interpretation. Still other effects of modulation occur with then-clauses that
themselves express only a possibility, such as “if Hillary runs then she may win”.
Experiments have corroborated that modulation blocks the construc-
tion of models (Quelhas, Johnson-Laird, & Juhos, 2010). Consider these two
conditionals translated from the Portuguese:
If the dish is lasagne then its basis is pasta.
If the cake is made of eggs then it can be suspiro.
For the first sort of conditional, participants allow that the dish can be pasta
but not lasagne. But, for the second sort of conditional, they do not allow
that the cake can be suspiro but not made of eggs—all Portuguese know
that suspiro is made from eggs. The two sorts of conditionals yield appro-
priately different patterns of inference.
Consider the inference:
Luisa didn’t play soccer.
Therefore, if Luisa played music then she didn’t play soccer.

The conditional conclusion refers to three possibilities in which Luisa
played, respectively:
Music ¬ Soccer
¬ Music ¬ Soccer
¬ Music Soccer
Hence, the conclusion refers to the possibility to which the premise refers,
and so the inference is valid. Yet, most people reject it. Why? One answer
is that it is unacceptable because it throws information away, that is, its
conclusion also refers to an alternative possibility that conflicts with the
premise:
¬ Music Soccer
This conflict, as Orenes and Johnson-Laird (2012) argued, may deter indi-
viduals from drawing the inference. If so, then modulation that blocks the
conflicting model should yield an acceptable inference, for example:
Luisa didn’t play soccer.
Therefore, if Luisa played a game then she didn’t play soccer.
The conditional now refers to just two possibilities in which Luisa played:
A game ¬ Soccer
¬ A game ¬ Soccer
The conditional can’t refer to the case in which Luisa didn’t play a game
but played soccer because soccer is a game. So, both the preceding possibili-
ties refer to the same possibility as the premise. In this case, a highly reliable
increase occurs in the percentage of participants who accepted the inference.
And analogous phenomena occur with inferences to disjunctive conclusions.
Another effect of modulation is to introduce spatial, temporal, or other
relations between the if-clause and the then-clause. As a consequence, indi-
viduals make different inferences (Quelhas et al., 2010). For example, given
these premises:
If Laura got the virus, then she infected Renato.
If she infected Renato, then he went to hospital.
Laura got the virus.
Participants tend to infer that Laura got the virus before Renato went to
hospital. But, given these premises:
If Cristina wrote the article, then Marco asked her to write it.
If Marco asked her to write it, then he met her at the meeting.
Cristina wrote the article.
Participants tend to infer that Cristina wrote the article after Marco met
her.The temporal inferences depend on the participants’ general knowledge
about the typical orders of events.
A subtle effect of temporal modulation is illustrated in the following
contrasting examples (Juhos, Quelhas, & Johnson-Laird, 2012). The first
example is:
If the author writes the book, then the publisher publishes it.
The author writes the book.
What follows?
Individuals tend to infer:
The publisher publishes it.
The second example is:
If the author writes the book, then the publisher publishes it.
The publisher publishes the book.
What follows?
Individuals tend to infer:
The author wrote the book.
The difference is that for the first inference, the participants tended to use
the present tense (the experiment was carried out in Portuguese), whereas
for the second inference, they tended to use the past tense. As in English,
which has no future tense, the present tense in Portuguese can be used
to refer to future events. The same phenomenon occurred in inferences
from disjunctions. The categorical premise accordingly establishes a refer-
ence time, and events prior to it are referred to in the past tense, and events
subsequent to it are referred to in the present tense. This sort of modula-
tion is tacit—participants are not usually aware of its effects—but it shows
that general knowledge influences the interpretation of conditionals and
disjunctions.
A crucial corollary of modulation concerns logical form. A typical for-
mal rule of inference is:
A or B.
Not-B.
Therefore, not-A.
This rule is applicable to any premises that have the corresponding logical
forms, which are transparent in logic, because they are defined by its gram-
mar. In language, however, logical forms are far from transparent, and no
algorithm exists to determine them because they are not just a matter of
grammar. They depend on the possibilities to which assertions refer. So too
does validity, and it therefore can be decided only on a case-by-case basis.
The model theory makes no use of logical form, but merely the grammatical
structure of sentences, and it uses meaning, reference, and knowledge to
modulate logical interpretations.
8. INDUCTION AND ABDUCTION

Modulation depends on knowledge, and so it is a bridge from deduc-
tion to induction. Inductive inferences yield specific conclusions, general-
izations, and, above all, explanations. They depend on knowledge and its
availability (Tversky & Kahneman, 1973). An induction may yield a true
conclusion; but it may not, even if its premises are true. The engineers in
charge of Chernobyl induced that the reactor was intact after the explosion.
Their inference was plausible because no nuclear reactor had ever melted
down before. But, they were wrong, and their delay in making the correct
inference cost lives. Induction is indeed risky.
Logic is “monotonic” in that further premises warrant further conclu-
sions, and no subsequent premise ever calls for a valid conclusion to be
withdrawn—not even its direct contradiction. At Chernobyl as in daily life,
individuals withdraw conclusions, even valid ones, in the light of subsequent
information.Their reasoning is “nonmonotonic”.They withdraw some con-
clusions because they are based on assumptions made by default. They infer,
say, that Fido has four legs because Fido is a dog and by default dogs have
four legs, but then they discover that poor old Fido has only three legs. This
process of withdrawing conclusions based on default assumptions is integral
to the model theory ( Johnson-Laird & Byrne, 1991). But, retractions also
occur in other cases.You believe, say, that if someone pulls the pistol’s trigger
then it will fire. Someone pulls the trigger. Y
et, the pistol does not fire. Hence,
there is a conflict between a valid inference from your beliefs—that the
pistol fires—and the incontrovertible fact that it doesn’t fire. So, you have to
withdraw your conclusion and modify at least one of your beliefs. Artificial
intelligencers have devised various systems of nonmonotonic reasoning to
deal with such cases, but these approaches have grown increasingly remote
from psychological plausibility (see Brewka, Dix, & Konolige, 1997). In fact,
at the heart of human performance is the abduction of explanations that

resolve inconsistencies (Johnson-Laird, Girotto, & Legrenzi, 2004). It is these
explanations that, as a by-product, yield revisions to beliefs.
An inconsistent set of assertions is a potentially serious matter in daily
life. For example, disasters at sea are often a consequence of a conflict
between a mariner’s mental model and reality (Perrow, 1984).The ability to
detect inconsistencies is accordingly one hallmark of rationality. Reasoners
can do it by trying to construct a single model of all the relevant informa-
tion. If they succeed, they evaluate the information as consistent; but if they
fail, they evaluate it as inconsistent (see Johnson-Laird et al., 2004, for cor-
roboratory evidence). Once they have detected an inconsistency, they can
use their knowledge to try to explain it. The rest of this section focuses on
such explanations, that is, abductions.
The basic units of explanations are causes and their effects. In the case of
an inconsistency, the effect makes possible the facts of the matter. Accord-
ing to the model theory, causation refers to what is possible and to what
is impossible in the co-occurrence and temporal sequence of two events
(Frosch & Johnson-Laird, 2011; Goldvarg & Johnson-Laird, 2001). A com-
puter program implementing this account constructs mental models of the
premises, as in the pistol example, detects the inconsistency, and uses its
models of causal relations to build a chain resolving the inconsistency, for
example, a person emptied the pistol and so there were no bullets in the pistol
( Johnson-Laird et al., 2004). Such an explanation is bound to repudiate at
least one previous belief, which reasoners can modify to refer to a situation
that was once possible, but that did not occur, as in the counterfactual con-
ditional, if a person hadn’t emptied the pistol and there were bullets in the pistol then
the pistol would have fired (see Byrne, 2005). Experimental evidence showed
that individuals are usually able to create such explanations, which tend
to refute the conditional premise ( Johnson-Laird et al., 2004). Individuals
rate the cause and effect as more probable than either the cause alone or
the effect alone—a fallacy in which a conjunction is wrongly judged to be
more probable than its constituents (Tversky & Kahneman, 1983).
A study of abduction ( Johnson-Laird et al., 2004) examined such infer-
ences as:
If a pilot falls from a plane without a parachute then the pilot dies.
This pilot didn’t die. Why not?
Some participants made a valid deduction:
The pilot didn’t fall from a plane without a parachute.
But, other participants made explanatory abductions, such as:
The pilot fell into a deep snowdrift and so wasn’t hurt.

The plane was on the ground and he [sic] didn’t fall far.
The pilot was already dead.
An inadvertent demonstration of the imaginative power of human abduc-
tions used pairs of sentences chosen at random from pairs of stories, also
chosen at random (see Johnson-Laird, 2006, chap. 14). The result was pairs
of sentences such as:
Celia made her way to a shop that sold TVs.
Maria had just had her ears pierced.
The participants’ task was to describe “what is going on” in such scenar-
ios. To the experimenters’ surprise, the participants were usually able to
comply. The task was easier in another condition in which the sentences
were edited minimally to ensure that they both referred to the same
individual:
Celia made her way to a shop that sold TVs.
She had just had her ears pierced.
Typical examples of the participants’ responses in this case were:
She’s getting reception in her earrings and wanted the shop to investigate.
She wanted to see herself wearing earrings on closed-circuit TV.
She won a bet by having her ears pierced, using money to buy a new TV.
What was striking was how rarely individuals were stumped for an expla-
nation. Human reasoners are adept at abductions—they outperform any
existing computer program. Their explanatory ability underlies supersti-
tions ( Johnson-Laird, 2006, chap. 14). It also underlies science, but scientists
test putative explanations: they search for counterexamples.
A long-standing view of a rational reaction to inconsistency is encapsu-
lated in William James’s remark: “[The new fact] preserves the older stock
of truths with a minimum of modification, stretching them just enough
to make them admit the novelty” ( James, 1907, p. 59). Cognitive scien-
tists have often defended the same view (e.g. deKleer, 1986; Gärdenfors,
1992; Harman, 1986; cf. Elio & Pelletier, 1997, for results to the contrary).
Naive individuals, however, are much more concerned to explain incon-
sistencies because explanations can help them to decide what to do. They
readily sacrifice minimalism for this goal. For instance, when reasoners are
asked what follows from inconsistent premises, they spontaneously offer
an explanation that resolves the inconsistency, and they judge that such
explanations are more probable than revisions to the premises that restore
consistency (Khemlani & Johnson-Laird, 2011). Once they have formu-
lated such an explanation, a striking phenomenon occurs. It becomes
harder for them to detect the inconsistency in comparison with cases in
which instead of explaining the inconsistency, they rate which assertion

is more surprising. They seem to have explained the inconsistency away,
perhaps by reinterpreting the generalization in the premises as hold-
ing by default so that it is less vulnerable to contrary facts (Khemlani &
Johnson-Laird, 2012b). In sum, reasoners are able to resolve inconsisten-
cies. They tend to do so by using knowledge to abduce causal models that
explain the origins of the conflicts. This reasoning usually makes sense of
the inconsistency, although on some occasions it fails to yield any expla-
nation whatsoever.
9. PROBABILITIES: EXTENSIONAL AND INTENSIONAL

Induction is often uncertain, and uncertainty implies probability. Indi-
viduals who know nothing of the probability calculus happily infer proba-
bilities. How they make such inferences should be part of a unified theory of
reasoning. Following Tversky & Kahneman (1983), we distinguish between
extensional reasoning in which the probability of an event is inferred from
the different mutually exclusive ways in which it can occur and nonexten-
sional (intensional) reasoning in which the probability of an event is inferred
from some relevant evidence or index. In principle, extensional reasoning
is deductive, whereas nonextensional reasoning is inductive, and much of
it, as Tversky & Kahneman (1973, 1983) showed, depends on heuristics. In
daily life, probabilistic reasoning may mix extensional and nonextensional
processes. Hence, we explain how the model theory applies to both of them.
Mental models represent possibilities, and so simple extensional inferences
can be made on the assumption that each possibility is equiprobable bar-
ring evidence to the contrary (Johnson-Laird, Legrenzi, Girotto, Legrenzi, &
Caverni, 1999). The probability of an event is accordingly the proportion
of models in which it holds.The theory allows that models can also be tagged
with numerals denoting their probabilities. Similar principles underlie other
theories of probabilistic reasoning (e.g. Falk, 1992; Shimojo & Ichikawa,
1989). The model theory, however, assigns equiprobability, not to events but
to models of events. Classical probability theorists, such as de Laplace (1995);
(originally published in 1819), advocated an analogous principle of “indiffer-
ence”, but ran into difficulty because events can be partitioned in different
conflicting ways (Hacking, 1975). Mental models merely reflect the probabili-
ties that the individual constructing them assigns to events: different individu-
als can therefore partition events in different ways without self-contradiction.
A simple extensional problem (from Johnson-Laird et al., 1999) is:
In the box, there is a green ball or a blue ball or both.
What is the probability that both the green and the blue balls are there?
The mental models of the premise are:
Green
Blue
Green Blue
Hence, the equiprobability principle predicts correctly that individuals will

tend to estimate the probability of green and blue as 1/3. But, a telltale sign
of mental models is that individuals succumb to illusory inferences about
probabilities. Here is an example:
There is a box in which there is at least a red marble or else there is
a green marble and there is a blue marble, but not all three marbles.
What is the probability that there is a red marble and a blue marble in
the box?
The mental models of the premise represent two possibilities:
Red
Green Blue
They imply that the red and blue marbles cannot occur together, and so
their probability is zero. And most people make this estimate. However, the
disjunction means that when its first clause is true, its second clause is false,
and it can be false in three ways:
Red Green ¬ Blue
¬ Green Blue
¬ Green ¬ Blue
When the second clause of the premise is true, the first clause is false:
¬ Red Green Blue
Hence, there are four distinct possibilities for what’s in the box, and, on the
assumption of equiprobability, the probability of green and blue is, not zero,
but 25%. The participants in an experiment performed much better with
the control problems than with the illusory problems of this sort (Johnson-
Laird et al., 1999).
Turning to nonextensional reasoning, consider this question about a

unique event:
What is the probability that Hillary Clinton is elected US President in 2016?
Some psychologists argue that such probabilities are meaningless because
probabilities concern only the natural frequencies with which events occur
(e.g. Cosmides & Tooby, 1996). However, naive reasoners are happy to
make estimates for unique events, and, as we have observed, their estimates
correlate reliably over different contents, ranging from politics to climate
(Khemlani et al., 2012). The psychological mystery about such estimates
is what mental processes underlie them, and, in particular, where do the
numbers come from?
We proposed a dual process theory (see Section 3) in which the intuitive
system given, say, the question about Hillary, adduces evidence, such as: Hillary
was a very effective Senator; and many effective Senators have become President.
It uses a mental model representing this evidence to construct a primitive
non-numerical representation of a degree of belief in the proposition. This
iconic representation is akin to the following sort of diagram:
| −−−^ |
in which the left vertical represents impossibility, the right vertical rep-
resents certainty, and the pointer at the end of the line corresponds to
the strength of the particular belief, such as, Hillary will be elected President.
The intuitive system can translate this representation into the sorts of
description that a non-numerate individual would use, such as: “it’s as
likely as not”.
The deliberative system can map the degrees of belief represented in an
icon into a numerical estimate. Because this system has access to working
memory, it can carry out proper arithmetical operations. It can also try to
keep track of the complete joint probability distribution (the JPD). Given
two unique events, such as the election of Clinton in 2016 and the Demo-
crats gaining control of Congress, the JPD consists in the set of probabilities
for each possible combination of the affirmations and negations of the rel-
evant propositions:
Hillary is President & Democrats control Congress 35%
Hillary is President & not (Democrats control Congress) 30%
Not(Hillary is President) & Democrats control Congress 15%
Not(Hillary is President) & not(Democrats control Congress) 20%
The JPD provides all the information needed to estimate any probability
concerning the domain. There are many different sets of probabilities from
which the values of the JPD can be inferred. For instance, if you know the
values of the three probabilities in each of the following triples then, in
principle, you can infer that values of the probabilities in the JPD, where,
say, A denotes “Hillary is President” and B denotes “Democrats control
Congress”:
P(A), P(B), P(A and B)
P(A), P(B), P(A or B, or both)
P(A), P(B), P(A|B)
The last of these triples includes P(A|B), which is the conditional probabil-
ity of B on the assumption that A occurs.
Granted the limited ability of the intuitive system to carry out loops
of operations (see Section 3), it is capable of only a small number of
primitive analogs of arithmetical operations of the sort found in infants
(Barth et al., 2006; Dehaene, 1997; Xu & Spelke, 2000) and adults in
non-numerate cultures (Gordon, 2004). It can add two pointers, subtract
one from another, take their mean, and multiply a proportion signified
by one pointer by another—all within the bounds between certainty and
impossibility and all in crude error-prone ways. The theory accordingly
postulates that to estimate the probability of a conjunction of events,
reasoners should tend to split the difference between them, but some
may take the proportion of a proportion. The latter is a more complex
operation (in terms of Kolmogorov complexity, see Li & Vitányi, 1997),
and so it should tend to be used less often. Reasoners should likewise
make analogous inferences in estimating conditional and disjunctive
probabilities.
We implemented the intuitive and deliberative systems in a computer
model and tested its predictions in experiments (Khemlani et al., 2012).
The results showed that the participants concurred in the rank orders of
their estimates of the probabilities of unique events. For example, they
agreed that the US is more likely to make English the official language
of the country (the average estimate was a probability of 46%) than to
adopt an open border policy (an average estimate was a probability of
15%). Hence, they are to some extent relying on knowledge and men-
tal processes in common. They tended to estimate the probability of a
conjunction by taking the mean of their estimates of the probabilities of
its conjuncts. This tendency was even evident in the overall means, for
example, their mean estimate of the conjunction of the US adopting an
open border policy and making English the official language was 26%, a
value falling between their mean estimates of the two conjuncts. It yields a
violation of the JPD, that is, the negative probability in the third conjunc-
tion shown here:
English Open borders: 26%
English ¬ Open borders: 20%
¬ English Open borders: -11%
¬ English ¬ Open borders: 65%
Violations of the JPD, however, were smaller when the conjunction came
last as opposed to first in the sequence of judgments. When it was last,
the participants had already made numerical estimates of the probabili-
ties of its conjuncts, and so they could use a deliberative procedure, such
as taking a proportion of a proportion. This method is appropriate only
for independent events, but a prior study established that they were not
independent.
The model theory of probabilities dispels some common misconcep-
tions. Probabilistic reasoning isn’t always inductive. Extensional estimates
can be deductively valid, but they can also yield illusory values. Likewise,
nonextensional estimates of unique events depend on intuitions, and the
resulting violations of the JPD suggest that the probability calculus is not
native to human cognition. Individuals simulate events, but their restricted
repertoire of intuitive methods leads them into error.
10. MENTAL SIMULATIONS AND INFORMAL

PROGRAMS
Is there one sort of thinking that depends on mental simulation
and that cannot be explained in any other terms? In our view, there is.
It is the thinking that underlies the creation of algorithms and computer
programs. Expert programming is an intellectual discipline that depends
Figure 1.1 The railway domain with an example of an initial configuration in which a set
of cars is on the left side (a) of the track, the siding (b) can hold one or more cars while
other cars are moved to the right side of the track (c).
on knowledge of programming languages. Hence, our studies focused

on how nonprogrammers formulate algorithms in everyday language
(Khemlani & Johnson-Laird, 2013). To make the task easy, the programs
concerned the railway domain shown in Figure 1.1. The participants had
to rearrange the cars in a train on the left track using the siding so that
they arrived on the right track in the required new order. In the termi-
nology of automata theory, the siding acts as a “stack” on which to store
cars temporarily. Items on the siding move back to the left track, which
therefore also functions as a stack. Automata with two stacks are equiva-
lent to Universal Turing machines. Hence, if cars can also be added and
removed, the railway serves as a general-purpose computer (Hopcroft &
Ullman, 1979).
Most people can solve rearrangement problems in the railway
domain. They use a simple variant of “means-ends” analysis in which
they work backward from the required goal, invoking operations rel-
evant to reducing the difference between the current state and the goal
(e.g. Newell, 1990, Newell & Simon, 1972). For rearrangement prob-
lems, they need only envisage each successive car in the goal. Suppose,
for instance, they have to rearrange the order ABCD into ACBD. The
starting state is:
ABCD[ ]
where the square brackets denote the contents of the siding, which is
empty at the start. Their immediate goal is to get D to the far end of the
right track:
[ ]…D
So, they move D from left to right track:
ABC[ ]D
The next partial goal is to get B to the right track, and so they need to move
C out of the way onto the siding:
AB[C]D
Now, they can move B to the right:
A[C]BD
They move C off the stack:
AC[ ]BD
The next move is intriguing. They should move both A and C together
from left to right track. But, if reasoners perseverate, they may move only C
to the right track. Their solution won’t be minimal because they then have
to make a separate move of A to right track. Our initial study investigated
all 24 possible rearrangements of four cars, and the participants easily solved
each of them, but they did tend to perseverate: every participant made one
or more unnecessary moves.
In the principal experiment, the participants, who were not program-
mers, had to formulate algorithms for three sorts of rearrangement: reversals
in which, say, ABCDEFGH on left track becomes HGFEDCBA on right
track; palindromes in which, say, ABCDDCBA becomes AABBCCDD, and
parity sorts in which, say, ABCDEFGH becomes ACEGBDFH, that is, cars
in odd-numbered positions precede those in even-numbered positions.
Each solution calls for recursion, that is, a loop of operations. Primitive
recursion in the theory of recursive functions corresponds to a loop car-
ried out for a given number of times, a so-called “for-loop”, whereas mini-
mization corresponds to a loop carried out while a given condition holds,
a “while-loop” (see Rogers, 1967). While-loops are more powerful than
for-loops because only they can compute certain functions. Indeed, when
a while-loop is entered, there may be no way to determine how many
times it will repeat before it yields an output or whether it will ever halt
to yield an output. But, how do nonprogrammers formulate informal
algorithms? The task isn’t deductive: they can deduce the consequences
of a program, but they can’t create it using deduction alone (Kitzelmann,
Schmidt, Mühlpfordt, & Wysotzki, 2002). Likewise, they don’t rely on
probabilities any more than they do for Sudoku puzzles (Lee et al., 2008).
The one viable method is to simulate a solution to a problem, observe
what happens in the simulation, and translate these observations into a
description. The simulation depends on a kinematic sequence of men-
tal models representing successive states of the world, real or imaginary
(Johnson-Laird, 1983, chap. 15).
Let’s examine the process in more detail. The first step is to solve two
different examples of the relevant rearrangement problem. Without two
examples differing in numbers of cars, rearrangements are ambiguous. The
solution to reversing a train of four cars is as follows:
ABCD[ ], A[BCD], [BCD]A, B[CD]A, [CD]BA, C[D]BA, [D]CBA,
D[ ]CBA, [ ]DCBA
As this protocol illustrates, only three sorts of move are possible, and they
occur in these summaries of simulations that solve reversals of trains of four
and five cars:
S3 R1 L1 R1 L1 R1 L1 R1
S4 R1 L1 R1 L1 R1 L1 R1 L1 R1
where “S3” means move three cars to the siding from left track, “R1” means
move one car to right track from left track, and “L1” means move one car
to left track from the siding. The second step uses the two summaries to
work out the loop of moves they contain and any moves before or after it
(pace Miller, 1974,1981; Pane, Ratanamahatana, & Myers, 2001). The loop
in the simulations above is (R1 L1). But, how many times should it be iter-
ated? There are two ways to answer this question, depending on whether
reasoners are formulating a while-loop or a for-loop. The simpler way is to
observe the conditions in the simulation when the loop halts, which are
respectively:
D[ ]CBA
E[ ]DCBA
The condition that halts the loop is that no cars are left on the siding,
and so the while-loop should continue as long as the siding isn’t empty.
The alternative is to compute the number of times that a for-loop should be
executed. It calls for the solution of a pair of simultaneous linear equations
to obtain the values of a and b in:
Number of iterations = a × length of the train + b.
The final step maps the structure of the solution into an informal descrip-
tion. We implemented this entire process in a computer program, which
constructs programs for any rearrangement problem based on a single loop.
It produces a for-loop and a while-loop in Lisp and translates the while-
loop into informal English. Each of these functions solves any instance of
the relevant class of rearrangements. Table 1.1 presents its solutions for the
three sorts of rearrangement: reversals, palindromes, and parity sorts.
If individuals use simulation to devise algorithms, then they should tend
to use while-loops rather than for-loops because it is easier to observe the
halting condition of a while-loop than to solve simultaneous equations.
The overall difficulty of formulating an algorithm should depend on its
Kolmogorov complexity, which is the length of its shortest description in
a given language, such as Lisp (Li & Vitányi, 1997). A good proxy is the
number of instructions. In Table 1.1, the functions for reversals and palin-
dromes call for four instructions, whereas parity sorts call for five instruc-
tions. Within a given level of complexity, another factor should also affect
difficulty: the mean number of operands (i.e., cars) per move. This mea-
sure distinguishes reversals, which have 1.38 operands per move for eight
cars, from palindromes, which have 1.75 operands per move for eight cars.
Hence, the three sorts of problem should increase in difficulty from reversals
through palindromes to parity sorts.
Toward a Unif ied Theory of Reasoning
Table 1.1 Loops for Computing Minimal Solutions to Three Sorts of General Problem: Reversals, Palindromes, and Parity Sorts Using “for”-
loops and “while”-loops and Their Informal Description (from the Output of the Computer Program mReasoner for Abducing them)
While-loops
For-loops Lisp Informal English

a) Reversals (e.g., ABCDEFGH ⇒ HGFEDCBA)
(setf track (S (+ (* 1 len) -1) track)) (setf track (S (+ (* 1 len) -1) track)) Move one less than the cars to siding.
(loop for i from 1 to (+ (* 1 len) -1) (loop while (> (length (second track)) 0) While there are more than zero cars
do (setf track (R 1 track)) do (setf track (R 1 track)) on siding.
(setf track (L 1 track))) (setf track (L 1 track))) Move one car to right track.
(setf track (R 1 track)) (setf track (R 1 track)) Move one car to left track.
Move one car to right track.
b) Palindromes (e.g., ABCDDCBA ⇒ AABBCCDD)
(setf track (S (+ (* ½ len) -1) track)) (setf track (S (+ (* ½ len) -1) track)) Move one less than half the cars to
(loop for i from 1 to (+ (* 1/2 len) -1) (loop while (> (length(first track)) 2) siding.
do (setf track (R 2 track)) do (setf track (R 2 track)) While there are more than two cars
(setf track (L 1 track))) (setf track (L 1 track))) on left track.
(setf track (R 2 track)) (setf track (R 2 track)) Move two cars to right track.
Move one car to left track.
Move two cars to right track.
(Continued)
31
32
Table 1.1 Loops for Computing Minimal Solutions to Three Sorts of General Problem: Reversals, Palindromes, and Parity Sorts Using “for”-
loops and “while”-loops and Their Informal Description (from the Output of the Computer Program mReasoner for Abducing them)—(Cont’d)
While-loops
For-loops Lisp Informal English

c) Parity sorts (e.g., ABCDEFGH ⇒ ACEGBDFH)
(loop for i from 1 to (+ (* ½ len) -1) (loop while (> (length (first track)) 2) While there are more than two cars
do (setf track (R 1 track)) do (setf track (R 1 track)) on left track.
(setf track (S 1 track)) (setf track (S 1 track))) Move one car to right track.
P. N. Johnson-Laird and Sangeet S. Khemlani

(setf track (R 1 track)) (setf track (R 1 track)) Move one car to siding.
(setf track (L (+ (* ½ len) -1) track)) (setf track (L (+ (* ½ len) -1) track)) Move one car to right track.
(setf track (R (+ (* ½ len) 0) track)) (setf track (R (+ (* ½ len) 0) track)) Move one less than half the cars to
left track.
Move half the cars to right track.
Our experiment corroborated the predictions. Individuals were able to

create informal algorithms, even though they had no access to the railway
domain while they carried out the task. They formulated algorithms for
the three sorts of problems, once for trains of eight carriages, and once for
trains of any length, in a counterbalanced order. Performance with trains
of eight cars was at ceiling, but with trains of any length, it corroborated
the predicted trend in accuracy and in time. Likewise, the participants used
many more while-loops than for-loops. The use of while-loops correlated
with accuracy, whereas the use of for-loops had a negative correlation
with accuracy. Differences in ability were striking: the best participant
was correct on every problem, whereas the worst participant was correct
for only a third of the eight-car problems and for none of problems with
trains of any length.
The ability to create algorithms is useful in daily life: loops of opera-
tions are ubiquitous in everything from laying a table to shutting down
a nuclear reactor. Intelligent individuals are able to carry out this task,
and our results corroborated an account that bases their thinking on the
ability to make simulations. It is difficult to see how else they could cre-
ate programs.
11. TOWARD A UNIFIED THEORY

We have now described the main principles of the model theory,
illustrated them in various sorts of reasoning, and outlined programs
implementing special cases of the theory. But, these programs are a dozen
separate pieces, often employing ad hoc notations, and so an urgent task
is to integrate them. We have therefore begun to integrate the disparate
parts within a single unified theory, implementing it in a large-scale
computer program, mReasoner (available at http://mentalmodels.princ-
eton.edu/models/mreasoner/). The problem is to bring together various
sorts of reasoning (e.g. relational, sentential, modal, causal, quantifica-
tional) with various sorts of task (e.g. formulating conclusions, evaluating
given conclusions, evaluating consistency) in a way that predicts various
sorts of phenomena (e.g. accuracy, latency, effects of modulation). Our
aim here was to describe the more important insights that have emerged
so far.
A mental model can represent the spatial relations among a triangle,
circle, and square (as in Section 4), but the size of the figures, their distance
apart, and so on, may not be intended to represent their real sizes or dis-
tances apart. Analogous issues occur with mental models of other sorts of
assertion. For instance, a quantified assertion, such as:
Some of the actors are bakers
has the following sort of iconic model shown in this diagram of four
individuals:
Actor Baker
Actor Baker
Actor
Baker
The numbers of mental tokens in this case are not intended to represent
the actual numbers of actors or bakers. Only the overlap between the two
sets is iconic. When reasoners search for an alternative model of a set of
premises, they can modify all but the essentials of a model. So, how does
the system keep track of the essentials? A single uniform answer is that
it relies on the meanings of assertions. Hence, mReasoner uses a gram-
mar, a lexicon, and a parser to construct representations of meanings, that
is, intensional representations (for an account of their construction, see
Khemlani, Lotstein, & Johnson-Laird, submitted for publication). They are
then used to build extensional representations, that is, mental models. Both
sorts of representations are crucial in reasoning, and to illustrate this point,
we consider the intuitive and deliberative systems in reasoning from quan-
tified assertions.
The model theory treats the intensions of quantified assertions as rela-
tions between sets (see Boole, 1854; Cohen & Nagel, 1934).The advantages
of this treatment are twofold. First, it dovetails with a long-standing treat-
ment of quantifiers in the model theory (Johnson-Laird, 1983, chap. 15):
a set is represented iconically as a set of mental tokens, and a quantified
assertion is represented as a relation between such sets. Second, it works for
all quantifiers in natural language, including those such as “more than half
of the artists”, which cannot be defined using the quantifiers that range
over entities in logic (Barwise & Cooper, 1981). Here are some illustrative
examples of this treatment of quantifiers, which the intensions of assertions
capture:
All A are B A B (A is included in B.)
Some A are B A B (Intersection of A and

B is not empty.)
No A is a B A B= (Intersection of A and
B is empty.)
Some A are not B A−B (Set of A that are not

B is not empty.)
Most A are B |A B|> |A− B| (Cardinality of intersection >

cardinality of A that are not B.)
More than half the |A B|> |A|/2 (Cardinality of intersection >

A are B ½ of cardinality of A.)
A corollary is that determiners, such as “most”, have parameters specifying

such matters as the minimal cardinality of A, the cardinality of the quantifier,
“most A”, constraints on the relation between the two cardinalities, and
so on.
The intuitive system can construct a model of an assertion and update
the model according to subsequent assertions. Hence, given the premises of
a syllogism:
Some of the actors are bakers.
All the bakers are colleagues.
It updates the model above to:
Actor Baker Colleague
Actor Baker Colleague
Actor
Baker Colleague
In general, affirmative premises are added so as to minimize the number
of distinct sorts of individual, whereas negative premises are added so as to
maximize the number of distinct sorts of individual.
Once the intuitive system has an initial model, it can draw a conclusion
establishing a new set-theoretic relation, that is, a relation that is not asserted
in the premises. In the past, the model theory has eschewed heuristics, but
it now embodies them to frame both the quantifier in the conclusion (its
mood) and the order of the terms that occur in it: “actors” and “colleagues”
(its figure). When two premises differ in mood, one of them dominates the
other in determining the mood and figure of the initial conclusion. The
order of dominance reflects two principles governing valid inferences:
• A negative premise can yield only valid conclusions that are negative.
• Within both negative and affirmatives, a particular premise—one based,
for instance, on “some”—can yield only valid conclusions that are par-
ticular.
The resulting order of dominance for syllogisms is accordingly:
Some _ are not _ > No _ are _ > Some _ are _ > All _ are_
In our example, the first premise is dominant: “artists” is its subject, and so
“artists” is the subject of the conclusion, and the term in the other premise,
“colleagues”, is in the predicate of the conclusion, that is, “some of the art-
ists are colleagues”. Analogous principles apply to other sorts of premise.
They account for the well-known figural effect that occurs in syllogistic
reasoning, for example, the tendency to infer the conclusion above rather
than the converse,“some colleagues are artists” (see, e.g. Bucciarelli & Johnson-
Laird, 1999). The order of dominance is the same as the order invoked in
Chater and Oaksford (1999) from probabilistic considerations. But, since
our principles derive from valid inferences, and yield only conclusions that
hold in initial mental models, they do not depend on probabilities.The heu-
ristics operate without storing any information in working memory, and so
they are rapid, but fallible.
The deliberative system makes a recursive search for alternative mod-
els falsifying an initial conclusion. When the system finds a counterex-
ample, it formulates a new conclusion if one is possible or else declares
that no definite conclusion follows about the relation between the end
terms. It searches for counterexamples using the same operations as
did participants working with external models in the form of cutout
shapes (Bucciarelli & Johnson-Laird, 1999): adding a new individual to
a model, breaking an individual into two, and moving a property from
one individual to another. As a meta-analysis showed, the resulting the-
ory embodied in mReasoner outperforms all current theories of syllogis-
tic reasoning (see Khemlani & Johnson-Laird, 2012c; Khemlani et al.,
submitted for publication).
The implementation of the unified theory has so far focused on unique

probabilities and quantified assertions. It draws its own conclusions from
quantified premises, evaluates given conclusions about what is necessary or
about what is possible, formulates counterexamples to putative conclusions,
and evaluates whether or not a set of quantified assertions is consistent. It
carries out these tasks using both the intuitive system, which builds initial
models, and the deliberative system, which searches for alternative models.
Hence, it is able to predict human inferences. It provides the beginnings of
a unified computational account, which we are extending to accommodate
sentential and relational reasoning.
12. CONCLUSIONS
The psychology of reasoning would have been simpler if human
beings were logicians or probabilists. Logic and the probability calculus
are not native mental faculties but cultural discoveries. Some individu-
als master these technologies; some do not. And, in our culture, most
individuals have smatterings of them at best. As the model theory pre-
dicts, deductive and probabilistic inferences are difficult and fallible. An
awareness of the occurrence of errors led Aristotle and his intellectual
descendants to devise logical and probabilistic calculi. These technolo-
gies, however, are unlikely foundations for human reasoning. So, what is?
We have argued that it is mental simulation. Reasoners build models of
premises and base their inferences on them. This view seems undeniable
for reasoning that creates informal algorithms. The evidence we have
presented shows that it applies also to all the main domains and tasks of
reasoning, from deductions based on sentential connectives to inductions
about the probabilities of unique events. But, its manifold applications are
a source of its main weakness—its potential disintegration into a bunch
of separate subtheories. Their unification is viable because each subthe-
ory is constrained by the main principles of the theory. What is much
harder is to implement a computer program that predicts the responses
that reasoners make to any inferential task, but mReasoner is a step toward
that goal.
ACKNOWLEDGMENTS
This research was supported by a National Science Foundation Grant No. SES 0844851 to
the first author to study deductive and probabilistic reasoning and by a National Research
Council Research Associateship to the second author. We are grateful to more colleagues
than we can name here, but many of them can be found among the references.
REFERENCES
Bar-Hillel,Y. (1964). Language and information processing. Reading, MA: Addison-Wesley.
Barnes, J. (Ed.), (1984). The complete works of Aristotle. Princeton, NJ: Princeton University
Press.
Barres, P., & Johnson-Laird, P. N. (2003). On imagining what is true (and what is false).
Thinking & Reasoning, 9, 1–42.
Barrouillet, P., & Lecas, J.-F. (1998). How can mental models theory account for content
effects in conditional reasoning? A developmental perspective. Cognition, 67, 209–253.
Barrouillet, P., & Lecas, J.-F. (1999). Mental models in conditional reasoning and working
memory. Thinking & Reasoning, 5, 289–302.
Barrouillet, P., Grosset, N., & Lecas, J. F. (2000). Conditional reasoning by mental models:
chronometric and developmental evidence. Cognition, 75, 237–266.
Barth, H., La Mont, K., Lipton, J., Dehaene, S., Kanwisher, N., & Spelke, E. S. (2006).
Nonsymbolic arithmetic in adults and young children. Cognition, 98, 199–222.
Barwise, J., & Cooper, R. (1981). Generalized quantifiers and natural language. Linguistics and
Philosophy, 4, 159–219.
Bauer, M. I., & Johnson-Laird, P. N. (1993). How diagrams can improve reasoning. Psychological
Science, 4, 372–378.
Boole, G. (1854). An investigation of the laws of thought. London: Macmillan.
Braine, M. D. S. (1978). On the relation between the natural logic of reasoning and standard
logic. Psychological Review, 85, 1–21.
Brewka, G., Dix, J., & Konolige, K. (1997). Nonmonotonic reasoning: An overview. Stanford, CA:
CLSI Publications, Stanford University.
Bryant, P. E., & Trabasso, T. (1971). Transitive inferences and memory in young children.
Nature, 232, 456–458.
Bucciarelli, M., & Johnson-Laird, P. N. (1999). Strategies in syllogistic reasoning. Cognitive
Science, 23, 247–303.
Bucciarelli, M., & Johnson-Laird, P. N. (2005). Naive deontics: a theory of meaning,
representation, and reasoning. Cognitive Psychology, 50, 159–193.
Byrne, R. M. J., & Johnson-Laird, P. N. (1989). Spatial reasoning. Journal of Memory & Language,
28, 564–575.
Byrne, R. M. J. (2005). The rational imagination: How people create alternatives to reality.
Cambridge, MA: MIT.
Carreiras, M., & Santamaría, C. (1997). Reasoning about relations: spatial and nonspatial
problems. Thinking & Reasoning, 3, 191–208.
Chater, N., & Oaksford, M. (1999). The probability heuristics model of syllogistic reasoning.
Cognitive Psychology, 38, 191–258.
Cohen, M. R., & Nagel, E. (1934). An introduction to logic and scientific method. London:
Routledge & Kegan Paul.
Cohen, L. J. (1981). Can human irrationality be experimentally demonstrated? Behavioral and
Brain Sciences, 4, 317–370.
Cosmides, L., & Tooby, J. (1996). Are humans good intuitive statisticians after all? R
e-thinking
some conclusions of the literature on judgment under uncertainty. Cognition, 58, 1–73.
Cosmides, L., & Tooby, J. (2005). Neurocognitive adaptations designed for social exchange.
In D. M. Buss (Ed.), Evolutionary psychology handbook (pp. 584–627). New York: Wiley.
Craik, K. (1943). The nature of explanation. Cambridge: Cambridge University Press.
Dehaene, S. (1997). The number sense. Oxford, UK: Oxford University Press.
deKleer, J. (1986). An assumption-based TMS. Artificial Intelligence, 28, 127–162.
Elio, R., & Pelletier, F. J. (1997). Belief change as propositional update. Cognitive Science, 21,
419–460.
Evans, J. St. B.T. (2008). Dual-processing accounts of reasoning, judgment and social cogni-
tion. Annual Review of Psychology, 59, 255–278.
Evans, J. St. B.T. (2012). Questions and challenges for the new psychology of reasoning.
Thinking & Reasoning, 18, 5–31.
Falk, R. (1992). A closer look at the probabilities of the notorious three prisoners. Cognition,
43, 197–223.
Frosch, C. A., & Johnson-Laird, P. N. (2011). Is everyday causation deterministic or probabilistic?
Acta Psychologica, 137, 280–291.
García-Madruga, J. A., Moreno, S., Carriedo, N., Gutiérrez, F., & Johnson-Laird, P. N. (2001).
Are conjunctive inferences easier than disjunctive inferences? A comparison of rules and
models. Quarterly Journal of Experimental Psychology, 54A, 613–632.
Gärdenfors, P. (1992). Belief revision: an introduction. In P. Gärdenfors (Ed.), Belief revision
(pp. 1–20). Cambridge, UK: Cambridge University Press.
Girotto, V., Mazzocco, A., & Tasso, A. (1997). The effect of premise order in conditional
reasoning: a test of the mental model theory. Cognition, 63, 1–28.
Goldvarg, Y., & Johnson-Laird, P. N. (2000). Illusions in modal reasoning. Memory & Cogni-
tion, 28, 282–294.
Goldvarg,Y., & Johnson-Laird, P. N. (2001). Naive causality: a mental model theory of causal
meaning and reasoning. Cognitive Science, 25, 565–610.
Goodwin, G. P., & Johnson-Laird, P. N. (2005). Reasoning about relations. Psychological
Review, 112, 468–493.
Goodwin, G. P., & Johnson-Laird, P. N. (2008). Transitive and pseudo-transitive inferences.
Cognition, 108, 320–352.
Gordon, P. (2004). Numerical cognition without words: evidence from Amazonia. Science,
306, 496–499.
Hacking, I. (1975). The emergence of probability. Cambridge: Cambridge University Press.
Harman, G. (1986). Change in view: Principles of reasoning. Cambridge, MA: MIT Press,
Bradford Book.
Henle, M. (1962). On the relation between logic and thinking. Psychological Review, 69,
366–378.
Hopcroft, J. E., & Ullman, J. D. (1979). Formal languages and their relation to automata. Reading,
MA: Addison-Wesley.
Inhelder, B., & Piaget, J. (1958). The growth of logical thinking from childhood to adolescence.
London: Routledge, Chapman & Hall.
Jahn, G., Knauff, M., & Johnson-Laird, P. N. (2007). Preferred mental models in reasoning
about spatial relations. Memory & Cognition, 35, 2075–2087.
James, W. (1907). Pragmatism—a new name for some old ways of thinking. New York: Longmans,
Green.
Jeffrey, R. (1981). Formal logic: Its scope and limits (2nd ed.). New York, NY: McGraw-Hill.
Johnson-Laird, P. N., & Byrne, R. M. J. (1991). Deduction. Hillsdale, NJ: Erlbaum.
Johnson-Laird, P. N., & Byrne, R. M. J. (2002). Conditionals: a theory of meaning, pragmatics,
and inference. Psychological Review, 109, 646–678.
Johnson-Laird, P. N., & Hasson, U. (2003). Counterexamples in sentential reasoning. Memory
& Cognition, 31, 1105–1113.
Johnson-Laird, P. N., & Savary, F. (1995). How to make the impossible seem probable. In J. D.
Moore & J. F. Lehman (Eds.), Proceedings of the seventeenth annual conference of the cognitive
science society. Mahwah, NJ: Erlbaum.
Johnson-Laird, P. N., & Savary, F. (1999). Illusory inferences: a novel class of erroneous
deductions. Cognition, 71, 191–229.
Johnson-Laird, P. N., Byrne, R. M. J., & Schaeken, W. S. (1992). Propositional reasoning by
model. Psychological Review, 99, 418–439.
Johnson-Laird, P. N., Legrenzi, P., Girotto,V., Legrenzi, M., & Caverni, J.-P. (1999). Naive prob-
ability: a mental model theory of extensional reasoning. Psychological Review, 106, 62–88.
Johnson-Laird, P. N., Girotto, V., & Legrenzi, P. (2004). Reasoning from inconsistency to
consistency. Psychological Review, 111, 640–661.
Johnson-Laird, P. N., Lotstein, M., & Byrne, R. M. J. (2012). The consistency of disjunctive
assertions. Memory & Cognition, 40, 769–778.
Johnson-Laird, P. N. (1975). Models of deduction. In R. Falmagne (Ed.), Reasoning: Represen-
tation and process (pp. 7–54). Springdale, NJ: Erlbaum.
Johnson-Laird, P. N. (1983). Mental models. Cambridge: Cambridge University Press.
(Cambridge, MA: Harvard University Press).
Johnson-Laird, P. N. (2004).The history of mental models. In K. Manktelow & M. C. Chung
(Eds.), Psychology of reasoning:Theoretical and historical perspectives (pp. 179–212). New York:
Psychology Press.
Johnson-Laird, P. N. (2006). How we reason. New York: Oxford University Press.
Juhos, C., Quelhas, C., & Johnson-Laird, P. N. (2012). Temporal and spatial relations in sen-
tential reasoning. Cognition, 122, 393–404.
Kahneman, D. (2011). Thinking, fast and slow. New York, NY: Farrar, Strauss, Giroux.
Khemlani, S., & Johnson-Laird, P. N. (2009). Disjunctive illusory inferences and how to
eliminate them. Memory & Cognition, 37, 615–623.
Khemlani, S., & Johnson-Laird, P. N. (2011). The need to explain. Quarterly Journal of
Experimental Psychology, 64, 276–288.
Khemlani, S., & Johnson-Laird, P. N. (2012a). The processes of inference. Argument and
Computation, 1–17. (iFirst).
Khemlani, S., & Johnson-Laird, P. N. (2012b). Hidden conflicts: explanations make inconsis-
tencies harder to detect. Acta Psychologica, 139, 486–491.
Khemlani, S., & Johnson-Laird, P. N. (2012c). Theories of the syllogism: a meta-analysis.
Psychological Bulletin, 138, 427–457.
Khemlani, S., & Johnson-Laird, P.N. (2013). Mental simulation and the construction of informal
algorithms. In M. Knauff M. Pauen, N. Sebanz, & I. Wachsmuth (Eds.), Proceedings of
the 35th Annual Conference of the Cognitive Science Society. Austin, TX: Cognitive
Science Society.
Khemlani, S., Lotstein, M., & Johnson-Laird, P. N. (2012). The probabilities of unique events.
PLoS ONE, 7, 1–9. (Online version).
Khemlani, S., Lotstein, M., & Johnson-Laird, P. N. A unified theory of syllogistic reasoning,
submitted for publication.
Khemlani, S., Orenes, I., & Johnson-Laird, P. N. (2012). Negation: A theory of its meaning,
representation, and use. Journal of Cognitive Psychology, 24, 541–559.
Kitzelmann, E., Schmidt, U., Mühlpfordt, M., & Wysotzki, F. (2002). Inductive synthesis
of functional programs. In J. Calmet B. Benhamou, et al. (Eds.), Artificial intelligence,
automated reasoning, and symbolic computation (pp. 26–37). New York: Springer.
Knauff, M., & Johnson-Laird, P. N. (2002).Visual imagery can impede reasoning. Memory &
Cognition, 30, 363–371.
Knauff, M., & Ragni, M. (2011). Cross-cultural preferences in spatial reasoning. Journal of
Cognition and Culture, 11, 1–21.
Knauff, M., Fangmeier, T., Ruff, C. C., & Johnson-Laird, P. N. (2003). Reasoning, models,
and images: behavioral measures and cortical activity. Journal of Cognitive Neuroscience, 15,
559–573.
Knauff, M. (2013). Space to reason: A spatial theory of human thought. Cambridge, MA: MIT Press.
Kosslyn, S. M. (1994). Image and brain. Cambridge, MA: MIT Press.
Kroger, J. K., et al. (2002). Recruitment of anterior dorsolateral prefrontal cortex in human
reasoning: a parametric study of relational complexity. Cerebral Cortex, 12, 477–485.
Kroger, J. K., Nystrom, L. E., Cohen, J. D., & Johnson-Laird, P. N. (2008). Distinct neural
substrates for deductive and mathematical processing. Brain Research, 1243, 86–103.
Krumnack, A., Bucher, L., Nejasmic, J., Nebel, B., & Knauff, M. (2011). A model for relational
reasoning as verbal reasoning. Cognitive Systems Research, 11, 377–392.
de Laplace, P.-S. (1995). Philosophical essay on probabilities. New York: Springer- Verlag.
(Originally published in 1819).
Lee, N. Y. L., Goodwin, G. P., & Johnson-Laird, P. N. (2008). The psychological problem of
sudoku. Thinking & Reasoning, 14, 342–364.
Legrenzi, P., Girotto, V., & Johnson-Laird, P. N. (2003). Models of consistency. Psychological
Science, 14, 131–137.
Leibniz, G. W. (1685, 1952). The art of discovery. In P. P. Weiner (Ed.), Selections—Gottfried
Wilhelm Leibniz. New York: Charles Scribners. (Originally published 1685).
Li, M., & Vitányi, P. (1997). An introduction to Kolmogorov complexity and its applications (2nd
ed.). New York: Springer-Verlag.
Miller, L. (1974). Programming by non-programmers. International Journal of Man–Machine
Studies, 6, 237–260.
Miller, L. (1981). Natural language programming: styles, strategies, and contrasts. IBM Systems
Journal, 20, 184–215.
Newell, A., & Simon, H. A. (1972). Human problem solving. Englewood Cliffs: Prentice-Hall.
Newell, A. (1990). Unified theories of cognition. Cambridge, MA: Harvard University Press.
Oaksford, M., & Chater, N. (2007). Bayesian rationality. Oxford: Oxford University Press.
Orenes, I., & Johnson-Laird, P. N. (2012). Logic, models, and paradoxical inferences. Mind &
Language, 27, 357–377.
Osherson, D. N. (1974). 1976). Logical abilities in children. (Vols. 1–4). Hillsdale, NJ: Erlbaum.
Pane, J. F., Ratanamahatana, C. A., & Myers, B. A. (2001). Studying the language and
structure in non-programmers’ solutions to programming problems. International Journal
of Human-Computer Studies, 54, 237–264.
Peirce, C. S. (1931–1958). Collected papers of Charles Sanders Peirce. ( Vols. 8). C. Hartshorne,
P. Weiss, & A. Burks, (Eds.), Cambridge, MA: Harvard University Press.
Perrow, C. (1984). Normal accidents: Living with high-risk technologies. New York: Basic
Books.
Polk, T. A., & Newell, A. (1995). Deduction as verbal reasoning. Psychological Review, 102,
533–566.
Quelhas, A. C., Johnson-Laird, P. N., & Juhos, C. (2010). The modulation of conditional
assertions and its effects on reasoning. Quarterly Journal of Experimental Psychology, 63,
1716–1739.
Reitman, W. R. (1965). Cognition and thought. New York: Wiley.
Rips, L. J. (1994). The psychology of proof. Cambridge, MA: MIT Press.
Rogers, H. (1967). Theory of recursive functions and effective computability. New York:
McGraw-Hill.
Schaeken,W. S., Johnson-Laird, P. N., & d’Ydewalle, G. (1996a). Mental models and temporal
reasoning. Cognition, 60, 205–234.
Schaeken, W. S., Johnson-Laird, P. N., & d’Ydewalle, G. (1996b). Tense, aspect, and temporal
reasoning. Thinking & Reasoning, 2, 309–327.
Shaw, P., Greenstein, D., Lerch, J., Clasen, L., Lenroot, R., Gogtay, N., et al. (2006). Intellectual
ability and cortical development in children and adolescents. Nature, 440, 676–679.
Shimojo, S., & Ichikawa, S. (1989). Intuitive reasoning about probability: theoretical and
experimental analyses of the “problem of three prisoners”. Cognition, 32, 1–24.
Sloman, S. A. (1996). The empirical case for two systems of reasoning. Psychological Bulletin,
119, 3–22.
Stanovich, K. E. (1999). Who is rational? Studies of individual differences in reasoning. Mahwah,
NJ: Erlbaum.
Stenning, K., & van Lambalgen, M. (2008). Human reasoning and cognitive science. Cambridge,
MA: MIT Press.
Störring, G. (1908). Experimentelle Untersuchungen über einfachen Schlussprozesse. Archiv
für die gesamte Psychologie, 11, 1–27.
Tenenbaum, J. B., & Griffiths, T. L. (2001). Generalization, similarity, and Bayesian inference.
Behavioral and Brain Sciences, 24, 629–640.
Tversky, A., & Kahneman, D. (1973). Availability: a heuristic for judging frequency and
probability. Cognitive Psychology, 5, 207–232.
Tversky, A., & Kahneman, D. (1983). Extensional versus intuitive reasoning: the conjunction
fallacy in probability judgment. Psychological Review, 90, 292–315.
Vandierendonck, A., Dierckx, V., & De Vooght, G. (2004). Mental model construction in
linear reasoning: evidence for the construction of initial annotated models. Quarterly
Journal of Experimental Psychology, A, 57, 1369–1391.
Verschueren, N., Schaeken, W., & d’Ydewalle, G. (2005). A dual-process specification of
causal conditional reasoning. Thinking & Reasoning, 11, 278–293.
Walsh, C., & Johnson-Laird, P. N. (2004). Co-reference and reasoning. Memory & Cognition,
32, 96–106.
Waltz, J. A., Knowlton, B. J., Holyoak, K. J., Boone, K. B., Mishkin, F. S., et al. (1999). A system
for relational reasoning in human prefrontal cortex. Psychological Science, 10, 119–125.
Xu, F., & Spelke, E. S. (2000). Large number discrimination in 6-month-old infants. Cognition,
74B, 1–11.
Yang, Y., & Johnson-Laird, P. N. (2000). Illusions in quantified reasoning: how to make the
impossible seem possible, and vice versa. Memory & Cognition, 28, 452–465.
CHAPTER TWO
The Self-Organization of Human

Interaction
Rick Dale*,1, Riccardo Fusaroli*,‡,§, Nicholas D. Duran*,†,
Daniel C. Richardson†
*Cognitive and Information Sciences, University of California Merced, Merced, CA, USA
†Division of Psychology and Language Sciences, University College London, London, UK
‡Interacting Minds Center, Aarhus University, Aarhus, Denmark
§Center for Semiotics, Aarhus University, Aarhus, Denmark
1Corresponding author: E-mail: [email protected] or cognaction.org/rick
Contents
1. Introduction: The “Centipede’s Dilemma” of Interaction Research 44
2. An Example Theoretical Debate and the Need for Integration 46
3. Self-Organization and Human Interaction 49
3.1. The Need to Integrate Accounts of Cognition in Linguistic Interaction 49
3.2. Dynamics, Self-organization, and All that Jazz 50
3.2.1. Complex System 51
3.2.2. Self-Organization 53
3.2.3. Synergies and the Reduction of Degrees of Freedom 54
3.3. Summary, Social Modulation, and Multimodal Coordination 55
3.3.1. Social Modulation of Cognitive Dynamics 56
3.3.2. Coordination, Complementarity, Synergies 56
4. Cognitive Dynamics under Social Constraints 57
4.1. Social Modulation of the Dynamics of Low-Level Visual Attention 57
4.2. Social Modulation of Higher Level Processes, Like Perspective-Taking 60
4.3. Perspective-Taking as Self-Organization under Social Constraint 63
5. Coordination, Complementarity, and Interactive Performance 68
5.1. Behavioral Synchrony and Interactive Alignment 69
5.2. An Alternative Model: Interpersonal Synergies 71
5.2.1. Complementarity 72
5.2.2. Interactional Patterns 73
5.3. Testing Models of Linguistic Coordination: Alignment and Synergy 74
5.4. Interpersonal Synergies: A Summary 77
6. Conclusion: Time for More Models 77
6.1. Summary77
6.2. Moving Forward: Models of These Processes 78
6.3. Surface Network Analysis, and Mechanistic Models 78
6.4. Conclusion84
Acknowledgments84
References84

44 Rick Dale et al.
Abstract
We describe a “centipede’s dilemma” that faces the sciences of human interaction.
Research on human interaction has been involved in extensive theoretical debate,
although the vast majority of research tends to focus on a small set of human behaviors,
cognitive processes, and interactive contexts. The problem is that naturalistic human
interaction must integrate all of these factors simultaneously, and grander theoretical
mitigation cannot come only from focused experimental or computational agendas.
We look to dynamical systems theory as a framework for thinking about how these
multiple behaviors, processes, and contexts can be integrated into a broader account
of human interaction. By introducing and utilizing basic concepts of self-organization
and synergy, we review empirical work that shows how human interaction is flexible
and adaptive and structures itself incrementally during unfolding interactive tasks, such
as conversation, or more focused goal-based contexts. We end on acknowledging
that dynamical systems accounts are very short on concrete models, and we briefly
describe ways that theoretical frameworks could be integrated, rather than endlessly
disputed, to achieve some success on the centipede’s dilemma of human interaction.
1. INTRODUCTION: THE “CENTIPEDE’S DILEMMA” OF

INTERACTION RESEARCH
Next time you have a conversation, pay close attention to what you
and your partner are doing. This self-consciousness can be a bit jarring.
Like the so-called “centipede’s dilemma”, attempting awareness of your
numerous cognitive processes and behaviors, and those of your conver-
sation partner, can quickly disrupt a natural flowing performance. Dialog
otherwise seems so easy (Garrod & Pickering, 2004). How do we do it? The
famous poem by Katherine Craster has a toad posing to a centipede, “Pray,
which leg moves after which?” The centipede goes about some introspec-
tion, attempting awareness of this coordination, only to find that she can no
longer move.
The same thing seems to happen if we do this in a conversation. The
information, and relevant cognitive mechanisms, that can bear on a conver-
sational performance probably outnumbers a centipede’s legs, especially if
you include the array of processes that are not available to conscious report.
How do we coordinate everything?
In this review article, we consider a fundamental and still unsolved puz-
zle faced by the fields that study human interaction. Much like consciously
focusing on ourselves while conversing, the scientific agenda itself also suf-
fers from a kind of centipede’s dilemma. Ongoing work tends to focus on
particular levels of analysis. For example, we know that people use similar
vocabulary during interaction, may tend to match in other linguistic styles,
The Self-organization of Human Interaction 45
or even take on similar bodily postures and movements. Through explora-

tion of these levels, there are many theoretical proposals of candidate cog-
nitive and social processes. Terminologies that populate these theories are
diverse, conceptually overlapping, and still don’t enjoy consensus definitions:
mirroring, simulating, coupling, entrainment, coordination, imitation, mim-
icry, alignment, synchrony, joint action, theory of mind, perspective-taking,
mutuality, accommodation, empathy, contagion, and more.
What is still lacking is a systematic agenda to uncover how these various
processes work together to bring about multimodal coordination between two
interacting people.The current agenda of isolating processes and developing
broad theoretical proposals from relatively circumscribed domains is some-
what like the centipede’s analysis of its performance leg by leg by leg.There
is now a heterogenous assemblage of experimental techniques and obser-
vational analyses and an associated array of diverse theoretical mechanisms
that have yet to be integrated. The vibrant debates that ensue focus almost
entirely on some subset of experimental paradigms, cognitive processes, or
social contexts. The result is an exciting, evolving, but fractured domain.
Admittedly, framing the problem in this way may seem overly grandiose.
But if a comprehensive theory of human interaction is our goal, then this
is a real puzzle to be solved. We allay any enticement (or skepticism) here:
we’re not going to solve the puzzle in this paper. We will, however, propose
one potential route to a solution.To do this, we look to concepts of adaptive
and self-organizing systems, drawing from the tradition of the “dynamical
systems framework” as it has come to be known in the cognitive sciences.
Importantly, this is a framework for thinking about the problem of coordina-
tion during linguistic interaction. It does not answer the question of what
the precise array of mechanisms is or the processes of their interaction. It
also does not, at least by necessity, replace extant theoretical proposals. Our
positive thesis is simple: we will argue that the dynamical systems frame-
work may help to integrate existing theories.
In what follows, we begin with more background on the debate about the
mechanisms underlying human interaction. We then introduce some funda-
mental concepts of dynamics.These concepts bring about some generic expec-
tations about how human interaction should be coordinated and structured.We
argue that, in fancy terminology, “self-organization into functional synergies”
should be (and is) evident in interactive data (Section 2).
We showcase some evidence for this in a review of empirical literature.
This background review focuses on two key aspects of linguistic interaction.
The first (Section 4) is that basic social variables can sharply modulate the
46 Rick Dale et al.
many behaviors involved in interaction, at several levels of analysis. We look

to low-level visual attention, and then higher level spatial perspective-taking,
during linguistic interaction. The second empirical review is a glance at
complementarity between behaviors of two people interacting, extending
current theories of alignment in dialog and looking to the usefulness of the
concept of “synergy” (Section 5).
Following this, in a concluding discussion, we relate dynamical sys-
tems to other theories (Section 6). We speculate on some ways that these
dynamic concepts could be pursued in other computational models as a
means of becoming more precise about what exactly is coordinating dur-
ing human interaction. As we articulate in this section, a common and
important objection to dynamical systems accounts is that they are weak on
identifying mechanisms: dynamical theorists argue “interactions dominate
cognition,” and focus almost exclusively on relatively indirect measurement
outcomes of that interaction; however, where there is interaction, there must
be things interacting. We address this issue in this final section by arguing
that an important way to move forward is to integrate useful concepts from
dynamics with some computational frameworks already exploring language
and complex cognition.
2. AN EXAMPLE THEORETICAL DEBATE AND THE

NEED FOR INTEGRATION
There are some prominent theoretical debates in the realms of dis-
course and psycholinguistics. One of the best known debates revolves
around how, and how much, human beings track information about one
another as they interact. Some theories posit a two-stage process, with
primacy given to egocentric (self-centered) processes, and more social
“other-centric” processes coming online only more slowly and strategically
(e.g. Barr, 2008; Keysar, Lin, & Barr, 2003; Lin, Keysar, & Epley, 2010). Other
theories posit a fundamental sensitivity to a conversation partner, with a
rich layering of common ground that emerges while two people talk (e.g.
Clark, 1996; Schober & Brennan, 2003; for recent discussion, see Brennan,
Galati, & Kuhlen, 2010; Brennan & Hanna, 2009; Brown-Schmidt & Hanna,
2011; Shintel & Keysar, 2009).
Other theoretical agendas in this debate have aimed to specify key cog-
nitive processes that permit one person to keep track of, or continually adapt
to, their conversation partner. Some of these accounts centralize a process
of multilevel “alignment,” which can be automatic and often nonconscious
and builds common representational states across individuals while they talk
(Garrod & Pickering, 2004; Pickering & Garrod, 2004). Some have taken
the suggestion of a human “mirror system” to be central, specifying core
social processes that must be in place for us to interact successfully (for
review, see Gallese, 2008). Recent accounts have articulated the important
role of executive function during conversation (Brown-Schmidt, 2009a,b),
of memory (Horton, 2005; Horton & Gerrig, 2005), and of the integration
of basic contextual parameters of an interaction (Brennan et al., 2010). Still
others have identified kinds of coordination, such as the emergent versus
nonemergent linguistic interaction that see different origins in activities
done jointly (Knoblich, Butterfill, & Sebanz, 2011).
We see at least three exciting characteristics to this growing litera-
ture. First, researchers in these areas are beginning to tap into the cogni-
tive mechanisms that might underlie social and linguistic interaction (e.g.
Brown-Schmidt, 2009a,b; Gambi & Pickering, 2011; Horton, 2005; Mehler,
Weiß, Menke, & Lücking, 2010; Pickering & Garrod, 2009; Reitter, Keller, &
Moore, 2011). This advances the valuable work on observational and con-
versation analysis that has shed great light on the structure of interaction
(Sacks, Jefferson, & Schegloff, 1995; Schegloff, 2007), but is not capable of
identifying the cognitive processes that drive it.1
Secondly, and relatedly, social cognitive neuroscience (e.g. Frith &
Frith, 2001; Van Overwalle, 2008) and related areas (e.g. imitation: Wang &
Hamilton, 2012) have begun to explore these basic mechanisms at the level
of the brain. The growth of this subfield of cognitive neuroscience has been
very rapid, with many programmatic proposals for studying the circuits
underlying social interaction (e.g. Cooper, Catmur, & Heyes, 2012; Dumas,
Chavez, Nadel & Martinerie, 2012; Hasson, Ghazanfar, Galantucci, Garrod,
& Keysers, 2012; Konvalinka & Roepstorff 2012; Wolpert, Doya, & Kawato,
2003).
Thirdly—and this should sound odd—researchers have come to embrace
the inherent social nature of language and to carry out investigations of
cognitive processing in more naturalistic circumstances (see Tanenhaus &
Brown-Schmidt, 2008 and Fusaroli & Tylén, 2012 for a review). The past
century has seen some fundamentally different assumptions for a scien-
tific understanding of language. For example, the classic conception of the
ideal speaker–hearer, perhaps useful in some circumscribed domains, is an
1 Although the brilliant corpus strategies used by researchers like Bard, Aylett, and others can reveal
substantial clues about cognitive mechanism through, for example, acoustic properties of what one
person says to another (see, e.g. Bard & Aylett, 1999; Bard et al., 2000).
48 Rick Dale et al.
assumption that has outlived any usefulness it may have had in understand-
ing how people actually use language in so wide a circumstance. Con-
versation analysis and discourse psychology have now been coupled with
sophisticated computational and behavioral methods such as natural lan-
guage processing and computational linguistics (Graesser, Swamer, & Hu,
1997), eye-tracking (Tanenhaus, Spivey-Knowlton, Eberhard, & Sedivy,
1995), automated body movement (Paxton & Dale, in press; Schmidt, Morr,
Fitzpatrick, & Richardson, 2012) and acoustic analysis (Oller et al., 2010;
Wyatt, Bilmes, Choudhury, & Kitts, 2008), dynamical systems methods
(Riley & Van Orden, 2005; Shockley, Santana, & Fowler, 2003), and more.
Language is a complex and multidimensional activity, and our understand-
ing of it—how it evolved, is learned, and is used—must come from inte-
grating such sophisticated methods in naturalistic circumstances, not only
from abstract assumptions about linguistic structure that rarely manifest
themselves except in preempirical intuitions.
Although these are exciting developments, we would argue that theoreti-
cal integration has been less emphasized. We can think of a few reasons why
this might be. For one, a researcher’s theoretical proposals are usually tied to
the specific contexts she or he studies. This is a natural feature of any scien-
tific explanation (see Cartwright, 1999), but it limits the generalizability of
the processes proposed. The very fact, for example, that language users can
be rendered relatively egocentric, or relatively “other-centric,” by experi-
mental design means that something more complex is going on cognitively
than simply the deployment of fixed architectures (see also Brennan et al.,
2010 for discussion). Another reason, in our opinion, is that the multidi-
mensional and “multimechanism” aspect of human interaction means that
traditional conceptions of cognitive explanations are fundamentally chal-
lenged. In such a complex circumstance, theories seem unlikely to succeed
by anchoring to small set of specific mechanisms, but rather to a context-
dependent integration of a wide variety of processes acting together. This is
what we mean by the “centipede’s dilemma”: there tends to be much local
and circumscribed analysis and much less cross-paradigm and intertheoreti-
cal synthesis.
We introduce one way that approaches integration. Specifically, we
look to the tools offered by what is often termed the “dynamical systems”
approach in cognitive science (Chemero, 2009; Port & Van Gelder, 1995;
Richardson, Dale, & Marsh, in press; Spivey, 2007; Thelen & Smith, 1994;
Turvey, 1990; more on this below). A dynamical approach to these phenom-
ena affords a variety of theoretical tools that embrace context-dependent
integration, adaptation, and process flexibility. In the study of basic cognitive

processing, significant debate has emerged about the usefulness of a so-called
“nonlinear interaction-dominant dynamic complex systems” approach (see
collection in Van Orden & Stephen, 2012), and whether it really adds much
above already present accounts such as constraint–satisfaction mechanisms
(e.g. Eliasmith, 2012). These are all very important concerns, and we will
address some of them in discussion below. However, this debate seems to
have been unfortunately influenced by overly radical and unrealistic theo-
retical commitments, and perhaps reactionary tendencies in commentators.
In many circumstances, a dynamical systems approach can be integrated in
telling ways with existing theories. We argue that there is great value in the
approach, with specific benefits to be gained from applying it to conversa-
tion. We begin by describing the theoretical framework in a highly intro-
ductory manner for those who still haven’t read much about it or have been
too skeptical to get into it.
3. SELF-ORGANIZATION AND HUMAN INTERACTION

3.1. The Need to Integrate Accounts of Cognition in
Linguistic Interaction
The empirical literature on conversation and relevant interaction can
stymie many theoretical proposals. A cursory glance at this empirical
literature shows a more complex story than is typically portrayed in any
single theory. The cognitive processes proposed to be centrally involved
in social interaction are numerous. In addition, they operate in a highly
context-dependent way. Depending on the experimental paradigm cho-
sen by the researcher, one can highlight some capacities over others. As
noted above, particular laboratory interaction tasks may produce behav-
ioral patterns indicative of egocentrism (Barr & Keysar, 2002; Keysar,
Barr, Balin, & Brauner, 2000; Keysar, Lin, & Barr, 2003; cf. Shintel &
Keysar, 2009); at the same time, putting two people who are highly
acquainted in an interaction may have a similar effect, of highlighting
egocentrism, since each person can make assumptions that they are likely
to be understood (e.g. Wu & Keysar, 2007a; whether this is an explicit
metacognitive assumption is unknown). However, when establishing
pointed moments of conversational disruption, a conversation partner’s
needs or abilities, or different cultural contexts, these egocentric tenden-
cies can become reversed (see, e.g. Wu & Keysar, 2007b; Brown-Schmidt,
2009a,b; Brown-Schmidt, Gunlogson, & Tanenhaus, 2008; Galati &
50 Rick Dale et al.
Brennan, 2010; Roche et al., submitted for publication; Tanenhaus &

Brown-Schmidt, 2008; see for review Brennan et al., 2010; Schober &
Brennan, 2003).
These are exciting avenues of investigation, with extremely clever experi-
ments destined to fuel this debate and discussion. Nevertheless, an inference
to the best set of unique capacities is not possible from a small set of experi-
ments or even from a whole literature that highlights specific designs. The
range of possible contexts of human interactions is simply too numerous
to do so. For these reasons, it is unlikely to be the case that conversational
performance and linguistic interaction, in whole, can be accounted for in
terms of a small single subset of mechanisms. Of course, not all theories
aim to be so comprehensive as they tend to focus on specific aspects of
social interaction. We would argue that, to achieve a more comprehensive
account of social interaction, an integration of these task contexts, and cogni-
tive capacities, is needed (cf. Brennan et al., 2010). But how can we hook
up differing accounts into an overall theoretical framework that can achieve
this integration? Here, we argue that a dynamical systems framework may
serve these questions in valuable ways. In the following section, we describe
what we mean by “dynamical systems account” and describe two basic, but
important, features: self-organization and synergies.
3.2. Dynamics, Self-Organization, and All that Jazz

It is widely known that the dynamical systems approach to cognition utilizes
some terminology unfamiliar to many cognitive scientists.2 This concern
has been expressed in many critiques. For example, a recent commen-
tary’s tongue-in-cheek title uses the comprehensive phrase “[n]onlinearly
coupled, dynamical, self-organized critical, synergistic, scale-free, exqui-
sitely context-sensitive, interaction-dominant, multifractal, interdependent
brain-body-niche systems” (Wagenmakers, Van der Maas, & Farrell, 2012).
These are legitimate concerns because identifying important new theo-
retical concepts inside an array of unfamiliar terms requires at least some
concrete aspects of the agenda (in fact, this is the important point expressed
in the above-mentioned commentary). To be fair, however, we could say
the same thing about classical information-processing accounts as they
emerged. One could construct such a title for virtually any theoretical
account as terms that make subtle distinctions or highlight particular
nuances are common in all theoretical domains of cognitive science (rich
2 This is a deliberately mild way of putting it. Others, such as anonymous reviewers of some journal
articles, have described the vocabulary as “Star Wars terminology”. Readers may have other examples.
vocabularies appear to be a feature of any domain of human expertise;

Tanaka & Taylor, 1991). For example, the classical cognitive approach is
the “truth-value preserving, hierarchically organized, discretely symbolic,
recursive, satisficing, structure-dependent, information-encapsulated modu-
lar” approach. Naturally, as Wagenmakers et al. demand, concrete models
help anchor such terms, and it is true that the dynamics approach needs
more of them (see, e.g. Kello, 2013, for some recent exciting progress; we
also discuss this in concluding Section 6).
So before getting lost in a wave of fancy new terms, we expend some
energy in this review article discussing why they are used. To do so, we
describe a simplified and shortened version of a dynamic self-organized
approach to cognition. We do this to present only the most general ideas
and avoid some detailed debate that has emerged even in these areas. Read-
ers may be surprised to discover that even among this tribe of cognitive sci-
ence, there is significant dispute from teeth-gnashing displays that threaten
abstract theoretical constructs to pleas to remain open minded about such
constructs (for reviews, see Chemero, 2009; Dale, 2008). Still, the core ideas
can be laid out readily in a short section, as we attempt here. Our goal was
to showcase the specific aspects of this account of cognition that seem help-
ful to understand conversation (see Richardson et al., in press, for a thor-
ough presentation of both theory and methods).
3.2.1. Complex System

We can take “dynamics” for granted here. The dynamical systems approach
takes the position that it is important to study the time-evolving properties
of systems.3 A dynamical system is simply one that is changing in time and
can (in some way) be modeled as such (mathematically, computationally, or
just conceptually). Instead, let’s start with the notion of a “complex system”.
The phrase itself seems highly relevant to our language abilities. Carruthers
(2002) refers to language as an “intersection” system because, among other
functions (e.g. complex thought), effective language use requires a wide
variety of mechanisms to successfully intersect. Several mechanisms have
already been implicated in theories of perspective-taking during dialog and
other aspects of conversation. These have included social memory traces
(Horton, 2005), memory for shared experiences (Galati & Brennan, 2010;
3 We will also take “system” for granted. No further words are offered on this.You have to start from
somewhere. Readers wishing to have pure operationalized definitions of all things can consult the
success of Rudolf Carnap’s early-20th-century attempt to do so. One of the authors would wager that
readers recognize this attempt in proportion to the success of it.
52 Rick Dale et al.
Wu & Keysar, 2007), social status adaptation (Duran & Dale, 2011), execu-
tive control (Brown-Schmidt, 2009a,b), priming processes and alignment
(Garrod & Pickering, 2004), the mirror neuron system (Gallese, 2008), for-
ward models of social and linguistic prediction (Pickering & Garrod, in
press), rich common ground representations (Clark, 1996), socially guided
attention (Kingstone, Smilek, Ristic, Friesen, & Eastwood, 2003), percep-
tuomotor linkages (Shockley, Richardson, & Dale, 2009), and even processes
at various linguistic levels such as perception of accent (Lev-Ari & Keysar,
2010) and lexical (Bortfeld & Brennan, 1997; Brennan & Clark, 1996;
Niederhoffer & Pennebaker, 2002) and syntactic choice (Branigan,
Pickering, & Cleland, 2000; Branigan, Pickering, Stewart, & McLean, 2000).
Treated as a system of intersecting mechanisms, our language capac-
ity appears quite complicated. In the parlance of researchers who embrace
dynamics and complexity science, this complex system is unlikely to be con-
trolled by a central “homunculus”. No theory of our language capacity
has proposed such a central executive that simultaneously integrates all of
these mechanisms. For example, emerging models of sentence processing
imply that, even at just this processing level, central processing cannot alone
account for our success, and an explanation must derive from exploring
the dynamic relationship between memory retrieval, working memory, and
focal attention (e.g. Lewis, Vasishth, & Van Dyke, 2006; McElree, 2006;
Raczaszek-Leonardi, 2010). In the domain of motor control, where

dynamical systems have been and continue to be highly influential, this is
sometimes referred to as “Bernstein’s problem” or the “degrees of freedom
problem” (Turvey, 1990). If the components making up our language sys-
tem are truly modular, there are simply too many ways in which our overall
language system can change, with each mechanism flailing about unto itself
unless it is somehow anchored to other processes around it. Put simply,
there are too many degrees of freedom in this system for it to be managed
by a single control process. These many proposed mechanisms must some-
how influence each other, directly and continually, in order for language to
function in naturalistic circumstances. In any one experiment, we focus on
a very deliberately narrowed set of controlling variables and identify their
influence on a very specific set of resultant behaviors. Such is the justifiable
nature of experimental science.
Naturalistic language performance seems very unlikely to be based on
a single control process. Somehow, our system integrates all of these com-
ponents simultaneously. There are, at present, a limited number of theo-
ries for how this is accomplished (although see, for the closest current
approximation of a grand theory, Pickering & Garrod, 2004, 2009, in press).

But this is how we wish to pose the problem: If language, in its naturalistic
context, is underlain by such a wide array of processes, then these processes
must somehow interact, mutually constrain each other, and act together
continually to produce coherent performance. Systems that do this—that
have a multitude of parts that mutually interact and constrain each other—
are often referred to as “complex systems” (see Gallagher & Appenzeller,
1999; articles therein, for discussion). The term is only meant to highlight
the problem of interdependency that must be present among the system’s
components for it to function.4
3.2.2. Self-Organization
So if there is not a control process that “calculates the positions” of all
mechanisms (working memory, social judgment, visual attention, etc.), then
there must be some other means by which we can understand how they
function together. A process that contrasts with the presence of a central
controller is self-organization. Without a central control process, the mecha-
nisms must mutually constrain each other to behave (in whole) as a stable
performance. There are plenty of natural examples that are often raised to
exemplify this concept (see Kauffman, 1996, for many examples). For exam-
ple, the behavior of a beehive, termite, or ant’s nest is not controlled by a
single entity but is a large self-organizing organism unto itself (see Seeley,
2010, and Richardson et al., in press, for more discussion).The same may be
true for human interaction.
This is often where things get heated between dynamical systems
researchers and other cognitive scientists. Isn’t working memory that exec-
utive controller? Clearly it cannot be because there is much more being
coordinated during conversation than just a handful of manipulated chunks
of information (and see note about sentence processing above). What about
process threading in working memory extended over time? Recent com-
putational models of complex cognitive control may be relevant here, but
even these require articulating the details of interacting components in the
system (see, e.g. Salvucci & Taatgen, 2008). Similarly, what we are suggesting
is that there may be “chains of influence” between processes of the cogni-
tive system that we tend not to explore. Whatever one’s favorite array of
4 Complex systems are also figured to involve interactions among components, which produce
collective higher level behaviors not reducible to properties of the components themselves. We
wish to avoid this debate here (“emergence” or “emergentism”), although it seems likely a natural
consequence of the perspective we describe here (see also Knoblich et al., 2011).
54 Rick Dale et al.
theoretical constructs or model formalisms, these processes must be work-

ing simultaneously to bring about stable cognitive performance, which seems
especially true for face-to-face human interaction.
So how does self-organization work? Often, when one is trying to con-
vince skeptical colleagues of the value of these concepts, already they may
point to two issues. First, perhaps this is trivial:“Okay, so you’re saying to put
it all together, great. That’s obvious.” Or, the process is far too vague to be
even worthy of consideration: “Okay, so you’re saying they work together,
but that isn’t telling us anything new, because we don’t know what processes
it relies on!”These critiques are entirely legitimate, but the devil is always in
the details. The researchers in the dynamics crowd have identified elegant
ways of understanding what goes on during the process of self-organization
(even in high-level cognition; see, e.g. Dixon, Stephen, Boncoddo, & Anastas,
2010). Many of these dynamical concepts are descriptions of form rather
than function.They are characterizations of what is taking place in the system
as it self-organizes. Here, we consider an important one. When a system of
many interacting components self-organizes, it undergoes a reduction in its
degrees of freedom.
3.2.3. Synergies and the Reduction of Degrees of Freedom

As noted above, a key issue raised in motor control decades ago by Nikolai
Bernstein was that in order for a human being to perform any coherent
action, a massive array of variables must somehow coalesce in order for it to
happen. In the 1940s, Bernstein was trying to understand how motor con-
trol harnesses the high number and complexity of the components of the
human body (Bernstein, 1967; Kelso, 2009; Latash, Scholz, & Schoner, 2007;
Turvey, 1990). There is no way we can micromanage each and every joint
and muscle at the same time. He introduced the idea of synergy: a function-
ally driven reduction of degrees of freedom, where components do not sim-
ply align, but also complement and compensate for each other. Instead of a
top–down microcontrol, he hypothesized that the different components get
coupled and constrain each other locally.
A classic example is Bernstein’s analysis of chisel and hammer. If we
want to strike a chisel with a hammer, this gives direction to and con-
strains the workings of our body. The exact timing and force of contraction
and relaxation of all the individual muscles in our hands, fingers, and arm
are locally regulated to comply with that overall goal and the unfolding
interaction with the environment. This intuition was tested empirically by
measuring the precision of movements at all relevant joints in a blacksmith’s
hammering of a chisel. The variability of the trajectory of the tip of the

hammer across a series of strikes turned out to be smaller than the vari-
ability of the trajectories of the individual joints on the hammering arm
(Bernstein, 1967; Latash, 2008). The joints are not acting independently but
correcting each other’s errors at the relevant timescale, to preserve func-
tion, thus supporting the idea that the function itself is the coordinating
principle. Importantly, when putting the hammer down and, say, grasping a
cup of tea, the very same joints and muscles will flexibly combine in very
different ways, thus stressing the functional, that is, task-oriented nature of
the synergy.5
The same idea may be true of human interaction. The array of mecha-
nisms described above do not merely interact. They must have interdepen-
dencies operating in a coherent fashion that organizes the system into a
lower dimensional functional unit, and possibly a much smaller number
of stable higher level behaviors, unexpectedly lower than what would be
anticipated from the complexity of the system’s composition. For exam-
ple, perhaps at the coarsest level of description in human interaction, one
could see stable modes in the form of arguing (Paxton & Dale, submitted
for publication) or flirting (Grammer, Kruck, & Magnusson, 1998) or joint
decision making (Fusaroli & Tylén, submitted) or giving directions (Cassell
et al., 2007; cf. the notion of “oral genres,” e.g. Busch, 2007). Beneath these
coarse-level quasistable characterizations of interaction, we have differ-
ent levels of coordination taking place. Within one interlocutor, whatever
components compose a cognitive system must work together to support
coherent individual performance; across two individuals, a similar process of
systematic reduction of degrees of freedom may organize interactions into
stable modes of functioning (Shockley et al., 2009).
3.3. Summary, Social Modulation, and Multimodal

Coordination
So far, we have argued that the study of interaction has faced a kind of
“centipede’s dilemma” that the field has specialized in specific experimental
paradigms, specific behavioral channels and social contexts but that it has
not integrated knowledge in a systematic way. Y et, in an important sense, the
cognitive system undoubtedly performs this integration during interaction.
5 Note that this has sparked decades of exciting work and debate in motor control. Although
Bernstein’s solution has, in many respects, become standard in broad strokes, how it is solved can
be the subject of some debate (see, among many, Latash et al., 2007; Newell, Broderick, Deutsch, &
Slifkin, 2003; Todorov & Jordan, 2002; Turvey, 2007).
56 Rick Dale et al.
One framework for thinking about this, which we have outlined in brief
in the previous section, is to import the concepts of self-organization and
synergies into this discussion. Our reasoning is that there are far too many
degrees of freedom available to a dyad during conversational performance
for the cognitive system to compute their activities all at once. In the fol-
lowing sections, we offer extensive empirical review, looking to two general
features of this issue of conversational performance.
3.3.1. Social Modulation of Cognitive Dynamics

The first is the fundamental role of the social in human conversation. Despite
the intrinsic social nature of language and conversation, there has been much
debate on whether and to what extent social variables, such as the belief states
and presence of another person, modulate the dynamics of performance.
The self-organization of some cognitive or conversational performance is
shaped by social variables in various ways. For example, the mere presence
of a person will greatly change behavior of the visual attentional system (as
monitored by eye-tracking). In addition, facts that one knows about a poten-
tial social partner may guide perspective-taking strategies. We explore this
in review of empirical literature. The dynamical systems approach suggests
the following basic insight: significant reorganization of interactive behav-
iors should occur under different social contexts. This sees interaction as a
process that is organizing itself around key variables, such as a social ones; it
also means that basic theoretical accounts emphasizing of “egocentric” or
“other-centric” processes may be a dialectical veil over the deeper flexibility
and self-organization taking place during human interaction.
3.3.2. Coordination, Complementarity, Synergies

The second is, even if we acknowledge the importance of social variables
on the way that cognitive processing unfolds, there must nevertheless be a
process of coordination among the channels in conversation. For example,
if one learns something new about a conversation partner, it may suddenly
shift one’s focus both in the content of what is being said and how one says
it. Here is where the concept of synergies becomes most important: two
people interacting in a joint task come to form their behaviors through
compensatory complementary behaviors. These behaviors influence one
another locally and incrementally, making the whole conversational perfor-
mance itself a kind of self-organizing synergy. We review the research sug-
gesting that this is the case and offer theoretical discussion that is meant to
supplement, not replace, current discussion of coordination and alignment.
4. COGNITIVE DYNAMICS UNDER SOCIAL

CONSTRAINTS
4.1. Social Modulation of the Dynamics of Low-Level
Visual Attention
It can be very lonely, being a participant in a cognitive psychology experi-
ment. If the experimenter wants to study memory, language processing, or
decision making, a common first step is to exclude as many social factors
as possible. The participant sits alone, typically, interacting with a computer,
and perhaps a researcher, whom they don’t know, follows a rigid script.
Like friction in a high school physics problem, social context is discarded
by initial simplifying assumptions in cognitive models. Both forces get in
the way of more important aspects of phenomena, the argument goes. But
both friction and social context are unavoidable in the world outside of the
laboratory. In fact, we argue that they are both are essential to understanding
many phenomena.The claim is not that what is discovered in a typical cog-
nitive laboratory is invalid because of an absence of social context. Rather,
we argue that perhaps there are interesting dynamic interactions between
cognitive processes and social context that occur all the time in the real
world (Hutchins, 1995a,b, 2010, 2011), but are left at the door of many a
cognitive laboratory.
In this section, we review more naturalistic experimental designs that
have looked to how social variables constrain the dynamics of low-level
visual attention using eye movements. We survey a range of experiments
that have been titrated according to the level of social context that they
entail. Each measures visual attention, in the form of eye movements, to
see how these varying levels of social context influence cognitive and
perceptual processing. In the first, the level of social context was high as
two people interacted with each other while having a conversation or
an argument. The researchers measured how their thoughts about each
other drove their eye movements around a scene or an empty screen.
In the second, a participant alone was asked to listen to the opinions of
one person talking on a screen, but the researchers found that their eye
movements were influenced by the presence and the identity of others
in the background. In the lowest level of social context, the researchers
gave participants an explicitly nonsocial task of looking at a set of pictures
while sat alone. The researchers studied the effect of introducing a mini-
mal social context to the task by telling them that another unseen person
58 Rick Dale et al.
was also looking at the same images at the same time. Across all of these
levels, the researchers find a pervasive effect of social context on low-level
visual processes.
When two people are engaged in a conversation with each other, they
can display great sensitivity to each others’ thoughts and beliefs. But what
happens at a lower level of social context? Here, we look at the case where
the participant is merely a spectator, watching a prerecorded video of a
group of people and listening to one of them give their opinion. A standard
cognitive approach might be to focus on the words of the speaker and how
they are processed by the participant. But Crosby, Monin, and Richardson
(2008) looked at the relevance of the other people in the video, the silent
bystanders who provided a social context.
In their experiment, participants watched a video of four people giving
their views on Stanford University’s admissions policies. All four members
of the “focus group” could be seen on screen at the same time, in a grid
arrangement of cubicles, and they all wore headphones so they could hear
what each other said. At one point, the speaker in the top right corner com-
plained that, “certain groups who come from less privileged backgrounds…
get an unfair advantage.” At this point, participants routinely fixated a man
on the bottom row of the screen who was black. It appeared that partici-
pants were sensitive to the fact that the speakers’ words, criticizing policies
of affirmative action which would typically benefit black Americans, might
be offensive to members of that group.
A parsimonious explanation, however, is that the participants sim-
ply noticed the ethnicity of everyone at the start, and the speaker’s words
simply activated a memory of one person on the screen. This memory
trigger launches an eye movement, as memory representations often do
(Richardson, Dale, & Tomlinson, 2009; Richardson & Spivey, 2000). But
Crosby et al. (2008) were able to rule out this “association hypothesis”. In
another condition, before the speaker began talking, an offscreen voice said
that the headphones of people on the bottom row were being turned off.
Then the participants saw exactly the same video. The association hypoth-
esis predicted that since the black member of the focus group was still on
screen, he would still attract a fixation. In this case, however, participants
barely looked at him when the potentially offensive remarks were made.
Participants were supposed to be simply listening to the speaker’s words. But
these results show that they were also keeping track of the social identity of
all the other people on screen, monitoring whether or not they could hear
the speaker and, presumably, anticipating how each might respond to the
speaker’s words. In other words, for the participants, the cognitive task of
processing speech was embedded in a social context.
Other paradigms too have shown an effect of “social tuning” (Shteynberg,
2010; Shteynberg and Galinsky, 2011). In these paradigms, stimuli are
explicitly identified as being relevant by other people and as a consequence
are processed selectively by individuals. Recently, researchers have adapted
the Simon task and inhibition of return paradigms, splitting these cognitive
tasks between pairs of participants. The behavior of the pairs is remarkably
similar to the individuals’, showing the same patterns of response interfer-
ence (Knoblich et al., 2011). This work suggests that when participant act
jointly with each other, in a very simple social context, they immediately
represent each others’ tasks and goals.
The final set of experiments attempt to reduce social context to it low-
est level. Our strategy was to take a simple perceptual task that participants
carry out alone or jointly with another person and to make the difference
between those conditions as small as possible (Richardson et al., 2012). Pairs
of participants were sat in opposite corners of a laboratory room, each look-
ing up at a screen while their gaze was tracked (see Figure 2.1). On each
trial of the experiment, they saw four images on screen for 8 s. Beforehand,
Figure 2.1 Example context in which two people were given different beliefs about
what their “socially copresent” partner could see. Imagines of differing emotional
valence were presented. (Adapted with permission from Richardson et al. (2012)).
60 Rick Dale et al.
they were either told that both they and their partner would be looking at
the same images or that they would be looking at images and their partner
would be looking at symbols. Participants could not see each other and
could not interact at all. Nevertheless, when they thought that their current
perceptual experience was being shared with another, their eye movements
were systematically changed.They looked more toward pictures with a neg-
ative valance than when looking alone. We believe that people are doing
so because they each believe that the other person is looking more at the
negative images. At least, when participants are told that this is a memory
task and that they will score more points if they recall the same pictures as
their partners, they too look more at negative images (unpublished data).
When the images are replaced by album covers, people will look more at
classical albums when they are looking jointly, and their partner (a confed-
erate in this case) walked in carrying a violin (unpublished data). Across all
these experiments, we have found that even a minimal social context—the
belief that an unseen other was seeing the same stimuli—was enough to
manipulate an individuals’ visual processing.
From a rich interactive conversation to listening to one person against
a backdrop of bystanders, to gazing at images alone, believing that someone
else is too, social context can have a pervasive influence on visual attention
(Risko & Kingstone, 2011; Laidlaw et al., 2011). Under even subtle circum-
stances, people take into account each other’s knowledge and visual context
to coordinate their gaze around an empty display.They anticipate each others’
responses to potentially offensive remarks. And even with the thinnest slice of
social context, when there is no interaction or contact between people, they
will still shift their gaze toward where they think each other is looking. Impor-
tantly, this process occurs when the variables are in the right arrangement—for
example, when the unintended recipient of the offensive remark can hear it.
4.2. Social Modulation of Higher Level Processes, Like

Perspective-Taking
Social variables can radically alter the dynamics of visual attention. Here,
we review recent research suggesting that these same dynamic changes
take place in relatively higher level cognitive processing: perspective-
taking. In pragmatic models of language processing, an essential compo-
nent of how people produce and comprehend language depends greatly
on communicative function. The social environment thus has a central
role in constraining how language is interpreted and used (Brennan
et al., 2010; Brennan & Hanna, 2009; Brown-Schmidt & Hanna, 2011;
Brown-Schmidt & Tanenhaus, 2008; De Jaegher, Di Paolo, & Gallagher,

2010; Hanna & Tanenhaus, 2004; Hanna, Tanenhaus, & Trueswell, 2003).
A critical source of constraint is in the common ground that may exist
between language users. Common ground corresponds to the shared
characteristics derived from the local context of being present with
another, such as viewing the same scene, to more global shared histories
that arise from being members of the same culture or speaking the same
language. This information is brought to bear when interpreting what
another says and when choosing words to speak (Clark & Krych, 2004;
Clark & Wilkes-Gibbs, 1986; Fussell & Krauss, 1992; Lockridge & Brennan,
2002; Schober & Brennan, 2003).
A central question is when common ground information is available in
language processing. As described earlier in this review, at one theoreti-
cal extreme is a view that people are primarily egocentric and that even
when common ground information is available for influencing a particular
interpretation, people initially rely on their own frame of reference or act
to minimize their own difficulty in processing (Epley, Keysar, Van Boven,
& Gilovich, 2004; Keysar, Barr, & Horton, 1998). Otherwise, as the argu-
ment goes, to integrate common ground early in processing would result
in increased cognitive effort and processing times (Barr, 2008; Keysar et al.,
2000). But such conclusions stand in contrast to other studies that show
people are quite capable of making rapid social judgments based on briefly
presented sources of social information, such as dispositional expressions
(Ambady, Bernieri, & Richeson, 2000) or gaze direction (Hanna &
Brennan, 2007; Teufel, Fletcher, & Davis, 2010). This social information can
also extend to belief attributions about another, such as another’s needs,
characteristics, or limitations that are initially present in an interaction
(Bortfeld & Brennan, 1997), or are emergent factors (Horton & Gerrig,
2002, 2005). Integrating common ground information does not necessarily
have to be a cognitively complex process, as simple attributes of another
can immediately constrain what and how something is interpreted. More-
over, such integration is commensurate with “incremental models” of lan-
guage processing. As words are encountered in a sentence, new evidence
is provided for the commitment, or abandonment, of a particular interpre-
tation. As a sentence unfolds, multiple interpretations are simultaneously
activated and competing for expression, with accruing evidence constrain-
ing possible interpretations (Seidenberg & MacDonald, 1999). Based on
this account, common ground information is tantamount to just another
source of potential constraint.
62 Rick Dale et al.
From a “traditional” dynamical systems perspective, what constitutes a

relevant constraint is similar, but is more connected to what can be directly
perceived from within an interactive social environment (Marsh, R ichardson,
Baron, & Schmidt, 2006; Marsh, Richardson, & Schmidt, 2009; Richardson,
Marsh, & Schmidt, 2005). Contrary to egocentric accounts of processing,
there is no intermediary “representational” stage where another’s intentions
are first calculated and then acted upon. Rather, the behavior and actions
of another hold immediate sway on how people respond to each other.
Such direct couplings are possible through the interactive context by which
individuals’ processing capabilities are reshaped by the presence and actions
of social partners (Ramenzoni, Davis, Riley, Shockley, & Baker, 2011; Riley,
Richardson, Shockley, & Ramenzoni, 2011). Such connectivity allows for
nimble social coordination that cannot be reduced to individual-level con-
tributions, but instead must be evaluated on the basis of the social unit,
where the emergence of meaning is reciprocally caused and maintained by
social partners during interaction (Marsh et al., 2009).
On the face of it, this traditional dynamical systems approach should
be closely aligned with language theories that allow for common ground
constraints to have immediate influence on moment-by-moment linguis-
tic processing. Yet, what constitutes common ground often corresponds to
simple inferential states based on what another knows or believes. Such
inferences are not easily integrated with a dynamical systems approach,
where interpersonal coordination is driven by the actions, or possibilities
for action, that are expressed and integrated by physically co-situated agents.
Rather, what is found in the social environment are opportunities for merg-
ing individual-level perceptuomotor systems into a collective system, with
evidence taken from joint action tasks where people rapidly converge on
patterns of coordinated synchronous movements (Knoblich et al., 2011;
Knoblich & Sebanz, 2006). Put simply, common ground invokes the cogni-
tive, and many dynamical systems theorists avoid this.
So if social interaction is grounded exclusively in coupled percep-
tuomotor systems, there appears to be little room for the role of infor-
mationally grounded sources in shaping language processes. To push the
boundaries of the dynamical approach, explanations need to go beyond
motor behaviors alone to more abstract properties of the social environment
(Chemero, 2009). Such attributional properties do not necessarily require
elaborate mental operations or representation, but instead can be thought
of as spontaneously elicited opportunities for social responding, embodied
from past histories of social interaction. As Schmidt (2007) describes, these
experiences elicit tendencies to respond in socially appropriate ways and are

sustained by cultural expectations and reinforced by the immediate social
context. This behavior is inextricably defined by the relationship between
social partners and “affords” opportunities for responding, even during lan-
guage comprehension.
One of the simplest belief attributions in a language task is whether a
communicative partner is an intentional agent (Gallagher, Jack, Roepstorff, &
Frith, 2002). Although most interactions provide situated cues to deter-
mine veridicality, there are scenarios, such as with computer-mediated
interactions, where people may be uncertain whether they are interact-
ing with someone real or simulated. When people are told beforehand the
true nature of personhood, their response orientations change in systematic
ways, even though the actions of the other remain the same (Nass, Fogg, &
Moon, 1996). For example, when a partner is thought to be simulated,
people are more likely to use language that is less complex, presumably
because of perceived communicative limitations of an artificial intelligence
(Branigan, Pickering, Pearson, McLean, & Brown, 2011).Thus, simple belief
attributions have the capacity to guide various modes of responding (Wilkes-
Gibbs & Clark, 1992). A challenge for dynamical systems is to explain how
such behavior arises through an emergent self-organized process in which
informational couplings within a social environment produce complex but
systematic behaviors (also see Di Paolo & De Jaegher, 2012; Richardson,
Marsh, & Schmidt, 2010).
4.3. Perspective-Taking as Self-Organization under Social

Constraint
Duran and Dale (in press) present one such attempt by employing a
dynamical simulation of response resolution in a task where participants’
beliefs were central in disambiguating utterances spoken by another. In this
response data, participants and a simulated agent were “connected” within a
virtual environment through an elaborate ruse in which the participant was
unaware of whether their partner was actually simulated (Duran, Dale, &
Kreuz, 2011).This omission allowed participants to form their own impres-
sions about the reality of their partner. For each trial, the task proceeded
with the simulated agent instructing the participants to select one of the
two objects in the shared environment (see Figure 2.2). The position of the
partner shifted from trial to trial, sometimes creating a situation in which
the intended referent was ambiguous. When both participants were in the
same location, perspectives were ostensibly shared, and an instruction such
64 Rick Dale et al.
2000
Them
Reaction time (ms)

1850
1700
1550
1400
0 90 180
You
Relative rotation (ego vs. other)
Figure 2.2 When participants are asked to retrieve a folder in an “ostensibly social” com-
puter task as shown on the left, they tend to exhibit different mental rotation functions,
predicted by whether they are taking theirs or another person’s perspective. When the
ostensible partner asks for a folder in ambiguous trials (left), participants taking the
other perspective exhibit a mental rotation function (right). Figure based on the designs
and typical mental rotation results found in Duran & Dale (in press), Duran et al. (2011),
and originally of course the classic Schober (1993).
as, “Grab the object on the right,” permitted straightforward identification.

However, when participants’ positions were across from each other, the same
instruction created an ambiguous referent as to whose “right” was the basis
for interpretation.That is, instruction receivers could either select the object
on their partner’s right, thereby taking into consideration the perspective of
the other, or they could select the folder on their own right, in what would
be an egocentric interpretation.
Despite an informationally situated social context, simple belief attri-
butions, in the form of whether a communicative partner was thought to
be real or simulated, the researchers were able to shape perspective-taking
strategies across participants. Specifically, when people believed that their
partner was simulated, other-centric responding was facilitated in three
key ways: (1) the likelihood for other-centrism increased, (2) responses
were faster within and across trials, and (3) decreased competition from
the egocentric response option. Such behavioral change was also seen in
a dynamical systems simulation that treated attributional factors as a con-
trol parameter within a low-dimensional attractor landscape of partly stable
perspective-taking modes. In this simulation, when a person takes the per-
spective of another, this response stabilizes through the graded accumula-
tion and competition of both other- and egocentric factors. Thus, contrary
to existing models of language processing where these factors are mostly
Figure 2.3 Perspective-taking can be modeled as a low-dimensional dynamic process.

Simulating this landscape allows qualitative fits to three timescales of human data: (i)
decisions, (ii) response times, and (iii) response dynamics (Duran & Dale, in press). See
Figure 2.4 for illustration. A full discussion of this lower order characterization of higher
dimensional dynamics can be found in Onnis and Spivey (2012).
independent components, here, they are explicitly allowed to interact from

the very start of processing to influence behavior over time (e.g. multipo-
tentiality).
To conduct the simulation, Duran and Dale (in press) borrowed from
the Haken–Kelso–Bunz model that was originally developed to capture
the relative coordination of bimanual motor movements in time and space
(Kelso, 1981, 1995; see Figure 2.3). This model has been extended to a
variety of domains, showing widespread commonalities between percep-
tual, cognitive, and motor systems (e.g. Engstrom, Kelso, & Holroyd, 1996;
Frank, Richardson, Lopresti-Goodman, & Turvey, 2009; Tuller, Case, Ding, &
Kelso, 1994; Van Rooij, Bongers, & Haselager, 2002; see Schmidt & Turvey,
1995; Chemero, 2009, for reviews). It draws from core principles of bistable
dynamics to allow complex behaviors to self-organize over time, with
responses unfolding within a low-dimensional attractor landscape. Thus,
perspective-taking during communication, much like in the previous
research, could be described as following coordinative dynamics similar to
those observed in perceptuomotor coupling.
There is a precedent for considering high-dimensional processes in
low-dimensional forms. Recently, Onnis and Spivey (2012) have advocated
for linking such means of modeling and visualizing systems. Assuming, for
example, that perspective-taking can be represented as a high-dimensional
66 Rick Dale et al.
neural process, akin to a population code (top row of Figure 2.3), one can
derive a direct visualization in lower dimensions of how the system is
transitioning between stable states (bottom row of Figure 2.3). Self-
organization into “other” or “ego” perspectives can be seen as traversing a
low-dimensional landscape.
In the current instantiation of the model, a control parameter in a poten-
tial function is set to initiate bistable attractor basins of other-centric or
egocentric interpretations. These basins are a reduction of system complex-
ity to a quantifiable and transparent outcome variable of the two perspective
types. In other words, perspective-taking is characterized as in a system of
substantially reduced degrees of freedom, a “lower dimensional” cognitive
space in which choices are made. The particular shape of each basin cor-
responds to the likelihood and speed in which the system can settle into a
particular response, with deeper and steeper basins indicating a stronger pull
and therefore more rapid stabilization. During the time course of a single
trial, “settling” occurs through nonlinear competition between landscape
shape, initial conditions (i.e. starting position in the landscape), and a subtle
noise impulse. When a response threshold is met, the control parameter is
adjusted, and a new trial is allowed to run. This is analogous to updating a
belief about one’s partner being real or simulated at the end of each trial,
with beliefs becoming stronger and more stable over time. In doing so, the
global characteristics of response choice stability are captured, as well as
the competition effects that influence the moment-by-moment processes
involved in response execution. By capturing the response dynamics also
exhibited by human participants, simple social constraints, in the form of
belief attributions, are essential pieces of information that bias a system’s
“perspectival” landscape and thus its eventual behavioral strategy (see exam-
ples in Figure 2.3).
Figure 2.4 shows the timing and response characteristics of data seen in
Duran and Dale (in press) and Duran et al. (2011). In the top left, a represen-
tative response histogram of this task is shown. The model can capture two
stable strategies, namely as attractors in the lower dimensional landscape.
Models of this kind naturally display organization in the form of two stable
“modes” or strategies. The bottom left panel shows that, over trials, humans
display a drop in reaction time across trials, but more so for ego versus
other reaction times. Finally, the panel to the right shows a zooming into
the very dynamics of responses. Subjects often show faster more direct ego
trajectories; if they are responding using the other-centric perspective, their
mouse movement trajectories tend to show more curvature.These are three
(a)
Across subjects
(c)
# subjects
Target Comp.
Other % Spatial strategy Ego
(b)
Other trajectory
Ego trajectory
Other subjects
RT
Ego subjects
Across trials Within trials

Figure 2.4 A representation of the three timescales of a task that the lower dimensional
model of Duran & Dale (in press) can capture.
interconnected timescales from the response distribution of participants

down to response times to the fine-grained dynamics of responses. Duran
and Dale (in press) show that setting one set of parameters can simultane-
ously model all three such timescales.
Perspective-taking in communication is fundamental to how language is
used and understood. This was certainly evident in the Duran et al. (2011)
study. Somewhat counterintuitively, participants were more likely to con-
sider the other’s perspective when they believed the other to be simulated.
This result makes sense when communication is viewed as a collaborative
process between language users.The goal of communication is to maximize
mutual understanding, and when one partner is unable or hard-pressed to
do so, the other will compensate by putting in increased effort, such as
engaging in other-centric perspective-taking behavior (Clark, 1996; Clark &
Krych, 2004; Clark & Wilkes-Gibbs, 1986; Goodwin, 2003). This ten-
dency emerges, although past histories of social interaction where peo-
ple actively attempt to establish mutual understanding. Attributions about
others’ abilities to cooperate are eventually embodied by language users and
brought to bear in responding, even in simple communicative scenarios
such as the one described above (also see Schober, 1993, 1995). When par-
ticipants think they are interacting with a simulation, where cooperation
68 Rick Dale et al.
by the other is not even possible, the “afforded” response is to assume the
other’s perspective in interpreting their ambiguous instructions.
Such spontaneous perspective-taking occurs despite enacting increased
cognitive demands. However, demand is minimized in communication
where assessments about a partner can be reduced to simple alternatives,
such as whether a conversational partner is very young or not or has a lan-
guage disorder that changes the goals of mutual understanding (Newman-
Norlund et al., 2009; Perkins & Milroy, 1997). Of course, reciprocal and
emergent constraints occur during the course of an interaction that sub-
sumes any individual-level sources of difficulty. The dyad operates as a unit
that collaboratively minimizes processing load, with success depending on
the level of coordination shared between the language users (Fusaroli et al.,
2012; Louwerse et al., 2012).
5. COORDINATION, COMPLEMENTARITY, AND

INTERACTIVE PERFORMANCE
The foregoing review suggests that the dynamics of both low- and
high-level cognition can be seen as responding to subtle social variables
present in the environment. Subtle changes to these variables can lead to
rapid changes in the organization of those processes (e.g. visual attention
radically changing when you learn something new about your partner).
Such rapid changes are hallmarks of self-organizing systems—the capacity
for rapid nonlinear change.
However, even below these subtle social variables, there are consider-
ably more dimensions that must be managed by a dyad, namely the array of
behavioral and cognitive possibilities during an interaction. As noted in the
introduction, language is increasingly acknowledged as a social coordination
device, a way of accomplishing otherwise difficult or impossible coordina-
tion of actions and cognitive processes (Clark, 1996; Fowler, Richardson,
Marsh, & Shockley, 2008; Fusaroli, Gangopadhyay, & Tylén, submitted for
publication; Fusaroli & Tylén, 2012; Galantucci, 2009; Galantucci & Sebanz,
2009; Hasson et al., 2012; Hutchins & Johnson, 2009; Louwerse, Dale, Bard, &
Jeuniaux, 2012; Pickering & Garrod, 2004; Tylén, Fusaroli, Bundgaard, &
Østergaard, in press; Tylén, Weed, Wallentin, Roepstorff, & Frith, 2010).
Through language, we can easily entertain a friend while waiting for her
bus to arrive, exchange words with a stranger in an elevator, coordinate
in carrying a heavy piano down a flight of stairs, negotiate the price of an
apartment, share information, and make joint decisions. However, as earlier
described, coordinating via language is a complex business. A great number

of studies have been dedicated to unveiling the crucial subtleties in coor-
dinating not only topics, lexical choices, and syntax but also gestures, gaze
management, head movements, and postural sways, which underlie conver-
sations (Goodwin, 2000, 2011; Louwerse et al., 2012). Such a multimodal
richness seems to imply serious uncertainty ( Jaeger, 2010) and cognitive
load (Garrod & Pickering, 2009) for the participants: how does a conversant
choose between all the possible linguistic and nonlinguistic behaviors on
so many different channels at once? How does a person focus his or her
attention and interact in a meaningful way? In other words, how can inter-
locutors seemingly effortlessly orchestrate all these dimensions (each level,
presumably, with its own numerous degrees of freedom) in tight intra- and
interpersonal coordination?
As a reaction to computationally heavy models of conversations, requir-
ing theory of mind and full accommodation of models of the other, there
has been a strong focus on low-demanding bottom-up models of lin-
guistic coordination, such as the model of interactive linguistic alignment
(Pickering & Garrod, 2004). After a brief presentation of this model, we will
argue that it should be integrated and complemented in the larger model
of interpersonal synergies, presenting evidence supporting this and a few
studies testing the prediction of the synergy model.
5.1. Behavioral Synchrony and Interactive Alignment

One intuitive way of reducing the complexity in interpersonal interactions
is to diminish the range of possible behaviors via a progressive adaptation
to each other. By becoming increasingly similar, the interlocutors greatly
simplify the cognitive load needed to interact with the other. Indeed,
there is strong evidence for behavioral mimicry (Chartrand & Van Baaren,
2009) and interactive linguistic alignment (Pickering & Ferreira, 2008;
Pickering & Garrod, 2004). Interacting human beings have been observed
to mimic each other’s posture, gestures, and other behaviors (Chartrand &
Van Baaren, 2009). A prototypical example of an experimental investiga-
tion of this kind of human unconscious mimicry is the “Chameleon effect”
(Chartrand & Bargh, 1999; Dijksterhuis & Bargh, 2001). In this experiment,
participants interacted with an unknown confederate in two consecutive
picture-describing sessions. In one session, the confederate either rubbed
her face or shook her foot while describing the pictures with the partici-
pants, while the second confederate performed the behavior that the first
confederate did not.The behavior of the participants, “secretly” recorded on
70 Rick Dale et al.
videotape, showed that participants shook their foot more in the presence
of the foot-shaking confederate and rubbed their faces more in the pres-
ence of the face-rubbing confederate. Debriefing indicated that participants
were unaware of their mimicry. Analogously, facial expressions, gestures, and
yawns have been observed to spread across interlocutors and around a room
(Louwerse et al., 2012; Platek, 2010).
Pickering and Garrod have argued that mimicry is commonly co-
opted in linguistic interactions through what is called “interactive linguistic
alignment”: interlocutors tend to imitate each other’s choice of linguis-
tic forms. Participants primed with a specific syntactic structure are more
likely to produce new sentences employing the same syntactic structure
under circumstances in which alternative nonsyntactic explanations could
be excluded (Bock, 1986; Branigan et al., 2000; Gries, 2005; Hartsuiker &
Westenberg, 2000; Levelt & Kelter, 1982; Pickering & Branigan, 1999;
Smith & Wheeldon, 2001; Szmrecsanyi, 2005, 2006). Analogously, topics
(Angus, Smith, & Wiles, 2012; Angus,Watson, Smith, Gallois, & Wiles, 2012)
and lexical choices (Brennan & Clark, 1996; Clark & Wilkes-Gibbs, 1986;
Garrod & Anderson, 1987; Garrod & Clark, 1993; Garrod & Doherty,
1994; Orsucci, Giuliani, & Webber, 2006; Orsucci, Giuliani, & Zbilut, 2004;
Orsucci, Walter, Giuliani, Webber, & Zbilut, 1997; Wilkes-Gibbs & Clark,
1992) tend to be imitated across interlocutors. Linguistic alignment can also
be found at more subtle levels of linguistic coordination: interlocutors align
accent and speech rate (Giles, Coupland, & Coupland, 1991). More recently,
a lot of effort has also been put in showing that the organization of pauses
in and between interlocutors’ speech and their average pitch, intensity, and
voice quality tend to become similar over time (De Looze & Rauzy, 2012;
Kousidis & Dorran, 2009; Lee et al., 2010; Lelong & Bailly, 2011; Levitan &
Hirschberg, 2011; Nishimura, Kitaoka, & Nakagawa, 2008; Pardo, Gibbons,
Suppes, & Krauss, 2011;Truong & Heylen, 2012; Vaughan, 2011). In a single
conversation, many of these channels will be aligned, as recently shown in
a massive study by Louwerse et al. (2012). These channels have been argued
not to be independent. On the contrary, aligning on one channel in many
cases seems to facilitate alignment on others. For instance, syntactic priming
is enhanced when the same lexical items or even just semantically related
ones are also repeated (Branigan, Pickering, & Cleland, 2000; Branigan,
Pickering, Stewart, et al., 2000; Cleland & Pickering, 2003).
Several mechanisms have been proposed to underlie these phenom-
ena: most researchers seem to agree on an unconscious priming mecha-
nisms, a “perception–action link” (Chartrand & Bargh, 1999; Dijksterhuis &
Bargh, 2001) or “structural priming” (Pickering & Ferreira, 2008), in other

words the overlapping between mechanisms involved in perceiving a
behavior and producing it, which implies that by perceiving a behavior the
participant preactivates the production of the same. Other authors prefer to
focus on the conscious aspects of alignment, where interlocutors try to take
each other into account, developing conceptual pacts on which words to
use (Brennan et al., 2010; Clark & Brennan, 1991).
Whatever the mechanisms at work, alignment within and across modali-
ties is a very effective way to reduce degrees of freedom—namely the pos-
sible behaviors from which to choose. Not only the other’s behavior can be
used as guide in how to behave but also the repertoire of possible behaviors
is reduced over time. The mechanism of alignment might be differently
motivated, but it is sensible to argue that once it is established, it plays an
important role, making linguistic interactions more manageable. However,
a few problems arise when we take it at face value as the fundamental
motor of linguistic coordination. A few studies are pointing out that not
all conversations contain the same amount of linguistic alignment (Healey,
Howes, & Purver, 2010; Reitter, Moore, & Keller, 2006) and that coordina-
tion might not rely on alignment across neighboring speech turns, but on
the contrary across many speech turns, thus escaping the tight temporal
constraints of automatic priming (Reitter & Moore, 2007). At a more intui-
tive level, a conversation constituted exclusively of reciprocal repetitions
should not strike anybody as a very productive one.
5.2. An Alternative Model: Interpersonal Synergies

As described earlier in this review, the reduction of degrees of freedom
is not a new problem. Bernstein (1967) proposed that functional units of
motor control are established through mutual constraint among the parts
of the body and motor control system, effectively reducing the degrees
of freedom of the system into “synergies”. Recently, Ramenzoni and col-
leagues (Ramenzoni et al., 2011; Ramenzoni, Riley, Shockley, & Baker,
2012; Riley et al., 2011) have been exploring interpersonal motor synergies.
They showed that in joint actions participants increasingly coordinate hands,
forearms, and torsos, forming reciprocally compensating synergies spanning
across individuals. While studies of interpersonal motor coordination are
not rare (Marsh et al., 2009; Richardson, Marsh, Isenhower, Goodman, &
Schmidt, 2007; Schmidt, Bienvenu, Fitzpatrick, & Amazeen, 1998; Schmidt
& Richardson, 2008), only very recently has this approach been applied
to linguistic coordination (Fusaroli, Raczaszek-Leonardi & Tylén, in press).
72 Rick Dale et al.
The rest of this section will argue that linguistic coordination is achieved
through interpersonal synergies, that is, through functionally driven reduc-
tion of the degrees of freedom involved in the interaction. This approach
does not dispense with alignment but introduces additional mechanisms—
complementarity and interactional patterns—and integrates alignment with
this dynamical inspired perspective.
5.2.1. Complementarity
Several studies have pointed out that interlocutors strive to complement
each other’s behavior to develop a structured conversation. For example,
turn-taking seems one of the most elementary examples of complementar-
ity: a remarkable—and seemingly universal (Sidnell & Enfield, 2012; Stivers
et al., 2009)—ability of humans to not do the same thing at the same time,
that is, stay quiet when the other speaks. Simultaneous starts are reported to
be surprisingly rare in dyadic conversations ( Jefferson, 1988), even if more
than 50% of the pauses between interlocutors are below the usual threshold
for reactions (300 ms). Wilson and Wilson (2005) have been developing a
model of turn-taking that explains this fine-tuned complementarity: the
beginning of an interaction sets up an oscillator in each of the interlocu-
tors’ cognitive systems establishing a shared frequency of speech rate (see
also Buder & Eriksson, 1999). This cyclic pattern governs the potential for
initiating speech at any given instant for both interlocutors. The interlocu-
tors, in other words, have to keep the same pace (alignment). However, if
the oscillators were simply entrained in phase, simultaneous starts would be
frequent.Therefore, the oscillators must be entrained in antiphase, giving the
participants both a common rhythm, constituted by speech rate and length
of comfortable pauses, and complementarity—readiness to take the floor
must be opposite at any given moment for speaker and hearer. This ability
seems to appear at a very early developmental stage (Gratier & Devouche,
2011; Murray & Trevarthen, 1985; Nadel, Carchon, Kervella, Marcelli, &
Réserbat-Plantey, 1999; Spurrett & Cowley, 2004; Warlaumont, 2012).
Recent work on conversations involving patients with speech impairment
further shows the importance of complementarity. Expert interlocutors—for
example, family members—tend to engage compensatory procedures to keep
the conversation fluent despite the impairment (Dressler, Buder, & Cannito,
2009; Goodwin, 2003, 2011; Wilkinson, Beeke, & Maxim, 2003). For exam-
ple, Goodwin reports on Chil, who, after having suffered a severe stroke, can
only speak three words: “yes,” “no,” and “and”. Despite this clear impairment,
Chil is able to engage in complex conversations by coordinating other people’s
utterances. Chil thus relies on different types of reciprocal compensatory moves

to restore the dialog: on the one hand, interlocutors have to actively produce
utterances completing and supporting Chil’s conversational moves. On the
other hand, Chil’s three words are relational ones: they do not communicate
much on their own, but make sense only in a conversational situation.Together
with a host of nonverbal means such as facial expressions and gesture, Chil
employs his minimal vocabulary to couple with the other interlocutors’ com-
municative activity. Relying on three words, he is able to coordinate, support,
supplement, and sometimes reject his interlocutors’ utterances (Goodwin, 2011).
Similarly, Dressler and colleagues (Dressler et al., 2009) have explored prosodic
patterns in conversations with aphasic patients. They report that conversation
with familiar interlocutors displays overall prosodic rhythms, which are much
more fluent and regular than conversations with unfamiliar interlocutors.
5.2.2. Interactional Patterns

Beyond this basic rhythm of interaction, conversation analysis has persua-
sively shown how speech turns are often organized in functionally struc-
tured sequences of turns, such as adjacency pairs: questions are ordinarily
responded to with an answer, not with another question; offers and invi-
tations are ordinarily followed by acceptances or declinations, and so on
(Schegloff, 1986).Turns and adjacency pairs are themselves not free-floating
entities, but often fulfill a role in larger interactional patterns, locally unfold-
ing routines that scaffold and constrain the possibilities of actions and inter-
pretation in joint activities (Clark, 1996; Levinson, 1983). Interactional
patterns are typically conceived of as normative static phenomena already
shared—or assumed to be shared—by interlocutors (Sacks, Schegloff, &
Jefferson, 1974; Schank & Abelson, 1977). The synergy approach, however,
implies that these elements are part of a dynamic context-sensitive interac-
tion. Interactional patterns vary in formality and flexibility from free and
relatively unconstrained conversation over the morning coffee to tightly
structured and sometimes even explicitly codified task-oriented conversa-
tions (Hutchins, 1995a, 1995b; Perry, 2010). Interactional patterns work to
reduce the overall degrees of freedom of the system in a functionally driven
way and enable a smoother flow of the interaction.
A number of recent studies indirectly show that ad hoc interactional pat-
terns emerge and are maintained in task-related interactions. In a version of
“the maze game” (Healey & Mills, 2006; Mills & Gregoromichelaki, 2010),
it was observed that, over the course of 12 games, participants radically
structured and shortened their linguistic exchanges from more than 150
74 Rick Dale et al.
turns to brief and efficient exchanges. Through a shared history of interac-

tion, the structure of their interaction is stabilized. This enabled participants
to smoothly produce and interpret highly elliptical and fragmentary utter-
ances without much negotiation or clarification. Extending this work, Mills
(2011) systematically investigated how these interaction patterns emerge
and spread in a small speech community. Each participant played a number
of games with shifting partners within a “community”. Then, in a critical
test trial, half of the participants were paired with a member from another
community. This perturbation seriously disrupted the interaction in the
affected groups. Participants were found to edit their utterances to a much
higher degree, were observed to explicitly acknowledge each other’s utter-
ances more often, and overall performed less accurately.The findings suggest
that interactional patterns emerge from a shared history of interaction and
come to implicitly constrain the degrees of freedom of the interlocutors,
diminishing ambiguity and supporting a smoother and more effective flow
of the coordination (for a more comprehensive discussion of these issues, cf.
Mills, in press, and Fusaroli, Raczaszek-Leonardi, & Tylén, in press).
5.3. Testing Models of Linguistic Coordination: Alignment

and Synergy
The review so far suggests that complementarity, in the form of system-
atized patterns of interaction between two people, is a crucial component
of human interaction. Three recent studies based on the same experimental
design (cf. Figure 2.5) have tried to test implications of the model (Bahrami
et al., 2010; Fusaroli et al., 2012; Fusaroli & Tylén, submitted; Fusaroli, Abney,
Bahrami, Kello, & Tylén, submitted). In the experiment, pairs of participants
were instructed to individually indicate in which of two brief visual displays
they had just been shown a contrast oddball. If their individual decisions
diverged, they were prompted to discuss and reach a joint decision. In order
for a pair to achieve a cooperative benefit, that is, to perform better than
the best of the individuals, they had to find ways of assessing and comparing
their individual levels of confidence so as to choose, on a trial-by-trial basis,
the decision of the more confident participant. In other words, they had to
develop an interactional pattern for accurately expressing confidence and
smoothly taking joint decisions relying on that.
This paradigm generated a corpus of task-oriented conversations—
which emphasizes the development over time of interactional patterns to
quickly solve the repeated tasks—as well as an accurate measure of coopera-
tive performance—to assess the efficacy of linguistic coordination. Different
Figure 2.5 Interactive perceptual detection task. (a) Participants both view noisy stim-
uli and can communicate regarding the presence of a target. (b) The sequence of events
in the task from stimulus presentation to the presentation of feedback. Trials began with
two stimulus intervals, which contained Gabor patches, with one of them appearing
quickly during one of the intervals, and participants had to guess which one. If their
decisions did not match, they were required to negotiate about it and come to an agree-
ment. (Image adapted with permission from Bahrami et al. (2010)).
aspects of linguistic coordination could be assessed: lexical, prosodic, and

acoustic production behaviors.
The first study investigated lexical alignment. As described earlier, this
notion of alignment predicts that the more people use the same words, the
better they will perform (“indiscriminate” lexical alignment). By contrast, a
model of coordination as synergy would predict that the alignment of con-
fidence expressions only—serving the interaction’s goals of sharing confi-
dence to make a joint decision—would correlate with performance. The
analysis did show prominent “indiscriminate” alignment in all pairs: inter-
locutors displayed a high probability of picking up and employing words
used by the other in the previous interaction. However, the more a dyad
indiscriminately repeated each other’s words, the lower the collective benefit
they gained from cooperation. Automatic linguistic alignment seemed to be
76 Rick Dale et al.
deleterious to coordination on the task. In contrast, the participants’ recip-

rocal, selective adaptation to vocabularies of expressing confidence (task-
motivated selective alignment), turned out to correlate positively with the
collective benefit gained from linguistic coordination (see Figure 2.6).
The second study (Fusaroli & Tylén, submitted) more systematically
compared linguistic repetitions at three levels: first repetition of triplets of
phonemes, second repetition of patterns of pitch, and finally repetition of
patterns of speech pause sequences. A model of coordination as alignment
would have the structure of repetitions across subjects predicting perfor-
mance, while a model of coordination as synergy would predict the structure
of repetitions at the interaction level—that is, not discriminating between
interlocutors—to be correlated with performance. In other words, a synergy
model would predict that the relevant coordination happens in interactional
patterns where it does not matter which interlocutor shares confidence and
which makes a decision, as long as somebody fills those roles in each joint
decision. Employing a combination of information theory and recurrence
plots (Marwan, Carmen Romano,Thiel, & Kurths, 2007), the authors quan-
tified these repetitions both across interlocutors and in the overall interac-
tion. The results show that the relevant coordination happens at the level
of interactional patterns, but not simply across interlocutors: The more the
interlocutors develop a regular pattern of lexical choices, pitch and speech
pause sequences, which repeats across joint decisions, no matter who is pro-
ducing its different parts, the better they perform. On the contrary, indices of
repetitions across interlocutors did not correlate with performance.
Figure 2.6 Results from the alignment of word usage during the task. The Collective
Benefit (y-axis) is a measure of how much dyads benefited from their interaction; local
linguistic alignment was a benefit (right plot); however, rampant widespread “indiscrim-
inate” alignment predicted a drop in joint performance (left plot). (Adapted with permis-
sion from Fusaroli et al. (2012)).
A third study further supported these findings and investigated the

temporal dimension of coordination, in other words, how synergies self-
organize over time (Fusaroli et al., submitted). The researchers showed
that individuals’ speech production displays scaling laws (lognormal dis-
tribution), which are a signature of behaviors constrained by the emerg-
ing dynamics of the interaction. Indexes of behavioral alignment (mutual
information) were shown to decrease over time, while indexes of more
complex multiscale coordination (complexity matching; West, Geneston,
& Grigolini, 2008) were shown to increase over time. Finally, the increase
in multiscale c oordination—and not its initial value—was shown to sig-
nificantly correlate with performance. In other words, the findings further
support the importance of the self-organization of the linguistic interac-
tion in shaping the behavior of its components. Crucially, it is also shown
that self-organization happens over time, increasing in strength and efficacy
while interlocutors adapt to each other developing cooperative routines.
5.4. Interpersonal Synergies: A Summary

The empirical evidence reviewed strongly suggests that the current focus on
interactive alignment as the main engine of coordination has to be integrated
in a more complex model of interpersonal synergies, which encompasses
complementary dynamics and the development of interactional patterns
(coordinative routines). This model makes predictions, which are already
supported in a handful of studies: only task-oriented alignment is effective
in fostering coordination; important aspects of the linguistic interactions
can be described only if we focus on the development of stable patterns of
interaction whose role can be indifferently filled in by one or the other of
the interlocutors; the self-organization of the interaction takes time, grow-
ing in strength and efficacy as the interlocutors adapt to each other, by both
aligning and complementing each other.
6. CONCLUSION: TIME FOR MORE MODELS

6.1. Summary
We have offered some discussion and review of how interaction can be
understood as a process of self-organization. First, we showed that social
variables when perceived, and when taking particular forms, can funda-
mentally change the cognitive processes and behaviors of a conversation
partner. These emerging patterns can be described in the form of “phase
transition,” where lower level systems become organized differently in a
78 Rick Dale et al.
manner that is shaped by these social variables. But how do these lower level
processes constrain each other and act together? Akin to the centipede’s
dilemma, rather than understanding the interaction “leg by leg by leg,”
we entertained the notion of a synergy between interacting human beings:
the behaviors—turn-taking and rhythms, use of particular words, emer-
gence of adjacency pairs, and so on—can be seen as an array of levels that
are mutually constraining, and dynamically evolving, as two people come
to form in an important way, a “unit of analysis,” and the interaction itself a
stable, if temporary, synergy itself. Perspective-taking might be seen as part
of this synergetic process, shifting from allo- to egocentric or vice versa, as
the interlocutors enact or develop coordinative routines.
6.2. Moving Forward: Models of These Processes

The two key features we have articulated mostly describe the form of inter-
action, rather than the underlying mechanisms that give way to it. This
is an issue raised often in discussion and critique of dynamical systems
approaches to cognition (e.g. Bechtel, 1997; Dale, 2010; Eliasmith, 1996,
2012; Wagenmakers et al., 2012). In fact, we described that one exciting
aspect of growing approaches to social interaction is that these approaches
factor in mechanism. One critique of the current review could be that
we have simply advocated for a wholesale integration of as much as can
be gleaned about mechanism—and this doesn’t really tell us much about
mechanism. We have advocated instead for conceptualizing human interac-
tion as a system that self-organizes and adapts to particular contexts, such as
social variables, and organizes itself through evolving local interactions, such
as in incremental contributions to a dialog, including even nonverbal chan-
nels, like winks and nods. These are important critiques, and they should be
addressed directly. So, we end this paper with a brief review of some model-
ing endeavors that will help to guide integration of many channels, helping
to solve the centipede’s dilemma.
6.3. Surface Network Analysis, and Mechanistic Models

One way to get at the synergies directly is to carry out integrative
analysis of “multimodal” (multiperson, multibehavior, multilevel) cor-
pora. The past decade has seen a growing agenda to build large-scale
corpora of human interactions, capturing a variety of interpersonal
behaviors, linguistic contributions, contextual variables, and so on (e.g.
Fusaroli et al., 2012; Louwerse et al., 2012). After such collection of
data, researchers often go about identifying the relationship among
particular variables, such as gestures and group collaboration or prosodic

contours in particular discourse situations. These agendas are impor-
tant for understanding interactions at particular levels of analysis—the
manner in which gestures are deployed, and in what context, and how
prosody may index particular modes of interaction.
The argument we have made is that it is a nontrivial mission, both
methodologically and theoretically, to discover the manner in which
these multiple behaviors, and cognitive processes, are integrated during
ongoing interaction. One way to do this “at the surface” is to translate
corpora into a form that allows the analysis of the temporal relationship
between behaviors. In other words, different behaviors such as nod-
ding, gestures, use of particular words, and so on, can be rendered into
analyzable time series. This was done by Louwerse et al. (2012), who, at
a rate of 250 ms, tracked patterns of synchrony between interlocutors
in a direction-giving dialog (e.g. participants tended to laugh and smile
together, nod one after the other at a particular timescale). Extraction
of time series would permit an exploration of the dynamic interaction
between different channels, and between people, and exploring how
these change over the course of an interaction. One way to do this is to
project the channels into a network structure, with nodes representing
the behaviors and edges representing their relationship (e.g. strength of
connection).
Consider the following hypothetical research scenario: investigating
bouts of human interaction along a set of four behaviors (A, B, C, and D) and
measuring these behavioral channels at 250 ms intervals. Such a hypotheti-
cal data set is presented in Figure 2.7. V
arious circumstances may arise dur-
ing interaction. The channels may exhibit only weak coincidental structure,
with each “degree of freedom” of this system being one of these channels.
However, if systems exhibited pure “synchrony,” then behavioral channels
across individuals serve to constrain each other. So, instead of 2 × 4 = 8
degrees of freedom in the interaction, we have only 4 since each channel
serves to constrain that in the other person.6 If a process of alignment were
to cascade across levels, as predicted in Pickering and Garrod (2004), for
example, we would have a continued shrinking of the degrees of freedom.
As displayed beneath the middle panel in Figure 2.8, that saturation would
6 Here, we are using “degree of freedom” in a very informal way, simply to specify whether a channel,
or set of channels, is “free to vary” or whether they constrain each other in some fashion. Of course,
network analysis can involve gradient aspects of these couplings, but we ignore this for simplicity here.
80
D
C
B
A
D
C
B
A
0 50 100 150 200 250 300 350 400 450 500
D
Person 2
C
B
Behaviors
D
Person 1
C
B
A
0 50 100 150 200 250 300 350 400 450 500
Time (250 ms interval)
Figure 2.7 Simulated point processes of behavioral events. “Person 1” and “Person 2” have four behavioral dimensions (A–D). In human data,
these would correspond to delimitable actions such as nods, laughter, or gesture (Louwerse et al., 2009, 2012). Across time, these events
occur and may serve to coordinate behavior both within and across modeled processes. In the top plot, all processes are random; in the bot-
tom, there is a greater probability of “alignment” (e.g. Person 1’s A occurring with Person 2’s A). This may not be evident by mere visualization,
Rick Dale et al.

but by inducing a network through (for example) temporal correlation, we can extract the interactive structure of the model (see Figure 2.8).
result in coupling across behaviors. We would effectively have only a single

degree of freedom as behaviors fluctuate now all together as one unit.
We know through extensive explorations described earlier in this
paper that speech, gesture, and other features of interaction will exhibit
coordination (e.g. see Louwerse et al., 2009). This can be identified as
clusters in the network that become tightly entrained over time. These
are portrayed in the rightmost panel in Figure 2.8. The degrees of free-
dom relevant to this interaction are now constrained by the number
of unconnected subgraphs (in Figure 2.8, top right). As an interaction
changes across time, the network structure may change, but the degrees
of freedom may stay the same (see lower right figure, “Transition”). We
could imagine this sort of thing occurring during face-to-face interac-
tion. Imagine two students discussing lecture briefly, which one of them
missed. In this bout of interaction, nodding and gesturing and speaking
may have a characteristic temporal interaction. However, if this part of
the conversation ended, and one asks the other for directions, suddenly
their gaze and gesture may take on that “clamped degrees of freedom”
property, while others may change.
This network analysis approach may serve as a powerful means of
visualizing and quantifying the “surface configurations” of an interac-
tion and providing clues to underlying mechanisms. The authors are
engaged in some early work exploring this possibility (Dale & Louwerse,
2012; Fusaroli et al., submitted; Paxton & Dale, in press; see also related
work in Bergmann & Kopp, 2009; Kopp, 2010). There are considerable
details in need of investigation if this agenda were to be carried out in
the naturalistic context (here we have only sketched this hypothetically
using random point processes). For example, what temporal functions
best characterize the linking between channels? Gesture and nodding (for
example) have a different timescale from, one would suspect, explanation
or querying, referred to as “dialog moves”. Another issue is what the appro-
priate measures are to determine that these channels are indeed coupled.
Methods such as vector autogression (Dixon & Stephen, 2012), Bayes nets
(Bergmann & Kopp, 2009), and related techniques (Shalizi, Camperi, &
Klinkner, 2007) may do a better job at capturing the cross-covariation
among so many channels.
This “surface network” analysis may be a useful way of proceeding to
extract the “hidden” degrees of freedom that are guiding the behavioral
structure of a conversation. Still, it is important to note that, in some ways,
this research agenda is already unfolding in some prominent projects.
Person 2
A B A B A B
82
D C D C D C
A B A B A B
D C D C D C
Person 1
A B A B
Saturation Transition
D C D C
A B A B
D C D C
Figure 2.8 An illustration of different graph (network) structures induced from different interrelationships among simulated point processes
(Person 1 and Person 2). On the top left, only very light gray edges reflect a weak connection across all pairs of nodes (no behaviors are
Rick Dale et al.

coupled). In the middle column, an illustrate of alignment (A’s, B’s, and so on, go correlate), with saturation in the behavior (cascading such
that behavioral events all occur together in a kind of synchronous multilevel alignment). In the rightmost column, a synergistic structure,
where there is occasional alignment, but amidst a variety of other interconnections that may fluctuate from moment to moment (e.g. top
panel: {APerson 2, DPerson 2, CPerson 2}, {APerson 1, BPerson 1, BPerson 2}, {DPerson 1, CPerson 1}, transitions to bottom panel: {DPerson 1, BPerson 1, APerson 2}, {APerson 1,
CPerson 1}, {DPerson 2, BPerson 2}).
For example, consider the case of the Augmented Multiparty Interaction

corpus (AMI), which tracked several simulated meetings at many behav-
ioral levels (not unlike the above simulation). Computer vision and
speech automation techniques permitted the extraction of a wide range
of behaviors among several people while discussing topics such as design-
ing artifact prototypes. From extraction of multiple channels, researchers
have been developing automated techniques for capturing argumentation
(Hakkani-Tur, 2009), the structure of the meeting (Murray, Renals, Car-
letta, & Moore, 2006), the emergence of particular emotions (Reidsma,
Heylen, & Ordelman, 2006), and so on. This work is beginning to lever-
age, in essence, the probabilistic relationship among multiple channels
during interaction. Perhaps, these probabilistic models will help solve the
centipede’s dilemma.
Once we have this surface structure, and potentially even an estimate
of the number of “freely moving parts” of an interaction, there is still the
open question of the specific cognitive processes that underlie the control
of these degrees of freedom. Lack of space precludes a detailed review,
but there are many exciting possibilities that may be pursued. In models
of reading and sentence processing, Miyake and colleagues have used
latent variable modeling to relate cognitive processing tendencies with
individual differences measures like the Wisconsin card sort and dual-
task batteries (Miyake et al., 2000). This individual differences approach,
through statistical modeling, may be useful in the interactive context to
identify the cognitive constraints on specific forms of interaction. This
agenda has begun in the work of Brown-Schmidt (e.g. 2009a,b) who
has identified the role of executive control in predicting the extent to
which one person is likely to integrate knowledge of another during
interaction. These are statistical models, but there are also some compu-
tational possibilities. For example, rational models and adaptive control
theory have allowed some researchers to tap into the dynamic relation-
ship among hypothesized cognitive processes and constraints to capture,
for example, reading and other language comprehension (Lewis et al.,
2006; Bicknell & Levy, 2010; Smith & Levy, 2008). It may be possible to
induce something akin to a Hidden Markov process, beneath the “sur-
face structure” we articulated above, and specifying in greater detail the
cognitive interactions taking place that control that surface behavior. As
noted earlier, mathematical development of these models in the motor
control literature has reached a very sophisticated level (e.g. Todorov &
Jordan, 2002).
84 Rick Dale et al.
6.4. Conclusion
We have not advocated for an approach that supplants existing theoretical
accounts of interaction. The role of memory (Horton, 2005), executive
control (Brown-Schmidt, 2009a,b), alignment and priming processes
(Pickering & Garrod, 2004), coordination and adaptation (Brennan et al.,
2010; Schober & Brennan, 2003), perceptuomotor coupling (Richardson
et al., 2009; Shockley et al., 2003), accessibility accounts (Barr & Keysar, 2002),
and so on, are all crucial for accounting for interacting persons. Although the
authors of this article may wrestle with each other on this grander point,
it seems instructive to proclaim that all of these theories have central contri-
butions to play in accounting for interaction. One major, and very simple,
reason for this could be advanced in the following way: These theories have
been usefully deployed in specific experimental contexts investigated by the
researchers who have advocated for them. This means that they have strong
empirical backing in some subset of human interactive situations; a corollary
of this is that they are predictive of human interaction in similar situations.
We have argued that it is time to integrate, to go beyond the centipede’s
dilemma, and gain an understanding of the manner in which processes coor-
dinate and act together. Motivated by basic concepts of self-organization
and synergy, we described a series of experiments that show the flexibil-
ity of human interaction. Under different social situations, low- and high-
level cognitive processes flexibly adapt. Under d ifferent task conditions, the
dyad self-organizes through local exchanges, incrementally emerging, that
develop whole new “synergies”. By exploring the structure and underlying
control mechanisms, perhaps an integration of these theories will be pos-
sible. We haven’t done this here, but we have provided some clues that seem
useful to us. We hope some readers feel the same way.
ACKNOWLEDGMENTS
This research was supported in part by NSF grants BCS-0826825 and a Minority Postdoctoral
Research Fellowship (to the third author), and in part by the Danish Council for Independent
Research - Humanities (FKK) project “Joint Diagrammatical Reasoning in Language”, and the
EUROCORES grant EuroUnderstanding “Digging for the Roots of Understanding”.
REFERENCES
Ambady, N., Bernieri, F. J., & Richeson, J. A. (2000). Toward a histology of social behavior:
judgmental accuracy from thin slices of the behavioral stream. In Mark P. Zanna (Ed.),
Advances in experimental social psychology (Vol. 32, pp. 201–271). Academic Press.
Angus, D., Smith, A., & Wiles, J. (2012). Human communication as coupled time series:
quantifying multi-participant recurrence. IEEE Transactions on Audio, Speech and Language
Processing, 20, 1795–1807.
Angus, D., Watson, B., Smith, A., Gallois, C., & Wiles, J. (2012). Visualising conversation
structure across time: insights into effective doctor-patient consultations. PLoS ONE, 7,
e38014. http://dx.doi.org/10.1371/journal.pone.0038014.
Bahrami, B., Olsen, K., Latham, P. E., Roepstorff, A., Rees, G., & Frith, C. D. (2010). Optimally
interacting minds. Science, 329, 1081–1085. http://dx.doi.org/10.1126/science.1185718.
Bard, E. G., & Aylett, M. P. (1999). The dissociation of deaccenting, givenness, and syntactic
role in spontaneous speech. In Proceedings of the 1999 International Conference on
Spoken Language Processing (pp. 1753–1756).
Bard, E. G., Anderson, A. H., Sotillo, C., Aylett, M., Doherty-Sneddon, G., & Newlands, A.
(2000). Controlling the intelligibility of referring expressions in dialogue. Journal of
Memory and Language, 42(1), 1–22.
Barr, D. J. (2008). Pragmatic expectations and linguistic evidence: listeners anticipate but do
not integrate common ground. Cognition, 18–40.
Barr, D. J., & Keysar, B. (2002). Anchoring comprehension in linguistic precedents. Journal of
Bechtel, W. (1997). Representations and cognitive explanations: assessing the dynamicist’s
challenge in cognitive science. Cognitive Science, 22(3), 295–318.
Bergmann, K., & Kopp, S. (2009). GNetIc–Using bayesian decision networks for iconic ges-
ture generation. In Intelligent virtual agents (pp. 76–89).
Bernstein, N. A. (1967). Coordination and regulation of movement. New York: Pergamon Press.
Bicknell, K., & Levy, R. (2010). A rational model of eye movement control in reading. In
Proceedings of the 48th annual meeting of the association for computational linguistics (pp. 1168–1178).
Bock, J. K. (1986). Syntactic persistence in language production. Cognitive Psychology, 18, 355–387.
Bortfeld, H., & Brennan, S. E. (1997). Use and acquisition of idiomatic expressions in refer-
ring by native and non‐native speakers. Discourse Processes, 23(2), 119–147. http://dx.doi.
org/10.1080/01638537709544986.
Branigan, H. P., Pickering, M. J., & Cleland, A. A. (2000). Syntactic co-ordination in dialogue.
Cognition, 75, 13–25.
Branigan, H. P., Pickering, M. J., Pearson, J., McLean, J. F., & Brown, A. (2011). The role of
beliefs in lexical alignment: evidence from dialogs with humans and computers. Cogni-
tion, 121(1), 41–57. http://dx.doi.org/10.1016/j.cognition.2011.05.011.
Branigan, H. P., Pickering, M. J., Stewart, A. J., & McLean, J. F. (2000). Syntactic priming
in spoken production: linguistic and temporal interference. Memory and Cognition, 28,
1297–1302.
Brennan, S. E., & Clark, H. H. (1996). Conceptual pacts and lexical choice in conversation.
Journal of Experimental Psychology. Learning, Memory, and Cognition, 22, 1482–1493.
Brennan, S. E., Galati, A., & Kuhlen, A. K. (2010). Two minds, one dialog: coordinating
speaking and understanding. Psychology of Learning and Motivation, 53, 301–344.
Brennan, S. E., & Hanna, J. E. (2009). Partner-specific adaptation in dialog. Topics in Cognitive
Science, 1(2), 274–291.
Brown-Schmidt, S. (2009a). Partner-specific interpretation of maintained referential prec-
edents during interactive dialog. Journal of Memory and Language, 61(2), 171–190.
Brown-Schmidt, S. (2009b). The role of executive function in perspective taking during
online language comprehension. Psychonomic Bulletin & Review, 16(5), 893–900.
Brown-Schmidt, S., & Tanenhaus, M. K. (2008). Real-time investigation of referential
domains in unscripted conversation: a targeted language game approach. Cognitive Sci-
ence, 32(4), 643–684. http://dx.doi.org/10.1080/03640210802066816.
Brown-Schmidt, S., Gunlogson, C., & Tanenhaus, M. K. (2008). Addressees distinguish shared
from private information when interpreting questions during interactive conversation.
Cognition, 107(3), 1122–1134.
Brown-Schmidt, S., & Hanna, J. E. (2011). Talking in another person’s shoes: Incremental
perspective-taking in language processing. Dialogue & Discourse, 2(1), 11–33.
86 Rick Dale et al.
Buder, E. H., & Eriksson, A. (1999). Time-series analysis of conversational prosody for the
identification of rhythmic units. Proceedings of the 14th international congress of phonetic sciences,
(Vol. 2, pp. 1071–1074). Retrieved from: http://www.ling.gu.se/∼anders/papers/1071.pdf.
Busch, M. W. (2007). Task-based pedagogical activities as oral genres: a systemic functional
linguistic analysis. ProQuest.
Carruthers, P. (2002). The cognitive functions of language. Behavioral and Brain Sciences,
25(06), 657–674.
Cartwright, N. (1999). The dappled world. Cambridge, UK: Cambridge University Press.
Chartrand, T. L., & Bargh, J. A. (1999). The chameleon effect: the perception-behavior link
and social interaction. Journal of Personality and Social Psychology, 76, 893–910.
Chartrand, T. L., & Van Baaren, R. (2009). Human mimicry. Advances in Experimental Social
Psychology, 41, 219–274.
Chemero, A. (2009). Radical embodied cognitive science. MIT Press.
Clark, H. H. (1996). Using language. Cambridge University Press.
Clark, H. H., & Brennan, S. E. (1991). Grounding in communication. In L. B. Resnick (Ed.),
Perspectives on socially shared cognition (pp. 127–149). Washington: American Psychological
Association.
Clark, H. H., & Krych, M. A. (2004). Speaking while monitoring addressees for understand-
ing. Journal of Memory and Language, 50(1), 62–81.
Clark, H. H., & Wilkes-Gibbs, D. (1986). Referring as a collaborative process. Cognition,
22(1), 1–39.
Cleland, A. A., & Pickering, M. J. (2003). The use of lexical and syntactic information in
language production: evidence from the priming of noun- phrase structure. Journal of
Memory and Language, 49, 214–230.
Cooper, R. P., Catmur, C., & Heyes, C. (2012). Are automatic imitation and spatial compat-
ibility mediated by different processes? Cognitive Science.
Crosby, J. R., Monin, B., & Richardson, D. (2008). Where do we look during potentially
offensive behavior? Psychological Science, 19(3), 226–228.
Dale, R. (2008).The possibility of a pluralist cognitive science. Journal of Experimental & Theo-
retical Artificial Intelligence, 20(3), 155–179.
Dale, R. (2010). Critique of radical embodied cognitive science. Journal of Mind and Behavior,
31, 127-140.
Dale, R., & Louwerse, M. M. (2012). Human interaction as a multimodal network structure.
Presented at the Conceptual Structures, Discourse, and Language.
Dale, R., Warlaumont, A. S., & Richardson, D. C. (2011). Nominal cross recurrence as a gen-
eralized lag sequential analysis for behavioral streams. International Journal of Bifurcation
and Chaos, 21, 1153-1161.
De Jaegher, H., Di Paolo, E., & Gallagher, S. (2010). Can social interaction constitute social
cognition? Trends in Cognitive Sciences, 14(10), 441–447.
De Looze, C., & Rauzy, S. (2012). Measuring speakers’ similarity in speech by means of
prosodic cues: methods and potential (pp. 1393–1396). Presented at the interspeech 2011.
Di Paolo, E., & De Jaegher, H. (2012). The interactive brain hypothesis. Frontiers in Human
Neuroscience, 6. http://dx.doi.org/10.3389/fnhum.2012.00163.
Dijksterhuis, A., & Bargh, J. A. (2001). The perception-behavior expressway: automatic effects
of social perception on social behavior. Advances in Experimental Social Psychology, 33, 1–40.
Dixon, J. A., & Stephen, D. G. (2012). Multi-scale interactions in dictyostelium discoideum
aggregation. Physica A: Statistical Mechanics and Its Applications. Retrieved from: http://
www.sciencedirect.com/science/article/pii/S0378437112006280.
Dixon, J. A., Stephen, D. G., Boncoddo, R., & Anastas, J. (2010). The self-organization of
cognitive structure. Psychology of Learning and Motivation, 52, 343–384.
Dressler, R. A., Buder, E. H., & Cannito, M. P. (2009). Rhythmic patterns during conversa-
tional repairs in speakers with aphasia. Aphasiology, 23, 731–748.
Dumas, G., Chavez, M., Nadel, J., & Martinerie, J. (2012). Anatomical connectivity influences
both intra- and inter-brain synchronizations. PLoS One, 7(5), e36414.
Duran, N. D., & Dale, R. (in press). Perspective-taking in dialogue as self-organization under
social constraints. New Ideas in Psychology.
Duran, N. D., Dale, R., & Kreuz, R. J. (2011). Listeners invest in an assumed other’s perspec-
tive despite cognitive cost. Cognition, 121, 22–40.
Eliasmith, C. (1996). The third contender: a critical examination of the dynamicist theory of
cognition. Philosophical Psychology, 9(4), 441–463.
Eliasmith, C. (2012). The complex systems approach: rhetoric or revolution. Topics in Cogni-
tive Science, 4(1), 72–77. http://dx.doi.org/10.1111/j.1756-8765.2011.01169.x.
Engstrom, D. A., Kelso, J. A., & Holroyd, T. (1996). Reaction-anticipation transitions in
human perception-action patterns. Human Movement Science, 15(6), 809–832.
Epley, N., Keysar, B., Van Boven, L., & Gilovich, T. (2004). Perspective taking as egocentric
anchoring and adjustment. Journal of Personality and Social Psychology, 87(3), 327.
Fowler, C. A., Richardson, M. J., Marsh, K. L., & Shockley, K. (2008). Language use,
coordination, and the emergence of cooperative action. In A. Fuchs & V. Jirsa (Ed.),
Coordination: Neural, behavioral and social dynamics. Springer.
Frank, T. D., Richardson, M. J., Lopresti-Goodman, S. M., & Turvey, M. T. (2009). Order
parameter dynamics of body-scaled hysteresis and mode transitions in grasping behavior.
Journal of Biological Physics, 35(2), 127–147.
Frith, U., & Frith, C. (2001). The biological basis of social interaction. Current Directions in
Psychological Science, 10(5), 151–155.
Fusaroli, R., Abney, D. H., Bahrami, B., Kello, C., & Tylén, K. (submitted). Conversation,
coupling and complexity: Matching Scaling Laws Predict Performance in a Joint Deci-
sion Task.
Fusaroli, R., Bahrami, B., Olsen, K., Roepstorff, A., Rees, G., Frith, C., & Tylén, K. (2012).
Coming to terms quantifying the benefits of linguistic coordination. Psychological science,
23(8), 931–939.
Fusaroli, R., Raczaszek-Leonardi, J., & Tylén, K. (in press). Dialogue as interpersonal synergy.
New Ideas in Psychology.
Fusaroli, R., & Tylén, K. (2012). Carving language for social interaction: a dynamic approach.
Interaction Studies, 13, 103–123.
Fusaroli, R., & Tylén, K. (submitted). Individual behavior, interactive alignment or interper-
sonal synergy? A model-comparison study on linguistic dialog.
Fussell, S. R., & Krauss, R. M. (1992). Coordination of knowledge in communication: effects
of speakers’ assumptions about what others know. Journal of Personality and Social Psychol-
ogy, 62(3), 378.
Galantucci, B. (2009). Experimental Semiotics: a new approach for studying communication
as a form of joint action. Topics in Cognitive Science, 1, 393–410.
Galantucci, B., & Sebanz, N. (2009). Joint action: current perspectives. Topics in Cognitive
Science, 1, 255–259.
Galati, A., & Brennan, S. E. (2010). Attenuating information in spoken communication: for
the speaker, or for the addressee? Journal of Memory and Language, 62(1), 35–51.
Gallagher, R., & Appenzeller, T. (1999). Beyond reductionism. Science, 284(5411), 79–79.
http://dx.doi.org/10.1126/science.284.5411.79.
Gallagher, H., Jack, A. I., Roepstorff, A., & Frith, C. D. (2002). Imaging the intentional stance
in a competitive game. Neuroimage, 16(3 pt. 1). Retrieved from: https://pure.au.dk/ws/
files/48455072/gallagher2002.pdf.
Gallese,V. (2008). Mirror neurons and the social nature of language: the neural exploitation
hypothesis. Social Neuroscience, 3(3–4), 317–333.
Gambi, C., & Pickering, M. J. (2011). A cognitive architecture for the coordination of utter-
ances. Frontiers in Psychology, 2.
88 Rick Dale et al.
Garrod, S., & Anderson, A. (1987). Saying what you mean in dialogue: a study in conceptual
and semantic co-ordination. Cognition, 27, 181–218.
Garrod, S., & Clark, A. (1993). The development of dialogue co-ordination skills in school-
children. Language and Cognitive Processes, 8, 101–126.
Garrod, S., & Doherty, G. (1994). Conversation, co-ordination and convention: an empiri-
cal investigation of how groups establish linguistic conventions. Cognition, 53, 181–215.
Garrod, S., & Pickering, M. J. (2004).Why is conversation so easy? Trends in Cognitive Sciences,
8(1), 8–11. http://dx.doi.org/10.1016/j.tics.2003.10.016.
Garrod, S., & Pickering, M. J. (2009). Joint action, interactive alignment, and dialog. Topics in
Cognitive Science, 1, 292–304.
Giles, H., Coupland, J., & Coupland, N. (1991). Contexts of accomodation. New York: Cam-
bridge University Press.
Goodwin, C. (2000). Action and embodiment within human situated interaction. Journal of
Pragmatics, 32, 1489–1522.
Goodwin, C. (2003). Conversation and brain damage. Oxford; New York: Oxford University Press.
Goodwin, C. (2011). Building action in public environments with diverse semiotic resources.
Versus, 112–113.
Graesser, A. C., Swamer, S. S., & Hu, X. (1997). Quantitative discourse psychology. Discourse
Processes, 23(3), 229–263. http://dx.doi.org/10.1080/01638539709544993.
Grammer, K., Kruck, K. B., & Magnusson, M. S. (1998). The courtship dance: patterns of
nonverbal synchronization in opposite-sex encounters. Journal of Nonverbal Behavior,
22(1), 3–29.
Gratier, M., & Devouche, E. (2011). Imitation and repetition of prosodic contour in vocal
interaction at 3 months. Developmental Psychology, 47, 67–76. http://dx.doi.org/10.1037/
a0020722.
Gries, S. T. (2005). Syntactic priming: a corpus-based approach. Journal of Psycholinguistic
Research, 34, 365–399. http://dx.doi.org/10.1007/s10936-005-6139-3.
Hakkani-Tur, D. (2009). Towards automatic argument diagramming of multiparity meetings.
In IEEE international conference on Acoustics, Speech and Signal Processing, 2009. ICASSP
2009 (pp. 4753–4756). Retrieved from: http://ieeexplore.ieee.org/xpls/abs_all.
jsp?arnumber=4960693.
Hanna, J. E., & Brennan, S. E. (2007). Speakers’ eye gaze disambiguates referring expressions
early during face-to-face conversation. Journal of Memory and Language, 57(4), 596–615.
Hanna, J. E., & Tanenhaus, M. K. (2004). Pragmatic effects on reference resolution in a col-
laborative task: evidence from eye movements. Cognitive Science, 28(1), 105–115.
Hanna, J. E., Tanenhaus, M. K., & Trueswell, J. C. (2003). The effects of common ground
and perspective on domains of referential interpretation. Journal of Memory and Language,
49(1), 43–61. http://dx.doi.org/10.1016/S0749-596X(03)00022-6.
Hartsuiker, R. J., & Westenberg, C. (2000). Word order priming in written and spoken sen-
tence production. Cognition, 75, 27–39.
Hasson, U., Ghazanfar, A. A., Galantucci, B., Garrod, S., & Keysers, C. (2012). Brain-to-brain
coupling: a mechanism for creating and sharing a social world. Trends in Cognitive Sciences,
16(2), 114–121. http://dx.doi.org/10.1016/j.tics.2011.12.007.
Healey, P. G. T., Howes, C., & Purver, M. (2010). Does structural priming occur in ordinary
conversation? Presented at the linguistic evidence.
Healey, P. G. T., & Mills, G. (2006). Participation, precedence and co-ordination in dialogue.
Presented at the 28th annual conference of the science society.
Horton, W. S. (2005). Conversational common ground and memory processes in language
production. Discourse Processes, 40(1), 1–35.
Horton, W. S., & Gerrig, R. J. (2002). Speakers’ experiences and audience design: knowing
when and knowing how to adjust utterances to addressees. Journal of Memory and Language,
47(4), 589–606.
Horton, W. S., & Gerrig, R. J. (2005). The impact of memory demands on audience design
during language production. Cognition, 96(2), 127–142. http://dx.doi.org/10.1016/
j.cognition.2004.07.001.
Hutchins, E. (1995a). How a cockpit remembers its speeds. Cognitive Science, 19, 265–288.
Hutchins, E. (1995b). Cognition in the Wild. Cambridge, Mass: MIT Press.
Hutchins, E., & Johnson, C. M. (2009). Modeling the emergence of language as an embod-
ied collective cognitive activity. Topics in Cognitive Science, 1, 523–546. http://dx.doi.
org/10.1111/J.1756-8765.2009.01033.X.
Hutchins, E. (2010). Cognitive Ecology. Topics in Cognitive Science, 2(4), 705–715.
Hutchins, E. (2011). Enculturating the Supersized Mind. Philosophical Studies, 152(3), 437–446.
Jaeger, T. F. (2010). Redundancy and reduction: speakers manage syntactic informa-
tion density. Cognitive Psychology, 61, 23–62. http://dx.doi.org/10.1016/j.cog-
psych.2010.02.002.
Jefferson, G. (1988). On the Sequential Organization of Troubles-Talk in Ordinary Conver-
sation. Social Problems, 35(4), 418–441.
Kauffman, S. (1996). At home in the universe: the search for the laws of self-organization and com-
plexity: the search for the laws of self-organization and complexity. Oxford University Press.
Kello, C. T. (2013). Critical branching neural networks. Psychological Review, 120, 230–254.
Kelso, J. A.S. (1981). On the oscillatory basis of movement. Bulletin of the Psychonomic Society,
18(63), 9.
Kelso, J. A. (1995). Dynamic patterns:The self-organization of brain and behavior. The MIT Press.
Kelso, J. A.S. (2009). Synergies: atoms of brain and behavior. Advances in Experimental Medicine
and Biology, 629, 83–91. http://dx.doi.org/10.1007/978-0-387-77064-2_5.
Keysar, B., Barr, D. J., Balin, J. A., & Brauner, J. S. (2000). Taking perspective in conversation:
the role of mutual knowledge in comprehension. Psychological Science, 11(1), 32–38.
Keysar, B., Barr, D. J., & Horton, W. S. (1998). The egocentric basis of language use: Insights
from a processing approach. Current Directions in Psychological Science, 7(2), 46–50.
Keysar, B., Lin, S., & Barr, D. J. (2003). Limits on theory of mind use in adults. Cognition,
89(1), 25–41.
Kingstone, A., Smilek, D., Ristic, J., Friesen, C. K., & Eastwood, J. D. (2003). Attention,
researchers! It is time to take a look at the real world. Current Directions in Psychological
Science, 12(5), 176–180. http://dx.doi.org/10.1111/1467-8721.01255.
Knoblich, G., Butterfill, S., & Sebanz, N. (2011). Psychological research on joint action: the-
ory and data. Psychology of Learning and Motivation-Advances in Research and Theory, 54, 59.
Knoblich, G., & Sebanz, N. (2006).The social nature of perception and action. Current Direc-
tions in Psychological Science, 15(3), 99–104.
Konvalinka, I., & Roepstorff, A. (2012). The two-brain approach: how can mutually interact-
ing brains teach us something about social interaction? Frontiers in Human Neuroscience. 6.
Kopp, S. (2010). Social resonance and embodied coordination in face-to-face conversation
with artificial interlocutors. Speech Communication, 52(6), 587–597.
Kousidis, S., & Dorran, D. (2009). Monitoring convergence of temporal features in spontane-
ous dialogue speech. In Conference papers (p. 1). Retrieved from: http://arrow.dit.ie/cgi/
viewcontent.cgi?article=1003&context=dmccon.
Laidlaw, K. E., Foulsham, T., Kuhn, G., & Kingstone, A. (2011). Potential social interactions
are important to social attention. Proceedings of the National Academy of Sciences, 108(14),
5548–5553.
Latash, M. L. (2008). Synergy. Oxford; New York: Oxford University Press.
Latash, M. L., Scholz, J. P., & Schoner, G. (2007). Toward a new theory of motor synergies.
Motor Control, 11, 276–308.
Lee, C., Black, M., Katsamanis, A., Lammert, A., Baucom, B., Christensen, A., et al. (2010).
Quantification of prosodic entrainment in affective spontaneous spoken interactions of
married couples (Vols. 793–796). Presented at the interspeech.
90 Rick Dale et al.
Lelong, A., & Bailly, G. (2011). Study of the phenomenon of phonetic convergence thanks to
speech dominoes. In Analysis of Verbal and Nonverbal Communication and Enactment.
The Processing Issues (pp. 273–286). Springer Berlin Heidelberg.
Lev-Ari, S., & Keysar, B. (2010).Why don’t we believe non-native speakers? The influence of
accent on credibility. Journal of Experimental Social Psychology, 46(6), 1093–1096.
Levelt,W. J. M., & Kelter, S. (1982). Surface form and memory in question answering. Cogni-
tive Psychology, 14, 78–106.
Levinson, S. C. (1983). Pragmatics. Cambridge Cambridgeshire; New York: Cambridge
University Press.
Levitan, R., & Hirschberg, J. (2011). Measuring acoustic-prosodic entrainment with respect
to multiple levels and dimensions (pp. 3081–3084). Presented at the interspeech 2011.
Lewis, R. L., Vasishth, S., & Van Dyke, J. A. (2006). Computational principles of working
memory in sentence comprehension. Trends in Cognitive Sciences, 10(10), 447–454.
Lin, S., Keysar, B., & Epley, N. (2010). Reflexively mindblind: using theory of mind to inter-
pret behavior requires effortful attention. Journal of Experimental Social Psychology, 46(3),
551–556.
Lockridge, C. B., & Brennan, S. E. (2002). Addressees’ needs influence speakers’ early syntactic
choices. Psychonomic Bulletin & Review, 9(3), 550–557.
Louwerse, M. M., Benesh, N.,Watanabe, S., Zhang, B., Jeuniaux, P., & Vargheese, D. (2009).The
multimodal nature of embodied conversational agents. In N. A. Taatgen, & H. van Rijn
(Eds.), Proceedings of the 31st Annual Conference of the Cognitive Science Society (pp. 1459–1463).
Louwerse, M. M., Dale, R., Bard, E. G., & Jeuniaux, P. (2012). Behavior matching in multi-
modal communication is synchronized. Cognitive Science, 36(8), 1404–1420.
Marsh, K. L., Richardson, M. J., Baron, R. M., & Schmidt, R. C. (2006). Contrasting
approaches to perceiving and acting with others. Ecological Psychology, 18(1), 1–38.
Marsh, K. L., Richardson, M. J., & Schmidt, R. C. (2009). Social connection through joint
action and interpersonal coordination. Topics in Cognitive Science, 1, 320–339.
Marwan, N., Carmen Romano, M., Thiel, M., & Kurths, J. (2007). Recurrence plots for the
analysis of complex systems. Physics Reports, 438, 237–329.
McElree, B. (2006). Accessing recent events. Psychology of Learning and Motivation, 46, 155–200.
Mehler, A.,Weiß, P., Menke, P., & Lücking, A. (2010).Towards a simulation model of dialogi-
cal alignment. In Proceedings of the 8th international conference on the evolution of language
(Evolang8), 14–17 April 2010, Utrecht (pp. 238–245).
Mills, G. (2011). The emergence of procedural conventions in dialogue. Presented at the 33rd
annual conference of the cognitive science society.
Mills, G. (in press). Dialogue in joint activity: complementarity, convergence and conven-
tionalization. New Ideas in Psychology.
Mills, G., & Gregoromichelaki, E. (2010). Establishing coherence in dialogue: sequentiality,
intentions and negotiation. Presented at the SemDial (PozDial).
Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., Howerter, A., & Wager, T. D.
(2000). The unity and diversity of executive functions and their contributions to com-
plex “frontal lobe” tasks: a latent variable analysis. Cognitive Psychology, 41(1), 49–100.
Murray, G., Renals, S., Carletta, J., & Moore, J. (2006). Incorporating speaker and discourse
features into speech summarization. In Proceedings of the main conference on human language
technology conference of the North American chapter of the association of computational linguistics
(pp. 367–374). : Retrieved from: http://dl.acm.org/citation.cfm?id=1220882.
Murray, L., & Trevarthen, C. (1985). Emotional regulation of interactions between two-
month-olds and their mothers. Social Perception in Infants, 177–197.
Nadel, J., Carchon, I., Kervella, C., Marcelli, D., & Réserbat-Plantey, D. (1999). Expectancies
for social contingency in 2 month olds. Developmental Science, 2, 164–173.
Nass, C., Fogg, B. J., & Moon,Y. (1996). Can computers be teammates? International Journal of
Human-Computer Studies, 45(6), 669–678.
Newell, K. M., Broderick, M. P., Deutsch, K. M., & Slifkin, A. B. (2003). Task goals and
change in dynamical degrees of freedom with motor learning. Journal of Experimental
Psychology: Human Perception and Performance, 29(2), 379–387.
Newman-Norlund, S. E., Noordzij, M. L., Newman-Norlund, R. D.,Volman, I. A.C., Ruiter,
J. P., Hagoort, P., et al. (2009). Recipient design in tacit communication. Cognition,
111(1), 46–54.
Niederhoffer, K. G., & Pennebaker, J. W. (2002). Linguistic style matching in social interac-
tion. Journal of Language and Social Psychology, 21(4), 337–360.
Nishimura, R., Kitaoka, N., & Nakagawa, S. (2008). Analysis of relationship between
impression of human-to-human conversations and prosodic change and its modeling
(pp. 534–537). Presented at the interspeech 2008.
Oller, D., Niyogi, P., Gray, S., Richards, J., Gilkerson, J., Xu, D., et al. (2010). Automated vocal
analysis of naturalistic recordings from children with autism, language delay, and typical
development. Proceedings of the National Academy of Sciences of the United States of America,
107(30), 13354.
Onnis, L., & Spivey, M. J. (2012). Toward a new scientific visualization for the language sci-
ences. Information, 3(1), 124–150.
Orsucci, F., Giuliani, A., & Webber, C. (2006). Combinatorics and synchronization in natural
semiotics. Physica A: Statistical Mechanics and Its Applications, 361, 665–676.
Orsucci, F., Giuliani, A., & Zbilut, J. (2004). Structure & coupling of semiotic sets. AIP confer-
ence proceedings (Vol. 742, pp. 83). : Retrieved from: http://pdfserv.aip.org/APCPCS/
vol_742/iss_1/83_1.pdf.
Orsucci, F.,Walter, K., Giuliani, A.,Webber, C. L., & Zbilut, J. P. (1997). Orthographic s tructuring
of human speech and texts: Linguistic application of recurrence quantification analysis. Arxiv
Preprint cmp-lg/9712010.
Pardo, J. S., Gibbons, R., Suppes, A., & Krauss, R. M. (2011). Phonetic convergence in college
roommates. Journal of Phonetics.
Paxton, A., & Dale, R. Frame-differencing methods for measuring bodily synchrony in con-
versation. Behavior Research Methods, in press.
Paxton, A., & Dale, R. Argument disrupts interpersonal synchrony, submitted for publication.
Perkins, L., & Milroy, L. (1997). Sharing the communicative burden: a conversation-analytic
account of aphasic/non-aphasic interaction. Multilingua, 16(2–3), 199–215.
Perry, M. (2010). Socially distributed cognition in loosely coupled systems. AI & Society, 1–14.
Pickering, M. J., & Branigan, H. P. (1999). Syntactic priming in language production. Trends
in Cognitive Sciences, 3, 136–141.
Pickering, M. J., & Ferreira,V. S. (2008). Structural priming: a critical review. Psychological Bul-
letin, 134, 427–459. http://dx.doi.org/10.1037/0033-2909.134.3.427.
Pickering, M. J., & Garrod, S. (2004).Toward a mechanistic psychology of dialogue. Behavioral
and Brain Sciences, 27, 169–190.
Pickering, M. J., & Garrod, S. (2009). Prediction and embodiment in dialogue. European
Journal of Social Psychology, 39(7), 1162–1168.
Pickering, M. J., & Garrod, S. An integrated theory of language production and comprehen-
sion. Behavioral and Brain Sciences, in press.
Platek, S. M. (2010). Yawn, yawn, yawn, yawn; yawn, yawn, yawn! The social, evolutionary
and neuroscientific facets of contagious yawning. Frontiers of Neurology and Neuroscience,
28, 107–112. http://dx.doi.org/10.1159/000307086.
Port, R. F., & Van Gelder, T. (1995). Mind as motion: explorations in the dynamics of cognition.
Cambridge, Mass: MIT Press.
Ramenzoni,V. C., Davis,T. J., Riley, M. A., Shockley, K., & Baker, A. A. (2011). Joint action in
a cooperative precision task: nested processes of intrapersonal and interpersonal coordi-
nation. Experimental Brain Research. Experimentelle Hirnforschung. Experimentation Cerebrale,
211, 447–457. http://dx.doi.org/10.1007/s00221-011-2653-8.
92 Rick Dale et al.
Ramenzoni,V. C., Riley, M. A., Shockley, K., & Baker, A. A. (2012). Interpersonal and intra-
personal coordinative modes for joint and single task performance. Human Movement
Science. http://dx.doi.org/10.1016/j.humov.2011.12.004.
Reidsma, D., Heylen, D. K. J., & Ordelman, R. J.F. (2006). Annotating emotions in meetings.
LREC 2006. Retrieved from: http://eprints.eemcs.utwente.nl/8381/.
Reitter, D., Keller, F., & Moore, J. D. (2011). A computational cognitive model of syntactic
priming. Cognitive Science, 35(4), 587–637.
Reitter, D., & Moore, J. D. (2007). Predicting success in dialogue. Presented at the proceedings of
the 45th annual meeting of the association of computational linguistics (Vol. 45, pp. 808–815).
Reitter, D., Moore, J. D., & Keller, F. (2006). Priming of syntactic rules in task-oriented dia-
logue and spontaneous conversation (pp. 685–690). Presented at the proceedings of the 28th
annual conference of the cognitive science society.
Richardson, M. J., Dale, R., & Marsh, K. L. Complex dynamical systems in social and per-
sonality psychology: theory, modeling and analysis. In handbook of research methods in
social and personality psychology, in press.
Richardson, D. C., Dale, R., & Tomlinson, J. M. (2009). Conversation, gaze coordination,
and beliefs about visual context. Cognitive Science, 33(8), 1468–1482. http://dx.doi.
org/10.1111/j.1551-6709.2009.01057.x.
Richardson, M. J., Marsh, K. L., Isenhower, R. W., Goodman, J. R., & Schmidt, R. C.
(2007). Rocking together: dynamics of intentional and unintentional interpersonal
coordination. Human Movement Science, 26, 867–891. http://dx.doi.org/10.1016/j.
humov.2007.07.002.
Richardson, M. J., Marsh, K. L., & Schmidt, R. C. (2005). Effects of visual and verbal inter-
action on unintentional interpersonal coordination. Journal of Experimental Psychology:
Human Perception and Performance, 31(1), 62.
Richardson, M. J., Marsh, K. L., Schmidt, R. C., & Richardson, M. J. (2010). Challenging the
egocentric view of coordinated perceiving, acting and knowing. The Mind in Context,
307–333.
Richardson, D. C., & Spivey, M. J. (2000). Representation, space and Hollywood Squares:
looking at things that aren’t there anymore. Cognition, 76(3), 269–295.
Richardson, D. C., Street, C. N. H., Tan, J.Y. M., Kirkham, N. Z., Hoover, M. A., & Cavana-
ugh, A. G. (2012). Joint perception: gaze and social context. Frontiers in Human Neuro-
science, 6. Retrieved from: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3388371/.
Riley, M. A., Richardson, M. J., Shockley, K., & Ramenzoni,V. C. (2011). Interpersonal syn-
ergies. Frontiers in Psychology, 2, 38. http://dx.doi.org/10.3389/fpsyg.2011.00038.
Riley, M. A., & Van Orden, G. C. (2005). Tutorials in contemporary nonlinear methods for the
behavioral sciences. : National Science Foundation.
Risko, E. F., & Kingstone, A. (2011). Eyes wide shut: implied social presence, eye tracking and
attention. Attention, Perception, & Psychophysics, 73(2), 291–296.
Roche, J., Jaeger,T. F., Dale, R., & Kreuz, R. J. Don’t rush the navigator: disambiguating strat-
egies are hard to establish but easier to maintain, submitted for publication.
Sacks, H., Jefferson, G., & Schegloff, E. A. (1995). Lectures on conversation. (Vol. 1). :Wiley. Online
Library. Retrieved from: http://onlinelibrary.wiley.com/doi/10.1002/9781444328301.
fmatter/summary.
Sacks, H., Schegloff, E. A., & Jefferson, G. (1974). A simplest systematics for the organization
of turn-taking for conversation. Language, 50, 696–735.
Salvucci, D. D., & Taatgen, N. A. (2008).Threaded cognition: an integrated theory of concur-
rent multitasking. Psychological Review, 115(1), 101.
Schank, R. C., & Abelson, R. P. (1977). Scripts, plans, goals, and understanding: an inquiry into
human knowledge structures. Hillsdale, N.J. New York: L. Erlbaum Associates. (distributed
by the Halsted Press Division of John Wiley and Sons).
Schegloff, E. A. (1986). The routine as achievement. Human Studies, 9, 111–151.
Schegloff, E. A. (2007). Sequence organization in interaction. A primer in conversation analysis.

(Vol. 1). : Cambridge University Press.
Schmidt, R. C. (2007). Scaffolds for social meaning. Ecological Psychology, 19(2), 137–151.
Schmidt, R. C., Bienvenu, M., Fitzpatrick, P. A., & Amazeen, P. G. (1998). A comparison
of intra- and interpersonal interlimb coordination: coordination breakdowns and cou-
pling strength. Journal of Experimental Psychology: Human Perception and Performance, 24,
884–900.
Schmidt, R. C., Morr, S., Fitzpatrick, P., & Richardson, M. (2012). Measuring the dynamics
of interactional synchrony. Journal of Nonverbal Behavior, 36(4), 263–279. http://dx.doi.
org/10.1007/s10919-012-0138-5.
Schmidt, R. C., & Richardson, M. J. (2008). Dynamics of interpersonal coordination.
In Coordination: Neural, behavioral and social dynamics (pp. 281–308).
Schmidt, R. C., & Turvey, M. T. (1995). Models of interlimb coordination—equilibria,
local analyses, and spectral patterning: comment on Fuchs and Kelso (1994). Journal of
Experimental Psychology: Human Perception and Performance, 21(2), 432–443. http://dx.doi.
org/10.1037/0096-1523.21.2.432.
Schober, M. F. (1993). Spatial perspective-taking in conversation. Cognition, 47(1), 1–24.
Schober, M. F. (1995). Speakers, addressees, and frames of reference: whose effort is mini-
mized in conversations about locations? Discourse Processes, 20(2), 219–247.
Schober, M. F., & Brennan, S. E. (2003). Processes of interactive spoken discourse: the role of
the partner. In A. C. Graesser, M. A. Gernsbacher & S. R. Goldman (Eds.), Handbook of
discourse processes (pp. 211–232). Lawrence Erlbaum.
Seidenberg, M. S., & MacDonald, M. C. (1999). A probabilistic constraints approach to lan-
guage acquisition and processing. Cognitive Science, 23(4), 569–588.
Shalizi, C., Camperi, M., & Klinkner, K. (2007). Discovering functional communities in
dynamical networks. Statistical Network Analysis: Models, Issues, and New Directions, 140–157.
Shintel, H., & Keysar, B. (2009). Less is more: a minimalist account of joint action in com-
munication. Topics in Cognitive Science, 35(2), 281–322.
Shockley, K., Richardson, D. C., & Dale, R. (2009). Conversation and coordinative structures.
Topics in Cognitive Science, 1(2), 305–319.
Shockley, K., Santana, M.V., & Fowler, C. A. (2003). Mutual interpersonal postural constraints
are involved in cooperative conversation. Journal of Experimental Psychology: Human Per-
ception and Performance, 29(2), 326–332.
Shteynberg, G. (2010). A silent emergence of culture: the social tuning effect. Journal of Per-
sonality and Social Psychology, 99(4), 683.
Shteynberg, G., & Galinsky, A. D. (2011). Implicit coordination: sharing goals with similar
others intensifies goal pursuit. Journal of Experimental Social Psychology, 47(6), 1291–1294.
Sidnell, J., & Enfield, N. (2012). Language diversity and social action. Current Anthropology,
53, 302–333.
Smith, N. J., & Levy, R. (2008). Optimal processing times in reading: a formal model and
empirical investigation. In Proceedings of the 30th annual conference of the cognitive science
society (pp. 595–600). Retrieved from: http://idiom.ucsd.edu/∼rlevy/papers/smith-
levy-2008-cogsci.pdf.
Smith, M., & Wheeldon, L. (2001). Syntactic priming in spoken sentence production—an
online study. Cognition, 78, 123–164.
Spivey, M. J. (2007). The continuity of mind. Oxford, UK: Oxford University Press.
Spurrett, D., & Cowley, S. J. (2004). How to do things without words: infants, utterance-
activity and distributed cognition. Language Sciences, 26, 443–466.
Stivers, T., Enfield, N. J., Brown, P., Englert, C., Hayashi, M., Heinemann, T., et al. (2009).
Universals and cultural variation in turn-taking in conversation. Proceedings of the
National Academy of Sciences of the United States of America, 106, 10587–10592. http://
dx.doi.org/10.1073/pnas.0903616106.
94 Rick Dale et al.
Szmrecsanyi, B. (2005). Language users as creatures of habit: a corpus-based analysis of persis-

tence in spoken English. Corpus Linguistics and Linguistic Theory, 1, 113–149.
Szmrecsanyi, B. (2006). Morphosyntactic persistence in spoken english: A corpus study at the intersec-
tion of variationist sociolinguistics, psycholinguistics, and discourse analysis. Berlin/New York:
Mouton de Gruyter.
Tanaka, J. W., & Taylor, M. (1991). Object categories and expertise: is the basic level in
the eye of the beholder? Cognitive Psychology, 23(3), 457–482. http://dx.doi.
org/10.1016/0010-0285(91)90016-H.
Tanenhaus, M. K., & Brown-Schmidt, S. (2008). Language processing in the natural world.
Philosophical Transactions of the Royal Society B: Biological Sciences, 363(1493), 1105.
Tanenhaus, M., Spivey-Knowlton, M. J., Eberhard, K. M., & Sedivy, J. C. (1995). Integra-
tion of visual and linguistic information in spoken language comprehension. Science,
268(5217), 1632–1634.
Teufel, C., Fletcher, P. C., & Davis, G. (2010). Seeing other minds: attributed mental states
influence perception. Trends in Cognitive Sciences, 14(8), 376–382.
Thelen, E., & Smith, L. B. (1994). A dynamic systems approach to the development of cognition and
action. Cambridge, MA: The MIT Press.
Todorov, E., & Jordan, M. I. (2002). Optimal feedback control as a theory of motor coordina-
tion. Nature Neuroscience, 5(11), 1226–1235.
Truong, K. P., & Heylen, D. K. J. (2012). Measuring prosodic alignment in cooperative task-
based conversations. Presented at the interspeech 2012.
Tuller, B., Case, P., Ding, M., & Kelso, J. A. (1994). The nonlinear dynamics of speech cat-
egorization. Journal of Experimental Psychology: Human Perception and Performance, 20(1), 3.
Turvey, M. T. (1990). Coordination. The American Psychologist, 45, 938–953.
Turvey, M. T. (2007). Action and perception at the level of synergies. Human Movement
Science, 26(4), 657–697.
Tylén, K., Fusaroli, R., Bundgaard, P., & Østergaard, S. (2013). Making sense together: a
dynamical account of linguistic meaning making. Semiotica, 194, 39–62.
Tylén, K., Weed, E., Wallentin, M., Roepstorff, A., & Frith, C. D. (2010). Language as a tool
for interacting minds. Mind & Language, 25, 3–29.
Van Orden, G., & Stephen, D. G. (2012). Is cognitive science usefully cast as complexity
science? Topics in Cognitive Science, 4, 3–6.
Van Overwalle, F. (2008). Social cognition and the brain: a meta-analysis. Human Brain
Mapping, 30(3), 829–858.
Van Rooij, I., Bongers, R. M., & Haselager, W. (2002). A non-representational approach to
imagined action. Cognitive Science, 345–375.
Vaughan, B. (2011). Prosodic synchrony in co-operative task-based dialogues: a measure of
agreement and disagreement. In 12th annual conference of the international speech communica-
tion association.
Wagenmakers, E. J., Van der Maas, H. L. J., & Farrell, S. (2012). Abstract concepts require
concrete models: why cognitive scientists have not yet embraced nonlinearly coupled,
dynamical, self-organized critical, synergistic, scale-free, exquisitely context-sensitive,
interaction-dominant, multifractal, interdependent brain-body-niche systems. Topics in
Cognitive Science, 4, 87–93.
Wang, Y., & Hamilton, A. F. C. (2012). Social top-down response modulation (STORM): a
model of the control of mimicry in social interaction. Frontiers in Human Neuroscience, 6.
Warlaumont, A. S. (2012). A spiking neural network model of canonical babbling development.
Presented at the 2012 IEEE International Conference on Development and Learning (ICDL).
West, B. J., Geneston, E. L., & Grigolini, P. (2008). Maximizing information exchange
between complex networks. Physics Reports, 468, 1–99.
Wilkes-Gibbs, D., & Clark, H. H. (1992). Coordinating beliefs in conversation. Journal of
Wilkinson, R., Beeke, S., & Maxim, J. (2003). Adapting to conversation. In C. Goodwin
(Ed.), Conversation and brain damage (pp. 59–89). : Oxford University Press.
Wilson, M., & Wilson, T. P. (2005). An oscillator model of the timing of turn-taking. Psycho-
nomic Bulletin & Review, 12, 957–968.
Wolpert, D. M., Doya, K., & Kawato, M. (2003). A unifying computational framework for
motor control and social interaction. Philosophical Transactions of the Royal Society of Lon-
don. Series B: Biological Sciences, 358(1431), 593–602.
Wu, S., & Keysar, B. (2007a).The effect of information overlap on communication effective-
ness. Cognitive Science, 31(1), 169–181.
Wu, S., & Keysar, B. (2007b). The effect of culture on perspective taking. Psychological science,
18(7), 600–606.
Wyatt, D., Bilmes, J., Choudhury,T., & Kitts, J. A. (2008).Towards the automated social analy-
sis of situated speech data. In Proceedings of the 10th international conference on ubiquitous
computing (pp. 168–171). Retrieved from: http://dl.acm.org/citation.cfm?id=1409658.

CHAPTER THREE
Conceptual Composition: The

Role of Relational Competition
in the Comprehension of
Modifier-Noun Phrases and
Noun–Noun Compounds
Christina L. Gagné1, Thomas L. Spalding
Contents
1. Introduction98
2. Modifier-Noun Phrases and Compounds as Expressions of Combined Concepts 100
3. Theoretical Framework: A Three-Stage Theory of Conceptual Combination 101
4. Evidence of the Modifier’s Role in Relation Suggestion 104
4.1. General Usage 104
4.2. Recent Usage 107
5. The Nature of Relations and the Nature of Relational Competition 108
5.1. Nature of Relations 108
5.1.1. Relational Information is Accessed via Conceptual Representations 108
5.1.2. Representation of the Relations 109
5.1.3. Level of Relational Abstraction 111
5.2. Nature of Relational Competition 111
5.2.1. Is Competition Because of Number of Competitors? 111
5.2.2. Is Competition due to Inhibition or Facilitation? 113
5.2.3. Is There Competition among Alternative Interpretations? 114
6. T he Role of Relation Competition in the Processing of Compounds that
Lack an Underlying Relation 115
7. Evaluation of Relational Interpretations 119
8. Elaboration of Combined Concepts Following Relation Selection 121
9. Summary124
10. Concluding Remarks 126
References127

98 Christina L. Gagné and Thomas L. Spalding
Abstract
Compositionality and productivity, which are the abilities to combining existing
concepts and words to create new concepts and phrases, words, and sentences, are
hallmarks of the human conceptual and language systems. Combined concepts are
formed within the conceptual system and can be expressed via modifier-noun phrases
(e.g. purple beans) and compound words (e.g. snowball), which are the simplest forms
of productivity. Modifier-noun phrases and compound words are often paraphrased
using a relation to connect the constituents (e.g. beans that are purple, ball made of
snow). The phrase or compound does not explicitly contain the underlying relation,
but the existence of the relation can be shown by manipulating the availability of the
relation and observing the effect on the interpretation of the phrase or compound. This
chapter describes how novel modifier-noun phrases and established compounds are
interpreted. We present a theoretical account of relational interpretation of combined
concepts and present the empirical evidence for the use of relational structures. We
then present the empirical evidence supporting our theoretical account’s specific pre-
dictions about how relational interpretations are selected and evaluated and how the
relational interpretation is elaborated to create a fully specified new concept.
1. INTRODUCTION
Concepts are mental entities that allow a person’s experience with
the world to be organized into meaningful units. The ability to store and
access conceptual information is vital to human cognition because many
tasks that people perform on a daily basis (such as classification, prediction,
reasoning, and communication) require the application of existing knowl-
edge to new situations. Indeed, concepts have been called the “building
blocks of cognition” (Solomon, Medin, & Lynch, 1999). Concepts are not
represented in isolation but rather are part of a complex and highly inter-
related conceptual system. Relational information plays a vital role in the
organization of the conceptual system (for an overview, see Medin, 1989) as
well as a central role in nominal compounding (e.g. Allen, 1978; Gleitman &
Gleitman, 1970; Kay & Zimmer, 1976; Levi, 1978). In particular, knowledge
about how concepts interrelate underlies conceptual composition, which is
a process through which new concepts are created by combining existing
concepts.
In English, combined concepts are often expressed as modifier-noun
phrases (e.g. chocolate recipe) or, as they become more commonly used, as
compound words (e.g. handbag). Teall (1892, p. 5) made the observation that
“in the English language it is very common to name a thing, or express
an attribute, or assert an action or manner of action by omitting minor or
connecting parts of a full expression, and using only the principal elements
Role of Relational Competition in Modifier-Noun Phrases and Noun–Noun Compounds 99
in more or less arbitrary association and frequently in inverted order”. For

example, box for hats can be expressed as hat box. In other words, linguistic
expressions of combined concepts typically omit the implicit underlying
relation. However, there is a long tradition in linguistic work on com-
pounding that points to the involvement of these underlying relations, and
combined concepts can often be paraphrased by expressing the underlying
implicit relation (e.g. flu virus is a virus that causes flu; Allen, 1978; Bisetto &
Scalise, 2005; Downing, 1977; Lees, 1960; Levi, 1978; Marchand, 1969; Scalise
& Bisetto, 2009; Warren, 1978). As Kay and Zimmer (1976, p. 29) point
out, “… there is a relation between the two nouns that the hearer must
supply” to comprehend the phrase. These relation-based modifier-noun
constructions are in contrast to copulative constructions (e.g. singer–
songwriter), which some linguists argue have a dual head structure in which the
two constituents are linked via the conjunctive and (Bauer, 2008; Bisetto &
Scalise, 2005; Booij, 2005; Olsen, 2004; Olsen & Van der Meer, 2001;
Scalise & Bisetto, 2009).
Psychological research examining the processing of modifier-noun
phrases and compounds can provide useful insight into the nature and use
of relational information within the conceptual system. Indeed, researchers
interested in conceptual combination have investigated the role of relational
information, although they vary in terms of the relative importance placed
on the use of this information. Some have suggested that relation process-
ing is one of two processes that can be used (Estes, 2003; Wisniewski, 1996,
1997). Wisniewski (1996), for example, argues that combined concepts are
interpreted by one of two fundamentally different processes: relation linking
or property mapping. Others (e.g. Gagné & Shoben, 1997; Spalding, Gagné,
Mullaly, & Ji, 2010) have argued that relation processing is the initial pro-
cess and that the derivation of properties follows the selection of a relation.
That is, relation linking and property derivation are not mutually exclusive.
Instead, the relation plays a key role in determining which properties can
be inferred (Gagné, 2000; Gagné & Murphy, 1996; Spalding et al., 2010).
However, regardless of one’s theoretical perspective on conceptual combi-
nation, it has been acknowledged that the issue of relational information is
highly relevant for developing a deeper understanding of the processing of
combined concepts.
This chapter reviews the literature on relational interpretation of
combined concepts and describes how people interpret and understand
modifier-noun phrases and compound words. In particular, we focus on
how relational interpretations arise and compete with each other during
comprehension. We begin by describing our theory of relational interpre-

tation of combined concepts in general terms, in particular focusing on
the somewhat different roles of the modifier and head constituents. We
then describe the evidence for relational interpretations and for the nature
of relations and relational competition during the interpretation process.
As it is important to understand the role of relational competition within
the larger context of the full interpretation process, we end by discussing
how the cognitive system evaluates and then elaborates possible relational
interpretations.
2. MODIFIER-NOUN PHRASES AND COMPOUNDS AS

EXPRESSIONS OF COMBINED CONCEPTS
Modifier-noun phrases and compound words can be used to express
combined concepts. Both types of linguistic expressions represent, in some
sense, the least complex way of combining words of human languages, in
that they are the midpoints between single words and sentences. In short,
they constitute the beginning of linguistic productivity.
Because both modifier-noun phrases and compound words are linguis-
tic representations of combined concepts, a commonality of structure and
processing between the two kinds of linguistic representation suggests itself,
namely that interpretation and access to the underlying meaning might
involve an attempt at computing a meaning involving the constituents. On
the one hand, this might seem reasonable as compound words generally
enter the language as novel modifier-noun phrases and only later take on
the status of a known compound word (at least in English; see Downing,
1977; Libben & Jarema, 2006).Thus, given that novel modifier-noun phrases
require computed meanings, it might not be surprising that something
about that meaning computation is maintained even after the phrase enters
the language as a compound word. On the other hand, from a linguistic
perspective, it might be thought quite surprising that such processing would
survive for compounds that are well known as it is generally believed that it
is easier to remember a meaning than to compute it (see Bertram, Laine, &
Karvinen, 1999, for an example of this claim with respect to morphologi-
cally complex words). Furthermore, it might be thought strange from a
conceptual perspective: the meaning of the known compound has to have
been computed previously, probably many times; so it could be stored as a
separate concept that just happens to be linguistically identified by a com-
pound word.
However, within the compound word literature, there is growing evi-

dence of meaning construction during the access and interpretation of
compound words (Fiorentino & Poeppel, 2007; Gagné & Spalding, 2009;
Gagné, Spalding, Figueredo, & Mullaly, 2009; Inhoff, Radach, & Heller,
2000; Ji, Gagné, & Spalding, 2011; Koester, Gunter, & Wagner, 2007). That
is, despite the fact that compound words are known, and at least potentially
stored in some form of mental lexicon, meanings based on the constitu-
ent words are constructed during processing. In short, the processing of a
compound word is highly sensitive to its constituent structure. In fact, as
we will discuss in Section 6, the meaning computation process seems to
be initiated even for opaque compounds (e.g. hogwash), those for which
the actual meaning cannot be computed from the constituents. Given the
commonality of underlying structure, and the evidence that compound
words undergo a meaning construction process during comprehension, we
believe that theories of conceptual combination must apply to both novel
and familiar combinations.
3. THEORETICAL FRAMEWORK: A THREE-STAGE

THEORY OF CONCEPTUAL COMBINATION
The competition among relations in nominals (CARIN) theory of
conceptual combination claims that relational information plays a critical
role in the processing of novel modifier-noun phrases (Gagné & Shoben,
1997; Spalding & Gagné, 2008) and was the first primarily relational
psychological theory of conceptual combination. In general, information
about how objects, people, and so on, interact in the world is used to select
a relation that links the constituent concepts during the formation of a
new concept. In particular, according to this theory, the availability of rela-
tional information varies from constituent to constituent, and the avail-
ability affects the ease with which combined concepts can be processed.
For example, people know that when chocolate is used as a modifier, the
compound can usually be paraphrased using a MADE OF relation but that
other relations such as FOR and ABOUT are also possible. Consequently,
some relations are more readily available than others, and this difference in
availability influences the time required to interpret a compound. Another
key claim of this theory is that relations compete with each other for selec-
tion. Consequently, it should be easier to interpret a compound that requires
a relation that is highly available (i.e. a relation that is a strong competitor)
than to interpret a compound that requires a relation that is less available
(i.e. a relation that is a weak competitor). Moreover, changing the avail-

ability of a relation (by manipulating prior context, for example) should
affect the ease of comprehension. According to the CARIN theory, relation
selection is most heavily influenced by relational information pertaining to
the modifier concept.
The relational interpretation competitive evaluation (RICE) theory
(Spalding et al., 2010) is an extension and refinement of the CARIN theory.
According to RICE, the interpretation of a compound is obtained through
a “suggest evaluate elaborate” process (Figure 3.1).
Like CARIN, the RICE theory posits that multiple relational structures
are constructed and evaluated as possible interpretations. These structures
compete with one another, and ease of interpretation depends on how
quickly a relation structure can be identified as the most likely candidate.
The RICE theory distinguishes between the assignment of a constituent to
a particular morphosyntactic role (e.g. to either the modifier or head noun
role) and the selection of a relational structure. According to this theory,
relational information about the constituents is accessed in the context of
the constituents’ morphosyntactic roles. Furthermore, the search for suit-
able relations is triggered after the constituents have been assigned to their
respective morphosyntactic roles, and relation availability is associated with
the concepts in their particular role (i.e. the relational availability for snow
in snow hill is based on snow’s previous use as a modifier and is unaffected
by snow’s history of use as a head). As in the CARIN theory, the rela-
tions are initially suggested with a strength proportional to the relations’
availability for the modifier, and then the appropriateness of the suggested
Figure 3.1 Schematic of the RICE (Relational Interpretation Competitive Evaluation) theory.
relations is evaluated. However, the RICE theory is more comprehensive

than the earlier theory, in that it more clearly identifies the roles played
by the modifier and head noun and suggests that the constituents are dif-
ferentially involved in relation suggestion and evaluation. In particular, the
modifier is more heavily involved in relation suggestion, while information
associated with both constituents is heavily involved in the relation evalua-
tion phase of interpretation.
In addition, the RICE theory removes CARIN’s claim that relation
availability is associated only with the modifier concept. Unlike CARIN,
RICE proposes that relational interpretations are evaluated using semantic
and pragmatic information, but also relational information, associated with
both the modifier and the head concepts. Thus, as with CARIN, an impor-
tant aspect of the evaluation process is the ability of the constituents to
function as arguments for a given relation structure. In particular, a constitu-
ent must fit the entailments required to fulfill a particular function within
a particular relation. Thus, the constituent snow can function as a modifier
in the MADE OF relation because it is a material. Likewise, the constituent
planet can function as a head in a LOCATIVE relation because planets can
be in a physical location. However, although each constituent must fit the
restrictions of its respective role, the restrictions are codetermined. Thus,
the MADE OF relation, when actually paired with a particular head noun,
requires that the modifier be a material, but not just any material. It must
be a material that is appropriate for the head noun. Snow sculpture satisfies
these restrictions, but snow hospital does not. This aspect of the evaluation
process relies on world knowledge. Gagné (2002) and Spalding and Gagné
(2008) use the example of mountain planet to illustrate the use of this type of
knowledge. The interpretation “planet-LOCATED mountain” is rejected
because, to name just one pragmatic restriction, planets are too large to be
located in the mountains. Levi (1978) refers to this type of knowledge as
extralinguistic knowledge, and she outlines several semantic and pragmatic
considerations that are used to determine the contextually most plausible
reference for a given compound (for further discussion concerning the
use of extralinguistic information, see Downing, 1977; Finin, 1980; Levi,
1978; Meyer, 1993; Štekauer, 2005, 2006, 2009). As mentioned above, unlike
the CARIN theory, however, RICE also assumes that relational informa-
tion about the head plays a role during this evaluation stage. In particular,
in addition to the semantic and pragmatic factors listed above, a relation
should be more easily (and positively) evaluated when it is commonly used
with the head constituent.
Finally, once a suggested relation has been selected and evaluated as

appropriate, there is an elaboration stage in which the full meaning of the
compound is developed.The relational gist interpretation (e.g. man MADE
OF snow for snowman) must then be elaborated so that a snowman is known
to be cold rather than warm. This information is not explicitly in the gist
but is derived based on that gist. Importantly, the nature of the elaboration
depends on the relational gist and is not derived just from the contents
of the modifier and head concepts: a man made of snow is inanimate and
white, but a man for snow is animate and likely has a shovel. Clearly, the gist
makes a difference in how the elaboration proceeds. One interestingly open
point concerns the extent to which the relational gist is elaborated and
whether elaboration proceeds automatically when the combined concept
is interpreted or only when needed (see Spalding et al., 2010 and Gagné &
Spalding, 2011, for related discussions).
4. EVIDENCE OF THE MODIFIER’S ROLE IN

RELATION SUGGESTION
Both the CARIN and the RICE theories predict that relation avail-
ability for the modifier influences conceptual combination. The investiga-
tions of the role of relation availability have generally focused on two aspects
of experience with relations that were expected to affect relation availability.
First, a person’s experience with the ways in which a given constituent
tends to be used in combinations, and in particular what relations tend to
be used with that constituent, should affect how easily they can understand
new combinations. We refer to this factor as general usage. Second, the ease
of interpretation should be affected by what relations have most recently
been used with the constituent. We refer to this factor as recent usage. Both
streams of investigation point to the modifier constituent playing a role in
relation suggestion.
4.1. General Usage

One factor that influences the competitiveness of a particular relation is
knowledge about how likely it is to be used with the modifier constituent.
For example, because paper is often used in a compositional sense (MADE
OF) when it is in the modifier position, this relation will be a stronger
competitor than will other relations. Thus, a combination paper X will be
interpreted more quickly when the phrase requires the MADE OF relation
than when it requires a relation that is a weaker competitor. This prediction
was initially tested using information about which relations are most likely
to be used with constituents of novel combinations.To estimate the compet-
itiveness of various relations for particular constituents, Gagné and Shoben
(1997) created a set of potential novel compounds by crossing 91 modifiers
and 91 head nouns and each pairing was evaluated in terms of whether it
had a sensible literal interpretation. Each of the 3239 sensible phrases were
classified in terms of relational categories that were based on Levi’s (1978)
set of thematic relations. Levi’s categories are general and are based on her
hypothesis that all complex nominals are derived from underlying semantic
structures from which a predicate (e.g. CAUSE, HAVE, MAKE, FOR) is
deleted in the linguistic expression.
Next, the frequency with which each modifier and head noun appeared
with the various relations was determined. For example, of the 38 phrases in
which plastic was used as a modifier, 28 used the noun MADE OF modifier
relation, seven used the noun ABOUT modifier relation, two used the noun
DERIVED FROM modifier relation, and one used the modifier CAUSES
noun relation. These distributions were used to identify which relations
were highly competitive and which were less competitive for an individual
modifier or head noun. Relations were considered a good competitor for a
particular constituent if that relation was among the set of relations that was
used for the majority (60% or more) of all phrases using the constituent in
question. All other relations were considered weaker competitors.
To evaluate whether ease of comprehension was affected by relation com-
petitiveness, three experimental conditions were created. HH (High-High)
items had an underlying relation that was among the set of highly competitive
relations for both the modifier and the head, HL (High-Low) items were used a
relation that was among the set of highly competitive relations for the modifier
only, and LH (Low-High) items used an underlying relation that was among
the set of highly competitive relations for the head noun only. In Experiment
1, the individual words were controlled in that the identical words were used in
all three experimental conditions. In Experiment 2, the relation was controlled
such that there was an equal number of each relation type in each condition.
Nonsense filler items (e.g. plastic rain and scarf soda) were included. Each item
was presented on a computer screen, and participants indicated, by pressing a
key, whether the item had a sensible literal interpretation.
Both experiments demonstrated that it is easier to indicate that a phrase
has a sensible interpretation when the underlying relation is highly com-
petitive for the modifier than when it is a weak competitor; the HH and
HL conditions yielded faster response times than did the LH condition.
The results also indicated that the competitiveness of the relation with respect
to the head noun constituent did not influence response time; the HH and
HL conditions did not differ from each other. These two findings were
observed both in the analysis of variance (ANOVA), which used a dichoto-
mous measure (H vs L) of a relation’s competitiveness, as well as in a regres-
sion analysis, which used a continuous measure of a relation’s competitiveness.
Although there has been concern that the results reported in Gagné
and Shoben (1997) were because of the familiarity and plausibility of the
phrase rather than because of relation availability (Wisniewski & Murphy,
2005), regression analyses show this not to be the case (Gagné & Spalding,
2006). Objective familiarity (i.e. frequency) did not differ across the three
conditions. Also, even after subjective familiarity and plausibility were statis-
tically controlled, items with highly competitive relations took less time to
interpret than did items with less competitive relations. In addition, Gagné
and Spalding (2006) found that subjective familiarity ratings were affected
by relation availability; noun–noun phrases were more likely to be viewed as
familiar when preceded by an item with the same relation than by an item
with a different relation. If subjective familiarity were the causal factor, it
would not have been influenced by the manipulation of relation availability.
Similarly, sense/nonsense judgments (in which the person is instructed to
respond “sense” if they can think of a plausible meaning for the phrase) are
also affected by manipulations of relational availability (e.g. Gagné, 2001).
Hence, neither familiarity nor plausibility are reasonable explanations for
the results of Gagné and Shoben (1997).
Another concern has been that relation frequencies based on Gagné
and Shoben’s (1997) relational distributions might not be consistent with
relation frequencies calculated using an actual corpus (Maguire, Devereux,
Costello, & Cater, 2007). Contrary to this claim, Spalding and Gagné (2008)
found that there is a close relationship (r = 0.87) between the relation
frequencies derived by Gagné and Shoben and those based on the British
National Corpus.
A third concern might be that relational effects such as those demon-
strated by Gagné and Shoben (1997) would only occur when the com-
binations are interpreted out of context. Gerrig and Bortfeld (1999), for
example, have argued that interpreting combined concepts in isolation
could lead to a distorted view of the process of interpretation. Gagné
and Spalding (2004a), however, showed that the modifier-based relation
frequency effects are maintained, even when the combined concepts are
embedded in a supportive context. Novel modifier-noun phrases were
placed in either a neutral context (i.e. one that did not provide the inter-
pretation for the combination) or a supportive context (i.e. a context that
implied the interpretation). Phrases were read more quickly in a supportive
context than a neutral context, but the supportive context did not eliminate
the effect of modifier relation frequency; sentences containing a combina-
tion using a high-frequency relation were read more quickly than were
sentences containing a combination using a lower frequency relation.These
findings suggest that contextual information might be used in the verifica-
tion of potential interpretations but that it does not override the effect of
competing relations. That is, the context might suggest possible referents
for the phrase and that information might be used to evaluate which of the
suggested relations is most feasible in the current context.
In sum, the strength of a relation for the modifier constituent of a novel
phrase affects the ease of processing that phrase, in or out of context, and
even when other factors such as familiarity are controlled.
4.2. Recent Usage

Another factor that influences a relation’s competitiveness is recent experi-
ence with the constituent concepts. Gagné (2001) and Gagné and Shoben
(2002) found a robust relational priming effect: it takes less time to make a
sense/nonsense judgment to a novel combination (e.g. student vote) when
it is preceded by a combination with the same modifier and same relation
(e.g. student accusation) than when preceded by a combination using a dif-
ferent relation (e.g. student car). Gagné (2002) also demonstrated relation
priming when the modifier was semantically similar in the prime and target
combinations (e.g. scholar accusation as a prime for student vote). These find-
ings suggest that processing the prime phrase alters the competitiveness of
the relation used to interpret that phrase, and this change in the relative
availability of the relation influences the ease of processing a subsequent
phrase using the same modifier. However, a relation-priming effect was not
observed when the head noun was in common between the prime and the
target (Gagné, 2001, 2002; Gagné & Shoben, 2002), except when the target
combination had two equally plausible relations (Gagné & Shoben, 2002),
so that the evaluation was primarily a choice between those two relations
(this aspect of the evaluation phase is discussed further in Section 7).
The ease of processing lexicalized compounds is similarly affected
by recent exposure to a prime containing the same first constituent. In
Gagné and Spalding (2004b), the target items were established compounds
that had been selected from the Brown corpus (Francis & Kučera, 1982).
Prime items were constructed for each target compound (e.g. snowball); the
same-relation prime (e.g. snowfort) used the same relation (MADE OF) as
the target and the different-relation prime (e.g. snowshovel) used a differ-
ent relation. As was the case for novel compounds, sense/nonsense judg-
ments and lexical decision latencies were faster when the target had been
preceded by the same-relation compound than when it was preceded by
the different-relation compound. This relation-priming effect in estab-
lished compounds has been replicated several times (e.g. Gagné et al., 2009;
Spalding & Gagné, 2011). Relational priming was also observed when the
primes were restricted to existing compounds (Gagné & Spalding, 2009); so
the relational p riming effect for compounds is not limited to experiments
in which a large number of novel combinations are presented. In sum,
recent relational usage affects the interpretation of both novel and familiar
combinations, strongly suggesting a continuity of processing between novel
and familiar combinations.
5. THE NATURE OF RELATIONS AND THE NATURE

OF RELATIONAL COMPETITION
The research reviewed in the preceding sections provides strong
evidence that relational information affects the interpretation of combined
concepts, whether novel or familiar, whether modifier-noun phrases or
compound words. In this section, we discuss in more depth the nature of
the relations used for relational interpretation and the nature of relational
competition.
5.1. Nature of Relations

The RICE theory makes three main assumptions about the relations.
5.1.1. Relational Information is Accessed via

Conceptual Representations
First, RICE assumes that the relations used in relational interpretations are
in some way accessed through or connected with the conceptual represen-
tations, rather than the lexical representations, of the constituents. Because
relational information is associated with the conceptual representations, sim-
ilar relational effects should be likely to occur across (at least some) languages.
Indeed, the influence of relation availability has been demonstrated with
Mandarin (Ji & Gagné, 2007) and Indonesian (Storms & Wisniewski, 2005)
compounds and phrases, in addition to English. Furthermore, Gagné (2002)
showed relation priming when the modifier constituent of the prime com-
bination was highly semantically similar to the modifier constituent of the
target, in addition to when the modifier constituents were identical. Hence,
identical lexical items are not necessary to demonstrate relational priming.
However, when the prime modifier constituent is semantically unrelated
to the target, Gagné (2001) found no relational priming (see also Gagné,
Spalding, & Ji, 2005, for several failures to find relational priming when the
prime and target modifier constituents are unrelated). In addition, Gagné
et al. (2009) investigated relation priming when the repeated constituent
was either in the same position between prime and target or in a differ-
ent position (e.g. a given constituent served as the modifier for both prime
and target or moved from the modifier position in the prime to the head
position in the target). Relational priming was observed only when the
repeated constituent was in the same position in the prime and target.Thus,
relational access appears to depend on both the concept underlying the
constituent and the role (i.e. modifier or head) that the concept is playing
in the combination.
5.1.2. Representation of the Relations

Second, although RICE assumes that access is in some way dependent
on the constituents, it makes no claim with respect to the representation
of the relations themselves. Because relations are inherently parts of other
structures, it is not clear whether the relations that are used in relational
interpretations require separate representations or whether they are recov-
ered for use from their existence within existing relational interpretations.
Clearly, people have representations of semantic information that is similar
to these relations. For example, people must have a concept of causality,
but the question is whether this separate concept of causality is necessar-
ily implicated in the activation of the relation CAUSE during conceptual
combination. Although there have been claims that the relations are sepa-
rate representations, and not tied to any constituent, the demonstrations of
general relational effects have not been convincing. If the relations were
represented separately, and not accessed via constituents, one would expect
general differences in ease of processing, based on general characteristics
of the relations, such as their overall frequency or semantic complexity.
Shoben and Medin (as reported by Shoben, 1991) were among the first to
investigate relational effects in conceptual combination. They proposed that
relations differ in complexity and that this variation in complexity should
influence ease of processing. For example, causal relations (noun CAUSES
modifier; modifier CAUSES noun) appear to have more primitives than the
MADE OF relation. However, Shoben and Medin found no support for
this complexity hypothesis: there were no systematic differences in process-
ing time that could be attributed to relational complexity.
Estes (2003) proposed that the relations have separate representations
and, consequently, they could be primed when the constituents of the
primes and targets were completely unrelated. Estes (2003) claimed to
demonstrate relation priming without access via the constituents. How-
ever, Gagné et al. (2005) showed that the relational effect observed in
Estes (2003) was confounded with an extremely strong semantic prim-
ing effect. For example, swimming flippers primed rugby shoes more than
did road construction. However, it is clear that swimming and rugby are both
sports, and flippers and shoes are also o
bviously semantically related. Thus,
the same-relation primes were semantically similar to the targets, in addi-
tion to being relationally similar. This is problematic for Estes’ conclusion
because Gagné (2002) had previously demonstrated relation priming fol-
lowing the presentation of a prime containing a semantically similar modi-
fier; so it was already clear that lexically identical constituents were not
required for relation priming. Indeed, as Gagné et al. showed, this semantic
similarity between primes and targets was common in Estes’ materials,
leading to a highly significant difference in semantic similarity between
the same-relation pairs and the different-relation pairs. Gagné et al. failed
to find relational priming in four experiments when the prime-target
semantic similarity was controlled. This outcome is consistent with the
results of Experiment 6 of Gagné (2001) in which relation priming was
not observed when only the relation, but neither constituent, was in com-
mon between the prime and the target, as well as the finding that repeating
only the head constituent was not sufficient to result in relation priming
(Gagné, 2001, 2002; Gagné & Shoben, 2002).
In subsequent work, Estes and Jones (2006) claimed to have demon-
strated relation priming without access via the constituents, while control-
ling for semantic similarity between the primes and the targets. However,
they used only one relation in all their experimental target items; thus, an
unusually large proportion of the “sense” target items were a MADE OF
relation, while the filler items were of many different relations. Hence, a
problem arises with this data. If the prime is a MADE OF, then the target
item was more likely to require a “sense” response, than if the prime has
some other relation. Being able to predict the correct response to the target
from the prime is, of course, a critical problem in a priming paradigm
(Estes and Jones did have sense–sense filler items, but they varied in relation;
so they did not provide a complete control for this particular problem).Thus,
again, the purported general relation-priming effect is likely not because of
relation priming, per se. We should perhaps note here that some degree of
facilitation may occur with only a repeated relation under some circum-
stances because the relations do carry semantic information and seman-
tic information can affect relational interpretation (e.g. Spalding & Gagné,
2007). However, it appears that relation priming without at least strong
similarity of the constituent concepts is unlikely in the standard relational
priming paradigm as a number of attempts to demonstrate such priming
have failed, as discussed above.
5.1.3. Level of Relational Abstraction

Third, the RICE theory is not committed to a particular set of relations,
nor to relations at a particular level of abstraction. For the purposes of mak-
ing empirical predictions, the set of relations used in experiments meant to
test the theory (e.g. in constructing the experimental materials) have been
based on Levi (1978). The theory, however, is not reliant on relations at any
particular level of generality; the experimental results indicate that this set
of relations at this level of abstraction is sufficiently specific to give rise to
reliable and consistent effects, which is the most important consideration in
terms of testing the theory. In general, we suggest that relations are hierar-
chically organized (as are concepts themselves) in level of generality. Thus,
chihuahuas are not identical to beagles, but both are still usefully character-
ized as dogs as they share much of their meaning both with each other and
with other dogs. Similarly, there might be subrelations of a particular rela-
tion such that, for example, HAS-PART and HAS-POSSESSION are both
examples of a HAS relation. It may be that relations, like concepts, have a
preferred level of abstraction. Another possibility is that the initial interpre-
tation may be at a more abstract level, and the relation may be refined and
made more specific as the combined concept is elaborated. These specula-
tions await empirical test.
5.2. Nature of Relational Competition

5.2.1. Is Competition Because of Number of Competitors?
The evidence discussed thus far indicates that relations compete for s election.
But there are many forms of competition. The CARIN theory (Gagné &
Shoben, 1997) mathematically instantiated competition by using a ratio that
has the frequency (represented as a proportion) of the selected relation in
the numerator and the sum of the frequency of the three strongest competi-
tors and the selected relation in the denominator (Gagné & Shoben, 1997).
An exponential function is applied to each proportion. Although Maguire
et al. (2007) have suggested that this ratio does not embody competition,
Spalding and Gagné (2008) have demonstrated that the ratio makes sharp
distinctions between items on the basis of the number of competing rela-
tions that are stronger than the selected relation and that these distinctions
become sharper with increasing numbers of competitors.The original anal-
ysis (in Gagné & Shoben, 1997) found that including more than the three
strongest relations in the denominator of the strength ratio did not improve
the fit of the model; therefore, the number of stronger competitors was
based on the three strongest relations in a given modifier’s relational distri-
bution. Spalding and Gagné showed that three mathematical models that
instantiate sensitivity to number of stronger competitors fit the data better
than a model that includes only frequency of the selected relation. In addi-
tion, a model that simply includes the number (0–3) of stronger competitor
relations fits the Gagné and Shoben data as well as the original Gagné and
Shoben mathematical model. The model that simply has the number of
stronger competitors also fits the Gagné and Shoben (1997) data as well as
(or better than) the rank of the selected relation (Experiment 1, r = 0.47 for
rank and r = 0.45 for number of competitors, and Experiment 3, r = 0.25
for rank and r = 0.36 for number of competitors). Thus, as in Gagné and
Shoben’s analyses and models, including more stronger competitors (i.e.
ranks beyond 3) did not improve the fit of the model.
If the process of suggesting, evaluating, and eliminating relations is serial,
then using more competitors (i.e. rank) should have been a much better
predictor than just the number of stronger competitors as it should have
been able to pick up any variance associated with the competitors beyond
the third. However, the analysis shows that competitors beyond the third
have little impact on ease of processing. Overall, then, the data are con-
sistent with a parallel process in which the number of strong competitors
matters greatly, but the number of weak competitors (even if still stronger
than the required relation) does not matter as much. This might be because
such weak competitors require very little negative evidence from the head
to rule them out, while strong competitors require substantial negative
evidence from the head. As shown in Spalding et al. (2010) and discussed
below, the head plays a large role in the evaluation of suggested relations,
and perhaps the relations beyond the third competitor have little impact
because the information from the head has been brought online by this
point in processing, supporting the required relation. Thus, perhaps low-

frequency relations (for either constituent) are hard to accept and easy to
reject. This pattern would also make sense on a parallel processing account
in which the amount of activation to a given relation is a function of its
frequency for the modifier and in which the difficulty of ruling out an
activated relation is related to its level of activation.
5.2.2. Is Competition due to Inhibition or Facilitation?

The relational priming results described in Section 4.2 give rise to an inter-
esting question about the nature of the priming effect and thus about the
nature of relational competition effects in general. T he differences in response
times following the same-relation prime conditions vs the different-relation
prime conditions could have two sources. First, relation priming could be
because of the availability of the required relation being increased because
of recent exposure to a combination using that relation. Second, it could be
because of the different-relation prime becoming more available and thus
becoming a more successful competitor, which then slows the selection of
the required relation. The fact that the effect of relational availability seems
to be driven by strong competitors to the required relational interpretation
(Spalding & Gagné, 2008) suggests that one might expect that the relation-
priming effect would be primarily because of the activation of a compet-
ing relation by the prime rather than because of facilitation of the target
relation.
To examine this issue, Spalding and Gagné (2011) replicated the relation-
priming effects, but included a baseline modifier-only condition along with
the same-relation and different-relation conditions. For example, for the
target snowball (ball MADE OF snow), the baseline prime was snow, the
same-relation prime was snowfort (fort MADE OF snow) and the different-
relation prime was snowshovel (shovel FOR snow). Presenting only the mod-
ifier should activate the modifier’s relational distribution, but will not activate
any particular relation above its normal baseline within the distribution. In
all three conditions, the modifier is seen and thus this design controls for
priming because of lexical repetition from the prime to the target.
If the effect is pure facilitation, then the same-relation condition should
be fast and the different-relation condition and the baseline should both be
slow. If the effect is due purely to inhibition, then the different-relation con-
dition should be slow and the same-relation condition and baseline both
should be fast. If there is both facilitation and inhibition, then the same-
relation condition should be fast, the different-relation condition should be
slow, and the baseline should be in the middle. The results were consistent
with inhibition. In both experiments, recent presentation of a compound
using a different relation made it more difficult to interpret the target com-
pound, whereas responses in the same-relation and modifier-only prime
conditions did not differ.
5.2.3. Is There Competition among Alternative Interpretations?

In the previous sections, our operationalizations of relation competition
have been based on the relations’ use with individual constituents. However,
competition can also be defined in terms of alternative relational inter-
pretations at the level of the whole phrase/compound. According to the
RICE theory, language users must identify a particular interpretation when
a phrase or compound is encountered, but in doing so, multiple interpre-
tations are considered. These alternative interpretations influence ease of
processing. Gerrig (1989) measured the interpretation time of familiar
compounds (e.g. foot race) that were presented in a discourse context that
required either the conventional interpretation (e.g. “a race on foot”) or
an innovative interpretation (e.g. “a race the length of King Louis’ foot”).
The results indicated that the conventional meaning interfered with the
derivation of an innovative meaning, in that phrases that were more readily
interpreted with their conventional reading were more difficult to interpret
with the innovative reading.
While it may be unsurprising that an established meaning can interfere
with an innovative interpretation, by our view, an innovative interpretation
should also be able to interfere with an established meaning. Libben (1998)
presents a case in which such interference might be occurring. L ibben pro-
vides examples where a person with mixed aphasia produced the literal
meaning of opaque compounds, rather than the established meaning. For
example, the person paraphrased blueprint as “a print that is blue” and belly
button as “a button in your stomach”. Perhaps, this only occurs with aphasics;
perhaps, it only occurs because the established meaning is “lost” and there-
fore unavailable to the person.
However, it appears that non-impaired individuals also demonstrate
interference from innovative meanings: recent exposure to an innovative
meaning for a familiar compound increases the difficulty of selecting the
conventional meaning Gagné, Spalding, & Gorrie (2005). A familiar com-
pound (e.g. bug spray) was presented as part of a sentence that was consis-
tent with either the conventional reading (e.g. “Because it was a bad season
for mosquitoes, Debbie made sure that every time she went outside, she
wore plenty of bug spray”) or the innovative reading (e.g. “As a defense
mechanism against predators, the Alaskan beetle can release a deadly bug
spray”). Immediately after viewing the sentence, the participants saw the
phrase (e.g. bug spray) with a definition that was based on either the con-
ventional interpretation (e.g. “a spray for bugs”) or the innovative interpre-
tation (e.g. “spray produced by bugs”) and indicated whether the definition
was a plausible interpretation for the phrase. When the sentence used the
established meaning, the conventional definition was judged plausible 89%
of the time. However, when the sentence used the innovative meaning, the
conventional definition was judged plausible only 64% of the time. In terms
of response time, participants took longer to indicate that the conventional
reading was plausible when the sentence supported the innovative read-
ing than when it supported the conventional reading. We should note that
these results occurred even though the participants were told explicitly that
the definition did not have to be the best definition but that they should
indicate “yes” if the definition was plausible.Thus, the competition between
the relations was not restricted to a case in which the interpretations were
explicitly pitted against each other, as would be the case if the participants
were told to select the best definition. These findings indicate that compe-
tition from the innovative interpretation made it more difficult to accept
the conventional interpretations. This finding is highly consistent with the
fact that relation priming for established compounds is primarily a result of
increasing the availability of a relation other than the one required for the
target item (Spalding & Gagné, 2011), as previously discussed.
6. THE ROLE OF RELATION COMPETITION IN

THE PROCESSING OF COMPOUNDS THAT LACK
AN UNDERLYING RELATION
The investigations of familiar compounds described in Section 4.2
had, as their primary purpose, the demonstration that manipulations of
relation availability affect interpretation even of well-known compounds.
Beyond this specific purpose, the results strongly suggest that understand-
ing compound processing in general requires the application of what we
know about conceptual combination. In particular, the results above suggest
considering all compounds as being subjected to a meaning construction
process as a relatively obligatory aspect of comprehension.
Although the relation-priming results described above provide good evi-
dence for the meaning construction approach (and relational interpretation,
in particular) in compound word processing, opaque compounds provide

a particularly stringent test for the idea that meaning construction is rel-
atively obligatory for comprehension. Opaque compounds are those for
which the meaning cannot be computed from the meanings of the con-
stituents, as in, for example, the compound hogwash. Similarly, constituents
that contribute to the meaning of the compound (e.g. snow or ball in snow
ball) are semantically transparent, while those that do not (e.g. hog or wash in
hogwash) are semantically opaque. Critically, a person simply cannot deter-
mine the meaning of hogwash from hog and wash. Thus, such compounds
provide an extremely strong test of the idea that meaning construction
is attempted for all compounds. Such compounds also provide a strong
theoretical contrast between meaning construction approaches such as the
RICE theory and the more common conjunctive activation approach to
compound processing.
The meaning construction approach is currently a minority position
in the literature on compound processing as most current psycholinguistic
theories assume that compound processing proceeds primarily by a ctivating
the constituents’ stored representations, which are used to activate the com-
pound representation (e.g. Libben, 1998; Schreuder & Baayen, 1995, 1997;
Taft, 2003, 2004; Taft & Kougious, 2004; Zwitserlood, 1994; Zwitserlood,
Bolwiender, & Drews, 2005). That is, theories that rely on conjunctive acti-
vation posit that the meaning of a compound is retrieved via activation
from constituent representations, rather than actively computed via a mean-
ing construction process.
In general, activation approaches to compound processing account for
effects of semantic transparency by assuming that semantically opaque con-
stituents are unable to pass any activation to the semantic representation of
the constituent, while semantically transparent constituents are able to do
so. Thus, constituents may (or may not) contribute to the semantic access
of the compound to the extent that they pass activation to the compound’s
semantic representation. For example, in Libben’s (1998) theory, activation
of lexical representations of compounds results in either the activation (in
the case of transparent constituents) or the inhibition (in the case of opaque
constituents) of the corresponding semantic representations (see also Libben,
Gibson, Yoon, & Sandra, 2003). Zwitserlood (1994) proposed that there
are no links between the lexical representation of opaque constituent(s)
and the corresponding semantic representation, and, consequently, at the
semantic level, opaque compounds behave like monomorphemic words.
In sum, the conjunctive activation does not involve the active construction
of a compound’s meaning based on the constituents. Instead, the already-

existing semantic representation of the compound is activated via the con-
stituent representations. Thus, effects of constituents’ semantic transparency
on compound processing are explained by the spread of activation among
the stored representations of the constituents and the compound.
The RICE theory, on the other hand, suggests that the semantic rep-
resentation corresponding to an activated lexical representation is acti-
vated regardless of whether the constituent is transparent and that meaning
construction is always attempted. This approach, then, faces the burden of
explaining the existence of semantic transparency effects in compound pro-
cessing. We believe that the influence of the activated semantic representa-
tions of the constituents can be obscured because of conflict resulting from
the computed meanings based on these representations. As with novel com-
binations and transparent compounds, if the established meaning and the
computed meaning (or meanings) become available within the same time
frame, then the system must evaluate which meaning is most likely, and the
meanings are likely to compete with each other.
The research discussed above has demonstrated such competition
effects in the case of transparent compounds and phrases. In the case of
opaque compounds, by definition, the constructed meaning conflicts
with the conventional meaning, which should introduce processing costs
because this conflict must be resolved as the system attempts to settle
on one meaning. On average, there should be less competition between
the lexicalized meaning and the computed meaning for transparent com-
pounds than for opaque compounds because there is more consistency
among these meanings for transparent compounds (again, by the defini-
tion of opaque and transparent compounds). Ji et al. (2011) and Ji (2008)
found that opaque compounds were slower than matched transparent
compounds, consistent with the claim that opaque compounds face some
form of penalty in processing. Furthermore, Ji et al. found that experi-
mental manipulations that aided morphological decomposition (and,
thereby, aided semantic composition) slowed the processing of opaque
compounds, but did not slow the processing of transparent compounds.
Thus, the penalty faced by the opaque compounds appears to be one that
is specifically related to meaning construction. Additionally, the frequency
of the first constituent differentially influenced response time for trans-
parent and opaque compounds; the processing of transparent compounds
was helped by having a high-frequency first constituent, but opaque
compounds were hindered by having a high-frequency first constituent.
That is, the more available the constituent representation, the more
difficult it was to process an opaque compound, as would be expected if
the processing of opaque compounds entailed an attempt at meaning con-
struction that would interfere with the established meaning. In general,
these results are highly consistent with the idea that meaning construction
occurs for both opaque and transparent compounds. However, this process
produces a meaning that strongly conflicts with the conventional meaning
of opaque compounds.
Gagné and Spalding (submitted for publication) investigated whether
transparent and opaque compounds are sensitive to potentially competing
relational interpretations, as suggested by Ji et al. (2011).To answer this ques-
tion, Gagné and Spalding used the lexical decision data for transparent and
opaque compounds from Ji et al. and attempted to predict response time
using the diversity of relational interpretations that are possible for those
compounds. In this case, the potentially competing relations were assessed at
the level of the whole phrase using a possible relations task and using a com-
pletely different group of research participants. In the possible relations task,
participants are presented with a compound consisting of two words and are
asked to pretend that they are learning English and know each of the two
words, but have never seen the words used together. Then, they are asked
to select the most likely interpretation of the words when used together.
The participants select one relation from a list of relations derived from
Levi (1978). Relational diversity was computed for each compound simply
by counting the number of relations that were chosen by any participant
for a given compound. This relational diversity measure was a significant
predictor of lexical decision response time for both transparent and opaque
compounds. Thus, although there was no obvious reason for participants
in Ji et al.’s lexical decision task to be considering the possible relational
interpretations of the compounds, the ease of making the lexical decision
was affected by how many possible alternative interpretations exist for the
compound. In particular, the more possible relations there were, the slower
the lexical decision response time was, as predicted by the view that com-
peting relational interpretations slow processing. That such an effect occurs
even for opaque compounds (i.e. those compounds where the established
meaning cannot be computed from the constituents’ meanings) is strong
evidence that the meaning construction process is likely to be automatically
initiated when compound words are encountered. Furthermore, this mean-
ing construction process appears to be generally relational in nature so that
there seems to be a very strong continuity between the processing of novel
modifier-noun phrases and well-established compound words, as suggested

by the RICE theory.
7. EVALUATION OF RELATIONAL INTERPRETATIONS

Although the main focus of this chapter is the role of relational com-
petition in conceptual combination, it is also important to understand how
relational competition fits into the larger interpretive process. In this sec-
tion, we consider the evaluation stage of relational interpretation, and in the
next we consider elaboration. Most of the research reviewed to this point
would suggest that the relation availability for the head constituent has little
impact on relation selection. Indeed, this is what Gagné and Shoben (1997)
concluded. However, the RICE theory proposes that both semantic and
relational information associated with the head noun is used to evaluate the
relational interpretations suggested by the modifier (Spalding et al., 2010).
It is clear that semantic (and pragmatic) information associated with
the head affects relation selection. First, as initially suggested by Gagné and
Shoben (1997), semantic information associated with the head noun has a
strong influence on what relational interpretation is chosen for a combined
concept. Thus, the phrase mountain planet is not interpreted as a planet in
the mountains (despite the very strong likelihood that a combined concept
with mountain as the modifier will be interpreted using a locative rela-
tion) because planets are too big to be located in mountains. As Spalding
et al. (2010) pointed out (following Gagné & Shoben, 1997), nearly any
interpretation in which the relation is infrequent for the modifier is likely
to involve information (either semantic or relational) associated with the
head affecting the choice of relation. Otherwise, it is hard to see why one
would ever interpret a combined concept in a way that is low in frequency
for the modifier. Research specifically focused on the semantics of the head
(Spalding & Gagné, 2007) has shown that activating semantic properties
of the head in isolation can affect the relational interpretation chosen for
a combined concept. For example, noting that some machines are fragile
biases people to interpret clay machine as “a machine made of clay,” while
noting that some machines are sturdy biases people to interpret clay machine
as a “machine for clay”. In other words, activating semantic information
about the head that is consistent (or inconsistent) with a particular relational
interpretation makes that interpretation more (or less) likely compared
with activating semantic information about the head that is neutral with
respect to the relational interpretation (e.g. noting that some machines are
expensive). In some sense, these results should not be terribly surprising as

the relations that are used in relational interpretation are themselves largely
semantic in nature. Hence, prior activation of semantic information should
and does affect relation selection.
According to RICE, the lack of relational effects associated with the
head constituent in most previous research is because of the fact that tasks
such as the sense/nonsense judgment are largely driven by the modifier sug-
gesting multiple relations as possible interpretations. In fact, relational infor-
mation associated with the head has also been shown to affect the relational
interpretation of combined concepts. This has been shown, for example,
using relational priming of ambiguous combined concepts (e.g. brick factory
could be factory that makes bricks or a factory made of brick) by Gagné
and Shoben (2002), where the interpretation is a choice between two rela-
tively strong relational interpretations. However, unambiguous combined
concepts tend to show no effect of the relational distribution of the head
when the task is a sense/nonsense judgment (e.g. Gagné & Shoben, 1997).
Indeed, this asymmetry between the modifier and the head is one of the
most striking aspects of the early research on relational interpretation of
combined concepts (Spalding et al., 2010). Nevertheless, the head-based
priming effect with ambiguous combinations shows that head-based rela-
tional information does exist and can be used. The modifier-head asym-
metry appears to be largely and perhaps entirely because of the fact that
the sense/nonsense task is particularly sensitive to the modifier’s role as the
suggester of relations, probably because the sense/nonsense task requires the
complete process of interpretation of the phrase. Thus, the sense/nonsense
task results tend to strongly reflect what happens at the initial relation sug-
gestion phase of interpretation when the suggested relations are compet-
ing with each other on the basis of their strength for the modifier. If this
explanation is correct, then a task that isolates, or at least emphasizes, the
evaluation phase of relational interpretation should show relational effects
associated with the head.
Spalding et al. (2010) showed that the head’s relational distribution
strongly affects relational interpretation for unambiguous combined con-
cepts when the task is relation verification. In the relation verification task,
a combined concept is presented with its relational interpretation (e.g. snow
ball = ball made of snow) and the participants are asked whether this is a good
interpretation. Thus, when the task does not require the person to suggest
relational interpretations, but simply to verify that the presented relation
is appropriate, then the head’s relational distribution shows robust effects
(as does the modifier’s relational distribution). Spalding et al. also dem-
onstrated a robust head-based relational priming effect for unambiguous
combined concepts with the relation verification task. Finally, Spalding
et al. showed that the particular form of relational competition found in
the sense/nonsense task (i.e. the particular sensitivity to stronger competi-
tor relations, as discussed above) did not play a strong role in the relation
verification task. In other words, in the evaluation of an individual relation,
it is only the required relation’s strength (for both the modifier and the
head) that seems to play a role in relation verification response time. The
strength of other relations appears to play little role in the verification task.
We should note that this complex pattern of asymmetries between the
modifier and the head, and the differences among the different phases of
the relational interpretation process, are not consistent with any other exist-
ing theory of conceptual combination (see the discussion in Spalding et al.,
2010, pp. 304–307).
8. ELABORATION OF COMBINED CONCEPTS

FOLLOWING RELATION SELECTION
One major weakness in the literature on relational interpretation of
combined concepts has been the lack of detailed understanding of the pro-
cessing during interpretation.The RICE theory and the associated research
went some way toward remedying this situation with respect to the rela-
tion suggestion and verification phases, as well as elucidating the relation-
ship between novel conceptual combinations and familiar compounds.
The elaboration phase of RICE, which describes how we move beyond
the relation-based gist, however, is underdeveloped. Understanding this
elaboration process is critical to creating a complete theory of conceptual
combination and will also contribute to the understanding of noun phrase
comprehension in general.
If combined concepts are interpreted via the relational process described
in the RICE theory, then the elaboration must proceed from the gist. For
example, consider the combination snowman. If the gist is “man MADE OF
snow,” then the elaboration proceeds on that basis. But, if the gist is “man
FOR snow,” then the elaboration proceeds quite differently. A man made
of snow would be inanimate and white, while a man for snow would be
animate and have a shovel. On the other hand, if combined concepts are
elaborated via a feature transference process (such as is posited in schema-
based theories of conceptual combination; e.g. Wisniewski, 1997), then the
relational gist would seem to arise largely as an interpretation of the newly

created concept with its new features because otherwise there would be no
explanation for why the features are consistent with the relational gist. In
short, given that the features are consistent with the relational gist, it seems
very likely either that the features are derived from the gist or that the gist
is an interpretation of the newly created set of features, although it is pos-
sible that some other process gives rise to both features and gist. In terms of
the time-course of processing, data reported in Gagné (2000) indicate that
relation-based interpretations (e.g. mountain bird, which refers to a bird that
lives in the mountains) take less time than do property-based interpreta-
tions (e.g. robin canary, which refers to a canary that has a red breast). This
result strongly suggests that the inclusion of the features does not precede
the creation of the relational gist, which in turn suggests that elaboration
follows relation selection.The data also indicated that people show a prefer-
ence for using paraphrases that reflect the underlying relation, rather than
the properties of the newly formed concept. In both cases, it appears that
the elaboration of a combined concept proceeds from the relational gist
interpretation.
Two important points arise from this characterization of the elaboration
of combined concepts. First, the elaboration appears to involve a process of
reasoning from the interpretation, rather than a more mechanistic merging
or blending of features from the constituent concepts’ representations. This
view of elaboration fits well with recent work on the “modification effect”
(Connolly, Fodor, Gleitman, & Gleitman, 2007; Gagné & Spalding, 2011;
Jönsson & Hampton, 2008). In the modification effect paradigm, partici-
pants judge the likely truth of a feature for a modified noun (e.g. baby ducks
have webbed feet) or for the noun itself (e.g. ducks have webbed feet). The result
is that the judgments are lower for the modified noun than for the noun
(Connolly et al., 2007; Gagné & Spalding, 2011; Jönsson & Hampton, 2008).
While the empirical result is robust, three different theoretical explanations
have arisen. Connolly et al. claimed that the ratings are lower because the
features of modified nouns have to be learned from the world, rather than
inherited from the head noun. Jönsson and Hampton claimed that the fea-
ture was directly inherited from the head and then adjusted because it is
unlikely for the modifier. Gagné and Spalding showed that the effect occurs
for combinations with non-word modifiers (e.g. flag ducks), which does not
fit well with either explanation: the feature would never be learned for a
non-word combination, and the inheritance view predicts either perfect
inheritance or no inheritance at all. Gagné and Spalding propose that the
modification effect is driven by inferential processing at the point of deci-

sion about the feature and that the effect itself is because of participants’
expectation that the modified noun should be similar to, yet contrast with,
the noun: if there was no contrast, why modify the noun?
Second, one possible implication of such an understanding of elabora-
tion would be that judgments about features (even for unmodified con-
cepts) are also driven by processes other than simply inspecting the contents
of the concept. For example, we have shown (Gagné & Spalding, 2007) that
creating a relational interpretation of a combined concept can change the
likelihood of a feature for the head concept itself; interpreting the phrase
peeled apple makes it more difficult to affirm that apples are red. Note here,
that it is not, in general, a property of peeled things to remove redness—
instead, it is the interpretation (i.e. an apple that has been peeled), along
with the color of the apple beneath the peel, that gives the color of the
peeled apple. Even this, though, by most theories of conceptual combina-
tion, should not make it more difficult to verify that apples are red. After
all, the head noun’s conceptual representation should be unaffected and its
color feature intact. So, what is it that makes this verification more difficult?
Gagné and Spalding suggested that the effect is driven by the fact that the
interpretation of the phrase peeled apples causes the activation of represen-
tations of particular apples (namely, apples that have been peeled) and that
these highly activated apples, because they are not red, inhibit the verifica-
tion that apples in general are red.
These results on elaboration, in turn, might suggest that the relation
between concepts and features is potentially rather different than that in
most theories of concepts. In particular, it might support a view of features
closer to that of classical philosophical approaches to the relationship (in
particular, a relation of predication; Spalding & Gagné, in press) rather than
the kind of “containment” metaphor that is common in modern psychol-
ogy (see, e.g. the discussion in Laurence & Margolis, 1999). The point here
is that, in classical philosophical approaches, properties were predicated of
a term via reasoning and that while such predications could become well
remembered and habitual, this did not entail that the feature predicated
became part of the concept. The Aristotelian–Thomistic view of features,
for example, is much more like the type of theory identified by Laurence
and Margolis (1999) as connecting features to the concept via inference
(Spalding & Gagné, in press). In the Aristotelian–Thomistic view, features
(i.e. properties and accidents) are not part of the concept, per se. The fact
that most individual dogs have legs as parts does not mean that the concept
DOG has LEGS as a part, nor does the fact that people know that dogs have
legs mean that their concept of DOG must include LEGS as a part. Instead,
the features are themselves generally concepts linked to the (target) concept
in the act of predication. Aristotle often refers to this as combining and
dividing. Putting together two concepts (combining; e.g. dogs have legs)
or denying their combination (dividing; e.g. dogs do not have hard drives)
is a separate cognitive act from those involved in creating the concept or
in using the concept to represent the thing to the mind. The predication
can be remembered and can become habitual, but it is not part of the con-
cept. Critically, in such views, the properties predicated of concepts need
not combine in any mechanistic way, but are highly amenable to reason-
ing about whether the property would continue to be true of a modified
concept.
One point that has not yet been investigated much is the extent to which
elaboration is immediate and relatively automatic or neither immediate nor
automatic. The results of Gagné and Spalding (2007) that it becomes more
difficult to say that apples are red following a presentation of peeled apples
suggest at least some relatively immediate and automatic elaboration. On
the other hand, the results of Gagné and Spalding (2011) suggest that some
elaboration is not automatic and might only occur as needed (e.g. to answer
a particular question or to integrate with a new context). Furthermore, if
elaboration proceeds from the relational gist, then the elaboration of com-
bined concepts will presumably have strong similarity to other areas of text or
discourse processing. In particular, elaboration of combined concepts would
have to be highly similar to the processing of full noun phrases. As Wisniewski
(1996) points out, elaboration will also often involve selecting different senses
of modifiers (see also Mullaly, Gagné, Spalding, & Marchak, 2010) and prop-
erty terms, which will raise further complexities of processing. Clearly, the
elaboration of combined concepts is highly complex and is much less well
understood than relation selection. However, investigating elaboration may
provide important insight into a number of cognitive processes, including
reasoning, text or discourse processing and the relationship between concepts
and features. In short, this is a promising avenue for further research.
9. SUMMARY
This chapter reviews the recent literature on the relational interpreta-
tion of modifier-noun phrases and compound words. Section 1 argues for
the importance of conceptual composition processes in understanding how
the conceptual system as a whole operates. Section 2 presents evidence that

novel modifier-noun phrases and known compound words may be com-
prehended via a shared meaning construction process and thus that theo-
ries of conceptual composition should apply to both novel modifier-noun
phrases and known compounds.
In Section 3, we present a description of the RICE theory (summa-
rized in Figure 3.1), which is a theory of both modifier-noun phrase and
compound word processing, based on an underlying common process of
conceptual combination. RICE provides a theoretical framework for under-
standing relational interpretation of modifier-noun phrases and compounds
as a three stage process in which (1) relational interpretations are suggested
with a strength proportional to the relation’s availability for the modifier,
(2) these relational interpretations compete to be selected and are evalu-
ated with respect to both semantic information and relational availability
for both the modifier and the head noun, and (3) the selected relation is
elaborated beyond the relational gist interpretation.
Section 4 reviews the literature showing that the ease of comprehending
modifier-noun phrases and compound words is strongly affected by rela-
tional availability of the modifier. Such effects arise whether that availability
is based on long-term prior usage or recent usage, whether the phrases/
compounds are novel or well known, and whether they are presented in or
out of context.
Section 5 shows that the competition among relational interpreta-
tions takes on a specific character such that the primary determinant of
ease of processing is the number of competitors that are more available
(for the modifier) than the required relation. Consistent with this finding,
modifier-based relation-priming effects (i.e. repeated relations lead to faster
processing than unrepeated relations) are shown to result primarily from
competition from the different relation in the unrepeated condition, rather
than facilitation in the repeated relation condition. Finally, relational com-
petition effects are shown to occur at the level of the whole interpretation,
as well as at the level of the individual constituents.
Section 6 shows that relational competition effects can be seen even for
opaque compounds—those that lack an underlying relational interpreta-
tion entirely. For example, lexical decision RT’s for opaque compounds
can be (partially) predicted by the number of possible relational interpreta-
tions that the opaque compound could be construed as having. The more
such possible interpretations, the slower the lexical decision judgment, sug-
gesting competition among the possible interpretations and the established
meaning of the opaque compound. Such effects strongly suggest that the
relational interpretation process is more or less obligatory when modifier-
noun phrase or compound word structures are encountered.
In Section 7, we describe research that used a relation verification task
to isolate the evaluation process and focus it on a single relational inter-
pretation. This research demonstrates effects of the relation’s availability
for both the modifier and the head and contrasts these results with the
results of experiments using sense/nonsense judgments. This section shows
two critical points. One, although the effects of relational availability for
the head can be hidden in the sense/nonsense task, those effects are real.
Second, this result strongly suggests that the effect of multiple competitor
relations arises in the evaluation stage, even though the effects are related to
the relational availability for the modifier.
Section 8 provides a discussion of how the elaboration of the relational
interpretations might proceed. In particular, the idea that the elaboration
proceeds from the relational gist via reasoning processes is explored and
related to some other aspects of the literature on concepts.
10. CONCLUDING REMARKS

Understanding how people comprehend modifier-noun phrases
and compound words, and the combined concepts that underlie them,
is an important step to understanding the productivity that is a hallmark
of human cognition. It would seem that combining two words with a
relatively fixed syntactic structure should be straightforward. Indeed, it is
relatively simple from the perspective of the person doing the task. Within
about 1 s, the person has comprehended the phrase or compound, even
if it is completely new to them. Yet, this simplicity is also deceptively
complex. We have seen that the processing itself makes use of information
about the distribution of relations with each constituent, with the seman-
tics and pragmatics associated with each constituent, and with an even
broader range of information that can be brought to bear via reasoning
processes in (at least beginning) the elaboration beyond the relational gist
interpretation. Other reasoning processes are likely to arise in elaboration
well beyond the 1 s that it takes for comprehending the combination and
creating the relational gist. This seemingly simple use of the conceptual
and language systems provides an outstanding opportunity to learn about
human cognition.
REFERENCES
Allen, M. R. (1978). Morphological investigations. University of Connecticut. (Ph.D. Dissertation).
Bauer, L. (2008). Dvandva. Word Structure, 1, 1–20.
Bertram, R., Laine, M., & Karvinen, K. (1999). The interplay of word formation type, affixal
homonymy, and productivity in lexical processing: evidence from a morphologically
rich language. Journal of Psycholinguistic Research, 28, 213–226.
Bisetto, A., & Scalise, S. (2005). The classification of compounds. Lingue E Linguaggio, 4,
319–332.
Booij, G. (2005). Compounding and derivation: evidence for construction morphology. In
W. Dressler, D. Kastovsky, O. E. Pfeiffer & F. Rainer (Eds.), Morphology and its demarcations
(pp. 109–132). Amsterdam/Philadelphia: John Benjamins.
Connolly, A. C., Fodor, J. A., Gleitman, L. R., & Gleitman, H. (2007). Why stereotypes don’t
even make good defaults. Cognition, 103, 1–22.
Downing, P. (1977). On the creation and use of English compound nouns. Language, 53,
810–842.
Estes, Z. (2003). Attributive and relational processes in nominal combination. Journal of
Estes, Z., & Jones, L. L. (2006). Priming via relational similarity: a COPPER HORSE is
faster when seen through a GLASS EYE. Journal of Memory and Language, 55, 89–101.
Finin, T. W. (1980). The semantic interpretation of compound nominals. Urbana-Champaign:
University of Illinois. (Ph.D. Dissertation).
Fiorentino, R., & Poeppel, D. (2007). Compound words and structure in the lexicon.
Language and Cognitive Processes, 22, 953–1000.
Francis, W. N., & Kučera, H. (1982). Frequency analysis of English usage: Lexicon and grammar.
Boston: Houghton Mifflin.
Gagné, C. L. (2000). Relation-based combinations versus property-based combinations: a
test of the CARIN theory and the dual-process theory of conceptual combination.
Journal of Memory and Language, 42, 365–389.
Gagné, C. L. (2001). Relation and lexical priming during the interpretation of noun-noun
combinations. Journal of Experimental Psychology: Learning, Memory, and Cognition, 27,
236–254.
Gagné, C. L. (2002). Lexical and relational influences on the processing of novel compounds.
Brain and Language, 81, 723–735.
Gagné, C. L., & Murphy, G. L. (1996). Influence of discourse context on feature availability
in conceptual combination. Discourse Processes, 22, 79–101.
Gagné, C. L., & Shoben, E. J. (1997). Influence of thematic relations on the comprehension
of modifier-noun combinations. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 23, 71–87.
Gagné, C. L., & Shoben, E. J. (2002). Priming relations in ambiguous noun-noun combina-
tions. Memory & Cognition, 30, 637–646.
Gagné, C. L., & Spalding, T. L. (2004a). Effect of discourse context and modifier rela-
tion frequency on conceptual combination. Journal of Memory and Language, 50,
444–455.
Gagné, C. L., & Spalding, T. L. (2004b). Effect of relation availability on the interpretation
and access of familiar noun-noun compounds. Brain and Language, 90, 478–486.
Gagné, C. L., & Spalding, T. L. (2006). Relation availability was not confounded with
familiarity or plausibility in Gagné and Shoben (1997): comment on Wisniewski and
Murphy (2005). Journal of Experimental Psychology: Learning, Memory, and Cognition, 32,
1431–1437. discussion 1438–1442.
Gagné, C. L., & Spalding, T. L. (2007). The availability of noun properties during the
interpretation of novel noun phrase. Mental Lexicon, 2, 241–260.
Gagné, C. L., & Spalding, T. L. (2009). Constituent integration during the processing of
compound words: does it involve the use of relational structures? Journal of Memory and
Language, 60, 20–35.
Gagné, C. L., & Spalding, T. L. (2011). Inferential processing and meta-knowledge as the
bases for property inclusion in combined concepts. Journal of Memory and Language, 65,
176–192.
Gagné, C. L. & Spalding, T. L. Relation diversity for opaque and transparent English
compounds. In F. Rainer, W. Dressler, F. Gardani, H. C. Luschutzky (Eds.), Current issues
in linguistic theory. John Benjamins, submitted for publication.
Gagné, C. L., Spalding, T. L., Figueredo, L., & Mullaly, A. C. (2009). Does snow man prime
plastic snow? The effect of constituent position in using relational information during
the interpretation of modifier-noun phrases. Mental Lexicon, 4, 41–76.
Gagné, C. L., Spalding, T. L., & Gorrie, M. C. (2005). Sentential context and the interpreta-
tion of familiar open-compounds and novel modifier-noun phrases. Language and Speech,
48, 203–221.
Gagné, C. L., Spalding, T. L., & Ji, H. (2005). Re-examining evidence for the use of
independent relational representations during conceptual combination. Journal of

Gerrig, R. J. (1989). The time-course of sense creation. Memory & Cognition, 17, 194–207.
Gerrig, R. J., & Bortfeld, H. (1999). Sense creation in and out of discourse contexts. Journal
of Memory and Language, 41, 457–468.
Gleitman, L. R., & Gleitman, H. (1970). Phrase and paraphrase: Some innovative uses of language.
New York: Norton.
Inhoff, A. W., Radach, R., & Heller, D. (2000). Complex compounds in German: interword
spaces can facilitate segmentation but hinder assignment of meaning. Journal of Memory
and Language, 42, 23–50.
Ji, H. (2008). The influence of morphological complexity on word processing. Department of
Psychology, University of Alberta. (Ph.D. dissertation).
Ji, H., & Gagné, C. L. (2007). Lexical and relational influences on the processing of Chinese
modifier-noun compounds. Mental Lexicon, 2, 387–417.
Ji, H., Gagné, C. L., & Spalding, T. L. (2011). Benefits and costs of lexical decomposition
and semantic integration during the processing of transparent and opaque English
compounds. Journal of Memory and Language, 65, 406–430.
Jönsson, M. L., & Hampton, J. A. (2008). On prototypes as defaults (Comment on Connolly,
Fodor, Gleitman and Gleitman, 2007). Cognition, 106, 913–923.
Kay, P., & Zimmer, K. E. (1976). On the semantics of compounds and genitives in English.
In Proceedings of the sixth California Linguistics Association. San Diego: San Diego State
University.
Koester, D., Gunter, T. C., & Wagner, S. (2007). The morphosyntactic decomposition and
semantic composition of German compound words investigated by ERPs. Brain and
Language, 203, 64–79.
Laurence, S., & Margolis, E. (1999). Concepts and cognitive science. In E. Margolis &
S. Laurence (Eds.), Concepts: Core readings. Cambridge, MA: A Bradford Book, MIT Press.
Lees, R. B. (1960). The grammar of English nominalizations. Bloomington, IN: Indiana
University.
Levi, J. N. (1978). The syntax and semantics of complex nominals. New York: Academic Press.
Libben, G. (1998). Semantic transparency in the processing of compounds: consequences for
representation, processing, and impairment. Brain and Language, 61, 30–44.
Libben, G., Gibson, M., Yoon, Y. B., & Sandra, S. (2003). Compound fracture: the role of
semantic transparency and morphological headedness. Brain and Language, 84, 50–64.
Libben, G., & Jarema, G. (2006). The representation and processing of compound words. New York:
Oxford University Press.
Maguire, P., Devereux, B., Costello, F., & Cater, A. (2007). A reanalysis of the CARIN theory
of conceptual combination. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 33, 811–821.
Marchand, H. (1969). The categories and types of present-day English word-formation: A synchronic-
diachronic approach. Wiesbaden: Otto Harrassowitz.
Medin, D. L. (1989). Concepts and conceptual structure. American Psychologist, 44, 1469–1481.
Meyer, R. (1993). Compound comprehension in isolation and in context: the contribution of
conceptual and discourse knowledge to the comprehension of German novel noun–noun compounds.
Tübingen: Niemeyer.
Mullaly, A. C., Gagné, C. L., Spalding, T. L., & Marchak, K. A. (2010). Examining ambig-
uous adjectives in adjective-noun phrases: evidence for representation as a shared
core-meaning with sense specialization. Mental Lexicon, 5, 87–114.
Olsen, S. (2004). The case of copulative compounds. In A. ter Meulen & W. Abraham (Eds.),
The composition of meaning, from lexeme to discourses (pp. 17–37). Amsterdam: Benjamins.
Olsen, S., & Van der Meer, G. (2001). Coordination in morphology and syntax: the case of
copulative compounds. In A. ter Meulen (Ed.), Making sense: From lexeme to discourse.
Festschrift for Werner Abraham (pp. 87–101). Dorderecht: Kluwer.
Scalise, S., & Bisetto, A. (2009). The classification of compounds. In R. Lieber &
P. Stekauer (Eds.), Oxford handbook of compounding (pp. 34–53). Oxford: Oxford
University Press.
Schreuder, R., & Baayen, R. H. (1995). Modeling morphological processing. Hillsdale, NJ, England:
Lawrence Erlbaum.
Schreuder, R., & Baayen, R. H. (1997). How complex simple words can be. Journal of Memory
and Language, 37, 118–139.
Shoben, E. J. (1991). Predicating and nonpredicating combinations. In P. J. Schwanenflugal
(Ed.), The psychology of word meanings (pp. 117–135). Hillsdale, NJ: Erlbaum.
Solomon, K. O., Medin, D. L., & Lynch, E. (1999). Concepts do more than categorize. Trends
in Cognitive Sciences, 3, 99–105.
Spalding,T. L., & Gagné, C. L. (2007). Semantic property activation during the interpretation
of combined concepts. Mental Lexicon, 2, 25–47.
Spalding, T. L., & Gagné, C. L. (2008). CARIN theory reanalysis reanalyzed: a comment
on Maguire, Devereux, Costello, and Cater (2007). Journal of Experimental Psychology:
Learning, Memory, and Cognition, 34, 1573–1578.
Spalding, T. L., & Gagné, C. L. (2011). Relation priming in established compounds: facilita-
tion? Memory & Cognition, 39, 1472–1486.
Spalding, T. L., & Gagné, C. L. Concepts in Aristotle and Aquinas: implications for current
theoretical approaches. Journal of Theoretical and Philosophical Psychology, in press.
Spalding, T. L., Gagné, C. L., Mullaly, A. C., & Ji, H. (2010). Relation-based interpretations
of noun-noun phrases: a new theoretical approach. In S. Olson (Ed.), New impulses in
word-formation (Linguistische Berichte Sonderheft 17) (pp. 283–315). Hamburg: Buske.
Storms, G., & Wisniewski, E. J. (2005). Does the order of head noun and modifier explain
response times in conceptual combination? Memory & Cognition, 33, 852–861.
Štekauer, P. (2005). Meaning predictability in word formation: Novel, context-free naming units.
Amsterdam: John Benjamins Publishing Company.
Štekauer, P. (2006). On the meaning predictability of novel context-free converted naming
units. Linguistics, 44, 489–539.
Štekauer, P. (2009). Meaning predictability of novel context-free compounds. In R. Lieber
& P. Štekauer (Eds.), The Oxford handbook of compounding (pp. 272–298). Oxford, UK:
Oxford University Press.
Taft, M. (2003). Morphological representation as a correlation between form and
meaning. In E. Assink & D. Sandra (Eds.), Reading complex words: Cross-language studies
(pp. 113–137). Amsterdam: Kluwer Academic.
Taft, M. (2004). Morphological decomposition and the reverse base frequency effect.
Quarterly Journal of Experimental Psychology, 57, 745–765.
Taft, M., & Kougious, P. (2004).The processing of morpheme-like units in monomorphemic
words. Brain and Language, 90, 9–16.
Teall, F. (1892). English compound words and phrases. New York: Funk & Wagnalls.
Warren, B. (1978). Semantic patterns of noun-noun compounds. Goteborg: Acta Universitatis
Gothoburgensis.
Wisniewski, E. J. (1996). Construal and similarity in conceptual combination. Journal of
Wisniewski, E. J. (1997).When concepts combine. Psychonomic Bulletin and Review, 4, 167–183.
Wisniewski, E. J., & Murphy, G. L. (2005). Frequency of relation type as a determinant of
conceptual combination: a reanalysis. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 31, 169–174.
Zwitserlood, P. (1994). The role of semantic transparency in the processing and representa-
tion of Dutch compounds. Language and Cognitive Processes, 9, 341–368.
Zwitserlood, P., Bolwiender, A., & Drews, E. (2005). Priming morphologically complex
verbs by sentence contexts: effects of semantic transparency and ambiguity. Language and
Cognitive Processes, 20, 395–415.
CHAPTER FOUR
List-Method Directed Forgetting

in Cognitive and Clinical Research:
A Theoretical and Methodological
Review
Lili Sahakyan1, Peter F. Delaney, Nathaniel L. Foster, Branden
Abushanab
Department of Psychology, University of North Carolina at Greensboro, Greensboro, NC, USA
Contents
1. Introduction133
2. List-Method DF: Design and Measurement 134
2.1. Basic Design Issues 134
2.2. Multilist Designs in DF 136
2.3. Additional Dependent Measures in DF 137
3. Our Framework of List-Method DF 138
3.1. Older Ideas about the Causes of DF 138
3.2. A Two-Factor, Process-Based Account 139
4. Forgetting is a Strategic Decision 140
4.1. Importance of Controlled Strategies 141
4.2. Why do Some People Fail to Engage in DF? 142
4.3. Beliefs about Memory Developed during the Experiment 143
4.4. Preexisting Beliefs about Intentional Forgetting 144
4.5. Forgetting Strategies and Their Relative Effectiveness 146
5. Context Change as an Explanation for DF Impairment 148
5.1. Mental Context-Change Paradigm and Its Relation to DF 149
5.2. Recognition Tests and Indirect Tests of Memory 150
5.3. Mental Reinstatement of Context 153
5.4. Cuing with Context Cues vs Category Cues 154
5.5. Encoding Strength: Item vs Context Strength 155
5.6. Contextual Variability 156
5.7. Extralist Cued Recall 157
5.8. Importance of L2 Learning 158
5.9. Related Items Across Lists 159
5.10. Retrieving L2 before L1 160
5.11. Delay Effects 161
5.12. Individual Differences in DF as a Function of Working Memory Capacity 162
5.13. Reexposure or Retrieval of L1 Items Affects Subsequent Memory 162
5.14. Summary164
132 Lili Sahakyan et al.
6. Areas of Disagreement Across Studies 165

6.1. Serial Position Effects 165
6.2. List Differentiation in DF: Source vs Intrusion Errors 166
7. Strategy Change Explains DF Benefits 167
7.1. The Advent of the Dual-Process Account 167
7.2. A Large-Scale Replication Study 169
7.3. Test Order and the Benefits of DF 170
7.4. Summary170
8. Implications for Clinical Populations 172
8.1. Reinterpreting Clinical List-Method DF Results as Reflecting Context 173
8.1.1. Attention-Deficit Hyperactivity Disorder 173
8.1.2. Schizophrenia 174
8.1.3. Generalized Anxiety Disorder 175
8.1.4. Depression 175
8.2. F uture Directions for Clinical Research Using the List Method 176
8.3. Improving Research Design for Clinical Studies of List-Method DF 178
8.3.1. Inclusion of a Separate Remember Condition 178
8.3.2. Controlling Test Order and Measuring Output Order 179
8.3.3. Scaling Effects 180
8.3.4. Floor Effects 180
8.4. S ummary and Linking Clinical to Nonclinical Research on DF 181
9. Concluding Thoughts 181
Acknowledgments182
References182
Abstract
The primary purpose of this chapter is to provide an up-to-date review of the
twenty-first century research and theory on list-method directed forgetting (DF) and
related phenomena like the context-change effect. Many researchers have assumed
that DF is diagnostic of inhibition, but we argue for an alternative, noninhibitory
account and suggest reinterpretation of earlier findings. We first describe what DF is
and the state of the art with regard to measuring the effect. Then, we review recent
evidence that brings DF into the family of effects that can be explained by global
memory models. The process-based theory we advocate is that the DF impairment
arises from mental context change and that the DF benefits emerge mainly but perhaps
not exclusively from changes in encoding strategy. We review evidence (some new to
this paper) that strongly suggests that DF arises from the engagement of controlled
forgetting strategies that are independent of whether people believed the forget cue
or not. Then we describe the vast body of literature supporting that forgetting strategies
result in contextual change effects, as well as point out some inconsistencies in the DF
literature that need to be addressed in future research. Next, we provide evidence—
again, some of it new to this chapter—that the reason people show better memory
after a forget cue is that they change encoding strategies. In addition to reviewing
List-Method Directed Forgetting in Cognitive and Clinical Research 133
the basic research with healthy population, we reinterpret the evidence from the litera-
ture on certain clinical populations, providing a critique of the work done to date and
outlining ways of improving the methodology for the study of DF in special populations.
We conclude with a critical discussion of alternative approaches to understanding DF.
1. INTRODUCTION
Forgetting is often a passive process that happens regardless of our con-
scious intentions. However, this is not always the case; forgetting sometimes
is an active process of intentionally engaging strategies that reduce memory
for material that interferes with one’s current goals. To this end, forget-
ting can be an adaptive process. For example, to use software that has been
recently updated, the user must forget the prior ways of doing things to
avoid errors. Similarly, we are often told “never mind, forget what I just said”
following exposure to erroneous information.
Intentional forgetting has been studied extensively in the laboratory via
directed forgetting (DF) manipulations, which produce impaired mem-
ory for material following an instruction to try to forget. In DF studies,
participants study some information and are subsequently told to forget
certain portions of it. The list-method instructs people to forget an entire list
of earlier studied items, whereas the item-method instructs people to forget or
remember on item-by-item basis. Both methods demonstrate forgetting of
unwanted information on demand, although the way by which forgetting
is accomplished differs between procedures.
This chapter primarily centers on list-method DF, partly because the
mechanisms proposed to explain it have been a topic of active debate in the
recent decade, and partly because the presumed connection with inhibitory
processes has made the paradigm an attractive diagnostic tool for investigat-
ing clinical and special populations.We provide an up-to-date review of DF
and related phenomena (e.g. the context-change paradigm) with a focus on
studies conducted since 2000. There is already an excellent edited volume
describing twentieth century DF research (Golding & MacLeod, 1998).
We will begin by describing methodological issues in list-method DF.
The designs used to study DF in cognitive research have evolved to keep
pace with theoretical and methodological developments, and so we will
review the difficulties posed by earlier designs and suggest better-controlled
procedures. We will then outline our broad theory of DF and describe the
large number of studies supporting that broad theory, as well as areas of cur-
rent controversy in DF. Even researchers who are familiar with our previous
manifestos on this topic may find this section interesting, as our view has
evolved and expanded even since 2010, and because we will present pre-
viously unpublished data that bear on a number of issues. Finally, we will
explore how the theory can inform studies of individual differences, espe-
cially in clinical populations, and make recommendations for conducting
such studies.
2. LIST-METHOD DF: DESIGN AND MEASUREMENT

2.1. Basic Design Issues
DF is generally assessed in terms of two separate components of the effect:
the costs and the benefits.To understand these terms, we first need to explain
how most DF studies are set up, and then we will explain the usual findings.
The most frequently used design is the two-list design, which presents
two lists of items (usually words) to study for a later memory test. The first
list we will call L1 and the second list we will call L2. Following L1, a cue is
presented either to keep remembering L1 or to forget L1. Because in the
two-list design the cue always occurs between L1 and L2, we sometimes
refer to L1 as the precue items and L2 as the postcue items, which is helpful
when considering designs that include more lists (see below for a discussion
of these designs). There are a variety of ways to present the forget cue, but
a typical instruction is to try to forget L1 because it was “just for practice”.
The remember cue provides a control group against which to assess the
effects of the forget instruction, and it typically involves telling people to
keep remembering L1 because it was only the first half of the study material.
A filler task is usually inserted after learning L2 to reduce recency effects.
Afterward, participants receive a recall test for both lists, including items that
they were told to forget. In free recall, the forget instruction reduces recall of
L1 items relative to the remember group—a phenomenon called the costs of
DF. Similarly, the forget instruction produces better memory of L2 items rel-
ative to the remember group—a phenomenon called the benefits of DF (for
reviews, see Bäuml, 2008; Bjork, Bjork, & Anderson, 1998; MacLeod, 1998).
Whether participants are instructed to recall L1 or L2 first varies across
studies. Some studies have required participants to recall L1 first and then
recall L2 (e.g. Delaney & Sahakyan, 2007; Foster & Sahakyan, 2011; Mulji &
Bodner, 2010; Pastötter & Bäuml, 2007; Spillers & Unsworth, 2011). Other
studies have fully counterbalanced the test orders (e.g. Kimball & Bjork,
2002; Lehman & Malmberg, 2009, 2011a; Pastötter, Kliegl, & Bäuml, 2012;
Sahakyan & Delaney, 2010; Zellner & Bäuml, 2006), and some studies
have allowed participants to recall all items in any order (Golding & G ottlob,
2005; Joslyn & Oakes, 2005; Minnema & Knowlton, 2008; Wessel &
Merckelbach, 2006; Zellner & Bäuml, 2006).
One obvious concern is that differences in output order might affect
DF through output interference (Anderson, 2005). Output interference is
the finding that retrieving some items from a set first reduces the likelihood
of retrieving the rest of the set (Dong, 1972; Roediger, 1974; Roediger &
Schmidt, 1980; Tulving & Arbuckle, 1966). In DF, when test order is not
controlled people might be inclined to begin recall with L2 items due to
recency, causing output interference on L1 items. Given that L2 is better
remembered in the forget group than the remember group, there may be
greater output interference on L1 in the forget group. Indeed, when left
to recall lists in any order, participants prefer to initiate recall with postcue
remember items, thereby resulting in an effect of output order (Golding &
Gottlob, 2005). However, a recent meta-analysis of list-method DF stud-
ies that included both forget and remember groups did not support this
notion; initiating retrieval with L2 did not exaggerate the magnitude of
DF costs, although it made it easier to detect the DF benefits (Pastötter
et al., 2012). Although these results suggest that output order affects the
benefits more than the costs (at least with unrelated items across the lists),
we nevertheless suggest fully controlling the test order of each list to obtain
more “pure” measures of recall uncontaminated by prior retrieval. Output
interference could become a particularly thorny issue if the lists contain a
mixture of items (e.g. neutral and emotional words), which could create
different degrees of output interference if the test order is left uncontrolled.
Another procedural inconsistency across DF studies is whether a control
group is included in the design. In the past, many DF paradigms have failed
to include a control group wherein both L1 and L2 presentations are fol-
lowed by a remember cue. In such paradigms, the magnitude of forgetting
was quantified by comparing recall performance for the forget list (L1)
relative to the remember list (L2). This measure is referred to as the “R–F”
measure. The justification for not including a control group is based on the
notion that the only ostensible difference between L1 and L2 is the type of
cue received. However, there are several confounding variables that high-
light the necessity for a control group. First, heightened recall performance
for L2 items could be attributed to a recency effect, making it difficult to
assume that L2 gains are a result of L1 forgetting. Second, increased acces-
sibility for L2 items may be attributed to learning effects. That is, following
L1 exposure, participants may be more adapted to the procedure, leading
them to shift their encoding strategies for L2 items. Finally, reduced recall
performance for L1 items may be attributed to retroactive interference
accrued from L2 items. In other words, at the time of recall, L1 accessibility
is reduced due to the interference built up from L2 encoding. Given the
number of alternative explanations for reduced L1 recall and heightened
L2 recall, it is challenging to attribute these findings solely to DF with-
out comparing recall performance to a control group that was required to
remember both lists.
2.2. Multilist Designs in DF

A variation on the traditional two-list procedure is the use of a three-list
design, which was put forth to reduce potential confounds in the two-list
procedure (Lehman & Malmberg, 2009). The sources of these confounds
are as follows. First, L2 presentation is typically followed by a filler task,
whereas L1 presentation is not, which means that L2 memory could be
penalized in comparison with L1. For example, participants can covertly
rehearse the end of L1 during the beginning of L2 encoding, increasing
memory for the last few L1 items, while hindering encoding of the first
few L2 items. Additionally, in a traditional two-list design, L2 is subjected
to proactive interference from L1, but L1 receives no interference from a
prior list. Therefore, the two-list approach may benefit L1 and harm L2.
To control these factors, in a three-list design, participants study three lists,
each followed by a brief filler activity of equal duration before a forget or
a remember cue is presented. For half of the participants, the forget cue
is presented after L2, whereas the remaining half are told to remember all
lists. At the time of final test, half of the participants are required to recall
either L2 or L3 (L1 is never tested). Note that the three-list design could
lead to potential floor effects on L2 if presentation rates are not substantially
increased (e.g. Lehman & Malmberg, 2011a). Both the two-list and the
three-list designs manipulate the forget/remember cue between-subjects.
An alternative approach is the use of a four-list design, where the forget/
remember cue is varied within-subjects (e.g. Zellner & Bäuml, 2006). In this
procedure, two lists are assigned to the first block, and two lists are assigned
to the second block. The forget cue occurs either during the first block
(e.g. after presentation of L1) or during the second block (e.g. after pre-
sentation of L3). Participants complete the forget condition (F, R) for the
first block and the remember condition (R, R) for the second block, or
vice versa. Memory for both lists is tested after each block, with counter-
balanced test order of the lists. This procedure produces findings similar
to the traditional two-list version. One caveat concerns a potential shift

to better encoding strategies after the first block has been completed
and memory for both lists has been tested (Delaney & Knowles, 2005;
Sahakyan & Delaney, 2003). With that in mind, there may be some order
effects depending on whether forgetting is inferred from the RR–FR
order or the FR–RR order. For example, the size of the DF impair-
ment may be exaggerated in the FR–RR order because by the time of
L3, participants may adopt better encoding strategies on L3. Although
Bäuml et al. typically do not obtain session order effects, faster presenta-
tion rates that are employed in some of their studies may mask such effects
because with faster rates there is leaving little room for employing a bet-
ter encoding strategy than rehearsal. An order effect has been reported
in one study that used a slow presentation rate and obtained larger DF
when the forget cue occurred in the first block compared to the third
block (Hanczakowski, Pasek, & Zawadska, 2012). Therefore, researchers
should be mindful of potential order effects that may arise from use of the
four-list, within-subjects approach to DF.
2.3. Additional Dependent Measures in DF

DF has been assessed primarily in terms of the number of correctly recalled
or recognized items from each list (e.g. costs/benefits). In its early days, it
was inferred through an even less-sensitive measure (e.g. “R–F” difference).
However, recently researchers have strived to include a wider array of depen-
dent measures with the goal of constraining the interpretations. Such mea-
sures include investigations of serial position functions (Geiselman, Bjork, &
Fishman, 1983; Lehman & Malmberg, 2009; Pastötter & Bäuml, 2010;
Pastötter et al., 2012; Sahakyan & Foster, 2009; Sheard & MacLeod, 2005),
intrusion errors (Lehman & Malmberg, 2009; Sahakyan & Delaney, 2010;
Spillers & Unsworth, 2011), recall latencies, which indicate an average
time point in the recall period when responses were emitted (Spillers &
Unsworth, 2011; Unsworth, Spillers, & Brewer, 2012), first response func-
tions, which provide a measure of where in the list participants initiate their
recall (e.g. Lehman & Malmberg, 2009), and conditional response prob-
abilities, which indicate how people transition between responses during
recall (e.g. Unsworth et al., 2012). The use of multiple measures provides a
more detailed assessment of the recall process, thereby facilitating a deeper
understanding of the mechanisms underlying list-method DF. Together,
they coalesce to provide a broader perspective on list-method DF, which in
turn can be used to guide and advance future research.
3. OUR FRAMEWORK OF LIST-METHOD DF

3.1. Older Ideas about the Causes of DF
The earliest theory of DF was the selective rehearsal account (e.g. Bjork, 1970,
1972), which was proposed at the time when not much was known about
the distinction between the list-method and item-method DF. According to
this view, the costs arise because participants stop rehearsing to-be-forgotten
items in response to the forget cue, and they devote their rehearsal and mne-
monic activity more effectively to to-be-remembered items, producing the
benefits. This works well in explaining item-method DF, but its suitability
for explaining list-method DF started to wane when it became clear that
the list-method and item-method were differentially sensitive to recogni-
tion and indirect tests of memory (e.g. Basden, Basden, & Gargano, 1993;
MacLeod, 1999). Additionally, the presence of significant DF in incidental
learning (Geiselman et al., 1983; Sahakyan & Delaney, 2005, 2010; Sahakyan,
Delaney, & Goodmon, 2008) made it increasingly difficult for the selective
rehearsal account to encompass list-method DF, where it was soon supple-
mented by the retrieval inhibition account (Bjork, 1989; Geiselman et al., 1983).
According to the retrieval inhibition account, when participants are
instructed to forget L1, they initiate an inhibitory process that suppresses or
deselects that list so as to facilitate the learning of subsequent lists.Therefore,
L1 memory suffers from inhibition, and L2 memory benefits because inhib-
ited L1 items do not cause proactive interference on L2.
The term inhibition is interpreted differently among DF researchers.
Some interpret inhibition as suppressing the activation level of L1 items
(e.g. Barnier, Conway, Mayoh, Speyer, Avizmil, & Harris, 2007; Conway,
Harries, Noyes, Racsmány, & Frankish, 2000; Racsmány & Conway, 2006;
Racsmány et al., 2008); others propose that items reside in memory at their
full strength as evidenced by indices of memory other than free recall, but
retrieval of the L1 episode is inhibited (e.g. Bjork, 1989; Bjork & Bjork,
1996), and recently some have argued that the L1 context may be inhib-
ited (e.g. Anderson, 2005; Bäuml, 2008; Pastötter & Bäuml, 2010). Multiple
meanings of the term inhibition can be misleading, especially to research-
ers outside of the memory field who employ the DF paradigm as a test of
inhibitory function in various populations. Equating an empirical pattern of
findings with an underlying inhibitory mechanism is unwarranted because
mechanisms other than inhibition could lead to impaired access (e.g. context
change and blocking).
Historically, in the mid-to-late 1980s, inhibition was used to explain the

discrepant findings in item-method and list-method DF. However, noth-
ing was assumed about the nature of inhibition or the conditions under
which inhibition should be more or less evident. The unspecified nature of
inhibition makes it difficult to think of it as a mechanism that explains DF
as opposed to providing a description of empirical findings. Later, in the
1990s, inhibition was developed into a full theoretical mechanism explain-
ing retrieval-induced forgetting (RIF) (e.g. Anderson, 2003; Anderson,
Bjork, & Bjork, 1994), which is a memory impairment phenomenon that
occurs at the item-level, unlike the list-method DF, which occurs at the
list-level. Nowadays inhibition account makes testable predictions for RIF,
but it fails to make predictions for DF because it is unclear what exactly
is being inhibited in DF. Just about any outcome can be interpreted to be
consistent with inhibition, which significantly reduces its appeal as a theory
of DF. That DF is frequently reviewed together with other inhibitory para-
digms (e.g. RIF and think-no-think procedure) may have contributed to
inadvertent blending of the use of the term inhibition across paradigms (e.g.
Anderson, 2003, 2005; Depue, 2012).
We take issue with the notion that the “L1 episode” is suppressed as a
result of the forget cue. Someone not well-versed in the nuances of mem-
ory theory is likely to interpret “inhibition of L1 episode” as analogous to
“inhibition of items within that episode”. However, the L1 episode con-
tains more than just the studied items within that list. It contains contextual
information, which distinguishes that particular episode from other similar
episodes (e.g. the L2 episode).Thus, an episode is a complex term that refers
to both the items and the context within which those items were experi-
enced. We argue that there is no evidence for inhibition of items, and we
further argue that there is no evidence for inhibition of context. Instead, we
explain the impaired accessibility to L1 items by invoking the notion of a
mismatch between the retrieval cues being used to search the memory and
the contents of memory.
3.2. A Two-Factor, Process-Based Account

Figure 4.1 shows how we currently understand list-method DF. Assuming
a two-list design, the forget cue takes place after L1 and triggers two types
of reflective processing. First, it triggers a decision about how to comply
with the forget instruction. We argue that the primary mechanism produc-
ing forgetting involves a change of mental context between the two lists,
and that some strategies (but not all) bring about this contextual change.
Figure 4.1 A two-factor, process-based framework for list-method directed forgetting.
Context change also contributes somewhat to the benefits by reducing

proactive interference on L2, but this effect is often small and difficult to
detect, particularly when there are only two lists. Second, the forget cue
stops ongoing rehearsal processes on L1, which gives participants time to
think about what they should do to learn L2. This leads some participants
to change study strategies and consequently improve L2 learning. We think
that this strategy-change component accounts for the improved memory
for postcue material.
Sections 4–6 provide the rationale for each of these mechanisms and the
empirical support for them. Section 4 discusses the importance of controlled
strategies for obtaining DF, and how metacognitive beliefs affect the deci-
sion to deploy those strategies. Section 5 details the mental context-change
account and Section 6 explains the strategy-change account. We will there-
fore wait to explain in detail why we think these mechanisms provide a
better explanation for list-method DF phenomena than earlier accounts do.
4. FORGETTING IS A STRATEGIC DECISION

In a series of studies from our lab, we have explored what people do
in response to the forget cue. Specifically, we have accumulated a wealth of
information regarding how participants interpret and comply with forget
instructions. People may think that they already have forgotten, and there-
fore fail to deploy any strategy to forget. Likewise, they may not know what
kind of strategy to deploy in order to forget, and therefore do nothing.
Such results are important not only for demonstrating that forgetting is an
active process that requires conscious activity in order to forget, but also for
explaining individual differences—including between-group differences—
in the ability to forget unwanted information.
4.1. Importance of Controlled Strategies

Many studies in our lab have revealed that obtaining DF requires that par-
ticipants engage in controlled behaviors to reduce accessibility of unwanted
information. For example, Foster and Sahakyan (2011) collected retrospec-
tive verbal reports regarding what participants did in response to the for-
get instruction, including whether they believed/trusted the forget cue. In
a group of 80 forget-group participants, 25% reported “doing nothing”,
whereas the remaining participants reported a variety of strategies, which we
collectively termed as “doing something”. Nearly 34% of the forget-group
participants reported “not believing” the forget cue. Surprisingly, however,
there was virtually no correlation (φ = 0.08) between forget-cue belief and
whether participants engaged in controlled strategies in order to forget. Some
participants trusted the forget cue but did nothing in response to it, whereas
others engaged in forgetting strategies despite not trusting the cue. It seemed
like the latter group decided to “play along” despite having doubts about the
truthfulness of the forget cue. Most importantly, L1 recall in the forget group
was entirely determined by whether participants engaged in controlled strat-
egies or not, whereas believing in the forget cue was completely irrelevant.
When the forgetting strategy (doing something vs nothing) and the belief
in the forget cue (yes vs no) were simultaneously entered as predictors of L1
recall in the forget group, only the strategy variable was a significant predic-
tor; the belief variable was not. The results are summarized in Figure 4.2.
Instructions to forget in the real world are sometimes explicit. For exam-
ple, if construction is blocking a geographical route, one must temporarily
“forget the original route” in favor of a more contextually appropriate route
(e.g. Golding & Keenan, 1985). However, an implicit instruction to forget
may be just as relevant, such as forgetting a prior credit card number to
facilitate memory for a recently issued card. We compared the effectiveness
of explicit and implicit forget cues and showed that the magnitude of DF is
unaffected by the nature of the forget instruction (Foster & Sahakyan, 2011).
Regardless of whether forgetting comes from implicit needs of the environ-
ment, or explicit prompts of others, it leads to similar behavioral effects.
These findings highlight several key points about DF and suggest new
directions for future research. First, similar to prior research, which shows
0.45
Remember Forget-did something Forget-did nothing
0.40
0.35
Proportion list 1 recall
0.30
0.25
0.20
0.15
0.10
0.05
0.00
"yes" "no"
Believing the forget cue
Figure 4.2 Directed forgetting impairment as a function of belief in the cue and
controlled forgetting strategies. (New figure based on the data reported in Foster and
Sahakyan (2011)).
that intention to learn, per se, is not important for remembering, but rather
what people do to learn the information is what matters (Hyde & Jenkins,
1973; Mandler, 1967; Postman, 1964), the findings of Foster and Sahakyan
(2011) demonstrate that merely believing the forget cue is not sufficient
for demonstrating DF; what matters is what people do in response to the
forget cue (see also Mulji & Bodner, 2010). Second, because believing the
forget cue is not critical for obtaining DF, future research should be less
concerned about staging an event to make the forget cue more b elievable
(e.g. simulating a computer crash in the middle of the experiment), and
should instead be more vigilant about reporting whether participants
engaged in DF or not. For example, the absence of DF in certain condi-
tions or certain populations could simply be linked to “doing nothing”.
Finally, u
nderstanding why some participants engage in DF whereas others
do not should become a priority for future research as it will help to pro-
vide insights for understanding DF in special populations.
4.2. Why do Some People Fail to Engage in DF?

Are “do nothing” participants simply unmotivated to be in an experiment?
If so, the remember group should recall much more than they do. In fact,
though, the “do nothing” people show recall levels comparable to that of
the remember group. Maybe some people simply cannot formulate a strat-
egy in a brief period of time between the lists. Most experiments deliver
the forget cue and move on to the second list right away. Indeed, some of
the participants in our experiment reported “I didn’t do anything in particular,
there wasn’t enough time,” or “I wanted to forget but didn’t know how to forget,
so I didn’t do anything,” suggesting that extending the postcue time might
increase the chances of obtaining DF or produce larger-than-normal effects.
These reports also indicate that one may want to forget but be unable
to formulate a strategy. It is essential to be mindful of these issues when
interpreting DF performance among special populations, who may have
difficulty self-initiating a controlled strategy.
Metacognitive beliefs could also play a role in why some participants fail
to engage in DF. For example, if participants have a preexisting belief that
their memory is not good even before they attempt to memorize the mate-
rial in an experiment, then they may be less likely to engage in DF because
they may believe that they already forgot. Indeed, a study in our lab with
older adults revealed this problem (Sahakyan, Delaney, & Goodmon, 2008).
Many older adults, in response to the forget cue, spontaneously volunteered
that they had already forgotten. Not surprisingly, they did not engage in
DF. We consequently modified DF instructions that emphasized the need
to engage in forgetting despite the beliefs that they already forgot, and we
found that such instructions were more conducive for obtaining DF with
older adults than were the standard instructions.
4.3. Beliefs about Memory Developed during the Experiment

Participants could also develop beliefs about their memory during the exper-
iment as they gain experience memorizing the materials. Such beliefs might
also play a role in whether they engage in intentional forgetting. For e xample,
if some participants do not feel confident about their ability to remember
L1 after attempting to memorize it, they may see no need to engage in
DF. Support for this hypothesis comes both from older adults (Sahakyan,
Delaney, & Goodmon, 2008) and from a study involving mainly college
adults (Sahakyan, Delaney, & Kelley, 2004). After presenting L1, we asked
college participants to predict how many words from that list they would
be able to recall during the final test (i.e. an aggregate judgment of learning
( JOL)). Participants’ predictions ranged from the minimum of 20% (equiva-
lent to three words) all the way to 80% (equivalent to 12 words). After
providing the JOLs, they received a forget or remember cue, followed by L2.
Figure 4.3 Directed forgetting impairment as a function of cue and aggregate predic-
tions of list 1 learning. (New figure based on the data reported in Sahakyan, Delaney, and
Kelley (2004)).
Figure 4.3 shows L1 recall during the final test as a function of JOLs
and the cue. As the figure demonstrates, DF impairment was larger when
participants anticipated recalling many L1 items than when they antici-
pated recalling few L1 items. Although we did not collect forgetting strat-
egy reports in that study, it is possible that the more people expected to
remember from the first list, the more likely they were to engage in DF. In
contrast, if they felt they could not remember L1 that well, they may have
been less likely to engage in DF. Individual differences in JOLs could be
correlated with other variables that are involved in this effect (e.g. higher
intelligence or working memory capacity), but these data are consistent
with the idea that beliefs about memory could contribute to list-method
DF (for a demonstration of how item-by-item JOLs affect item-method DF,
see Foster & Sahakyan, 2012).
4.4. Preexisting Beliefs about Intentional Forgetting

In an unpublished study, Foster and Sahakyan investigated whether pre-
existing beliefs about whether it is possible to intentionally forget some-
thing might play a role in who engages in DF and who does not. People
who have beliefs that it is impossible to forget something after it has been
learned might be less likely to attempt DF. W
e asked participants who were
participating in a departmental mass screening “Do you think it is possible to

make yourself forget something after you have learned it?” The answer required a
yes/no response. Out of 130 participants who responded to this question,
49% indicated “yes”, and 51% indicated “no”. These participants were later
invited to take part in a DF experiment in our lab (they were unaware of
the nature of the experiment). At the end of the experiment, they indi-
cated what they did in response to the forget instruction. Their responses
were coded as “doing something” or “doing nothing” (similar to Foster &
Sahakyan, 2011). Replicating previous studies, 29% of the forget group
reported “doing nothing”. Interestingly, we found no correlation between
peoples’ beliefs about whether it is possible to intentionally forget and
whether they engaged in DF (φ = −0.04). Most importantly, the magnitude
of DF impairment was virtually identical among those who believed it is
possible to intentionally forget and those who did not believe it is possible.
What mattered for DF was whether participants engaged in deliberate strat-
egies to forget, replicating previous work of Foster and Sahakyan (2011).
The results are shown in Figure 4.4.
Overall, studies from our lab have consistently found that DF does not
emerge among participants who report “doing nothing” in response to the
0.45
Forget-did something Forget-did nothing Remember
0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0.00
"It is possible to intentionally forget" "It is NOT possible to intentionally forget"
pre existing beliefs

Figure 4.4 Directed forgetting impairment as a function of preexisting beliefs about
intentional forgetting and controlled forgetting strategies. (Based on unpublished data
collected by Foster and Sahakyan.)
forget cue. Engaging in DF does not appear to be linked to whether par-

ticipants trusted the forget cue, or whether they think it is possible to forget
things once they have been learned. A better understanding of why some
healthy college adults do not engage in DF is warranted, because it could
help explain DF performance in special populations. Failures to obtain DF
with certain populations may not necessarily reflect an inability to inhibit.
Instead, they may indicate that certain populations may select to do “noth-
ing” in response to the forget cue, but the reasons for such decisions could
be more complex and not necessarily linked to their ability to purposefully
forget. They may be unable to formulate a strategy, or they may not see a
need to engage in forgetting.
4.5. Forgetting Strategies and Their Relative Effectiveness

In recent years, Nathan Foster and Lili Sahakyan collected many strategy
reports from participants in list-method DF studies. We did not analyze
them in published papers due to smaller sample sizes. For this review, we
combined several studies to compile a dataset of 290 participants (with 145
in the forget group and 145 in the remember group) to investigate whether
DF depends on forgetting strategy or not. In each study, participants studied
15 unrelated nouns per list at 4 s/item, with the remember or forget cue
between the lists, but no other break/delay. A short filler task followed L2,
and then L1 was tested before L2. Then participants in the forget group
were asked, “What did you do to forget L1? If you clearly remember what
you did, please write that down. If you do not remember clearly, please do
not make up anything at this point. If you did not do anything in particular
to forget the words, or if you specifically tried to remember them, please
indicate this.”
The responses were grouped into six categories, and recall was assessed as
a function of different strategies. The results are summarized in Figure 4.5.
Reports of “stopping repeating L1 words,” or “engaging in diverting
thoughts” are more unambiguously interpretable compared to reports like
“stopping thinking about L1 words,” “clearing one’s head,” or “focusing
on the upcoming list,” because it is less clear how they went about those
processes. Thus, interpreting the relative effectiveness of these strategies
requires that we be mindful that some of these strategies may not be
mutually exclusive.
It is nevertheless clear from Figure 4.5a that to obtain DF impairment,
participants had to engage in some controlled behaviors; doing nothing in
response to the forget cue did not produce DF impairment. Additionally,
Directed forgetting impairment

0.50 (N = 145 in forget, and N = 145 in remember)
0.45
0.40
0.35
0.30
0.25
10% 31% 100%
0.20
0.15 21% 8%
12%
0.10
19%
0.05
0.00
Stop Stop thinking Focus on Diverting Clearing head Nothing Remember
rehearsing upcoming thoughts
Directed forgetting benefits

(N = 145 in forget, and N = 145 in remember)
0.55
0.50
0.45
0.40
0.35
0.30
21% 8%
0.25 19% 10% 31%
0.20 12% 100%

0.15
0.10
0.05
0.00
Stop Stop thinking Focus on Diverting Clearing head Nothing Remember
rehearsing upcoming thoughts
Figure 4.5 Directed forgetting costs (a) and benefits (b) as a function of forgetting
strategies reported by forget-group participants in a dataset containing N = 145 forget-
group participants and N = 145 remember-group participants. The black bar represents
the remember condition, and the gray bars represent the subgroups of the forget con-
dition broken down by reported strategy. The numbers inside the bars indicate the per-
centage of participants in each condition.
engaging in diversionary thoughts produced the biggest DF impairment

compared to other strategies, some of which did not even produce DF
(e.g. clearing one’s head). Finally, despite the variability in the size of the
DF impairment as a function of various strategies, nearly all strategies led
to DF benefits, including “doing nothing” to forget (Figure 4.5b). We will

revisit this issue in Section 7, where we specifically discuss the mechanisms
producing DF benefits.
5. CONTEXT CHANGE AS AN EXPLANATION FOR DF

IMPAIRMENT
Once forgetting is strategically initiated, we believe that it takes place
because of changes in mental context—an explanation first proposed by
Sahakyan and Kelley (2002). Our thinking was greatly influenced by formal
models of episodic memory, according to which people encode both the
content of the item during learning and various contextual features or attri-
butes that are present in the background (e.g. Anderson & Bower, 1972;
Estes, 1955; Gillund & Shiffrin, 1984; Hintzman, 1988; Howard & Kahana,
2002; Mensink & Raaijmakers, 1988; Tulving, 1983). Contextual informa-
tion includes environmental, spatial–temporal, emotional, and mental states
in which the item is experienced. Context cues play a particularly important
role in free recall because the initial retrieval attempt is solely guided by the
context cues that participants use to search their memory (e.g. Howard &
Kahana, 2002; Mensink & Raaijmakers, 1988). The success of retrieval is a
function of the overlap between the cues used to search memory and the
contents of memory. Recall suffers when there is a low overlap between the
contextual features present at encoding and the contextual features present
at test (e.g. Godden & Baddeley, 1975).
Based on these ideas, Sahakyan and Kelley (2002) proposed a mental
context-change account of DF, according to which the forget instruction
encourages participants to abandon the contextual cues that were prevalent
during L1 encoding and to sample new contextual cues for L2 encoding,
thereby segregating the two lists as separate events. The DF impairment
arises because retrieval context better matches L2 than L1 encoding con-
text, producing forgetting of L1 items in the forget group. The original
context-change account further explained that DF benefits arise because
of reduced proactive interference on L2 due to contextual differentiation
(Sahakyan & Kelley, 2002). We will argue later that changes in encoding
strategy produce a larger impact than the reduced proactive interference in
most studies, but that the reduced proactive interference effect is likely a real
effect that is difficult to assess through overall recall rates.
The rationale for proposing the mental context-change account was
triggered by the reports of participants who indicated engaging in distracting
thoughts in order to forget. Figure 4.5 confirms that this strategy leads
to the largest DF impairment (although other strategies could also con-
tribute to DF). We reasoned that engaging in diversionary thoughts could
change peoples’ mental context between L1 and L2. Recently, Lehman
and Malmberg (2009) developed a computational model of DF within the
framework of search of associative memory, by capitalizing on our notion
of context change between the lists, and the difficulty reinstating L1 context
during the test in the forget group.
In this section, we review the findings from various conditions that were
predicted to interact with list-method DF based on the context-change
mechanism along with other theoretically relevant findings.
5.1. Mental Context-Change Paradigm and Its Relation to DF

In the process of empirically testing the context-change account of DF,
Sahakyan and Kelley invented what has come to be known as the mental
context-change/diversion paradigm. This paradigm involves asking participants
to engage in diversionary thoughts following L1 learning, and examining
the impact of such activity on memory. Several years of investigation have
demonstrated that mental context-change produces forgetting much like
the explicit instructions to forget. A variety of tasks were shown to create
forgetting, such as imagining being invisible (Sahakyan & Kelley, 2002),
thinking about the childhood home (e.g. Sahakyan & Delaney, 2003), day-
dreaming about vacations (e.g. Delaney, Sahakyan, Kelley, & Zimmerman,
2010), or wiping a computer monitor (e.g. Mulji & Bodner, 2010). In con-
trast, other tasks do not lead to forgetting, such as solving arithmetic prob-
lems (Sahakyan & Delaney, 2003), counting forward or backward (Sahakyan,
Delaney, & Goodmon, 2008), briefly chatting with an experimenter (Aslan &
Bäuml, 2008), or waiting quietly (Sahakyan & Kelley, 2002).Why some tasks
create mental context-change whereas others do not is not fully clear and
requires future research.
Whenever context-change effects were found, they were found irre-
spective of encoding strategies (Sahakyan & Delaney, 2003), across individ-
ual differences in working memory capacity (Delaney & Sahakyan, 2007),
across age-related differences (Aslan & Bäuml, 2008; Sahakyan, Delaney, &
Goodmon, 2008), across serial position effects (Sahakyan & Foster, 2009),
and the boundary conditions that determine whether DF is obtained, such
as the need for L2 learning (Pastötter & Bäuml, 2007; but see Unsworth
et al., 2012). Thus, the mental context-change paradigm produces many
behavioral effects that are similar to DF. Interestingly, the retrieval dynamics
in the mental context-change paradigm were shown to be similar to the

environmental context-change paradigm across several dependent measures
including recall latencies, serial position effects, first response functions, and
the patterns of response transitions during recall captured through con-
ditional response probability functions (Unsworth et al., 2012). Thus, the
mental context-change paradigm produces behavioral effects similar to not
only DF but also the environmental context-change paradigm.
Some recent findings, however, have revealed differences across DF
and the mental context-change paradigm. For example, although Pastötter
and Bäuml (2007) reported that L2 learning is critical for obtaining both
DF and mental context-change effects, Unsworth et al. (2012) obtained
significant forgetting using a mental context-change paradigm without
involving L2 learning. Several methodological differences exist between
these two studies, most notably the list length, which was much longer in
the Unsworth et al. study (40 items) than is typically employed in the DF
studies.The use of such long lists may affect the likelihood of reinstating the
context of the items from the beginning of the list because those items were
studied long ago. Future research should consider more rigorous investiga-
tion of the list-length factor, including examining whether long lists might
also lead to a DF effect without the need for L2 learning.
5.2. Recognition Tests and Indirect Tests of Memory

The majority of list-method DF studies failed to obtain a DF effect on a
standard recognition test (e.g. Basden et al., 1993; Bjork & Bjork, 2003;
Block, 1971; Conway et al., 2000; Elmes, Adams, & Roediger, 1970;
Geiselman et al., 1983; Gottlob & Golding, 2007; Smith, Barresi, Gross,
& 1971; Racsmány, Conway, Garab, & Nagymáté, 2008; Reitman, Malin,
Bjork, & Higman, 1973; Whetstone, Cross, & Whetstone, 1996; Zellner
& Bäuml, 2006). Although two relatively recent studies obtained DF
benefits on L2 using longer study lists than were previously employed
(e.g. Benjamin, 2006; Sahakyan & Delaney, 2005), they too failed to obtain
DF impairment. In contrast to the list-method, the item-method DF effect
was always obtained on recognition tests (e.g. Basden et al., 1993; MacLeod,
1999). Implicit memory tests have also been shown to be insensitive to the
list-method DF, including both word stem and fragment completion tests
(Basden et al., 1993; Bjork & Bjork, 1996; but see Koppel & Storm, 2012),
tests of primed word association or general knowledge (Basden & Basden,
1996, 1998), and lexical decision (Racsmány & Conway, 2006). In all these
tasks, forget items show priming that is equivalent to remember items.
Because performance on the forget and the remember items did not
differ on recognition tests (or on implicit tests), the inhibitory account pro-
posed that those items were released from inhibition by virtue of being
represented during the test (e.g. Bjork, 1989; Bjork & Bjork, 1996, 2003;
Geiselman et al., 1983; Zellner & Bäuml, 2006). Although this explanation
is still rather popular, we think of it more as a description rather than a spe-
cific assumption or an explanation offered by theory. The release of inhibi-
tion was proposed mainly to reconcile the discrepant findings across the list
and the item methods. However, there is nothing a priori in the inhibitory
account to propose a release from inhibition on certain tests. For example,
assumptions of release are not made for the RIF phenomenon, which was
discovered much later than DF, and which is usually observed on recogni-
tion and lexical decision tests (e.g. Aslan & Bäuml, 2011; Hicks & Starns,
2004; Soriano, Jiménez, Román, & Bajo, 2009; Spitzer & Bäuml, 2007,
2009; Veling & van Knippenberg, 2004; Verde, 2004). If list-method DF
was observed on various memory tests just like the item-method, assump-
tions of release would not be made and the inhibitory interpretation of DF
would probably not even be proposed.
We have interpreted DF from the perspective of our contextual account,
and have a different take on the null effects of DF in various direct and indi-
rect tests. For example, the absence of DF in implicit memory tests is con-
sistent with the contextual account. Implicit tests do not require retrieval of
contextual information because they do not direct attention to the original
study episode. Since they do not require reinstating prior episodic context, we
would not expect to observe DF on such tests. Environmental context effects
also are not found on implicit tests (e.g. Parker, Gellatly, & Waterman, 1999).
The absence of DF in recognition is consistent with a body of literature
that shows that environmental context effects typically are not detected in
recognition (e.g. Godden & Baddeley, 1975). That said, the meta-analysis of
that literature revealed that under certain conditions, environmental con-
text effects do emerge in recognition, and those effects are larger when the
encoding processes are primarily nonassociative in nature (for a review, see
Smith & Vela, 2001). For example, a number of studies using nonwords or
unfamiliar faces as stimuli detected environmental context effects in recog-
nition (e.g. Dalton, 1993; Krafka & Penrod, 1985; Macken, 2002; Malpass &
Devine, 1981; Russo, Ward, Geurts, & Scheres, 1999; Smith & Vela, 1992).
Building on those findings, we predicted and obtained DF in recogni-
tion using nonword stimuli, despite not obtaining such effects using words
(Sahakyan, Waldum, Benjamin, & Bickett, 2009).
In another experiment, we obtained DF in recognition using word

stimuli by manipulating whether the test required distinguishing the targets
from similar or dissimilar distractors (Sahakyan et al., 2009, Experiment 3).
Similar distractors involved plurality-reversed versions of the target items,
whereas dissimilar distractors were novel words. We hypothesized that plu-
rality discrimination would engage more direct retrieval of contextual
information during the test, enabling us to detect DF in recognition even
with word stimuli. Indeed, the results confirmed our predictions: There was
no DF in the dissimilar distractor condition, but there was DF in the similar
distractor condition.
In the same year, Lehman and Malmberg (2009) reported DF in rec-
ognition of word stimuli using an exclusion/inclusion manipulation. They
reasoned that typical recognition tests ask participants to endorse any item
studied during the experiment, regardless of which list it appeared on
(i.e. inclusion condition). Under such conditions, participants do not have
to rely on context cues to differentiate one study episode from another, and
they can rely on item familiarity as the basis of recognition. In contrast, in
an exclusion condition, participants have to endorse only the words from
a designated list (either L1 or L2), and to reject the words that come from a
different list (or new words). To perform such a task accurately, participants
have to rely more heavily on context cues to differentiate one list from another.
Overall, the exclusion recognition is closer to what is needed to perform free
recall than inclusion recognition is, and indeed Lehman and Malmberg (2009)
predicted and obtained robust DF under exclusion instructions.
Plurality discrimination, exclusion recognition, and nonword recogni-
tion all rely on retrieval of contextual information, and hence all these
conditions showed DF in recognition. Recently, Racsmány, Conway, Garab,
and Nagymate (2008) used a remember/know procedure and also obtained
DF in recognition, but only for “remember” responses. Although one might
be tempted to conclude that DF impairs all types of contextual informa-
tion, the results from associative recognition further refine the nature of
contextual information that is important in DF. Specifically, Hanczakowski
et al. (2012) did not obtain DF on an associative recognition test, despite it
being a recollection-laden test. Although there was no effect in associative
recognition, within the same study, DF affected list discrimination judg-
ments. Therefore, the authors argued that DF does not lead to a generalized
impairment in recollection of all contextual details. Instead, DF affects pri-
marily global contextual information that differentiates L1 from L2, without
necessarily impairing interitem associations between the words within the
list. Finally, Gottlob and Golding (2007) demonstrated that despite the lack
of list-method DF in the item recognition, the forget instruction impaired
source memory for color and case of L1 items.
Overall, the reports of significant list-method DF in item recognition
suggest that the typical conditions of recognition rather than the recogni-
tion test itself were the reasons for the observed lack of DF in previous
research. When stimuli or test conditions encouraged greater utilization
of global contextual information, DF was observed in recognition. These
results challenge the inhibitory interpretation because it does not a priori
predict the conditions under which the release can or cannot be obtained.
In contrast, these results fit well with the context account. As the need to
retrieve context increases, the likelihood of seeing list-method DF in item
recognition or recognition of source also increases.
One of the earlier explanations of DF involved selective rehearsal, and
some researchers lately acknowledged that this account should not be pre-
maturely dismissed for intentional learning (e.g. Benjamin, 2006; Sheard &
MacLeod, 2005). Considering that people often report stopping rehearsal in
response to the forget cue and that engaging in this type of strategy produces
DF impairment, we briefly discuss the results of recognition findings from
the perspective of selective rehearsal.The absence of DF in recognition tests
is particularly problematic for the selective rehearsal account because recog-
nition and recall are sensitive to elaborative rehearsal (Geiselman & Bjork,
1980), and recognition is even more sensitive than recall to rote rehearsal
(Benjamin & Bjork, 2000; Craik & Watkins, 1973).Thus, regardless of nature
of the rehearsal process involved in DF, terminating rehearsal of to-be-
forgotten items should have negative consequences for recognition memory.
We note that recognition is typically not at ceiling despite the short lists
used in many DF studies, and even longer lists do not lead to DF impair-
ment in recognition (Benjamin, 2006; Sahakyan & Delaney, 2005). Thus,
any explanation that attempts to revive the selective rehearsal account must
successfully deal with this limitation, and should also be able to explain why
recognition effects emerge under the specific conditions described above.
5.3. Mental Reinstatement of Context

The context-change account assumes that participants rely on the recent
L2 context to probe their memory for L1, and because that context mis-
matches the L1 context in the forget group, it leads to impaired access to L1
items. The context account predicts that if participants could reinstate L1
context at the time of test, this should reduce DF.
The environmental context-change studies imply that participants do

not automatically reinstate the original environmental context when they
are tested in a new context, otherwise we would not be observing forget-
ting (e.g. Godden & Baddeley, 1975). Either participants do not attempt
contextual reinstatement in those studies, or they attempt to reinstate the
original context but are unsuccessful at it. We suspected that reinstatement
of mental context (i.e. internally generated context) might be even more
challenging than reinstatement of environmental context. To make L1 con-
text more memorable and distinctive, we played the music from Star Wars
at the start of the experiment. Before final test, Sahakyan and Kelley (2002)
asked participants to mentally reinstate the episodic context that was preva-
lent during L1 encoding by guiding them through similar types of steps as
previously employed in the environmental context literature (e.g. Smith,
1979). Reinstatement of L1 context did not eliminate the DF effect. How-
ever, it significantly reduced it both in the mental context-change condition
and in the forget group, suggesting that the interaction between the con-
tents of memory and the contextual cues used during the test is a critical
ingredient for obtaining DF.
If L1 items were inhibited as a result of DF, then why did they recover
from contextual reinstatement not just in the context-change group but
also in the forget group? To explain the recovery, the inhibitory account
would have to assume that context plays an important role in retrieval,
and that changing or reinstating the context at the time of final test has
consequences for memory. Because these are general claims made by many
formal theories of memory, it is unclear what is gained by postulating an
inhibitory mechanism if it must still explain the results by resorting to the
contextual mechanisms.
5.4. Cuing with Context Cues vs Category Cues

Recent research manipulated the structure of the study list combined with
specific ways of cuing memory at the time of test, and predicted the factors
that should eliminate DF vs preserve it (Lehman & Malmberg, 2011a). Par-
ticipants studied two lists. L2 always contained unrelated items, but L1 con-
sisted of unrelated items for some participants vs items drawn from a single
category for others. On the test, memory was probed either with a context
cue (e.g. “retrieve the items from L1”) or using the category cue (e.g. “use
the word clothing to help you retrieve the words from L1”). The authors
predicted that if DF impairment is due to context change, then it should
be reduced with category cues, which will reinstate the specific L1 context
more than will more general temporal context cues (e.g. Raaijmakers &
Shiffrin, 1981). Successful initial recall would in turn provide a good retrieval
cue for subsequent items. Replicating prior work, DF was obtained with
unrelated lists. There was also significant DF with the categorized list as
long as the test cue was just the temporal context. In contrast, when a cat-
egory cue was used to test memory, the costs were eliminated, as predicted.
Lehman and Malmberg’s (2011a) findings highlight the importance of the
interaction between retrieval cues and the contents of memory at test: The
same categorical list can either show DF or be released from DF depending
on the type of retrieval cues utilized during the test. The more specific the
cue is to the L1 study context, the less likely DF is to be observed.
Note that another study conducted by Wilson, Kipp, and Chapman
(2003) obtained a different outcome with categorical lists, arguing that
categorical lists were resistant to DF. However, before the study of each
list, Wilson et al. (2003) drew special attention to the fact that the lists in
their study were either unrelated or categorical in nature.Their test instruc-
tions were not described in sufficient detail to infer what type of retrieval
cues were given to participants. Thus, methodological differences between
the two studies may be responsible for the discrepant findings. Note that
Sahakyan (2004) also obtained intact DF using different categorical items
across the lists while probing memory with temporal context cues.
5.5. Encoding Strength: Item vs Context Strength

Does deeper/elaborate encoding of material protect against DF? Sahakyan
and Delaney (2003) had participants encode both lists either via the story
mnemonic or via rote rehearsal. Despite robust differences in overall memory
performance, they obtained equivalent DF impairment in both encoding
conditions, suggesting that interactive encoding does not protect against DF.
Our subsequent research suggests that certain ways of strengthening
items make them more susceptible to DF than other ways of strengthening
(Sahakyan, Delaney, & Waldum, 2008). Relying on research by Malmberg
and Shiffrin (2005), who suggested a more complex view of encoding
strength by differentiating it into an item strength and a context strength,
we derived and tested predictions of the contextual account of DF. Spe-
cifically, Malmberg and Shiffrin (2005) suggested that certain strengthening
manipulations increase the strength of the items without affecting their
associations with the list-context (e.g. depth of processing or extrastudy
time), whereas other manipulations lead to strengthening of both item and
contextual information in the memory trace (e.g. spaced presentations).
Based on their findings, we predicted that spaced items should suffer more
when context changes because item-to-context associations are stronger for
spaced items. Indeed, our study showed that spaced items showed larger DF
impairment than did massed items. Similar results have also been obtained in
an environmental context-change study (Isarida & Morii, 1986). In contrast
to the spacing manipulation, we predicted that the depth of processing or
the extrastudy time manipulations should not interact with DF, and indeed
the results confirmed our predictions. Even though items studied for longer
duration were remembered better than items studied for shorter duration,
they showed equivalent impairment from DF. Taken together, these results
suggest that what determines interactions with DF is not the strength of the
items per se, but rather the strength of item-to-context associations. Along
the same lines, Waldum and Sahakyan (2012) found that statements incon-
gruent with one’s attitude show greater DF impairment than do congru-
ent statements. Using a variety of memory paradigms, they confirmed that
greater DF of incongruent statements was driven by stronger associations
with episodic context, entirely consistent with the context account.
5.6. Contextual Variability

In an unpublished study, Sahakyan and Foster crossed DF with manipula-
tion of context variability (CV), which is the number of different semantic
contexts in which a given item occurs (e.g. Dennis & Humphreys, 2001;
Steyvers & Malmberg, 2003).The rationale for their study was that although
low-CV items have a memory advantage in both recognition and free
recall, they tend to suffer more from environmental context manipula-
tions between study and test (Marsh, Meeks, Hicks, Cook, & Clark-Foos,
2006). Marsh et al. (2006) argued that during encoding, low-CV items form
stronger item-to-context associations compared to high-CV items due to a
smaller fan of preexisting contextual associations. Therefore, when context
changes between the study and the test, low-CV items suffer more than
high-CV items do.
Consequently, we expected that DF should also affect low-CV items
more than high-CV items. We tested 46 participants randomly assigned
to forget or remember groups. Each participant studied two lists of 14
items selected from Steyvers and Malmberg’s (2003) appendix. Half of
the items within each list were low-CV and the other half was high-CV
words. Because CV correlates with word frequency (Steyvers & Malmberg,
2003), we controlled word frequency by selecting the items from the low-
frequency category similar to what was done by Marsh et al. (2006). During
Directed forgetting impairment and context variability

0.45
Forget
0.40
Remember
0.35
Proportion list 1 recalled
0.30
0.25
0.20
0.15
0.10
0.05
0.00
"low" Context variability "high"
Figure 4.6 Directed forgetting impairment as a function of context variability of s tudied
items. (Based on unpublished data.)
the test, L1 was always tested before L2. Figure 4.6 summarizes the findings.
DF impairment was much greater for low-CV items than for high-CV
items, consistent with the predictions of the context account.
5.7. Extralist Cued Recall

Most list-method DF studies employ free recall tasks partly because rec-
ognition and indirect memory tests are insensitive to DF. Sahakyan and
Goodmon (2010) used a different testing procedure that allowed them to
contrast the predictions of the inhibitory and context accounts. Specifi-
cally, participants studied two lists of items (called targets). The test pro-
vided cues that were meaningfully related to the targets, but were never
studied with them (hence the name extralist cued recall). For example, the
studied word planet could be tested with the cue universe. Extralist cued
recall is affected both by target characteristics and by cue characteristics
(for reviews, see Nelson, McKinney, Gee, & Janczura, 1998; Nelson &
McEvoy, 2005), and it provides an excellent opportunity to test the
opposite predictions of the inhibitory account and the context account
within the same study.
Sahakyan and Goodmon (2010) conducted five experiments manipu-

lating different sources of implicit associative strength. Crucially, some of
their experiments varied the implicit strength of the targets while holding
the cue strength constant, whereas other experiments held the implicit
strength of targets constant while varying the strength levels of the test
cues. Across all experiments, they found greater DF from the strong con-
ditions compared to the weak conditions, regardless of whether strength
referred to the targets or the test cues. This pattern of findings is exactly
what is predicted by the contextual account: Presentation of the extralist
cue should not diminish the importance of recovering contextual infor-
mation during the test. Reinstating the context helps to differentiate the
studied item from other items implicitly activated by the test cue. According
to the PIER model which explains extralist cued recall (Nelson, Goodmon, &
Ceo, 2007), context cues and extralist cues combine together to elicit recall
of the target.Thus, the context account makes a strong prediction that there
should be greater effects of DF in the strong than weak conditions regard-
less of whether the study varied the target strength level or the cue strength
level.
In contrast, the inhibitory view assumes that items can be released from
inhibition with provision of certain cues (e.g. copy cues of the items on the
recognition test). Hence, the inhibition account would predict that stronger
test cues would be more successful at releasing the targets from inhibition
than weaker cues, and hence DF should be smaller not greater in the condi-
tions where cue strength is high. The results did not support this prediction.
The findings of Sahakyan and Goodmon (2010) suggest that items/targets
were not inhibited at the level of individual item representation because
there is no way for the cognitive system to determine which targets should
be inhibited more and which targets should be inhibited less until a test cue
is provided. The effects in their study were instead entirely driven by the
combination of test cues and context cues, consistent with predictions of the
context account.
5.8. Importance of L2 Learning

A number of studies revealed boundary conditions that are needed to observe
DF impairment. For example, replicating prior work by Gelfand and Bjork
(1985; cited in Bjork, 1989), Pastötter and Bäuml (2007) showed that in the
absence of L2 learning, one does not observe DF impairment. They later
showed that DF is reduced with decreasing length of L2 (Pastötter & Bäuml,
2010). Finally, Conway et al. (2000) as well as Macrae, Bodenhausen, Milne,
and Ford (1997) demonstrated that divided attention during L2 learning by

a concurrent task reduces DF impairment.
According to the inhibitory view, in the absence of L2 learning, there
is no need to inhibit L1 items because inhibition is an adaptive mechanism
invoked to reduce interference (e.g. Conway et al., 2000; Barnier et al.,
2007). When there is no L2 learning, there is no need to invoke inhibi-
tion. The inhibitory view thus assumes that L1 items intrude during L2
learning and hence must be inhibited (e.g. Conway & Fthenaki, 2003;
Racsmány & Conway, 2006). However, in the domain of RIF, it is
generally assumed that simply studying a list of items should not trigger
inhibitory processes; inhibition should be triggered only in response to
competitive retrieval (e.g. Anderson, 2003; Storm & Levy, 2012). It is not
fully clear why the mere study of L2 would trigger competitive retrieval
of L1 items.
According to the context-change account, DF arises not simply because
of context change between the lists, but because participants experience
difficulty reinstating the context of the previous list. In the absence of L2
learning, reinstating the context of the only list that was studied is not
an issue because it was the only list that was studied (e.g. Jang & Huber,
2008). Thus, according to the context-change view, one would not expect
to observe DF. In contrast, in the presence of L2, participants have to rely on
context cues to distinguish between the two lists and reinstate the context
of the previous list. Also, in the environmental context literature, the effects of
the first-order paradigms, where a single list is studied and tested either in the
same or a different context, are much smaller compared to the second-order
paradigms that involve multiple lists (e.g. Eich, 1985; Fernandez & Glen-
berg, 1985; for a review of first- and second-order paradigms, see Bjork &
Richardson-Klavehn, 1989). We interpret the impact of L2 length from a
similar perspective. Enhanced length of L2 contributes to greater distanc-
ing of L1 context from the time of test and, as it gets more distant, the
reinstatement of that context would also get more difficult, explaining why
shorter L2 reduces DF. Finally, we interpret the results of divided atten-
tion as reflecting attentional demands of context encoding and context
change. We expand on this view in the section discussing working memory
interactions.
5.9. Related Items Across Lists

Several studies have shown that when items are related across the lists, DF
impairment is eliminated (e.g. Barnier et al., 2007, Experiment 5; Conway et al.,
2000; Sahakyan & Goodmon, 2007). These findings are not consistent with an
inhibitory viewpoint that assumes suppression at the level of items. Specifically,
one would assume that when the two lists involve related items, then L1 items
may be more likely to come to mind during L2 learning and require greater
inhibition compared to when the two lists are unrelated. However, the results
show that related items across the lists reduce rather than increase DF.
Although items across the lists may be related to each other in a num-
ber of ways, our research showed that when L2 items remind participants
of L1 items due to backward associations (e.g. chip← chisel), DF is elimi-
nated, but when L1 items are related to L2 items via forward associations
(e.g. chip→ wood), DF remains intact (Sahakyan & Goodmon, 2007; for
arelated result using the item procedure, see Golding, Long, & MacLeod,
1994). We proposed several mechanisms by which related items could
reduce DF. First, reminding could reinstate the context of L1 during L2
learning, thereby preventing or reducing contextual differentiation between
the lists. Second, reminding could initiate retrieval of L1 items during L2
learning, thereby linking L1 items to the context of both lists. This would
enhance the probability of their retrieval even if L2 context is used as the
main retrieval cue during the test. Finally, reminding could strengthen L1
items via retrieval during L2, and enhanced strength of L1 items could also
make them resistant to DF.
5.10. Retrieving L2 before L1

The meta-analysis suggests that output order does not affect the magni-
tude of DF impairment (Pastötter et al., 2012). In other words, whether
L1 is tested first in the recall sequence, or whether it is tested after L2, the
magnitude of DF impairment remains invariant. This is inconsistent with
the inhibitory explanation because if L2 is tested first, then according to
the inhibitory account, L1 items should be inhibited to allow retrieval of
L2. Thus, the DF impairment should be larger when L1 is tested after L2
than when it is tested first in the recall sequence. However, Pastötter et al.
(2012) showed that retrieving L2 first does not increase the magnitude of
forgetting.
An even more extreme manipulation was employed by Basden, Basden,
and Morales (2003), who gave multiple retrieval trials on L2 before the final
test. Despite multiple retrieval trials on L2, Basden, Basden, and Morales
(2003) and Basden, Basden, and Wright (2003) did not observe negative
effects of such retrieval on L1 memory, even when their lists consisted of
categorical items from the same taxonomy. Overall, these results suggest
that retrieving L2 before L1 has no detrimental effects on DF impairment,

contrary to what would be predicted by the inhibitory account.
5.11. Delay Effects

There is a scarcity of published studies that have examined DF after a delay.
If DF impairment arises from the mismatch of the study and test contexts,
then the impairment should dissipate with the passage of time because with
delay, context no longer would favor L2 context over L1, and thus one
would expect spontaneous recovery of L1 (e.g. Mensink & Raaijmakers,
1988). Similar predictions would be made also by the inhibitory account
because items should not remain in an inhibited state indefinitely. We
found only one published study that employed a delay manipulation with
a full-DF design (Shapiro, Lindsey, & Krishan, 2006). The delay involved
15 min, and participants were tested either immediately, or after delay (but
never in both). The immediate test revealed the costs and the benefits of
DF. In contrast, delay eliminated DF costs, but did not affect the benefits,
which remained preserved. Identical results were reported by Liu Xun in his
unpublished dissertation (2001). The presence of significant L2 benefits is
consistent with the encoding-based interpretation of DF benefits (Pastötter &
Bäuml, 2010; Sahakyan & Delaney, 2005) because stronger encoding of L2
items should be preserved relatively in delay (e.g. MacLeod, 1975). Thus,
the dissociating effect of delay is consistent with the dual-factor account
of DF.
Two other delay studies were briefly discussed in two chapters (Basden &
Basden, 1998; MacLeod, Dodd, Sheard, Wilson, & Bibi, 2003), and
they reported significant DF in a delayed condition. However, neither
study included a remember control group, and DF was inferred from
the R–F measure. Given that delay preserves the benefits (Liu, 2001;
Shapiro, Lindsey, & Krishnan, 2006), the difference between the two lists
would probably still be significant after a delay. Most important, the brief
description of the methods in both studies makes it difficult to know
whether the same participants took part in both the immediate and the
delayed test, or whether delay was varied strictly between subjects. This
could be a critical detail given the effects of testing on retention (e.g.
Roediger & Karpicke, 2006). If the same participants participated in both
conditions, then the immediate test could preserve DF on the delayed
test. In other words, significant DF on the delayed test in some studies
may be due to a testing effect. More research is needed to fully investigate
delay effects in DF.
5.12. Individual Differences in DF as a Function of Working

Memory Capacity
Another feature of the contextual change account is that attentional control
has been theoretically linked to effective use of context (e.g. Unsworth &
Engle, 2008). Therefore, one might expect that people with higher work-
ing memory capacity, which is an index of attentional control, would show
larger effects of the forget cue and of the context-change task, as they would
be better able to manipulate context. Indeed, this appears to be the case
in several recent studies. Delaney and Sahakyan (2007) showed that in a
typical two-list design, both L1 and L2 memory were positively correlated
with working memory capacity. (This effect disappeared when everyone
was instructed to make up a story using all the words on the list). However,
working memory capacity was negatively correlated with L1 memory in the
forget and context-change conditions in two experiments. In other words,
the DF and context-change manipulations produced substantially more
forgetting for high-span than for low-span participants. Consistent with
the context-change account, both the context-change task and the forget
task interacted with working memory capacity in the same manner. This
basic result has been replicated with adults and children (Aslan, Zellner, &
Bäuml, 2010), and in some reports, the benefits of DF were also linked to
working memory capacity (Soriano & Bajo, 2007).
The effects of dividing attention during L2 (e.g. Conway et al., 2000)
are consistent with the notion that context encoding requires attentional
resources. Dual task could interfere with context encoding during L2,
thereby giving contextual advantage to L1 compared to L2. It could also
make the two lists sufficiently differentiated even in the remember group.
Note that Conway et al. (2000) did not include a condition where atten-
tion was divided during both lists. Thus their attentional manipulation was
confounded with the list.
5.13. Reexposure or Retrieval of L1 Items Affects Subsequent

Memory
Goernert and Larson (1994) examined how part-set cuing affects list-
method DF and found that when participants were provided with a subset
of L1 items on the final test to use as cues for retrieval, their recall of
remaining L1 items showed improvement in the forget group, whereas
it suffered in the remember group compared to the noncued condition.
These effects were replicated and extended by Bäuml and Samenieh
(2010, 2012a, 2012b), who showed that both part-set cuing and also selec-
tive retrieval of a subset of L1 items can enhance recall of the remain-
ing L1 items in the forget group, while the same manipulations impaired
recall in the remember group. Similar effects were obtained in the men-
tal context-change paradigm (Bäuml & Samenieh, 2012a, 2012b). Prior
studies also showed that reexposure to L1 items on an intervening test
can release subsequent DF depending upon the nature of the intervening
test. The typical design of these studies involves (1) giving an initial test,
(2) reexposing all or a subset of items from both lists on an intervening
test, and (3) administering a final recall test. The intervening tests included
recognition (Basden et al., 1993; Basden et al., 2003; Bjork & Bjork, 1996),
word fragment completion (Basden et al., 2003; Bjork & Bjork, 1996),
implicit free association test (Basden et al., 1993), or a lexical decision
test intermixing all studied items with unstudied items and nonwords
(Racsmány & Conway, 2006). Although initial recall revealed significant
DF in these studies, the intervening tests did not (neither recognition,
nor lexical decision, nor fragment completion showed evidence of DF).
Importantly, the final recall findings differed depending upon the nature
of the intervening task. When the intervening test involved a recognition
test that included a subset of L1 items as lures, DF was released on a final
recall test. In contrast, intervening word fragment completion or lexical
decision tests did not release DF.
We interpret these findings to suggest that retrieval of contextual infor-
mation associated with the study episode is critical to reinstating access to
L1 items (similar arguments have been made by Bauml et al. for the part-set
cuing and part-set retrieval effects in the forget group). The fragment com-
pletion test is an indirect memory test because it does not direct the subject
to go back to the original learning episode, and it is considered mainly data
driven. Likewise, the lexical decision task does not require reinstating prior
episodic context.The intervening recognition test, however, is a direct mea-
sure of memory because the subject is instructed to think back to the learn-
ing episode and make a decision about a specific item (note that Bjork &
Bjork, 1996 included an exclusion instruction on the intervening test of
recognition, which would further facilitate retrieval of contextual informa-
tion).Thus, DF on the final free recall test is reduced when participants have
to think back to the original study episode and reinstate its episodic context.
In contrast, DF is unaffected by indirect tests that do not make any connec-
tion to the earlier study episode.
5.14. Summary
The evidence that contextual change underlies DF impairment comes from
a wide range of experimental sources.The idea came about because partici-
pants in DF studies self-reported thinking of something else, and subsequent
research confirmed that this strategy was effective in producing forgetting.
Furthermore, context is used to explain many other findings in memory, so
it is not invented specifically to explain DF, and can be modeled in context-
based formal models of memory. Among the findings that are predictable
from a context viewpoint are that recognition does not show DF unless it
relies on retrieval of contextual information; that mentally reinstating the
original study context releases DF; that indirect memory tests do not show
DF because they do not redirect back to the context of the study episode;
that items more strongly associated with episodic context show greater DF;
that cuing the items with stronger extralist cues leads to larger rather than
smaller DF; that having strong reminders on L2 can reduce DF; that having
no L2 eliminates DF; that categorized items can be susceptible to DF or not
depending on whether the test cue involves context cue or a category cue.
In our view, the account is strongly supported by the existing data, and any
future theory should strive to not only explain the effects that the context-
change account predicts but also make novel predictions that cannot be
handled by the current theory.
Shortly after we proposed the context-change account of DF, Anderson
(2005) proposed that the context-change mechanism is compatible with an
inhibitory interpretation of DF if inhibition is assumed to be the mecha-
nism by which context change takes place. According to Anderson’s inter-
pretation, inhibition is a flexible mechanism that can be targeted at different
levels of representation, including at a context representation. We are reluc-
tant to embrace this notion because we do not see it as advancing our
theoretical understanding of DF. It redescribes context change with new
terms (i.e. inhibition of context) without specifying a new theory of inhi-
bition of context. There is currently no evidence to suggest that context
is inhibited as opposed to being changed across the lists in DF. Impaired
memory of L1 items does not constitute any evidence of inhibition of
context or indeed any inhibition at all (MacLeod et al., 2003). Equating
mental context change with inhibition of context implies that any time we
change what we are thinking about, we engage in inhibition. By this view,
everything becomes inhibition, and thus impossible to know what does not
involve inhibition. Inhibition of context also complicates the explanation of
DF because it implies two mechanisms, not one. If we assume that context

serves as retrieval cue, then when one changes from context A to context B,
context match becomes poorer for the previous context A. Adding inhibi-
tion would be imposing a second mechanism that assumes that changing
from context A to context B requires inhibition of context A.To the best of
our knowledge, no one has offered an inhibitory theory of environmental
context-dependent memory (e.g. Godden & Baddeley, 1975).The impaired
memory arising from the context change is instead explained in terms of
the mismatch of cues being used to search memory and the contents of
memory, which leads to reduced accessibility.
6. AREAS OF DISAGREEMENT ACROSS STUDIES

While reviewing the empirical findings, two broad areas of inconsis-
tencies emerged as in need of further investigation and more careful analyses.
These include L1 serial position findings, and the source misattribution and
intrusion error findings.
6.1. Serial Position Effects

We identified five published DF studies that examined serial position func-
tions in DF (Geiselman et al., 1983; Lehman & Malmberg, 2009; Pastötter &
Bäuml, 2010; Pastötter et al., 2012; Sahakyan & Foster, 2009). Some of
these studies used serial positions as a tool for testing various DF theories,
whereas other studies plotted those functions without formally analyzing
them. Although the studies agree that the effect of the forget cue on L2
is mostly evident at the start–middle of the L2 serial position curve, their
results regarding L1 functions differed to some extent. For example, studies
that analyzed serial position functions did not obtain greater DF costs from
the primacy region of L1 curve (e.g. Pastötter & Bäuml, 2010; Pastötter
et al., 2012; Sahakyan & Foster, 2009), whereas studies that reported serial
position curves without formal analyses (or deviation measures) obtained
greater DF from the primacy regions of L1 curves (e.g. Geiselman et al.,
1983; Lehman & Malmberg, 2009). Lehman and Malmberg (2009) argued
that context is more prominent at the start of the list, which could explain
greater forgetting from the primacy region. A study reported by Sheard
and MacLeod (2005) also obtained greater DF costs from L1 primacy
region, although their study failed to obtain overall DF impairment.
Thus, some studies obtain greater DF impairment in the primacy region
whereas others do not. Note that Sahakyan and Foster (2009) obtained
equivalent DF impairment across L1 serial position curves of performed
actions phrases which typically do not produce primacy effects, suggesting
that the magnitude of DF impairment may not be linked to the primacy
effects. Considering these controversies and the methodological differences
that could have given rise to these findings, more careful attention is needed
to serial position effects in DF research.
6.2. List Differentiation in DF: Source vs Intrusion Errors

Some DF studies found that the forget cue impaired source discrimination
judgments in recognition (e.g. Bjork & Bjork, 2003; Gottlob & Golding,
2007; Hanczakowski et al., 2012), whereas we explained these findings in
terms of a bias to attribute L1 items to L2 during the recognition test
(e.g. Sahakyan & Delaney, 2005). Regardless of whether this is a bias effect or
a true source confusion effect, these findings overall suggest that DF impairs
list differentiation. Somewhat different conclusions are reached when DF
is examined via intrusion errors during free recall or list categorization
errors during free recall (e.g. Geiselman et al., 1983; Golding & Gottlob,
2005; Lehman & Malmberg, 2009; Sahakyan & Delaney, 2010; Spillers &
Unsworth, 2011). Some of these studies have found enhanced list differen-
tiation as evidenced by reduced intrusion errors, whereas others have found
the opposite. Although intrusion errors are notoriously difficult to study,
they could nevertheless be informative because they could indicate the type
of retrieval cues that participants rely on during free recall. For example, if
people make many intrusions from a wrong list, this might suggest problems
with the sampling stage, indication that they use “wrong retrieval cues” that
better match one list than the other list (e.g. Spillers & Unsworth, 2011).
As list differentiation in DF is studied further, it is important to keep
in mind that source errors and intrusion errors may not necessarily lead
to the same conclusions because they represent opposite sides of the same
coin. Source errors refer to the ability to retrieve the context given the
item, whereas intrusion errors are more diagnostic of the cuing property
of context, referring to the ability to retrieve the item given the context
cue. If item-to-context and context-to-item associations are not necessarily
equivalent in strength, or if ability to retrieve the context given the item as
opposed to retrieving the item given the context were not identical, then
the findings from source errors and intrusion errors may not necessarily
produce the same empirical outcomes.
7. STRATEGY CHANGE EXPLAINS DF BENEFITS

Sahakyan and Delaney (2003) were the first to propose that the costs
and benefits of DF might be dissociable, and that they might be explained
by different mechanisms. A critical piece of evidence for their proposal
was that it is possible to dissociate the costs and benefits (e.g. Benjamin,
2006; Conway et al., 2000; Joslyn & Oakes, 2005; Pastötter et al., 2012;
Sahakyan & Delaney, 2005, 2010; Shapiro et al., 2006; Spillers & Unsworth,
2011; Zellner & Bäuml, 2006). How could that be true if a single pro-
cess accounted for both results? While not all researchers agree what the
two processes are, most now agree that separate mechanisms are needed to
explain costs and benefits.
7.1. The Advent of the Dual-Process Account

The Sahakyan and Delaney (2003) dual-process account explains the ben-
efits by the strategy-change mechanism. According to this explanation, the
benefits of DF emerge because the forget cue encourages participants in the
forget group to adopt a more elaborate encoding strategy for L2 items com-
pared to those in the remember group. Thus, L2 benefits reflect a strategic
encoding enhancement in the forget group.
We initially proposed this explanation because we evaluated partici-
pants’ verbal reports of study strategies from Sahakyan and Kelley’s (2002)
study, which revealed that many participants adopted a better study strat-
egy on L2 than L1, and the forget group was almost twice as likely as
the remember group to switch encoding strategies (38% vs 22% in the
Sahakyan and Kelley study).Therefore, in Sahakyan and Delaney (2003), we
first reanalyzed Sahakyan and Kelley’s (2002) data by statistically control-
ling the strategies, and we found that the effect of cue (F vs R) was much
smaller compared to the effect of strategy switch, which accounted for the
majority of variance in L2 recall. We also conducted two experiments that
experimentally controlled the encoding strategy, with the same outcome. In
Experiment 1, participants studied both lists using the same encoding strat-
egy, and we did not obtain the benefits because everybody suffered mas-
sive proactive interference. In Experiment 2, everybody was told to switch
study strategy between L1 and L2 (including the remember group), and
everybody benefited from such switch and escaped proactive interference.
In both experiments, we always obtained DF costs.
This is important to emphasize because we often hear from our col-

leagues that a switch of study strategies between the two lists may be part
of mental context change and could explain the DF impairment. Although
strategy switch could contribute to mental context change, it cannot explain
why the DF impairment was still obtained in Experiment 2, when every-
body switched strategies, including the remember group. According to this
reasoning, there should be no DF costs if everybody changes mental context
due to strategy shift. However, the costs are present regardless of whether
people retain the same strategy on both lists, or change the strategy across
the lists, which led us to propose the dual-process account. Studies investi-
gating neural bases of DF also provide support for the two-factor account,
and suggest different neural origins for the costs and the benefits of DF and
mental context change (Bäuml, Hanslmayr, Pastotter, & Klimesch, 2008;
Hanslmayr et al., 2012; Pastötter, Bäuml, & Hanslmayr, 2008).
In general, whenever we controlled encoding strategy in subsequent
studies, we never obtained the benefits (e.g. Sahakyan & Delaney, 2005,
2010; Sahakyan, Delaney, & Waldum, 2008). Along the same lines, when
participants performed simple action phrases during encoding (e.g. break
a toothpick), we did not obtain the benefits, but the same action phrases
encoded verbally without controlling for study strategy produced the ben-
efits (e.g. Sahakyan & Foster, 2009). Although we explain L2 benefits in
terms of encoding strategy switch, a related and similar account was pro-
posed by Pastötter and Bäuml (2010), who argued that the benefits arise
because the forget cue resets the encoding processes at the start of L2
learning.This proposal is analogous to the earlier selective rehearsal account,
with an exception that it reserves selective rehearsal mechanism strictly
for explaining L2 benefits (as opposed to the costs and the benefits as was
initially assumed by the selective rehearsal). The primary reason for propos-
ing a reset of encoding (as opposed to a strategy change) is that L2 serial
position curves typically show much larger DF benefits in the primacy
regions of the curve than elsewhere in the list (e.g. Geiselman et al., 1983;
Lehman & Malmberg, 2009; Pastötter & Bäuml, 2010; Sahakyan & Foster,
2009). Pastötter and Bäuml (2010) argued that a switch to a better encoding
strategy should be evident throughout the entire list, not just at the begin-
ning of L2. This interpretation of strategy change is reasonable, but might
be a bit extreme. Although many participants spontaneously report doing
something different to encode L2 (or in addition to what they were doing
to encode L1), it is quite possible that they do not persist with that strategy
throughout the entire L2 learning. For instance, often participants report
starting out with some strategy on L2, but further down the list they report
being overwhelmed and being unable to keep it up.This could explain why
the benefits of DF are more prominent in the early sections of L2 serial
position curves than in later sections.
7.2. A Large-Scale Replication Study

Our dataset of N = 290 participants (Figure 4.5) provides reports of strat-
egy in response to the forget cue, and also contains reports of participants’
encoding strategy for both lists.We therefore reevaluated the strategy-change
account in our dataset with a substantially larger sample. As Figure 4.5
demonstrates, despite using a variety of forgetting strategies with some
being more effective than others in leading to DF impairment, nearly all
strategies were associated with almost invariant magnitude of DF benefits.
Interestingly, even the group that reported doing nothing to forget also
showed the benefits. The “doing-nothing” group essentially shows a disso-
ciation, whereby they show no DF impairment compared to the remember
group, but somehow they show the benefits on L2 despite no forgetting of
L1. If the benefits are driven by strategy change, then it is quite possible that
these participants changed study strategy between the lists (perhaps in hopes
that doing so might help them forget L1). We coded participants’ encod-
ing strategies for both lists, and created a variable that captured whether
participants employed a more efficient strategy on L2 compared to L1. For
example, any mention of doing something in addition to what they were
doing during L1 encoding was coded as a strategy switch (e.g. rehearsing on
L1, and rehearsing plus also grouping words together on L2).
Confirming previous work from our lab, we found that 28% of the for-
get group switched to a better study strategy on L2 compared to 10% of the
remember group, X2 (1, N = 290) = 14.43, p < 0.001. Although the rates
of strategy switch were much lower in the University of North Carolina at
Greensboro sample than in the previous samples of Florida State University
and University of Florida participants, the basic pattern was similar—the
forget group was more likely than the remember group to switch to better
encoding on L2. Confirming the results of Foster and Sahakyan (2011), we
also found that within the forget group, there was no relation between the
likelihood of switching to a better study strategy on L2 and the likelihood
of engaging in the forgetting strategy such as doing something vs nothing
to forget L1 (φ = −0.03, p = 0.73).We therefore collapsed across that factor
in the forget group, and analyzed L2 recall using cue (forget vs remember)
and strategy switch (yes vs no) as predictors. Strategy switch was the only
significant predictor (p = 0.007). People who switched study strategy

recalled significantly more from L2 than those who did not change study
strategy.The main effect of cue was not quite significant (p = 0.06) despite a
large sample size.These findings largely replicate our previous research con-
ducted at another institution (Sahakyan & Delaney, 2003), and they suggest
that strategy switch provides a better explanation for the DF benefits than
does the cue. However, it is important to note that we have always obtained
cue effects both in prior research and in the current dataset, just not at the
conventional levels of significance. This suggests that there is probably some
escape from proactive interference that arises from context change. How-
ever, such effects are difficult to detect in overall recall rates (at least in a
two-list design), and they are much smaller than the encoding effects arising
from the strategy change.
7.3. Test Order and the Benefits of DF

One factor that could contribute to failures to detect DF benefits may be
linked to test order. A recent meta-analysis on 20 studies that reported data
on output order concluded that failure to detect DF benefits in many stud-
ies could be driven by the fact that L2 was tested after L1 in many prior
studies (Pastötter et al., 2012). In addition to meta-analyses, Pastötter et al.
(2012) conducted two studies, where recall order was manipulated, and they
showed that the benefits of DF are much larger when L2 is tested first in the
recall sequence than when it is tested after L1.
Although several studies from our lab controlled the encoding strategy
and tested L2 after L1 (Sahakyan & Delaney, 2003; Sahakyan, Delaney, &
Waldum, 2008), it is unlikely that our chosen test order could account
for the absence of benefits in those studies. Sahakyan and Delaney (2010)
counterbalanced test order while controlling study strategy for some par-
ticipants and not for others. Although their published data were collapsed
across test order, Figure 4.7 replots the data broken out by test order, and
shows that the presence or absence of DF benefits was driven entirely by
whether the study strategy was controlled or not; test order had no impact
on the findings.
7.4. Summary
Most researchers now agree that DF benefits on L2 arise from a different
mechanism than the costs. We argued that the underlying factor is the
strategy change triggered by the forget cue. As with the context-change
mechanism, the idea for this came from verbal reports given by participants
Intentional learning condition with study strategy control

0.60
Forget
0.50 Remember
Proportion recalled
0.40
0.30
0.20
0.10
0.00
List 1 (costs) List 2 (benefits) List 1 (costs) List 2 (benefits)
Testing L1 first Testing L2 first
Intentional learning condition without study strategy control

0.60
Forget
Remember
0.50
Proportion recalled
0.40
0.30
0.20
0.10
0.00
List 1 (costs) List 2 (benefits) List 1 (costs) List 2 (benefits)
Testing L1 first Testing L2 first

Figure 4.7 Directed forgetting costs and benefits as a function of test order, controlled
study strategy (a), and uncontrolled study strategy (b). (New figures based on data
reported in Sahakyan and Delaney (2010)).
in our experiments, and was supported by successful predictions in other

experiments. For example, controlling participants’ encoding strategy via
instructions reduced or eliminated the benefits, as does the use of incidental
learning procedures. This review also reports on a large-scale replication of
the original work establishing the benefits were due to strategy changes:
even though strategy switches are by no means common, they produce
such large benefits to memory that they show up at the group level. We
think that reducing build-up of proactive interference due to contextual
isolation (the mechanism originally proposed by Sahakyan & Kelley, 2002
to explain the benefits) may also be important in some cases, but we have at
best limited empirical evidence for that mechanism.
8. IMPLICATIONS FOR CLINICAL POPULATIONS

People with clinical disorders sometimes behave atypically on cogni-
tive tasks. Demonstrations that clinical disorders alter behavior in cognitive
tasks are useful because they can improve diagnosis (if the difference is
sufficiently robust), differentiate subtypes of disorders, and clarify what func-
tions are impaired in clinical populations. Cognitive difficulties can suggest
treatments that may remediate symptoms, or predict the kinds of everyday
problems that people may have. Of course, it is insufficient to know that a
particular cognitive task is affected by a clinical diagnosis; we must have a
theory as to the reason why the disorder affects behavior in the cognitive
task. Hence, our theories about why cognitive effects occur are important in
interpreting why clinical populations perform the task differently.
List-method DF is one cognitive task that people with clinical diagnoses
sometimes perform differently from other people. Until the past decade,
list-method DF deficits have generally been interpreted in terms of deficits
in inhibition (for earlier reviews of the DF literature in clinical populations,
see Cloitre, 1998; McNally, 2005). Significant DF is taken as evidence that
inhibition is intact, while difficulties in DF are taken as evidence for impair-
ments in inhibitory control. However, our review provides an alternative
interpretation. Some clinical populations may have problems with context
processing or context change. Although not everyone is convinced that the
case for inhibition is well-established even in the other tasks (e.g. MacLeod
et al., 2003; Raaijmakers & Jakab, 2013), many researchers assume that
impairment on tasks like the stop-signal paradigm and RIF reflect inhibi-
tory control, and that deficits on these tasks should be diagnostic of inhibi-
tory deficits. If so, then if DF is also inhibitory in nature, we would expect
that the same inhibitory deficits should occur there as well. We will review
evidence that is consistent with a context-based interpretation of clinical
difficulties in DF, and that other putatively “inhibitory” tasks do not show
the same pattern of deficits as DF, consistent with our theory. The notion
that DF impairments are linked to context processing (and not inhibition)
provides new ways of thinking about how DF elucidates clinical disorders.
Having shown that context may be important to clinical disorders, we
will next argue that many of the studies that have been conducted in the
clinical literature are lacking in rigor, and that improving the m
ethodological
rigor of future studies will be necessary to draw firmer conclusions from DF
research. We will provide specific advice for clinical researchers seeking to
use DF as a tool for understanding context processing in clinical disorders.
Without singling out any earlier study, we will suggest that many earlier
studies could be repeated with more recently developed methodologies
from the cognitive literature in order to yield more detailed conclusions
about disorders.
8.1. Reinterpreting Clinical List-Method DF Results as

Reflecting Context
Given that context is so important in memory theory, it is perhaps surpris-
ing that it is so underrepresented as an explanatory mechanism in clinical
problems. Our ability to link events to their context plays an important role
in understanding aging deficits, for example, where research suggests that
binding items to their context may be extremely difficult for older adults.
However, the same logic has rarely been applied to clinical disorders, where
one might expect quite a lot of traction to be gained from linking disorders
to mental context processing. In the next few sections, we will outline some
disorders where we think that the context view may make sense of disso-
ciations with other tasks that are assumed to reflect inhibition, and perhaps
provide new ways of understanding how context is implicated in clinical
problems.
8.1.1. Attention-Deficit Hyperactivity Disorder

An obvious place to begin the review is with attention-deficit hyperactivity
disorder (ADHD), because it is widely believed that ADHD is in part a
dysexecutive disorder (e.g. Barkley, 1997). A number of studies have argued
that there are inhibitory deficits in ADHD patients, because they perform
poorly in the think/no-think paradigm, stop-signal paradigm, and elsewhere
(for a brief review, see Depue, Burgess, Wilcutt, Ruzic, & Banich, 2010).
Given that pattern of deficits, inhibitory theorists would likely expect

that ADHD should impair list-method DF as well, but that does not
appear to be true. Gaultney, Kipp, Weinstein, and McNeill (1999) com-
pared children aged 8–15 who had ADHD and who did not. Using a stan-
dard list-method paradigm, they obtained both the costs and benefits of
DF with both groups. Thus, ADHD children apparently show normal DF.
White and Marks (2004) compared undergraduates with ADHD to
non-ADHD undergraduates. Participants studied two lists of items using
a procedure similar to Geiselman et al. (1983): Half of the items were
designated as “learn” items that participants were supposed to memorize for
a later test and the other half were “judge” items that they had to rate for
pleasantness without intending to memorize them. The non-ADHD stu-
dents replicated the results of DF costs reported by Geiselman et al. (1983)—
namely, both learned and judged words suffered from DF. The results of the
ADHD group, on the other hand, can be interpreted differently depending
on how they are analyzed. The authors concluded that “individuals in the
ADHD condition did not show reliable evidence of intentional forgetting.”
However, a reexamination of their data suggests a different conclusion: nor-
mal costs (11%) were present for the learn items among ADHD students,
although for judge items the costs were absent (−6%). The pattern with
judge words may be a false negative and may be worth following up in
future work given that several studies have obtained DF in incidental encod-
ing (e.g. Sahakyan & Delaney, 2005, 2010). The results with the learn words
are largely consistent with the Gaultney et al. (1999) finding of intact DF in
ADHD. In sum, while ADHD patients show deficits in other tasks that have
been linked to inhibition, DF appears to be largely intact in this population.
8.1.2. Schizophrenia
Impairments in DF sometimes co-occur with deficits in other inhibitory
tasks. For example, schizophrenia is thought to be associated with defi-
cits in memory and inhibitory control (for reviews, see Hoff & Krennan,
2002; Perlstein, Carter, Barch, & Baird, 1998), and therefore researchers who
believe that DF reflects inhibition should predict deficits in list-method
DF in schizophrenics. Several studies have tested this possibility. In our
judgment, these are among the best-conducted list-method DF studies in
the clinical literature, and should be viewed as models for others.
Racsmány et al. (2008) examined medicated schizophrenics. While the
control group showed robust costs and benefits, the schizophrenic patients
showed neither costs nor benefits. Their overall memory was also lower.
Importantly, in a second experiment, they obtained robust RIF effects,

suggesting dissociation between DF and RIF.
Soriano et al. (2009) examined medicated patients diagnosed with
schizophrenia, schizoaffective disorder, and schizophreniform disorder. They
were split according to whether they reported hallucinations on the P ositive
and Negative Syndrome Scale (PANSS). Although the omnibus statistics did
not always work out due to low power, the planned comparisons suggested
that the costs and benefits were present for both controls and nonhalluci-
nators, whereas the patients who were hallucinating showed neither. Both
groups of patients had impaired memory compared to controls.Thus, Soriano
et al.’s (2009) findings suggest that the presence or absence of hallucinations is
a critical variable in terms of whether schizophrenics show DF or not.
8.1.3. Generalized Anxiety Disorder

Few studies have found any executive functioning difficulties in generalized
anxiety disorder (GAD). A recent study, for example, looked at postpartum
women’s GAD symptomology and found no correlation with a range of
executive functions other than working memory capacity (Vadnais et al.,
2012). Thus, there is not much evidence to suggest that GAD is associated
with impaired inhibition. Nevertheless, a person with high anxiety may find
it aversive to “think of something else” and may have difficulty encoding or
shifting contexts. In other words, anxiety disorders may provide an e xample
of difficulties shifting context despite intact inhibitory abilities. A study
by Albu (2008) used the full design of the list method to examine GAD
patients and patients without anxiety difficulties.Their L1 consisted of anx-
iety-related words and L2 consisted of neutral words. The major problem
with the study is that the authors did not counterbalance which items were
assigned to the forget and remember conditions. Therefore, any differences
between the groups must be interpreted in light of possible differences due
to item effects (we will discuss item effects more fully in Section 8.3). The
non-GAD group showed robust costs (and, unusually, antibenefits). In con-
trast, the GAD group showed neither costs nor benefits.These findings sug-
gest that the GAD patients might have a deficit in context change despite
intact performance on other inhibitory tasks.
8.1.4. Depression
An even more interesting pattern would occur if a clinical problem improved
DF. A study by Power, Dalgeish, Claudio,Tata, and Kentish (2000) examined
subclinically dysphoric people, comparing people with low-Beck Depression
Inventory (BDI) scores to high-BDI scorers (dysphorics). In all their studies,

they used mixed lists of positive and negative adjectives. Their Experiments
1 and 2 are difficult to interpret, as they did not include a remember control
group or control for test order of the lists, but the “R–F” measure was posi-
tive and significant in both groups. However, their Experiment 3 used clini-
cally depressed, high anxious, and healthy controls, and included a remember
control group for each sample. The number of participants in each cell was
extremely small (in some cases as few as five people), making interpretation
difficult. Although most of their effects were in the correct direction for
obtaining the costs and benefits, nothing was significant except an anti-
forgetting effect for negative words among the depressed participants. This
study suggests that negative words may be difficult to forget for depressed
people, perhaps because they trigger rumination. Consistent with a bias to
ruminate on negative words, Experiment 2 found that dysphorics recalled
approximately equal numbers of positive and negative words, whereas the
low-BDI participants showed a positivity bias and favored positive words.
Indeed, negative words are difficult for depressives to update in working
memory, which has led some authors to call them “sticky”—difficult to turn
attention away from (Joormann, Levens, & Gotlib, 2011).
The Power et al. (2000) study therefore does not provide a clean test of
whether depressives show greater DF than the controls. A recent poster by
Lehman and Malmberg (2011b) provided a progress report on their large-
scale study examining depressed peoples’ DF, and suggested that depressives
may be better able to change context. They used the full list-method DF
design with two lists of neutral words, and found that clinically depressed
people showed larger DF impairment than the control participants. Their
argument was that depressed patients may have a built-in mechanism for
producing more context change: depressive rumination.Although Joormann
and Tran (2009) studied people high on rumination in a list-method study,
they used a median split on a rumination scale, and did not find higher BDI
scores among their ruminators. Hence, their failure to obtain enhanced DF
among ruminators (using a no-control-group “R–F” statistic) may not be
surprising.
8.2. Future Directions for Clinical Research Using

the List Method
In sum, ADHD produces impaired executive functioning and poor per-
formance on tasks thought to reflect inhibitory control, but nonetheless
show intact DF. Contrastingly, GAD patients show impaired list-method
DF without clear executive or inhibitory deficits, suggesting a double dis-

sociation between inhibitory/executive measures and DF. Schizophrenia
produces impairments on both putatively inhibitory tasks and on DF (but
see Racsmány et al., 2008 for dissociations between DF and RIF). Finally,
depressed patients may show larger than normal DF, despite some evidence
that they sometimes have impaired working memory. Taken together, the
clinical studies are consistent with our argument that the basis of list-
method DF is not inhibition but contextual change, and that disorders
that affect the likelihood of effectively thinking of something else affect
list-method DF.
These studies all focused on whether or not DF occurs in a p articular
population, but the obvious next step is to ask why DF is impaired. In
this chapter, we have argued that DF is caused by context change, but also
that context change is under volitional control. Impaired DF might reflect
inability to deploy an effective strategy for forgetting. Perseverative difficul-
ties could make it difficult to stop ongoing processing and deploy forget-
ting strategies, for example. It may be useful to include measures of what
people do when attempting to forget (Foster & Sahakyan, 2011), as some
people may fail to initiate context change spontaneously, but could do so
normally with instruction (e.g. in the aging literature; Sahakyan, Delaney, &
Goodmon, 2008). Alternatively, impaired DF could reflect difficulties in
context processing, as was suggested for people with low working memory
capacity (Section 5.12). Superior DF, likewise, could make sense if people
are more able than usual to deploy an effective forgetting strategy—a case
we will suggest might apply to depression (Lehman & Malmberg, 2011b).
One possible method of getting at whether generalized context-change
problems are typical of a disorder or whether it is a problem of initiating
context change on one’s own may be to use diversionary thought tasks like
imagining your childhood home (Section 5.1).
That depressives seem more effective than others at forgetting neutral
words (Lehman & Malmberg, 2011b) but not mixed lists of negative and
positive words (Power et al., 2000) further suggests that forgetting may be
dependent on the type of material being forgotten (McNally, Metzger,
Lasko, Clancy, & Pitman, 1998). Different types of material could trigger
thoughts that produce forgetting or could fail to do so. Studies contrasting
forgetting of neutral and disorder-related words may be informative in other
disorders as well, especially if they include appropriate remember control
groups to facilitate differentiating between generalized memory biases and
biases specific to forgetting.
8.3. Improving Research Design for Clinical Studies

of List-Method DF
One problem with interpreting studies on list-method DF is that many of
them are based on older research designs that provide data that are often
misinterpreted. Hence, our next task is to suggest list-method designs that
provide the best chance of accurately answering research questions about
clinical populations’ DF. Whether or not one accepts that DF is context-
based, it is a good idea to use a design that can discriminate different theoret-
ical possibilities.We have outlined our specific recommendations as Table 4.1.
8.3.1. Inclusion of a Separate Remember Condition

As noted in Section 2.1, a common design issue in DF studies is the absence
of a remember control group. In such studies, the costs and benefits of DF
have been assessed using the “R–F” measure, which can be a problem-
atic assessment for DF effects. This measurement issue is particularly prob-
lematic when assessing DF in clinical populations. Specifically, if a clinical
population is impaired on the costs but shows normal benefits, this would
be impossible to distinguish using the “R–F” measure. According to the
theory outlined in this chapter, smaller costs but normal benefits suggest
difficulties with the use of context. Intact costs but impaired benefits suggest
strategy-change issues.Thus, including a remember control group allows for
a systematic assessment of the costs and benefits of DF.
It is understandable that including remember control groups when patient
populations are scarce may be difficult. For this reason, many designs rely on
comparing “R–F” in healthy vs patient populations. We think these studies
are informative in that they indicate differences in DF between patients and
healthy controls. However, one has to be very careful about interpreting
them in terms of impaired inhibition (or even impaired context change),
because both costs and benefits are rolled into the “R–F” differences. A pos-
sible solution is to adopt the four-list procedure (Section 2.2). Especially if
Table 4.1 Suggestions for Improved List-Method Directed Forgetting Designs

1. Use a remember control group—or at least the within-subjects design that
includes both RR and FR lists
2. Test list 1 and list 2 separately—or at least calculate output percentiles
3. Consider output order—if a study mixes different types of items (e.g. trauma
and neutral words), examine output order effects
4. Avoid or acknowledge scaling effects—baseline memory may differ across groups
5. Avoid floor effects—use appropriate numbers of items in each category
one is interested in the costs, there is little evidence that the within-subjects
design distorts the effect, as reviewed earlier. It is important to counterbal-
ance the assignment of words to lists as well; at least one study obtained odd
results because they failed to ensure that the words assigned to the forget
condition were assigned to the remember condition for other participants,
and hence item effects emerged (Albu, 2008).The four-list procedure allows
the researcher to compute both costs and benefits using a powerful within-
subjects comparison, which is clearly advantageous. If there are concerns
about order effects, they can also be addressed using this design. If there are
serious order effects, one can also focus on only the first lists studied.
8.3.2. Controlling Test Order and Measuring Output Order

Another design issue is that most clinical studies have not controlled the
test order of L1 and L2. In Section 2.1, we discussed the problem of output
interference in the benefits. While research suggests that in healthy controls
the costs are relatively unaffected by output interference effects (Pastötter
et al. 2012), this may not necessarily be the case for patients.There could be
systematic differences between patients and healthy controls as to whether
they start recall from L1 or L2 for a variety of reasons.
Controlling test order may be particularly important when mixing sev-
eral different types of items on the same list, as when one is interested
in whether trauma or neutral words are harder to forget. Comparing the
relative recall rate of different item types will be complicated by output
interference, especially if different groups of participants output them in dif-
ferent orders. For example, if we mix trauma and neutral words on each list,
then patients might begin recall from all the trauma words, whereas healthy
controls might have no such bias. Such differences in output order would
produce varying levels of output interference on different types of items
across the groups, resulting in apparent differences in recall rates. Hence, we
suggest that output order of the lists be controlled whenever possible.
At the very least, one should examine output order as a possible expla-
nation for any observed results. One way to examine output order is to
calculate the average output positions.The easiest way to do this is to create
output percentiles, which examine the average position within the output
of the items of a given type (e.g. Sahakyan, Delaney, & Waldum, 2008). For
example, imagine we mix trauma and neutral words, and the output items
are in the order trauma, trauma, trauma, neutral, trauma, neutral, neutral. T
he out-
put percentile for trauma words would be (1 + 2 + 3 + 5)/(7 × 4) = 0.39.
For neutral words, it would be (4 + 6 + 7)/(7 × 3) = 0.81. Thus, output for
neutral words is occurring later than for trauma words for this participant.
Further, significant differences in output percentiles between groups would
be evidence for differences in output order of different types of items.
8.3.3. Scaling Effects

Another common problem when comparing groups of people is a scaling
effect (also known as baseline effects). Some groups of people show general
memory impairments, and hence have lower baseline memory. For example,
this problem is well-known in the literature on aging—older adults have
impaired recollection, and so they perform worse than younger adults on
free recall tests. In this case, the costs might be numerically smaller in older
adults than in younger adults due to reduced overall memory. If the costs are
a percentage of the original memory, then Analysis of V ariance (ANOVA)
might incorrectly suggest that the costs are smaller for older adults than for
younger adults. One solution is either to enhance the memory of the weaker
group, such as by giving longer presentation times, or to provide more
elaborate processing instructions before the forget cue, especially because
neither study time nor depth of processing interacts with DF (Sahakyan,
Delaney, & Goodmon, 2008; Sahakyan, Delaney, & Waldum, 2008). If the
baseline recall rates are closer in the remember control group, then the
differences in the costs may be more easily interpreted. Of course, if the
costs are completely absent, it is more difficult to argue for a scaling effect
(unless there is a floor or ceiling effect in the data).
8.3.4. Floor Effects

Many clinical studies also show floor effects in recall.The standard diagnosis
for floor effects is that the confidence interval around the mean captures
0%. Another good indicator of floor effects is if a significant percentage of
participants recall zero of a given type of item.When the range of responses
is very small (e.g. 0, 1, 2 or 3), then a lot of participants would be needed
to detect a meaningful difference in the means, as the measurement error is
likely to exceed the size of the effect. Whenever there is an apparent floor
effect, one should be prudent when interpreting mean differences, as the
estimate of the variance is likely to be deflated, producing false positives.
It is difficult to avoid floor effects in recall, but a good start is to avoid
very long lists; not more than 20 items are likely to yield reasonable recall
rates. Likewise, if the items are fractionated into different types of items,
such as traumatic and neutral words, then one should try to keep these
sublists from getting too short. We suggest a minimum of eight items per
category. Of course, this restricts the kinds of tests that can be run with a
single design, but it is better to get meaningful answers about one or two
different types of item than to get meaningless floor effects with a large
number of different types of item.
8.4. Summary and Linking Clinical to Nonclinical

Research on DF
While we have argued that the current theory can explain many of the
differences in DF among clinical populations, in some ways, studies with
special populations provide the best evidence for individual differences in
DF. Even if contextual change can account for DF impairment in everyday
people, it may be less successful among some clinical groups for a variety
of reasons. The absence of DF in some clinical populations may provide
clues as to why some seemingly healthy participants show little or no DF.
Likewise, if some clinical populations show more effective DF than healthy
participants, it may give us some insight into why some healthy participants
show larger-than-usual DF effects. For example, it may well be that people
who ruminate more or people who are fantasy-prone are better forgetters
even in the absence of any clinical symptoms.
Clinical research could also shed light on the controversies surrounding
what types of items are more or less forgettable. Some populations have
difficulties related to specific types of words, which could enable us to
better understand how differential processing interacts with the forget

cue. Thus, more insights about the nature of DF can be gained through
well-designed studies with clinical populations. Although basic research
provides a foundation for interpreting population differences, research with
special populations is equally necessary for the future development and
refinement of DF theories. We are hopeful that by incorporating better,
modern research designs that were developed in the cognitive literature to
investigate DF that deeper conclusions can be drawn about context pro-
cessing in various clinical populations. Repeating some of the earlier stud-
ies with these modern designs should be helpful in discriminating between
different theoretical possibilities, and shed light on the nature of several
important clinical disorders.
9. CONCLUDING THOUGHTS
Obtaining DF requires engaging in controlled strategies, and the deci-
sion to engage in those strategies may be mediated by beliefs about one’s
memory abilities or whether it is possible to intentionally forget. These

findings highlight the importance of attending to participants’ behavior and
beliefs in understanding why memory phenomena occur. The costs and
the benefits of DF can be observed or not depending on what participants
choose to do. A closer attention to participant strategies is warranted in DF.
The context account as laid out here explains a substantial amount of DF
data, while also serving as foundation for testable predictions regarding the
impact of different manipulations on DF.The context-change mechanism is
not uniquely suited to explain DF. Instead, the account brings DF into the
family of effects that are explained by existing memory models, allowing
us to make new testable predictions. These models and the rich empirical
foundation of the environmental context literature serve as a foundation for
new quantitative and qualitative predictions.
The traditional way of interpreting DF deficits in many special popula-
tions is that they have deficits in inhibitory control. The context account of
DF, however, suggests new interpretations, opening up new directions for
investigating context processing and/or contextual binding deficits in those
populations.
Finally, a commonsense notion is that disruptions are inherently bad for
memory. Work on DF and mental context-change paradigm suggests that
how disruptive interruptions depend on the contents of the interruption
and the extent to which they change mental context. Some interruptions
are easier to recover from than others. A major challenge that lies ahead is to
more fully understand why some tasks but not others create mental context
change—in other words, to create a theory of what mental context consists
of. We look forward to seeing how these theories develop in the next few
decades. Understanding how mental context is represented and updated
may turn out to be central to memory functioning in many more ways than
we yet realize.
ACKNOWLEDGMENTS
The authors are grateful to Colleen Kelley, Colin MacLeod, and Tanya Jonker for their feed-
back and constructive comments on the earlier draft of this chapter.
REFERENCES
Albu, M. (2008). Automatic and intentional inhibition in patients with generalized anxiety
disorder. Cognition, Brain, & Behavior, 12, 233–249.
Anderson, M. C. (2003). Rethinking interference theory: executive control and the
mechanisms of forgetting. Journal of Memory and Language, 49(4), 415–445.
Anderson, M. C. (2005). The role of inhibitory control in forgetting unwanted memories:
a consideration of three methods. In C. MacLeod & B. Uttl (Eds.), Dynamic cognitive
processes (pp. 159–190). Tokyo: Springer-Verlag.
Anderson, M. C., Bjork, R. A., & Bjork, E. L. (1994). Remembering can cause forgetting:
retrieval dynamics in long-term memory. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 20, 1063–1087.
Anderson, J. R., & Bower, G. H. (1972). Recognition and retrieval processes in free recall.
Psychological Review, 79(2), 97–123.
Aslan, A., & Bäuml, K. T. (2008). Memorial consequences of imagination in children and
adults. Psychonomic Bulletin & Review, 15(4), 833–837.
Aslan, A., & Bäuml, K. T. (2011). Individual differences in working memory capacity pre-
dict retrieval-induced forgetting. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 37(1), 264–269.
Aslan, A., Zellner, M., & Bäuml, K. T. (2010). Working memory capacity predicts listwise
directed forgetting in adults and children. Memory, 18(4), 442–450.
Barkley, R. A. (1997). Behavioral inhibition, sustained attention, and executive functions:
Constructing a unifying theory of ADHD. Psychological Bulletin, 121, 65–94.
Barnier, A. J., Conway, M. A., Mayoh, L., Speyer, J., Avizmil, O., & Harris, C. B. (2007).
Directed forgetting of recently recalled autobiographical memories. Journal of Experimen-
tal Psychology: General, 136(2), 301–322.
Basden, B. H., & Basden, D. R. (1996). Directed forgetting: further comparisons of the item
and list methods. Memory, 4(6), 633–653.
Basden, B. H., & Basden, D. R. (1998). Directed forgetting: a contrast of methods and inter-
pretations. In J. M. Golding & C. M. MacLeod (Eds.), Intentional forgetting: Interdisciplinary
approaches (pp. 139–172). Mahwah, NJ US: Lawrence Erlbaum Associates Publishers.
Basden, B. H., Basden, D. R., & Gargano, G. J. (1993). Directed forgetting in implicit and
explicit memory tests: a comparison of methods. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 19(3), 603–616.
Basden, B. H., Basden, D. R., & Morales, E. (2003).The role of retrieval practice in directed for-
getting. Journal of Experimental Psychology: Learning, Memory, and Cognition, 29(3), 389–397.
Basden, B. H., Basden, D. R., & Wright, M. J. (2003). Part-list reexposure and release of
retrieval inhibition. Consciousness and Cognition: An International Journal, 12(3), 354–375.
Bäuml, K. H. (2008). Inhibitory processes. In H. L. RoedigerIII (Ed.), Cognitive psychology of
memory (pp. 195–220). Oxford: Elsevier.
Bäuml, K.-H., Hanslmayr, S., Pastotter, B., & Klimesch, W. (2008). Oscillatory correlates of
intentional updating in episodic memory. NeuroImage, 41, 596–604.
Bäuml, K. T., & Samenieh, A. (2010). The two faces of memory retrieval. Psychological Science,
21(6), 793–795.
Bäuml, K.T., & Samenieh, A. (2012a). Influences of part-list cuing on different forms of episodic
forgetting. Journal of Experimental Psychology: Learning, Memory, and Cognition, 38(2), 366–375.
Bäuml, K. T., & Samenieh, A. (2012b). Selective memory retrieval can impair and improve
retrieval of other memories. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 38(2), 488–494.
Benjamin, A. S. (2006). The effects of list-method directed forgetting on recognition mem-
ory. Psychonomic Bulletin & Review, 13, 831–836.
Benjamin, A. S., & Bjork, R. A. (2000). On the relationship between recognition speed and
accuracy for words rehearsed via rote versus elaborative rehearsal. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 26(3), 638–648.
Bjork, R. A. (1970). Positive forgetting: the noninterference of items intentionally forgotten.
Journal of Verbal Learning and Verbal Behavior, 9, 255–268.
Bjork, R. A. (1972). Theoretical implications of directed forgetting. In A. W. Melton &
E. Martin (Eds.), Coding processes in human memory (pp. 217–325). Washington, D. C.:
V. H. Winston & Sons.
Bjork, R. A. (1989). Retrieval inhibition as an adaptive mechanism in human memory. In
H. Roediger & F. M. Craik (Eds.), Varieties of memory and consciousness: Essays in honour
of Endel Tulving (pp. 309–330). Hillsdale, NJ England: Lawrence Erlbaum Associates, Inc.
Bjork, E. L., & Bjork, R. A. (1996). Continuing influences of to-be-forgotten information.

Consciousness and Cognition, 5, 176–196.
Bjork, E. L., & Bjork, R. A. (2003). Intentional forgetting can increase, not decrease, residual
influences of to-be-forgotten information. Journal of Experimental Psychology: Learning,
Bjork, E., Bjork, R. A., & Anderson, M. C. (1998). Varieties of goal-directed forgetting. In
J. M. Golding & C. M. MacLeod (Eds.), Intentional forgetting: Interdisciplinary approaches
(pp. 103–137). Mahwah, NJ US: Lawrence Erlbaum Associates Publishers.
Bjork, R. A., & Richardson-Klavehn, A. (1989). On the puzzling relationship between envi-
ronmental context and human memory. In C. Izawa (Ed.), Current issues in cognitive
processes:The Tulane Flowerree symposium on cognition (pp. 313–344). Hillsdale, NJ England:
Lawrence Erlbaum Associates, Inc.
Block, R. A. (1971). Effects of instructions to forget in short-term memory. Journal of
Experimental Psychology, 89(1), 1–9.
Cloitre, M. (1998). Intentional forgetting and clinical disorders. In J. M. Golding & C. M.
MacLeod (Eds.), Intentional forgetting: Interdisciplinary approaches (pp. 395–412). Mahwah,
NJ US: Lawrence Erlbaum Associates Publishers.
Conway, M. A., & Fthenaki, A. (2003). Disruption of inhibitory control of memory follow-
ing lesions to the frontal and temporal lobes. Cortex: A Journal Devoted to the Study of the
Nervous System and Behavior, 39(4–5), 667–686.
Conway, M. A., Harries, K., Noyes, J., Racsmány, M., & Frankish, C. R. (2000). The dis-
ruption and dissolution of directed forgetting: inhibitory control of memory. Journal of
Craik, F. I., & Watkins, M. J. (1973). The role of rehearsal in short-term memory. Journal of
Verbal Learning and Verbal Behavior, 12(6), 599–607.
Dalton, P. (1993).The role of stimulus familiarity in context-dependent recognition. Memory
& Cognition, 21(2), 223–234.
Delaney, P. F., & Knowles, M. E. (2005). Encoding strategy changes and spacing effects in the
free recall of unmixed lists. Journal of Memory and Language, 52(1), 120–130.
Delaney, P. F., & Sahakyan, L. (2007). Unexpected costs of high working memory capacity
following directed forgetting and context change manipulations. Memory & Cognition,
35, 1074–1082.
Delaney, P. F., Sahakyan, L., Kelley, C. M., & Zimmerman, C. A. (2010). Remembering to
forget: the amnesic effect of daydreaming. Psychological Science, 21(7), 1036–1042.
Dennis, S., & Humphreys, M. S. (2001). A context noise model of episodic word recognition.
Psychological Review, 108, 452–478.
Depue, B. E. (2012). A neuroanatomical model of prefrontal inhibitory modulation of
memory retrieval. Neuroscience and Biobehavioral Reviews, 36(5), 1382–1399.
Depue, B. E., Burgess, G. C., Willcutt, E. G., Ruzic, L., & Banich, M. T. (2010). Inhibitory
control of memory retrieval and motor processing associated with the right lateral pre-
frontal cortex: evidence from deficits in individuals with ADHD. Neuropsychologia, 48,
3909–3917.
Dong, T. (1972). Probe versus free recall. Journal of Verbal Learning and Verbal Behavior, 11(5),
654–661.
Eich, E. (1985). Context, memory, and integrated item/context imagery. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 11(4), 764–770.
Elmes, D. G., Adams, C., & Roediger, H. L. (1970). Cued forgetting in short-term memory:
response selection. Journal of Experimental Psychology, 86(1), 103–107.
Estes, W. K. (1955). Statistical theory of spontaneous recovery and regression. Psychological
Review, 62(3), 145–154.
Fernandez, A., & Glenberg, A. M. (1985). Changing environmental context does not reliably
affect memory. Memory & Cognition, 13(4), 333–345.
Foster, N. L., & Sahakyan, L. (2011). The role of forget-cue salience in list-method directed
forgetting. Memory, 19, 110–117.
Foster, N. L., & Sahakyan, L. (2012). Metacognition influences item-method directed forget-
ting. Journal of Experimental Psychology: Learning, Memory, and Cognition, 38, 1309–1324.
Gaultney, J. F., Kipp, K., Weinstein, J., & McNeill, J. (1999). Inhibition and mental effort in
attention deficit hyperactivity disorder. Journal of Developmental and Physical Disabilities,
11(2), 105–114.
Geiselman, R. E., & Bjork, R. A. (1980). Primary versus secondary rehearsal in imagined
voices: differential effects on recognition. Cognitive Psychology, 12(2), 188–205.
Geiselman, R. E., Bjork, R. A., & Fishman, D. (1983). Disrupted retrieval in directed forget-
ting: a link with posthypnotic amnesia. Journal of Experimental Psychology: General, 112,
58–72.
Gillund, G., & Shiffrin, R. M. (1984). A retrieval model for both recognition and recall.
Psychological Review, 91(1), 1–67.
Godden, D. R., & Baddeley, A. D. (1975). Context-dependent memory in two natural envi-
ronments: on land and underwater. British Journal of Psychology, 66(3), 325–331.
Goernert, P. N., & Larson, M. E. (1994). The initiation and release of retrieval inhibition.
Journal of General Psychology, 121, 61–66.
Golding, J. M., & Gottlob, L. R. (2005). Recall order determines the magnitude of directed
forgetting in the within-participants list method. Memory & Cognition, 33(4), 588–594.
Golding, J. M., & Keenan, J. M. (1985). Directed forgetting and memory for directions to a
destination. American Journal of Psychology, 98, 579–590.
Golding, J. M., Long, D. L., & MacLeod, C. M. (1994).You can’t always forget what you want:
directed forgetting of related words. Journal of Memory and Language, 33, 493–510.
Golding, J. M., & MacLeod, C. M. (1998). Intentional forgetting: Interdisciplinary
approaches. Mahwah, NJ US: Lawrence Erlbaum Associates Publishers.
Gottlob, L. R., & Golding, J. M. (2007). Directed forgetting in the list method affects recogni-
tion memory for source. Quarterly Journal of Experimental Psychology, 60(11), 1524–1539.
Hanczakowski, M., Pasek, T., & Zawadska, K. (2012). Context-dependent impairment of
recollection in list-method directed forgetting. Memory.
Hanslmayr, S.,Volberg, G., Wimber, M., Oehler, N., Staudigl, T., HYartmann, T., et al. (2012).
Prefrontally driven downregulation of neural synchrony medicated goal-directed
forgetting. Journal of Neuroscience.
Hicks, J. L., & Starns, J. J. (2004). Retrieval-induced forgetting occurs in tests of item
recognition. Psychonomic Bulletin & Review, 11(1), 125–130.
Hintzman, D. L. (1988). Judgments of frequency and recognition memory in a multiple-trace
memory model. Psychological Review, 95(4), 528–551.
Hoff, A. L., & Krennan, W. S. (2002). Is there a cognitive phenotype for schizophrenia: the
nature and course of the disturbance in cognition? Current Opinion in Psychiatry, 15,
43–48.
Howard, M. W., & Kahana, M. J. (2002). A distributed representation of temporal context.
Journal of Mathematical Psychology, 46(3), 269–299.
Hyde, T. S., & Jenkins, J. J. (1973). Recall for words as a function of semantic, graphic, and
syntactic orienting tasks. Journal of Verbal Learning and Verbal Behavior, 12(5), 471–480.
Isarida, T., & Morii, Y. (1986). Contextual dependence of the spacing effect in free recall.
Japanese Journal of Psychology, 57(1), 20–26.
Jang,Y., & Huber, D. E. (2008). Context retrieval and context change in free recall: recalling
from long-term memory drives list isolation. Journal of Experimental Psychology: Learning,
Joormann, J., Levens, S. M., & Gotlib, I. H. (2011). Sticky thoughts: depression and rumi-
nation are associated with difficulties manipulating emotional material in working
memory. Psychological Science, 22(8), 979–983.
Joormann, J., & Tran, T. B. (2009). Rumination and intentional forgetting of emotional
material. Cognition and Emotion, 23(6), 1233–1246.
Joslyn, S. L., & Oakes, M. A. (2005). Directed forgetting of autobiographical events. Memory
& Cognition, 33(4), 577–587.
Kimball, D. R., & Bjork, R. A. (2002). Influences of intentional and unintentional forgetting
on false memories. Journal Of Experimental Psychology: General, 131, 116–130.
Koppel, R. H., & Storm, B. C. (2012). Unblocking memory through directed forgetting.
Journal of Cognitive Psychology.
Krafka, C., & Penrod, S. (1985). Reinstatement of context in a field experiment on eyewit-
ness identification. Journal of Personality and Social Psychology, 49(1), 58–69.
Lehman, M., & Malmberg, K. J. (2009). A global theory of remembering and forgetting from
multiple lists. Journal of Experimental Psychology: Learning, Memory, and Cognition, 35(4),
970–988.
Lehman, M., & Malmberg, K. J. (2011a). Overcoming the effects of intentional forgetting.
Memory & Cognition, 39(2), 335–347.
Lehman, M., & Malmberg, K. J. (2011b, November). Modeling intentional forgetting
in depressed participants. (Poster presentation at the 52nd annual meeting of the
Psychonomic Society, Seattle, WA).
Liu, X. (2001). On the dynamics of directed forgetting: facilitation and interference in the
updating of human memory. Dissertation Abstracts International, 61(11), 6159B.
Macken, W. J. (2002). Environmental context and recognition: the role of recollection and
familiarity. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28(1),
153–161.
MacLeod, C. M. (1975). Long-term recognition and recall following directed forgetting.
Journal of Experimental Psychology: Human Learning and Memory, 1, 271–279.
MacLeod, C. M. (1998). Directed forgetting. In J. M. Golding & C. M. MacLeod (Eds.),
Intentional forgetting: Interdisciplinary approaches (pp. 1–57). Mahwah, NJ: Erlbaum.
MacLeod, C. M. (1999).The item and list methods of directed forgetting: test differences and
the role of demand characteristics. Psychonomic Bulletin & Review, 6(1), 123–129.
MacLeod, C. M., Dodd, M. D., Sheard, E. D., Wilson, D. E., & Bibi, U. (2003). In opposition
to inhibition. In B. H. Ross (Ed.), The psychology of learning and motivation: Advances in
research and theory (Vol. 43, pp. 163–214). New York, NY US: Elsevier Science.
Macrae, C., Bodenhausen, G.V., Milne, A. B., & Ford, R. L. (1997). On regulation of recollec-
tion: the intentional forgetting of stereotypical memories. Journal of Personality and Social
Psychology, 72(4), 709–719.
Malmberg, K. J., & Shiffrin, R. M. (2005). The ‘One-Shot’ hypothesis for context storage.
Journal of Experimental Psychology: Learning, Memory, and Cognition, 31(2), 322–336.
Malpass, R. S., & Devine, P. G. (1981). Guided memory in eyewitness identification. Journal
of Applied Psychology, 66(3), 343–350.
Mandler, G. (1967). Organization and memory. In K. W. Spence & J. T. Spence (Eds.), The
psychology of learning and motivation. (Vol. 1). New York: Academic Press.
Marsh, R. L., Meeks, J., Hicks, J. L., Cook, G. I., & Clark-Foos, A. (2006). Concreteness and
item-to-list context associations in the free recall of items differing in context variability.
McNally, R. J. (2005). Directed forgetting tasks in clinical research. In A. Wenzel & D. C.
Rubin (Eds.), Cognitive methods and their application to clinical research (pp. 197–212).
Washington, DC: American Psychological Association.
McNally, R. J., Metzger, L. J., Lasko, N. B., Clancy, S. A., & Pitman, R. K. (1998). Directed
forgetting of trauma cues in adult survivors of childhood sexual abuse with and without
posttraumatic stress disorder. Journal of Abnormal Psychology, 107, 596–601.
Mensink, G., & Raaijmakers, J. G. (1988). A model for interference and forgetting. Psychologi-
cal Review, 95, 434–455.
Minnema, M. T., & Knowlton, B. J. (2008). Directed forgetting of emotional words. Emotion,
8(5), 643–652.
Mulji, R., & Bodner, G. E. (2010). Wiping out memories: new support for a mental context
change account of directed forgetting. Memory, 18, 763–773.
Nelson, D. L., Goodmon, L. B., & Ceo, D. (2007). How does delayed testing reduce effects of
implicit memory: context infusion or cuing with context? Memory & Cognition, 35(5),
1014–1023.
Nelson, D., & McEvoy, C. L. (2005). Implicitly activated memories: the missing links of
remembering. In C. Izawa & N. Ohta (Eds.), Human learning and memory: Advances in
theory and application: The 4th Tsukuba International Conference on memory (pp. 177–198).
Mahwah, NJ US: Lawrence Erlbaum Associates Publishers.
Nelson, D. L., McKinney, V. M., Gee, N. R., & Janczura, G. A. (1998). Interpreting the
influence of implicitly activated memories on recall and recognition. Psychological Review,
105(2), 299–324.
Parker, A., Gellatly, A., & Waterman, M. (1999). The effect of environmental context manip-
ulation on memory: dissociation between perceptual and conceptual implicit tests.
European Journal of Cognitive Psychology, 11(4), 555–570.
Pastötter, B., & Bäuml, K.-H. (2007). The crucial role of post-cue encoding in directed
forgetting and context-dependent forgetting. Journal of Experimental Psychology: Learning,
Pastötter, B., & Bäuml, K. (2010). Amount of postcue encoding predicts amount of directed
forgetting. Journal of Experimental Psychology: Learning, Memory, and Cognition, 36(1), 54–65.
Pastötter, B., Bäuml, K. H., & Hanslmayr, S. (2008). Oscillatory brain activity before and after
an internal context change—evidence for a reset of encoding processes. NeuroImage, 43,
173–181.
Pastötter, B., Kliegl, O., & Bäuml, K. T. (2012). List-method directed forgetting: the forget
cue improves both encoding and retrieval of postcue information. Memory & Cognition,
40(6), 861–873.
Perlstein, W. M., Carter, C. S., Barch, D. M., & Baird, J. (1998). The Stroop task and
attention deficits in schizophrenia: a critical evaluation of card and single-trial Stroop
methodologies. Neuropsychology, 12, 414–425.
Postman, L. (1964). Acquisition and retention of consistent associative responses. Journal of
Experimental Psychology, 67(2), 183–190.
Power, M. J., Dalgleish, T. T., Claudio, V. V., Tata, P. P., & Kentish, J. J. (2000). The directed
forgetting task: Application to emotionally valent material. Journal Of Affective Disorders,
57, 147–157.
Raaijmakers, J. G. W., & Jakab, E. (2013). Rethinking inhibition theory: On the problematic
status of the inhibition theory for forgetting. Journal of Memory and Language, 68, 98–122.
Raaijmakers, J. G. W., & Shiffrin, R. M. (1981). Search of associative memory. Psychological
Review, 88, 93–134.
Racsmány, M., & Conway, M. A. (2006). Episodic inhibition. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 32, 44–57.
Racsmány, M., Conway, M. A., Garab, E. A., Cimmer, C., Janka, Z., Kurimay,T., et al. (2008).
Disrupted memory inhibition in schizophrenia. Schizophrenia Research, 101(1–3),
218–224.
Racsmány, M., Conway, M. A., Garab, E. A., & Nagymáté, G. (2008). Memory awareness
following episodic inhibition. Quarterly Journal of Experimental Psychology, 61(4), 525–534.
Reitman, W., Malin, J. T., Bjork, R. A., & Higman, B. (1973). Strategy control and directed
forgetting. Journal Of Verbal Learning & Verbal Behavior, 12, 140–149.
Roediger, H. L. (1974). Inhibiting effects of recall. Memory & Cognition, 2(2), 261–269.
Roediger, H., & Karpicke, J. D. (2006).Test-enhanced learning: taking memory tests improves
long-term retention. Psychological Science, 17(3), 249–255.
Roediger, H. L., & Schmidt, S. R. (1980). Output interference in the recall of categorized
and paired-associate lists. Journal of Experimental Psychology: Human Learning and Memory,
6(1), 91–105.
Russo, R., Ward, G., Geurts, H., & Scheres, A. (1999). When unfamiliarity matters: chang-
ing environmental context between study and test affects recognition memory for
unfamiliar stimuli. Journal of Experimental Psychology: Learning, Memory, and Cognition,
25(2), 488–499.
Sahakyan, L. (2004). Destructive effects of ‘forget’ instructions. Psychonomic Bulletin & Review,
11(3), 555–559.
Sahakyan, L., & Delaney, P. F. (2003). Can encoding differences explain the benefits of directed
forgetting in the list-method paradigm? Journal of Memory and Language, 48, 195–201.
Sahakyan, L., & Delaney, P. F. (2005). Directed forgetting in incidental learning and recogni-
tion testing: support for a two-factor account. Journal of Experimental Psychology: Learning,
Sahakyan, L., & Delaney, P. F. (2010). Item-specific encoding produces an additional benefit
of directed forgetting: evidence from intrusion errors. Journal of Experimental Psychology:
Sahakyan, L., Delaney, P. F., & Goodmon, L. B. (2008). “Oh, honey, I already forgot that”:
strategic control of directed forgetting in older and younger adults. Psychology and Aging,
23, 621–633.
Sahakyan, L., Delaney, P. F., & Kelley, C. M. (2004). Self-evaluation as a moderating factor
in strategy change in directed forgetting benefits. Psychonomic Bulletin & Review, 11,
131–134.
Sahakyan, L., Delaney, P. F., & Waldum, E. R. (2008). Intentional forgetting is easier after two
‘shots’ than one. Journal of Experimental Psychology: Learning, Memory, and Cognition, 34(2),
408–414.
Sahakyan, L., & Foster, N. L. (2009). Intentional forgetting of action: comparison of list-
method and item-method directed forgetting. Journal of Memory and Language, 61, 134–152.
Sahakyan, L., & Goodmon, L. B. (2007).The influence of directional associations on directed
forgetting and interference. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 33, 1035–1049.
Sahakyan, L., & Goodmon, L. B. (2010). Theoretical implications of extralist probes for
directed forgetting. Journal of Experimental Psychology: Learning, Memory, and Cognition,
36(4), 920–937.
Sahakyan, L., & Kelley, C. M. (2002). A contextual change account of the directed forgetting
effect. Journal of Experimental Psychology: Learning, Memory, and Cognition, 28, 1064–1072.
Sahakyan, L., Waldum, E. R., Benjamin, A. S., & Bickett, S. P. (2009). Where is the forget-
ting with list-method directed forgetting in recognition? Memory & Cognition, 37(4),
464–476.
Shapiro, S., Lindsey, C., & Krishnan, H. (2006). Intentional forgetting as a facilitator for recalling
new product attributes. Journal of Experimental Psychology: Applied, 12(4), 251–263.
Sheard, E. D., & MacLeod, C. M. (2005). List method directed forgetting: return of the
selective rehearsal account. In N. Ohta, C. M. MacLeod & B. Uttl (Eds.), Dynamic
cognitive processes (pp. 219–248). Tokyo: Springer.
Smith, S. M. (1979). Remembering in and out of context. Journal of Experimental Psychology:
Human Learning and Memory, 5, 460–471.
Smith, E. E., Barresi, J., & Gross, A. E. (1971). Imaginal versus verbal coding and the
primary-secondary memory distinction. Journal of Verbal Learning and Verbal Behavior,
10(6), 597–603.
Smith, S. M., & Vela, E. (1992). Environmental context-dependent eyewitness recognition.
Applied Cognitive Psychology, 6, 125–139.
Smith, S. M., & Vela, E. (2001). Environmental context-dependent memory: a review and
meta-analysis. Psychonomic Bulletin & Review, 8, 203–220.
Soriano, M. F., & Bajo, M.T. (2007).Working memory resources and interference in directed-
forgetting. Psicológica, 28, 63–85.
Soriano, M. F., Jiménez, J. F., Román, P., & Bajo, M. T. (2009). Intentional inhibition in
memory and hallucinations: directed forgetting and updating. Neuropsychology, 23,
61–70.
Spillers, G. J., & Unsworth, N. (2011). Are the costs of directed forgetting due to failures
of sampling or recovery? Exploring the dynamics of recall in list-method directed
forgetting. Memory & Cognition, 39(3), 403–411.
Spitzer, B., & Bäuml, K. (2007). Retrieval-induced forgetting in item recognition: evidence
for a reduction in general memory strength. Journal of Experimental Psychology: Learning,
Memory, and Cognition, 33(5), 863–875.
Spitzer, B., & Bäuml, K. (2009). Retrieval-induced forgetting in a category recognition task.
Steyvers, M., & Malmberg, K. J. (2003). The effects of normative context variability on rec-
ognition memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 29,
760–766.
Storm, B. C., & Levy, B. J. (2012). A progress report on the inhibitory account of
retrieval-induced forgetting. Memory & Cognition, 40(6), 827–843.
Tulving, E. (1983). Elements of episodic memory. New York: Oxford.
Tulving, E., & Arbuckle, T.Y. (1966). Input and output interference in short-term associative
memory. Journal of Experimental Psychology, 72, 145–150.
Unsworth, N., & Engle, R. W. (2008). Speed and accuracy of accessing information in
working memory: an individual differences investigation of focus switching. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 34(3), 616–630.
Unsworth, N., Spillers, G. J., & Brewer, G. A. (2012). Dynamics of context-dependent recall:
an examination of internal and external context change. Journal of Memory and Language,
66(1), 1–16.
Vadnais, S. A., Behm, A., Laake, L. M., Lopez, N. M., Oddi, K. B., Wu, K., et al. (2012,
May). Executive function correlates of symptoms of specific anxiety disorders and major
depression. (Poster presented at the 24th Annual Convention of the Association for
Psychological Science). Chicago, IL.
Veling, H., & van Knippenberg, A. (2004). Remembering can cause inhibition: retrieval-
induced inhibition as cue independent process. Journal of Experimental Psychology:
Verde, M. F. (2004). The retrieval practice effect in associative recognition. Memory &
Cognition, 32, 1265–1272.
Waldum, E. R., & Sahakyan, L. (2012). Putting congeniality effects into context: investigating
the role of context in attitude memory using multiple paradigms. Journal of Memory and
Language, 66(4), 717–730.
Wessel, I., & Merckelbach, H. (2006). Forgetting “murder” is not harder than forgetting
“circle”: listwise-directed forgetting of emotional words. Cognition and Emotion, 20,
129–137.
Whetstone, T., Cross, M. D., & Whetstone, L. M. (1996). Inhibition, contextual segregation,
and subject strategies in list method directed forgetting. Consciousness and Cognition, 5,
395–417.
White, H. A., & Marks, W. (2004). Updating memory in list-method directed forgetting:
individual differences related to adult attention-deficit/hyperactivity disorder. Personality
and Individual Differences, 37, 1453–1462.
Wilson, S., Kipp, K., & Chapman, K. (2003). Limits of the retrieval-inhibition construct: list
segregation in directed forgetting. Journal of General Psychology, 130(4), 341–358.
Zellner, M., & Bäuml, K. (2006). Inhibitory deficits in older adults: list-method directed
forgetting revisited. Journal of Experimental Psychology: Learning, Memory, and Cognition,
32(2), 290–300.

CHAPTER FIVE
Recollection is Fast and Easy:

Pupillometric Studies of Face
Memory
Stephen D. Goldinger*,1, Megan H. Papesh†
*Department of Psychology, Arizona State University, Tempe, AZ, USA
†Department of Psychology, Louisiana State University, Baton Rouge, LA, USA
Contents
1. Introduction192
2. Recognition Memory 192
3. Models of Memory 193
3.1. Threshold-Based Models 193
3.2. Continuous Models 195
3.3. Hybrid Models 196
4. Estimating Recollection and Familiarity 198
4.1. The Process-Dissociation Procedure 199
4.2. The Remember/Know Paradigm 201
5. Pupillometry203
5.1. The Neurophysiology of the Pupillary Reflex 203
5.2. Task-Evoked Pupillary Responses 204
5.3. Pupillometry and Memory 206
6. Psychophysiological Correlates of Memory for Faces 210
6.1. PDP and Pupillometry 211
6.2. RK and Pupillometry 214
7. General Conclusions 215
References215
Abstract
In this chapter, we discuss several distinct conceptualizations of recognition memory,
and their treatment of the (putative) processes underlying recollection and familiar-
ity. We focus most closely on the concept of recollection, which many dual-process
memory models assume to be a relatively slow, controlled process, during which con-
textual details from encoding are brought to mind (i.e., retrieval of episodic detail). We
then introduce the use of pupillometry—continuously measuring pupil diameter dur-
ing task performance—as an efficient means to estimate effortful cognitive processes
during memory encoding and retrieval. We review evidence from three pupillometric

192 Stephen D. Goldinger and Megan H. Papesh
studies of face memory, providing evidence to suggest that recollection is not slow
and effortful, as is often assumed, but is instead fast and easy.
1. INTRODUCTION
Have you ever spotted a familiar colleague who had recently changed
hairstyles, or had perhaps shaved off his signature mustache? You may have
experienced a slightly unsettled feeling, asking “What’s different about you?”
Similarly, you may feel a seemingly inexplicable familiarity with someone
you encounter at the market, avoiding interaction with that person because
you cannot determine how, or even if, you know him. The latter situation
exemplifies Mandler’s (1980) famous “butcher on the bus” phenomenon
(wherein you see your local butcher on a city bus, out of context, and
cannot determine why he seems familiar), but both situations give rise to
a nagging feeling of memory. Despite this subjective experience, both sce-
narios actually represent failures of memory, as you are unable to successfully
retrieve information that you know to be stored in your memory. Often,
these memory failures go unresolved; you cannot name the person in the
market, but are certain that you know him. Contrast that situation with one
more easily attributed to the success of memory: you see your butcher at
the grocery store, and recognize him instantly and automatically.This expe-
rience is by far more common. Throughout the day, you routinely interact
with people whom you instantly identify, fluently recollecting their names,
recent conversations and personal details. Indeed, a great deal of stored
memory is immediately available, with no effort, and no particular “feelings
of memory” whatsoever. Although these memories (and memory failures)
are experienced as disparate events, their underlying cognitive processes are
nonetheless captured by two-component processes that are often said to
make up “recognition memory”, recollection of episodic details, and vague
feelings of familiarity. This chapter will address some of the characteristics of
recollection and familiarity, and will present recent pupillometric data sug-
gesting that certain long-accepted assumptions ought to be reconsidered.
2. RECOGNITION MEMORY
In broadly conceptualizing recognition memory, in particular the
experience of recollection, two dominant and opposing views have persisted
for decades, single- and dual-process models. Variations upon each model
exist, and each has been implemented with varying degrees of success, but
Recollection is Fast and Easy: Pupillometric Studies of Face Memory 193
for present purposes, they will be dichotomized by their theoretical treat-

ment of recollection. Dual-process models treat recollection as a threshold
process, and they are intuitively appealing: as demonstrated above, we often
have the subjective awareness of different memory states corresponding to
recollection and familiarity. Dual-process models assume that these processes
are separate, and in fact, independent from one another (in both cognitive
and neural terms). Whereas familiarity is said to be a fast, automatic process
(e.g. you do not control when items “feel” familiar), recollection is said to
be slow, and deliberate. Single-process models, on the other hand, assume
that recollection is essentially strong familiarity, and that both processes are
subsumed within a single, strength-driven, automatic process. Rather than
existing as separate entities, they represent a continuum; some memories
will yield the subjective experience of recollection; others will elicit feel-
ings of familiarity. In the ongoing debate about the nature of recognition
memory, single- and dual-process models are often used to set the stage for
strong inference; to accept one is often to reject the other.The various mod-
els encapsulated by these discrepant overarching theoretical frameworks all
assume that memory decisions are based on the retrieval of some degree of
evidence (for consistency, we will call that evidence “strength”; this is for
clarity, not a theoretical stance). Interested readers should see Yonelinas &
Parks (2007) for a fuller discussion of the evidence for and against each of
the models (and others) discussed below.
3. MODELS OF MEMORY
3.1. Threshold-Based Models
Threshold models are often traced back to Fechner’s psychophysical research
(Boring, 1929), as the assumption inherent in these models is that a single
“evidence” threshold must be exceeded before an item can be detected as
previously encountered. High-threshold model is perhaps the simplest of such
views: according to high-threshold model, memory is characterized by two
distributions (often visually depicted as square distributions, but this is not a
theoretically constrained assumption), one representing target item strength
and one representing foil item strength. Each of the item types, targets and
foils, is assumed to have some “resting level” of inherent strength; by virtue
of the study session, targets accumulate additional strength, which serves
to separate the hypothetical target and foil distributions (Figure 5.1). The
“threshold”, according to this model, is the upper end of the foil distribu-
tion; items to the right of this point are said to fall above threshold, and are
Figure 5.1 Hypothetical distribution of item strength for targets and foils in high-
threshold model. Note that, because of their recent exposure, targets are assumed to
have increased levels of strength.
Figure 5.2 Hypothetical distribution of item strength for targets and foils in high-low
threshold model. The lower end of the target distribution is the low-threshold and the
upper end of the foil distribution is the high-threshold.
recognized in an all-or-none manner.This does not, however, prescribe that

observers make memory decisions in reference to the threshold. Rather,
decisions are made by adopting a response criterion (shown as the vertical
solid line in Figure 5.1); any item with a strength value above that criterion
is judged “old”. As the criterion is moved to the left, the perceiver is said to
respond more liberally; both hits and false alarms increase. Conservative cri-
teria serve to decrease hits and false alarms. A critical assumption, however,
is that the observer’s criterion, which can move throughout an experiment,
dichotomizes the distributions into binary old/new judgments.
Similarly, high-low threshold model (sometimes called two-high threshold
model, see Hilford, Glanzer, Kim, & DeCarlo, 2002; Y onelinas, 2002) assumes
that memory is a categorical process, but that two thresholds guide memory
decisions, one high-threshold and one low-threshold. Specifically, this model
also proposes two overlapping distributions of item strength for targets and
foils, but, in contrast to standard high-threshold model, it assumes that the
foil distribution extends beyond the lower tail of the target distribution
(Figure 5.2), which allows for two thresholds. Like high-threshold models,
items with strength values higher than the upper tail of the foil distribution
are said to fall above threshold and are recognized as previously encoun-
tered. Unlike high-threshold models, high-low threshold models include
a method by which new items can be “recognized” as new (e.g. “I would

have remembered the word waffle because my dog’s name is Eggo.”). Any
new item with strength values below the lower limit of the target distribu-
tion is said to fall below threshold, and can be “recognized” as new. Observ-
ers still set a single criterion against which they make memory decisions,
just as in high-threshold model, but the cognitive operations are different.
3.2. Continuous Models

In contrast to threshold models, continuous models do not make the
assumption that recollection is a threshold, all-or-none process. Rather,
they posit a continuous stream of evidence, capable of eliciting a range
of recollected details. As summarized by Wixted (2007) and Wixted and
Mickes, (2010), continuous models are based on the principles of signal-
detection theory, and posit that memory decisions are made by comparing
the strength of the retrieved memory signal to a decision criterion. As in
standard signal-detection theory, this view proposes the theoretical existence
of two Gaussian distributions (Figure 5.3), one reflecting target strength and
one reflecting foil strength. As in threshold theories, targets are assumed to
have greater strength, owing to their recent presentation. Also consistent
with the threshold models, any item that yields item strength exceeding
the decision criterion is judged “old”, whereas items with lower strengths
are judged “new”. Performance in such models is typically described by
the signal-detection index d’, the observer’s sensitivity, which corresponds
to the difference between the mean of the target and foil distributions.
Although equal-variance models (Figure 5.3(a)) usefully illustrate general
signal-detection-based models (and, in fact, have been incorporated into
hybrid models, see Section 3.3 below), abundant evidence supports unequal
variance distributions, wherein the target distribution is wider than the
Figure 5.3 Single-process, signal-detection models of recognition memory. (a) Cor-

responds to equal-variance models and (b) corresponds to unequal-variance models.
(Adapted from Wixted (2007)).
foil distribution (right panel of Figure 5.1; see, Ratcliff, Sheu, & Gronlund,
1992; Wixted & Stretch, 2004). Critically, regardless of the distributional
assumptions, continuous models all propose that recognition decisions are
based on a concept of continuous memory strength (Wixted & Stretch, 2004),
not on a threshold memory process.
3.3. Hybrid Models

Contemporary threshold theories, such as dual-process signal detection the-
ory (DPSD, see Yonelinas, 1994, 2001), incorporate elements from signal-
detection theory while maintaining some of the characteristics of earlier
threshold models. In similar fashion, recent continuous theories, such as
continuous dual-process (CDP) signal-detection theory (Wixted & Mickes,
2010), incorporate aspects of dual-process theory, while maintaining the
core assumptions of single-process models.1
Hybrid threshold theories, henceforth, “dual-process theories”, are intui-
tively appealing. In the “butcher on the bus” example described above,
people have the strong sense that they are failing to recollect episodic detail;
instead the butcher is merely familiar.This feeling of unresolved knowledge
can be easily contrasted with more richly detailed memories, such as not
only recognizing your butcher but also remembering that he recently gave
you a great recipe for grilled salmon.This intuitive mnemonic dissociation,
that richly detailed memories feel different than relatively vague memories,
captures the nature of dual-process theories: rather than assume that a sin-
gle memory strength variable explains both types of memory, dual-process
theories propose that recognition is served by two distinct, independent
processes, recollection and familiarity ( Jacoby, 1991). Whereas recollection is
assumed to occur by consciously controlled processing, reflecting a per-
son’s ability to recall the specific details of the encoding event, familiarity
is said to operate quickly and automatically, reflecting a vague “feeling of
knowing” that the encoding event had occurred.
An early version of a hybrid dual-process model was the two-criterion
model, wherein familiarity-based decisions are made quickly on the basis of
1 CDP (Wixted & Mickes, 2010) combines select elements of dual-process theory with key aspects
of signal-detection theory, to yield a signal-detection-based model capable of explaining subjective

feelings of recollection versus familiarity. According to this model, separate (nonindependent)
recollection and familiarity components exist, but memory decisions are still based on a continuous
stream of evidence, composed of the additive strength from each component. Because of this, the
model does not suggest that recollection and familiarity are necessarily quantitatively or qualitatively
different, so the predictions are similar to those of Univariate Signal Detection Theory (UVSD) and
will not be discussed further.
whether they fall above a high criterion or below a low criterion (Atkinson
& Juola, 1973, 1974; as cited in Wixted, 2007). According to this model,
recollection is only initiated if the memory strength falls between the two
criteria, acting, in essence, as a backup process. Later,Yonelinas (1994, 2001)
provided another attempt to combine the two models, with the DPSD
model. According to DPSD, recollection is a high-threshold, categorical
process: It either occurs or does not, and if it does occur, it will always be
associated with high confidence. Familiarity, on the other hand, is gener-
ally viewed as a continuous, ahistorical memory-strength variable, capable
of ranging in strength from low to high (Mandler, 1980; Yonelinas, 1994).
Familiarity, being described by strength of the signal, is therefore compat-
ible with an equal-variance signal-detection model. DPSD differs from the
two-criterion model of Atkinson and Juola primarily in its order of opera-
tions (and its added quantitative detail). Whereas the two-criterion model
assumes that recollection is initiated as a backup for failures of familiarity,
DPSD assumes the reverse. That is, if recollection does not occur, responses
are based on familiarity. Regardless of the order in which the processes are
said to occur (in fact, there are many instantiations of dual-process theory
in which the processes are initiated in parallel), dual-process theories almost
universally consider recollection and familiarity to be separate, indepen-
dent processes with distinct quantitative and qualitative characteristics (see
Yonelinas, 2002 for a review).
The majority of evidence in favor of separate recollection and familiar-
ity processes comes from functional process dissociations, or manipulations that
affect the contributions of each system independently. For example, several
studies have indicated that responses based on familiarity are faster than those
based on recollection (e.g. Hintzman & Caulton, 1997; Hintzman, Caulton,
& Levitin, 1998). Furthermore, it has been shown that this “fast familiar-
ity” process increases false alarms immediately after the presentation of a
new item, but the false alarm probability drops off with increased response
time, reflecting slower recollective retrieval dynamics (e.g. Gronlund &
Ratcliff, 1989; Hintzman & Curran, 1994; Jacoby, 1999; McElree, Dolan,
& Jacoby, 1999). Using event-related potentials (ERPs), many studies have
also found distinct electrophysiological correlates for recognition memory
responses based on recollection versus familiarity (Curran, 2000; Duarte,
Ranganath, Winward, Hayward, & Knight, 2004; Guo, Duan, Li, & Paller,
2006; Klimesch et al., 2001) and some have found opposite effects on rec-
ollection and familiarity following hippocampal damage (Sauvage, Fortin,
Owens, Yonelinas, & Eichenbaum, 2008). These dissociations (and others)
strongly suggest that two separate neural substrates underlie recognition

memory.2
If recollection and familiarity are separate processes, then they likely also
have distinct temporal properties. In virtually all dual-process conceptualiza-
tions of recollection and familiarity, familiarity is agreed to be the “faster”,
more automatic, process. Evidence supporting this assumption comes from
experiments in which processing speed has been manipulated, such that
perceivers have more or less time to make memory decisions, or when
decision times are measured. For example, Hintzman & Curran (1994) pre-
sented participants with lists of singular and plural nouns to study for an
upcoming memory test. Test stimuli consisted of old items, completely new
items, or distractors that were related to studied items, but had their plu-
rality reversed (e.g. if WAFFLE was studied, then WAFFLES would be a
related distractor word). Assuming that participants must recollect the studied
item in order to reject a related distractor, Hintzman & Curran (1994) used
response times to infer the temporal characteristics of recollection. They
found that recollection was relatively slow: participants discriminated old
from new words approximately 120 ms faster than they could discrimi-
nate old words from the related distractors. This finding, that the familiarity
process terminates faster than the recollection process, has been observed
repeatedly (e.g. Gronlund, Edwards, & Ohrt, 1997; Hintzman et al., 1998),
and is the basis for the assumption that familiarity is fast and automatic,
while recollection is slow and deliberate. Despite this, other researchers have
documented instances in which “old” responses based on familiarity are
slower than responses based on recollection (e.g. Dewhurst, Holmes, Brandt,
& Dean, 2006; Duarte et al., 2004; Wheeler & Buckner, 2004).
4. ESTIMATING RECOLLECTION AND FAMILIARITY

Although several methods exist for estimating the contributions of
recollection and familiarity to recognition memory, the two most popular
are the process-dissociation procedure (PDP; Jacoby, 1991) and the remember/
know paradigm (RK;Tulving, 1985). PDP was designed to separate the influ-
ences of recollection and familiarity in recognition memory and to elu-
cidate the notion that task distinctions do not necessarily map onto process
2 Because the focus of this chapter is on the psychophysiological characteristics of putatively
recollection-based memory decisions, a review of the single- versus dual-process debate regarding
the neuroscience of is beyond the scope of this chapter. Interested readers are encouraged to consult
Squire and Wixted (2011 or Wais, Squire, and Wixted, 2009).
distinctions. Specifically, Jacoby (1991) suggested that researchers should not

assume that a task is capable of measuring the exact processes that it is
intended to measure, as results from different tasks may reflect the tasks, and
not the processes assumed by the researcher. To demonstrate this in recog-
nition memory, Jacoby (1991) developed a procedure that does distinguish
between processes, while using a constant task, so as to avoid the problem of
comparing results from separate tasks.
4.1. The Process-Dissociation Procedure

The major assumption underlying PDP is that recollection and familiarity
independently contribute to memory (e.g. Mandler, 1980), such that their
contributions can be estimated by putting responses based on them in
mutual opposition ( Jacoby, 1991). When automatic processes (i.e. familiar-
ity) oppose consciously controlled processes (i.e. recollection), researchers
can infer which is the more dominant contributor to recognition m emory.
By analyzing the results of complementary facilitation and interference
paradigms, one is able to numerically estimate the extent to which a utomatic
and controlled processes contribute to recognition responses, assuming a
process dissociation is observed.
Whereas recollection has been proposed as an intentional (consciously
controlled) use of memory ( Jacoby, 1991), and is therefore likely a relatively
slow process, familiarity contributes automatically (unconsciously), and on a
relatively fast time course. In the facilitation paradigm (henceforth referred
to as inclusion), participants are typically presented with two lists of stimuli
that differ in some concrete way (e.g. List A contains words that are heard
and List B contains words that are read).3 During a subsequent recognition
memory test, participants are presented with words from both lists, along
with new words. All previously encountered words, regardless of original
list, are to be called “old”. It is assumed that recollection and familiarity both
contribute to “old” judgments during the inclusion test. Based on Jacoby
(1991), the probability of issuing an “old” response to words that were heard
(h) during study would be OIh = Rh + Fh − RhFh, where “I” indicates the
inclusion test, and R and F represent the contributions of recollection and
familiarity, respectively.
Suppose that participants are then presented with another set of two
lists (again heard and read, separately) to study. During test, they are again
3 Note that Jacoby also modified the paradigm to test associative memory and stem completion
( Jacoby & Rhodes, 2006). Only the classic paradigm will be discussed here.
presented with new words, as well as those from the two study lists, and are
asked to make recognition responses. However, instead of responding “old”
to any previously encountered words, participants are now asked to exclude
specific items. In this example, assume that they are asked to exclude words
that were previously heard, responding “old” only to those words that were
read during study. According to Jacoby, the probability of issuing an “old”
response to a previously heard word is now given by OEh = Fh(1 − Rh),
where E denotes the exclusion test. Based on simple algebra, this can be
represented as OEh = Fh − RhFh. In words, the probability of incorrectly
responding “old” to a previously heard word during the exclusion test is
equal to the familiarity of that word minus the multiplied contributions
of recollection and familiarity for heard words. The contribution of recol-
lection can thus be estimated by Rh = OIh − OEh, which simply indicates
that recollection and familiarity both contribute to “old” responses in the
inclusion test, but correct responses to excluded items during the exclusion
test can only be based on recollection: Participants must recollect an item’s
presentation modality to reject it. Subtracting the influence of familiarity
from the combination of R and F leaves only R. Familiarity can thereby be
estimated as Fh = OEh/(1 − Rh).
Although the PDP method has considerable utility, one of the key
criticisms of PDP regards the definition of recollection. Specifically, rec-
ollection is a relatively vague term, as people can (arguably) experience
varying degrees of recollection. Participants can recall varying degrees of
information about List 1 and List 2 items. For example, if spoken words are
used, voice can be a critical detail and participants may recall only gender
or more detailed, speaker-specific information (as in Dodson, Holland, &
Shimamura, 1998). In PDP experiments, both memories would be classi-
fied as recollection, despite their varying degrees of specificity. To address
this, researchers may differentiate between diagnostic and nondiagnositic recol-
lection (Gruppuso, Lindsay, & Kelley, 1997; Mulligan & Hirshman, 1997;
Yonelinas & Jacoby, 1996), with diagnostic information reflecting memory
for the source of test items and nondiagnostic information reflecting other,
noncriterial, memorial information. From this, it is clear that the recol-
lection parameter in PDP only measures the overall amount of diagnostic
memorial information, ignoring a potentially vast amount of other recol-
lective experiences.
Additional criticisms of PDP regard the underlying assumptions, pri-
marily the assumption that recollection and familiarity independently con-
tribute to recognition memory (Curran & Hintzman, 1995, 1997; Joordens
& Merikle, 1993; Richardson-Klavehn, Gardiner, & Java, 1996; but see
Jacoby, 1998; Jacoby, Begg, & Toth, 1997). This issue has been extensively
debated and, like the single-/dual-process distinction, each side has inter-
preted evidence as consistent with their view. For example, Curran and
Hintzman (1995, 1997) argued that high correlations between recollection
and familiarity at the item level violate the PDP independence assump-
tion, whereas Jacoby et al. (1997) argued that the correlations are “trivial”
and likely reflect violations of the boundaries within which PDP must be
used. Because the estimates of recollection and familiarity provided by PDP
critically rely on their functional independence, equivocal empirical results
necessitate investigations via other paradigms.
4.2. The Remember/Know Paradigm

Another approach to dissecting recognition memory is the RK paradigm
(Tulving, 1985), which makes use of subjects’ subjective feelings of the rela-
tive specificity of their memory. Although Tulving (1985) initially intended
for the procedure to differentiate between states of awareness associated
with subjective experiences of memory, it has more recently been used to
support dual-process theories of recognition memory (Wixted & Mickes,
2010). In this paradigm, participants are assumed to appreciate why they
make old/new recognition decisions. If an item is judged “old”, there
are three possible routes to this decision; an item can be “remembered”,
“known” (henceforth R and K) or simply guessed (although not all studies
use a “guess” option). Jacoby, Yonelinas, & Jennings (1997) suggested that
remember responses reflect episodic retrieval, and the function of conscious
recollection, whereas know responses reflect familiarity, or the recognition
of an item’s status as “old” without concomitant recollection of its earlier
presentation (Yonelinas, 2002).
Results from several studies (e.g. Engelkamp & Dehn, 1997; Gardiner,
1988; Gardiner & Java, 1991; Gardiner, Java, & Richardson-Klavehn, 1996;
Gardiner & Parkin, 1990; Rajaram, 1993) indicate that different experimen-
tal manipulations selectively enhance or diminish remember/know response
frequencies, supporting Gardiner’s (1988) reports of functional dissociations
between R and K responses.These dissociations imply that recollection and
familiarity are distinct, independent processes, and their existence is typi-
cally interpreted within a dual-process framework (Gardiner, 2001, as cited
by Dunn, 2004).
This interpretation has been criticized, however, as several researchers
have proposed that the response types reflect confidence in memory more
than they reflect the function of two separate memory processes (Donaldson,
1996; Hirshman, 1998). Instructions to respond “remember” or “know”,
according to this view, are interpreted by participants as a requirement to
adopt a more conservative or liberal response criterion, respectively (Dunn,
2004). Donaldson (1996) approached the RK task from a single-process
viewpoint, and suggested that participants complete the task by adopting
two-decision criteria, one (high-criterion) for remember responses and one
(low-criterion) for know responses. He argued that, although participants
were issuing responses that appeared to reflect two different memory systems,
they were responding in line with their decision criteria, which reflected
memory strength, and not separate processes (cf. Knowlton & Squire, 1995).
Meta-analyzes of decades of data and critical tests of recent data have been
taken to support both single-process (Donaldson, 1996; Dunn, 2004, 2008;
Hirshman & Master, 1997; Wixted, 2007; Wixted & Stretch, 2004) and dual-
process models (Conway, Dewhurst, Pearson, & Sapute, 2001; Gardiner,
Ramponi, & Richardson-Klavehn, 1998, 2002; Yonelinas, 2002).
The hybrid CDP model (Wixted & Mickes, 2010) was developed in
response to the dual-process interpretation of RK data. Although the model
is entirely based upon signal-detection theory, it assumes that participants
are able to determine whether their memorial experiences are based on the
function of recollection or familiarity. In short, CDP assumes that recogni-
tion decisions reflect the combined influence of recollection and familiarity,
such that decisions are still based on a “strength of evidence” dimension, but
that participants have access to the source of the predominant strength. Evi-
dence for this model comes from RK tasks in which participants provide
both confidence estimates and source discriminations. During encoding,
participants study words presented at either the top or bottom of the screen,
in red or blue font. When prompted to recognize a test item (centrally
presented, in black), participants provide RK judgments, confidence esti-
mates, and responses regarding original location and color (source discrimi-
nations). Wixted and Mickes found that, although “know” responses can be
associated with a high degree of confidence, corresponding source accuracy
was lower, relative to “remember” responses.
As discussed above, neuropsychological evidence has consistently been
interpreted within a dual-process framework, and this is largely because
of the widespread use of RK in neuroimaging studies. In many experi-
ments, researchers have observed elevated activity in the hippocampus for
R judgments, and almost no activity for K judgments (e.g. Aggleton et al.,
2005; Eldridge, Engel, Zeineh, Bookheimer, & Knowlton, 2000; Holdstock,
Mayes, Gong, Roberts, & Kapur, 2005; Moscovitch & McAndrews, 2002;
Uncapher & Rugg, 2005; Verfaellie, Rajaram, Fossum, & Williams, 2008;
Yonelinas et al., 2002). One difficulty in interpreting these effects, however,
is that R judgments are typically associated with higher confidence, rela-
tive to K judgments (e.g. Dunn, 2004, 2008; Rotello & Zeng, 2008; Wixted
& Mickes, 2010; Wixted & Stretch, 2004), making it impossible to deter-
mine whether the effect arises from differences in fundamental processes,
or differences in retrieved strength. This “strength confound” was originally
pointed out by Wixted (2009), who noted that recollective detail is almost
never entirely absent from know judgments (Mickes,Wais, & Wixted, 2009).
In fact, when subjective retrieval strength was equated, Wais, Squire, &
Wixted (2009) observed similar levels of hippocampal activity during puta-
tively recollection-based and familiarity-based memories (see also Kirwan,
Wixted, & Squire, 2008; Wais, 2008; W ais,Wixted, Hopkins, & Squire, 2006).
In order to interpret RK data, it seems, one needs to take an additional step
and collect overt, metacognitive estimates of confidence.
5. PUPILLOMETRY
5.1. The Neurophysiology of the Pupillary Reflex
To further examine differences in processes that are said to be fast and auto-
matic, versus slow and deliberate, researchers can appeal to neuroimaging
evidence, as discussed above, or they can examine a recently rediscovered
correlate of cognitive activity, pupil diameter. Pupillometry, the measurement
of the diameters of the eyes’ pupils, has been used for centuries to examine
visual and cognitive processing (e.g. Fontana, 1765). Although it is well-
known that the pupils dilate in response to changes in ambient lighting, it is
less well-known that the pupils also dilate in response to nonvisual stimuli,
such as emotions and thoughts (Goldwater, 1972; Janisse, 1974; Loewenfeld,
1958). This distinction characterizes two independent types of pupillary
reflex, tonic changes, which occur in response to general factors, such as
emotional arousal, stress, and anxiety, and phasic changes, which occur fol-
lowing the onset of stimuli for cognitive processing (Karatekin, Couperus,
& Marcus, 2004). These cognitively evoked pupillary reflexes occur follow-
ing inhibition of the parasympathetic nervous system’s Edinger–Westphal
nucleus (Steinhauer, Siegle, Condray, & Pless, 2004), which is controlled by
the locus coeruleus–norepinephrine (LC–NE) system. The LC is a subcor-
tical brain system that contains the noradrenergic system, which is the sole
source of the neurotransmitter NE. This system plays a critical role in the
control of attention (Gilzenrat, Nieuwenhuis, Jepma, & Cohen, 2010; for a

review, see Laeng, Sirois, & Gredebäck, 2012). A role for the LC–NE system
in memory consolidation has been determined by documenting LC–NE
activity during memory retrieval (Sterpenich et al., 2006) and slow-wave
sleep (Eschenko & Sara, 2008). Relevant to the experiments we summa-
rize below, the LC–NE system is also critically involved in the pupillary
reflex (Koss, 1986). In combined single-cell recording and pupillometry
studies with monkeys, researchers have documented a tight correspon-
dence between pupillary reflexes and activity in cells within the LC–NE
system (Rajkowski, Kubiak, & Ashton-Jones, 1993; Rajkowski, Majczynski,
Clayton, & Ashton-Jones, 2004).
Further neurophysiological evidence for the influence of cognitive
processing on pupil size comes from examining the muscles that control
pupillary reflexes. Two muscles are known to control pupil dilation and
constriction, the dilator and the sphincter; these muscles are differentially
affected by activation in the sympathetic and parasympathetic systems
(Steinhauer et al., 2004). As noted above, inhibition of the parasympathetic
nervous system has been attributed to dilation resulting from cognitive pro-
cessing. This system also controls the activity of the sphincter muscle; when
the parasympathetic system is inhibited, activity on the sphincter muscle
decreases (Steinhauer et al., 2004).These autonomic pathways hold recipro-
cal connections with the central nervous system (CNS), so it has been sug-
gested that they can modulate, or be modulated by, CNS structures related
to cognition (Gianaros, Van Der Veen, & Jennings, 2004; Steinhauer et al.,
2004). Investigations into the neural mechanisms of successful learning and
memory in animals have revealed a close correspondence between accu-
rate performance and the involvement of the autonomic system (Croiset,
Nijsen, & Kamphuis, 2000). Such findings are paralleled by recent findings
from human experiments, wherein increased autonomic responses (e.g. skin
conductance) are positively correlated with memory strength for emotional
words (Buchanan, Etzel, Adolphs, & Tranel, 2006). Additionally, stimulation
of the vagus nerve (a parasympathetic pathway known to carry signals to
the brain) is associated with memory formation and consolidation (Clark,
Naritoku, Smith, Browning, & Jensen, 1999).
5.2. Task-Evoked Pupillary Responses

As noted, phasic changes in pupil diameter occur following the onset of
cognitive processing. These reflexes are observed independently of tonic
changes; in dark-adapted conditions, which inhibit the parasympathetic
system, the pupils reliably dilate in response to cognitive demand (Steinhauer

& Hakerem, 1992), leading them to be referred to as task-evoked pupillary
responses (TEPRs). Although Eckhard Hess is often credited for initiating
the psychological study of pupillary reflexes (cf., Hess, 1965; Hess & Polt,
1964; Hess, Seltzer, & Shlien, 1965), his research focused almost exclusively
on the pupillary reflex as it reflected “emotionality” (Hess, 1965, p. 46), a
tonic response. Since then, the “emotional” component of the pupillary
reflex and the “cognitive” component have been clearly dissociated (Partala
& Surakka, 2003; Stanners, Coulter, Sweet, & Murphy, 1979). Kahneman
and Beatty (1966) and Beatty & Kahneman (1966) are best known for initi-
ating interest in TEPRs, and even suggested that TEPRs reflect a “summed
index” of brain activity during cognitive processing. Their early work dem-
onstrated that pupil dilations are time-locked to cognitive processing, and
that differences between and within tasks are observable via pupillometry
(Kahneman, 1973; Kahneman, Beatty, & Pollack, 1967). For example, in a
digit recall task, as participants were given more numbers to retain, their
pupils became larger; as the digits were recalled, the pupils constricted
with each additional item (Kahneman & Beatty, 1966). Although pupil-
lometry fell out of favor for some time, it has been used to infer cognitive
effort in a variety of domains, such as lexical decision (Kuchinke, Võ,
Hofmann, & Jacobs, 2007), attention (Kahneman, 1973; Karatekin et al., 2004),
word processing (Papesh & Goldinger, 2012), working memory (Granholm,
Asarnow, Sarkin, & Dykes, 1996; Van Gerven, Paas, Van Merriënboer, &
Schmidt, 2004), face perception (Goldinger, He, & Papesh, 2009), general
cognitive processing (Granholm & Verney, 2004), and recognition memory
(Kafkas & Montaldi, 2011; Papesh, Goldinger, & Hout, 2012).
The appeal of pupillometry lies in its relative independence: phasic pupil-
lary reflexes are largely impervious to attempts to consciously manipulate
them (in fact, in our research, most participants assume that the eye-tracking
equipment is primarily to track the movement of their eyes across the screen).
Further, whereas many neuroimaging methods are time-consuming, expen-
sive, and require the use of tightly controlled environments, pupil measure-
ments can be obtained with relatively basic eye-tracking equipment (e.g.
eye-trackers recording at 50 Hz are capable of providing sensitive pupil-
lometric data), with fewer environmental and methodological restrictions.
Given the tight correspondences between activity in neuroanatomical areas
relevant for cognitive processing and dynamic changes in pupil size, pupil-
lometry represents an ideal, efficient means for estimating neural activity
during many cognitive events (Kahneman, 1973).
5.3. Pupillometry and Memory

In the study of memory, pupillometry can be likened to ERP waveforms
(Beatty, 1982); enlarged pupils are typically associated with increased cogni-
tive demand (Granholm & Steinhauer, 2004; Porter,Troscianko, & Gilchrist,
2007). Comparing neurophysiological measures across study and test has
been used to differentiate the neural activity associated with subsequently
remembered versus forgotten information in both functional Magnetic
Resonance Imaging (fMRI) (e.g. Ranganath et al., 2004) and ERP investi-
gations (e.g. Cansino & Trejo-Morales, 2008; Duarte et al., 2004; Guo et al.,
2006). The logic underlying such studies is that encoding should utilize the
same set of processes and neural substrates that are subsequently recruited
during successful retrieval. Moreover, the strength and type (e.g. recollec-
tion or familiarity) of memory should be observable from different patterns
of activation during both encoding and retrieval. In our recent research,
pupillometry was used to examine encoding and recognition effort, acting
in place of metacognitive confidence estimates.
Although early TEPR investigations were criticized on the grounds that
now-standard experimental controls were not implemented, recent work
incorporates strict experimental control to eliminate the unwanted influ-
ence of tonic reflexes (Võ et al., 2008). Because pupils dilate reflexively
to changes in luminance, color, or the spatial frequency composition of
the visual input, care must be taken to equate, as much as possible, stimu-
lus characteristics in experimental designs that utilize pupillometry (Porter
et al., 2007). Porter and Troscianko (2003) identified several methodologi-
cal approaches that minimize unwanted pupillary reflexes, including the
use of relatively low-stimulus contrast, avoiding colored stimuli, and using
relatively long-stimulus exposure durations. Goldinger and Papesh (2012)
recently added to this list of constraints by suggesting the use of relatively
long (e.g. 1000 ms or more) intertrial intervals (ITIs) and baseline-correction
procedures. Both suggestions guard against carryover effects, wherein the dif-
ficulty of trial n influences the waveform of trial n + 1. Recent work has
taken such precautions into careful consideration, including several relevant
studies on the pupillary reflex and memory.
Early work on the effects of memory on pupil size documented that
familiar items elicited increases in pupil diameter, relative to unfamiliar items.
For example, Gardner, Mo, & Borrego (1974) demonstrated that familiar
consonant–vowel–consonant (CVC) trigrams resulted in enlarged pupils,
relative to unfamiliar CVCs; Gardner, Mo, & Krinsky (1974) observed the
same effect with consonant trigrams. More recent research, incorporating

the methodological constraints cited above and motivated by the similarity
between pupillary waveforms and ERP waveforms, which are known to
reflect memorial processes (Dietrich et al., 2000; Johansson, Mecklinger, &
Treese, 2004), has also documented effects of long-term memory on pupil
dilation. Võ et al. (2008) had participants study a series of positively and
negatively valenced words, followed by a speeded old/new recognition test.
Although they found effects of emotionality (e.g. better memory for emo-
tional words and smaller pupils to negative words), the most relevant finding
was a “pupillary old/new effect”, wherein pupils were larger during test
trials leading to hits, relative to correct rejections. The authors interpreted
this effect within a dual-process framework (as in Yonelinas, 2001, 2002),
suggesting that the increased pupil size observed for hits was directly related
to the occurrence of recollection.
A similar conclusion was drawn by Kafkas and Montaldi (2011), who
investigated the nature of incidental memory for images using a modified
RK procedure. In their study, participants viewed images without instruc-
tion to remember them, and were later given a surprise memory test in
which they distinguished memories based on degrees of familiarity (e.g.
F1–F3) or a single recollection response. Although they did not observe an
effect differentiating recollection from familiarity, pupil size during encod-
ing differed based on subsequent memory strength: as subsequent mem-
ory increased, pupil diameter decreased. These findings are suggestive, but
should be interpreted with caution. Kafkas & Montaldi’s (2011) procedure
eliminated TEPRs during the encoding phase, leaving only tonic changes
free to vary.
In an investigation of phasic pupillary changes and memory strength,
Papesh et al. (2012) observed the opposite pattern. Participants in their
experiment intentionally studied a series of words spoken by multiple talk-
ers; using an intentional encoding process, and feature-matched words,
allowed them to eliminate tonic pupillary reflexes, leaving phasic reflexes
free to vary. Following a brief break, participants made confidence-based
old/new decisions (e.g. 1-very sure new to 7-very sure old) to old and new
spoken words. Old words were spoken in one of the three voices—the
studied voice, a familiar voice, or a new voice. Examining pupil size by sub-
sequent memory performance revealed a clear relationship between confi-
dence during test and pupil size during encoding: as subsequent confidence
increased, so too did pupil size. Or, put another way, pupil dilation during
learning predicted future memory performance, on an item-by-item basis.
Further, pupil sizes were also larger when the study and test voices matched,
relative to when they did not match. Together, these findings revealed that
during intentional encoding of spoken material, the degree of cognitive
effort devoted to encoding can accurately predict subsequent recognition
memory.
Papesh et al. (2012) interpreted their results within Whittlesea’s (1997)
Selective Construction and Preservation of Experience (SCAPE) frame-
work for recognition decisions. According to Whittlesea and Leboe (2000)
and Whittlesea and Williams (1998, 2001), recognition decisions are made in
two stages. The first stage is characterized by the production of mental states,
such that the perceptual material (e.g. the test item) is identified, elaborated,
and any missing information is “filled in” by bringing to mind associated
images or ideas (Neisser, 1967). In the second step, the perceiver evaluates the
fluency of the production process. Rather than directly evaluating the stim-
ulus and comparing memory strength to a decision criterion, the perceiver
evaluates the relative harmony of mind; how easily did the production pro-
cess end? If the production process was characterized by “extra” elaboration,
then the expectation of fluent processing is violated, and the perceiver is left
with a sense of memory failure (e.g. a sense of familiarity). Because Papesh
et al. (2012) essentially observed a pupillary reinstatement of cognitive pro-
cessing across study and test, they concluded that memory traces are com-
posed of rich, detailed events, and that memory decisions are characterized
by an assessment of how easily those details are brought to mind.
Otero, Weekes, & Hutton (2011) obtained a similar result, and proposed
that pupillometry can be used to reveal the strength of memories across
encoding and retrieval.To more closely examine the roles of recollection and
familiarity, they employed the RK procedure, asking participants to provide
remember/know judgments for all “old” decisions. Using emotionally neu-
tral words, they obtained a pupillary old/new effect that was stronger when
participants reported the subjective experience of recollection: pupils were
largest when participants reported remembering, and were smallest when
participants correctly rejected items. Responses based on familiarity (know
responses) yielded diameters that were reliably smaller than those accom-
panying remember responses, but they were not significantly different from
correct rejections. Although this pattern is consistent with Võ et al.’s (2008)
interpretation, that the pupillary old/new effect is driven by the cognitively
demanding experience of recollection, when Otero et al. (2011) manipu-
lated levels of processing during study, they observed larger diameters for
items processed in deep, relative to shallow, orientations. This finding argues
against the interpretation of recollection in the pupillary old/new effect.

Specifically, if recollection is cognitively demanding, which yields enlarged
pupils, encoding manipulations known to increase the memorability of
items should not also yield larger diameters, but they do. Otero et al. (2011)
interpreted their findings as reflecting the relative strength of memories.
Consistent with a single-process, signal-detection model (and, we argue,
consistent with the SCAPE framework), they observed that stronger mem-
ories yielded enlarged pupils, and that this effect was not limited to con-
scious recollection. Papesh (2012) observed a similar result. Across encoding
and retrieval of high-frequency and low-frequency words, stronger memo-
ries (and subsequently stronger memories) were associated with enlarged
pupils (Figure 5.4). Even during encoding, when the pupil measurement
was temporally divorced from memory and confidence decisions, pupil size
accurately tracked subsequent memory strength.
Interestingly, Otero et al. (2011) also observed a pupillary old/new effect
in false alarms; pupils were larger when items were falsely recognized, relative
to when they were called “new”. Subsequent work by Heaver and H utton
(2011) further suggested that these memory-based effects in the pupillary
reflex are automatic, and not capable of influence by cognitive control. In
addition to replicating the pupillary old/new effect, they found that this
Figure 5.4 Pupil diameter as a function of confidence and word frequency during a
recognition test. Error bars represent standard errors of the mean. (From Papesh (2012)).
(For color version of this figure, the reader is referred to the online version of this book.)
effect was impervious to instructions to feign amnesia or to call every test

item “new”, regardless of its old/new status.Across three experiments, Heaver
and Hutton (2011) observed larger pupil sizes during old, relative to new, test
trials, even when participants did not overtly judge the items as “old”.
If phasic pupillary changes are positively correlated with cognitive
demand, why is retrieving a memory associated with more “demand” than
correctly rejecting an experience as novel? This question underscores the
differences across studies such as Kafkas and Montaldi (2011), in compari-
son to those by Otero et al. (2011) and Papesh et al. (2012). Whereas Kafkas
and Montaldi (2011) suggested that the experience of future recollection is
associated with pupillary constriction, the latter researchers suggested that
pupil dilation is related to the strength of memories, and the success of
memory retrieval. It is difficult to fully compare these studies, however, due
both to differences in stimuli and in methodology. For example, whereas
Kafkas and Montaldi (2011) used incidental encoding instructions, Otero
et al. (2011) and Papesh et al. (2012) used intentional encoding instructions.
This methodological difference alone is capable of explaining the discrep-
ant findings, as incidental encoding does not involve the same degree of
demand as intentional encoding, and may in fact yield only tonic reflexes.
6. PSYCHOPHYSIOLOGICAL CORRELATES OF
MEMORY FOR FACES
Recently, we have begun to apply converging methods to more fully
investigate the pupillary old/new effect, and the possible indices of recollec-
tion and/or familiarity in the pupillary reflex.To do this, we have conducted
several experiments using the same stimuli, but different methods, each of
which is aimed at eliciting measures of recollection versus familiarity. By
using the same stimuli, we can begin to more fully appreciate the mne-
monic properties of the pupillary reflex.
Goldinger et al. (2009) reported an investigation of the own-race bias
(ORB) in face recognition, in which we used eye movements and pupil-
lometry to provide insight into a well-studied cognitive phenomenon. Asian
and White participants had their eye movements and pupil sizes monitored
as they encoded a series of Asian and White faces. Afterward, participants
discriminated old from new faces, and were labeled as “high performers”
(i.e. participants who were relatively good at recognizing other-race faces)
and “low performers” (participants who performed relatively poorly).
Goldinger et al. (2009) also examined pupillary waveforms during encoding
as a function of subsequent memory, and observed a strong race effect.

Specifically, when encoding own-race faces, participants’ pupils dilated

much less, relative to when they encoded other-race faces. More interest-
ingly, we also observed an effect of subsequent memory performance. High
performers devoted greater effort to encoding other-race faces; participants
who subsequently scored poorly to other-race faces had significantly less
pupil dilation during encoding. In fact, as the encoding trials progressed,
subsequently low-scoring participants gradually “withdrew effort”; they
began moving their eyes less vigorously (i.e. they gathered less information)
and their pupil sizes peaked at much lower diameters.These results are con-
sistent with those reported by Otero et al. (2011) and Papesh et al. (2012)
pupil dilation reveals the experience of memory, even during encoding.
6.1. PDP and Pupillometry

Although these results were suggestive, they do not speak to the experi-
ences of recollection versus familiarity. To investigate this issue, Papesh and
Goldinger (2011) conducted two follow-up studies using well-known
memory paradigms to separate the contributions of recollection and famil-
iarity, the PDP and RK paradigms described earlier. In both experiments,
all participants were White, as recruiting recent Asian immigrants is difficult,
and the stimuli had already been verified to result in clear ORBs across
both races. In our PDP experiment, 36 participants were divided into two
groups, the exclusion group (n = 19) and the inclusion group (n = 17). All
participants studied a series of 28 Asian and White faces for 5 s each; each
face was bordered in either blue or purple, which was randomly selected
on each trial with the constraint that each race has an equal number of
each color. During the subsequent recognition test, exclusion participants
were instructed to “exclude” (i.e. call “new”) certain faces based on their
previously seen borders (all test faces were presented without colored bor-
ders). Some participants were told to exclude any faces that were originally
presented with blue borders; others were instructed to exclude faces that
originally had purple borders. Participants in the inclusion group received
standard recognition instructions (i.e. call all studied faces “old”).
Papesh and Goldinger (2011) replicated the standard ORB, such that par-
ticipants were better able to recognize own-race faces, relative to other-race
faces. More important, we observed clear differences in pupillary waveforms
during encoding as a function of subsequent memory performance. Unlike
Goldinger et al. (2009), who classified participants based on their overall
memory performance, Papesh and Goldinger (2011) divided the waveforms
Figure 5.5 Baseline-corrected pupil diameter as a function of face race and subsequent
memory performance in the exclusion (a) and inclusion (b) groups. (For color version of
this figure, the reader is referred to the online version of this book.)
by memory type. Figure 5.5 shows pupillary results during face encoding.
In each panel, the data represent baseline-corrected pupil diameters, shown
across an 8 s span. Recall that faces were only visible for 5 s, which is why
there are characteristic drops in pupil diameter at the 6 s mark (pupillary
changes typically lag by about 500 ms, relative to stimulus changes). Although
it is not shown in these figures, we also impose variable ITIs to ensure that
pupil diameters return to their resting values before the next face is shown.
In Figure 5.5(a,b), separate functions are shown, representing the race of
the studied face (Asian, White) and the eventual recognition outcome (Hit,
Miss). Figure 5.5(a) shows data from the exclusion condition; Figure 5.5(b)
shows the inclusion condition. In both (a) and (b), similar findings are appar-
ent. First participants devoted greater effort, as operationalized by pupil size,
to encoding other-race faces, as the blue (Asian) functions are generally
higher than the black (White) functions. Second, and of particular interest,
the pupil functions representing future hits show a dramatic asymmetry:
when participants studied Asian faces, and would later produce recognition
hits, they devoted considerable effort during encoding (this is more easily
appreciated in the left-hand panel, but was true in both conditions), rep-
licating Goldinger et al. (2009). When face encoding would lead to future
misses, no differences were observed as a function of race. Most intriguing,
when participants studied own-race faces, and would later produce recogni-
tion hits, their pupils were actually constricted, relative to baseline.
The results in Fig. 5.5 suggest that successful face encoding requires dif-
ferent cognitive resources, depending on the face of the race. However, it
remains unclear how these results address the question of future recollection
versus familiarity. To more fully appreciate the role of recollection, consider
just the exclusion group, specifically the “excluded” faces. As reviewed earlier,
in a correct exclusion trial, the participant must first recognize the face as hav-
ing been studied. After that, she must also retrieve the original border color.
In other words, the participant must recollect the item in order to successfully
exclude it. On the other hand, if the participant fails to correctly exclude
the face, it can be inferred that there was a failure of recollection: either the
participant did not recollect the face, or she may have falsely remembered
the wrong border. Given these assumptions, researchers can infer that correct
exclusion trials are completed on the basis of recollection, whereas incor-
rect exclusion trials represent the function of familiarity, in the absence of
recollection. As shown in Figure 5.6, we observed significantly larger pupils
during encoding trials that led to incorrect exclusion.When participants would
subsequently experience detailed recollection, their pupil sizes were smaller,
relative to when they would experience vague feelings of memory.
From these results, Papesh and Goldinger (2011) suggested that, consistent
with anecdotal experiences of easily recognizing well-known acquaintances,
recollection is typically fast, automatic, and cognitively easy. Although our
earlier work suggested that pupil size nicely tracks future memory strength,
Figure 5.6 Baseline-corrected pupil diameters as a function of face race and subse-
quent memory performance in the exclusion study trials. (For color version of this fig-
ure, the reader is referred to the online version of this book.)
these PDP data suggested that some memories are too strong to elicit this
standard pattern; when a memory is retrieved easily, as SCAPE would argue,
it involves relatively little cognitive effort.
6.2. RK and Pupillometry

To further assess whether recollection is fast and automatic, Papesh and
Goldinger (2011) examined face memory using the RK procedure. As in
our PDP study, participants (N = 30) encoded a series of 28 Asian and White
faces (presented without borders) for an upcoming memory test, with stan-
dard RK instructions. As in the previous experiments, pupil size during
encoding revealed an ORB: participants devoted greater cognitive resources
to encoding other-race faces, relative to own-race faces. We first exam-
ined encoding activity on the basis of subsequent memory, conditionalized
by whether the participants demonstrated an ORB in their recognition
data. Approximately half the participants demonstrated the typical ORB.
Despite this, both groups of participants showed that encoding other-race
faces demanded more cognitive effort (Figure 5.7). Further, the peaks for
subsequent “know” responses were reliably higher than the peaks for sub-
sequent “remember” responses. This finding is consistent with the findings
from the PDP experiment. Specifically, when participants are going to have
a specific, detailed recollection, they devote less effort to the study trial. It is
almost as if they might think to themselves “Well, I’m definitely going to know
this later,” and they need not try as hard to commit it to memory. Although
the results are not shown in Figure 5.7, it is important to note that pupil
diameters were also recorded during the recognition test: we observed a
Figure 5.7 Baseline-corrected pupil diameters (during encoding) as a function of face

race and subsequent memory performance in the RK experiment. (a) Represents the
ORB group and (b) represents the No ORB group. (For color version of this figure, the
reader is referred to the online version of this book.)
clear pattern that hits designated as “remember” elicited less pupil dilation
than hits designated as “know”.
7. GENERAL CONCLUSIONS
In popular conceptualizations of recognition memory, two processes
are said to contribute to memorial decisions, recollection and familiarity. All
current dual-process models assume that recollection is a slow, deliberate,
and demanding cognitive process, whereas familiarity is fast and automatic.
Here, we hope to have shown evidence that recollection may not be as slow
and cognitively taxing as previously theorized. By examining pupil dilations,
as a proxy for cognitive effort, we observe several very clear patterns. Dur-
ing the creation of memories, those items that will eventually lead to the
experience of recollection (as determined by the PDP procedure or “R”
responses in RK) are characterized by less effort during encoding. In similar
fashion, those same items elicit less effort during later recognition testing.
Effort during encoding closely predicts effort during test: recollected items
have a privileged existence, being easily encoded and easily retrieved.
Indeed, our findings are consistent with evidence that recollection-
based responses often occur more quickly than familiarity-based responses
(Dewhurst et al., 2006; Duarte et al., 2004; Wheeler & Buckner, 2004). In
our research, we use a psychophysiological index of cognitive processing,
finding evidence that recollection is fast and relatively automatic. Although
pupil dilation is sensitive to the strength and detail of memories (e.g. Otero
et al., 2011; Papesh et al., 2012), across two experiments, we have shown that
creating and retrieving memories via recollection engenders less pupil dila-
tion, relative to creating and retrieving memories via familiarity. Rather than
being slow and deliberate, we suggest that recollection is fast and easy. After
all, you do not have to work very hard to recognize your spouse when you
get home from work. It might require more effort to recognize your local
butcher, if he is there too (or so we assume… it’s an empirical question).
REFERENCES
Aggleton, J. P.,Vann, S. D., Denby, C., Dix, S., Mayes, A. R., Roberts, N., et al. (2005). S paring
of the familiarity component of recognition memory in a patient with hippocampal
pathology. Neuropsychologia, 12, 1810–1823. http://dx.doi.org/10.1016/j.neuropsycho-
logia.2005.01.019.
Beatty, J. (1982). Task-evoked pupillary responses, processing load, and the structure of
processing resources. Psychological Bulletin, 91, 276–292. http://dx.doi.org/10.1037/0033-
2909.91.2.276.
Beatty, J., & Kahneman, D. (1966). Pupillary changes in two memory tasks. Psychonomic Sci-
ence, 5, 371–372.
Boring, E. G. (1929). A history of experimental psychology. New York: Century.
Buchanan, T. W., Etzel, J. A., Adolphs, R., & Tranel, D. (2006). The influence of autonomic
arousal and semantic relatedness on memory for emotional words. International Journal of
Psychophysiology, 61, 26–33. http://dx.doi.org/10.1016/j.ijpsycho.2005.10.022.
Cansino, S., & Trejo-Morales, P. (2008). Neurophysiology of successful encoding and retrieval
of source memory. Cognitive, Affective, and Behavioral Neuroscience, 8, 85–98. http://dx.doi.
org/10.3758/CABN.8.1.85.
Clark, K. B., Naritoku, D. K., Smith, D. C., Browning, R. A., & Jensen, R. A. (1999). Enhanced
recognition memory following vagus nerve stimulation in human subjects. Nature Neu-
roscience, 2(1), 94–98. http://dx.doi.org/10.1038/4600.
Conway, M. A., Dewhurst, S. A., Pearson, N., & Sapute, A. (2001). The self and recollec-
tion reconsidered: how a ‘failure to replicate’ failed and why trace strength accounts
of recollection are untenable. Applied Cognitive Psychology, 15, 673–686. http://dx.doi.
org/10.1002/acp.740.
Croiset, G., Nijsen, M. J. M. A., & Kamphuis, P. J.G.H. (2000). Role of corticotropin-releasing
factor, vasopressin and the autonomic nervous system in learning and memory. European
Journal of Pharmacology, 405, 225–234.
Curran, T. (2000). Brain potentials of recollection and familiarity. Memory & Cognition, 28,
923–938.
Curran, T., & Hintzman, D. L. (1995). Violations of the independence assumption in pro-
cess dissociation. Journal of Experimental Psychology: Learning, Memory, and Cognition, 21,
531–547. http://dx.doi.org/10.1037/0278-7393.21.3.531.
Curran,T., & Hintzman, D. L. (1997). Comparing retrieval dynamics in recognition memory
and lexical decision. Journal of Experimental Psychology: General, 126, 228–247. http://
dx.doi.org/10.1037/0096-3445.126.3.228.
Dewhurst, S. A., Holmes, S. J., Brandt, K. R., & Dean, G. M. (2006). Measuring the speed of
the conscious components of recognition memory: remembering is faster than knowing.
Consciousness and Cognition, 15, 147–162.
Dietrich, D. E., Emrich, H. M.,Waller, C.,Wieringa, B. M., Johannes, S., & Münte,T. F. (2000).
Emotion/cognition-coupling in word recognition memory of depressive patients: an
event-related potential study. Psychiatry Research, 96, 15–29. http://dx.doi.org/10.1016/
S0165-1781(00)00187-6.
Dodson, C. S., Holland, P.W., & Shimamura, A. P. (1998). On the recollection of specific- and
partial-source information. Journal of Experimental Psychology: Learning, Memory, and Cog-
nition, 24, 1121–1136. http://dx.doi.org/10.1037/0278-7393.24.5.1121.
Donaldson, W. (1996). The role of decision processes in remembering and knowing. Memory
& Cognition, 24(4), 523–533.
Duarte, A., Ranganath, C., Winward, L., Hayward, D., & Knight, R. T. (2004). Dissociable
neural correlates for familiarity and recollection during the encoding and retrieval of
pictures. Cognitive Brain Research, 18, 255–272.
Dunn, J. C. (2004). Remember-know: a matter of confidence. Psychological Review, 111(2),
524–542. http://dx.doi.org/10.1037/0033-295X.111.2.524.
Dunn, J. C. (2008). The dimensionality of the remember-know task: a state-trace
analysis. Psychological Review, 115, 426–446. http://dx.doi.org/10.1037/0033-
295X.115.2.426.
Eldridge, L. L., Engel, S. A., Zeineh, M. M., Bookheimer, S. Y., & Knowlton, B. J. (2000).
A dissociation of encoding and retrieval processes in the human hippocampus. Journal of
Neuroscience, 25, 3280–3286. http://dx.doi.org/10.1523/JNEUROSCI.3420-04.2005.
Engelkamp, J., & Dehn, D. M. (1997). Strategy and consciousness in remembering subject-
performed actions. Sprache & Kognition, 16, 94–109.
Eschenko, O., & Sara, S. J. (2008). Learning-dependent, transient increase of activity in norad-
renergic neurons of locus coeruleus during slow wave sleep in the rat: brain stem-cortex
interplay for memory consolidation? Cerebral Cortex, 18, 2596–2603. http://dx.doi.
org/10.1093/cercor/bhn020.
Fontana, F. (1765). Dei moti dell’iride. : Lucca: J. Giusti.
Gardiner, J. M. (1988). Functional aspects of recollective experience. Memory & Cognition,
16, 309–313.
Gardiner, J. M., & Java, R. I. (1991). Forgetting in recognition memory with and without
recollective experience. Memory & Cognition, 19, 617–623. http://dx.doi.org/10.3758/
BF03197157.
Gardiner, J. M., Java, R. I., & Richardson-Klavehn, A. (1996). How level of processing really
influences awareness in recognition memory. Canadian Journal of Experimental Psychology,
50, 114–122. http://dx.doi.org/10.1037/1196-1961.50.1.114.
Gardiner, J. M., & Parkin, A. J. (1990). Attention and recollective experience in recognition
memory. Memory & Cognition, 18, 579–583.
Gardiner, J. M., Ramponi, C., & Richardson-Klavehn, A. (1998). Experiences of remem-
bering, knowing, and guessing. Consciousness and Cognition, 7, 285–288. http://dx.doi.
org/10.1006/ccog.1997.0321.
Gardiner, J. M., Ramponi, C., & Richardson-Klavehn, A. (2002). Recognition memory and
decision processes: a meta-analysis of remember, know, and guess responses. Memory, 10,
83–98. http://dx.doi.org/10.1080/0965210143000281.
Gardner, R. M., Mo, S. S., & Borrego, R. (1974). Inhibition of pupillary orienting reflex by
novelty in conjunction with recognition memory. Bulletin of the Psychonomic Society, 3,
237–238.
Gardner, R. M., Mo, S. S., & Krinsky, R. (1974). Inhibition of pupillary orienting reflex by
heteromodal novelty. Bulletin of the Psychonomic Society, 4, 510–512.
Gianaros, P. J., Van Der Veen, F. M., & Jennings, J. R. (2004). Regional cerebral blood flow
correlates with heart period and high-frequency heart period variability during work-
ing-memory tasks: implications for the cortical and subcortical regulation of cardiac
autonomic activity. Psychophysiology, 41(4), 521–530. http://dx.doi.org/10.1111/1469-
8986.2004.00179.x.
Gilzenrat, M. S., Nieuwenhuis, S., Jepma, M., & Cohen, J. D. (2010). Pupil diameter tracks
changes in control state predicted by the adaptive gain theory of locus coeruleus function.
Cognitive, Affective, and Behavioral Neuroscience, 10, 252–269. http://dx.doi.org/10.3758/
CABN.10.2.252.
Goldinger, S. D., He,Y., & Papesh, M. H. (2009). Deficits in cross-race face learning: insights
from eye-movements and pupillometry. Journal of Experimental Psychology: Learning, Mem-
ory, and Cognition, 35, 1105–1122. http://dx.doi.org/10.1037/a0016548.
Goldinger, S. D., & Papesh, M. H. (2012). Pupil dilation reflects the creation and retrieval
of memories. Current Directions in Psychological Science, 21, 90–95. http://dx.doi.
org/10.1177/0963721412436811.
Goldwater, B. C. (1972). Psychological significance of pupillary movements. Psychological Bul-
letin, 77, 340–355. http://dx.doi.org/10.1037/h0032456.
Granholm, E., Asarnow, R. F., Sarkin, A. J., & Dykes, K. L. (1996). Pupillary responses
index cognitive resource limitations. Psychophysiology, 33, 457–461. http://dx.doi.
org/10.1111/j.1469-8986.1996.tb01071.x.
Granholm, E., & Steinhauer, S. R. (2004). Introduction: pupillometric measures of cognitive
and emotional processing. International Journal of Psychophysiology, 52, 1–6. http://dx.doi.
org/10.1016/j.ijpsycho.2003.12.001.
Granholm, E., & Verney, S. P. (2004). Pupillary responses and attentional allocation problems
on the backward masking task in schizophrenia. International Journal of Psychophysiology,
52, 37–52. http://dx.doi.org/10.1016/j.ijpsycho.2003.12.004.
Gronlund, S. D., Edwards, M. B., & Ohrt, D. D. (1997). Comparison of the retrieval of item
versus spatial position information. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 23, 1261–1274. http://dx.doi.org/10.1037/0278-7393.23.6.1261.
Gronlund, S. D., & Ratcliff, R. (1989). Time course of item and associative informa-
tion: implications for global memory models. Journal of Experimental Psychology:
Learning, Memory, and Cognition, 15, 846–858. http://dx.doi.org/10.1037/0278-
7393.15.5.846.
Gruppuso,V., Lindsay, D. S., & Kelley, C. M. (1997). The process-dissociation procedure and
similarity: defining and estimating recollection and familiarity in recognition memory.
Journal of Experimental Psychology: Learning, Memory, and Cognition, 23, 259–278. http://
dx.doi.org/10.1037/0278-7393.23.2.259.
Guo, C., Duan, L., Li, W., & Paller, K. A. (2006). Distinguishing source memory and item
memory: brain potentials at encoding and retrieval. Brain Research, 1118, 142–154.
Heaver, B., & Hutton, S. B. (2011). Keeping an eye on the truth? Pupil size changes associ-
ated with recognition memory. Memory, 19, 398–405. http://dx.doi.org/10.1080/0965
8211.2011.575788.
Hess, E. H. (1965). Attitude and pupil size. Scientific American, 212, 46–54.
Hess, E. H., & Polt, J. M. (1964). Pupil size in relation to mental activity during simple
problem-solving. Science, 143, 1190–1192.
Hess, E. H., Seltzer, A. L., & Schlien, J. M. (1965). Pupil response of hetero- and homosexual
males to pictures of men and women: a pilot study. Journal of Abnormal Psychology, 70,
165–168.
Hilford, A., Glanzer, M., Kim, K., & DeCarlo, L.T. (2002). Regularities of source recognition:
ROC analysis. Journal of Experimental Psychology: General, 131, 494–510.
Hintzman, D. L., & Caulton, D. A. (1997). Recognition memory and modality judgments:
a comparison of retrieval dynamics. Journal of Memory and Language, 37, 1–23. http://
dx.doi.org/10.1006/jmla.1997.2511.
Hintzman, D. L., Caulton, D. A., & Levitin, D. J. (1998). Retrieval dynamics in recognition
and list discrimination: further evidence of separate processes of familiarity and recall.
Memory & Cognition, 26, 449–462.
Hintzman, D. L., & Curran, T. (1994). Retrieval dynamics of recognition and frequency
judgments: evidence for separate processes of familiarity and recall. Journal of Memory and
Language, 33, 1–18. http://dx.doi.org/10.1006/jmla.1994.1001.
Hirshman, E. (1998). On the logic of testing the independence assumption in the process-
dissociation procedure. Memory & Cognition, 26, 857–859. http://dx.doi.org/10.3758/
BF03201168.
Hirshman, E., & Master, S. (1997). Modeling the conscious correlates of recognition mem-
ory: reflections on the remember-paradigm. Memory & Cognition, 25, 345–351. http://
dx.doi.org/10.3758/BF03211290.
Holdstock, J. S., Mayes, A. R., Gong, Q. Y., Roberts, N., & Kapur, N. (2005). Item recog-
nition is less impaired than recall and associative recognition in a patient with selec-
tive hippocampal damage. Hippocampus, 15, 203–215. http://dx.doi.org/10.1002/
hipo.20046.
Jacoby, L. L. (1991). A process dissociation framework: separating automatic from inten-
tional uses of memory. Journal of Memory and Language, 30, 513–541. http://dx.doi.
org/10.1016/0749-596X(91)90025-F.
Jacoby, L. L. (1998). Invariance in automatic influences of memory: toward a user’s guide for
the process-dissociation procedure. Journal of Experimental Psychology: Learning, Memory,
and Cognition, 24, 3–26. http://dx.doi.org/10.1037/0278-7393.24.1.3.
Jacoby, L. L. (1999). Ironic effects of repetition: measuring age-related differences in mem-
ory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 25, 3–22. http://
dx.doi.org/10.1037/0278-7393.25.1.3.
Jacoby, L. L., Begg, I. M., & Toth, J. P. (1997). In defense of functional independence: vio-
lations of assumptions underlying the process-dissociation procedure? Journal of
Experimental Psychology: Learning, Memory, and Cognition, 23, 484–495. http://dx.doi.
org/10.1037/0278-7393.23.2.484.
Jacoby, L. L., & Rhodes, M. G. (2006). False remembering in the aged. Current Directions in
Psychological Science, 15, 49–53. http://dx.doi.org/10.1111/j.0963-7214.2006.00405.x.
Jacoby, L. L., Yonelinas, A. P., & Jennings, J. M. (1997). The relation between conscious and
unconscious (automatic) influences: a declaration of independence. In J. D. Cohen &
J. W. Schooler (Eds.), Scientific approaches to consciousness. Carnegie Mellon symposia on cogni-
tion (pp. 13–47). Hillsdale, NJ: Erlbaum.
Janisse, M. P. (1974). Pupil size, affect and exposure frequency. Social Behavior and Personality,
2, 125–146. http://dx.doi.org/10.2224/sbp.1974.2 2.125.
Johansson, M., Mecklinger, A., & Treese, A. (2004). Recognition memory for emotional and
neutral faces: an event-related potential study. Journal of Cognitive Neuroscience, 16, 1840–
1853. http://dx.doi.org/10.1162/0898929042947883.
Joordens, S., & Merikle, P. M. (1993). Independence or redundancy? Two models of con-
scious and unconscious influences. Journal of Experimental Psychology: General, 122, 462–
467. http://dx.doi.org/10.1037/0096-3445.122.4.462.
Kafkas, A., & Montaldi, D. (2011). Recognition memory strength is predicted by pupillary
responses at encoding while fixation patterns distinguish recollection from familiarity.
Quarterly Journal of Experimental Psychology, 64, 1971–1989. http://dx.doi.org/10.1080/
17470218.2011.588335.
Kahneman, D. (1973). Attention and effort. New York: Prentice Hall.
Kahneman, D., & Beatty, J. (1966). Pupil diameter and load on memory. Science, 154, 1583–
1585. http://dx.doi.org/10.1126/science.154.3756.1583.
Kahneman, D., Beatty, J., & Pollack, I. (1967). Perceptual deficit during a mental task. Science,
157, 218–219. http://dx.doi.org/10.1126/science.157.3785.218.
Karatekin, C., Couperus, J. W., & Marcus, D. J. (2004). Attention allocation in the dual-task
paradigm as measured through behavioral and psychophysiological responses. Psycho-
physiology, 41, 175–185. http://dx.doi.org/10.1111/j.1469-8986.2004.00147.x.
Kirwan, C. B., Wixted, J. T., & Squire, L. R. (2008). Activity in the medial temporal lobe
predicts memory strength, whereas activity in the prefrontal cortex predicts recollec-
tion. Journal of Neuroscience, 28, 10541–10548. http://dx.doi.org/10.1523/JNEURO-
SCI.3456-08.2008.
Klimesch, W., Doppelmayr, M., Yonelinas, A., Kroll, N. E.A., Lazzara, M., Röhm, D., et al.
(2001). Theta synchronization during episodic retrieval: neural correlates of conscious
awareness. Cognitive Brain Research, 12, 33–38. http://dx.doi.org/10.1016/S0926-
6410(01)00024-6.
Knowlton, B. J., & Squire, L. R. (1995). Remembering and knowing: two difference expres-
sion of declarative memory. Journal of Experimental Psychology: Learning, Memory, and
Cognition, 21, 699–710. http://dx.doi.org/10.1037//0278-7393.21.3.699.
Koss, M. (1986). Pupillary dilation as an index of central nervous system α2-adrenoceptor
activation. Journal of Pharmacology Methods, 15, 1–19.
Kuchinke, L.,Võ, M. L.-H., Hofmann, M., & Jacobs, A. M. (2007). Pupillary responses dur-
ing lexical decisions vary with word frequency but not emotional valence. International
Journal of Psychophysiology, 65, 132–140.
Laeng, B., Sirois, S., & Gredebäck, G. (2012). Pupillometry: a window to the preconscious? Per-
spectives on Psychological Science, 7, 18–27. http://dx.doi.org/10.1177/1745691611427305.
Loewenfeld, I. E. (1958). Mechanisms of reflex dilation of the pupil: historical review and
experimental analysis. Documenta Ophthalmologica, 12, 185–448.
Mandler, G. (1980). Recognizing: the judgment of previous occurrence. Psychological Review,
87, 252–271. http://dx.doi.org/10.1037/0033-295X.87.3.252.
McElree, B., Dolan, P. O., & Jacoby, L. L. (1999). Isolating the contributions of familiar-
ity and source information to item recognition: a time course analysis. Journal of
Experimental Psychology: Learning, Memory, and Cognition, 25, 563–582. http://dx.doi.
org/10.1037//0278-7393.25.3.563.
Mickes, L., Wais, P. E., & Wixted, J. T. (2009). Recollection is a continuous process: implica-
tion for dual-process theories of recognition memory. Psychological Science, 20, 509–515.
http://dx.doi.org/10.1111/j.1467-9280.2009.02324.x.
Moscovitch, M., & McAndrews, M. P. (2002). Material-specific deficits in “remembering”
in patients with unilateral temporal lobe epilepsy and excisions. Neuropsychologia, 40,
1335–1342. doi:PII S0028-3932(01)00213-5.
Mulligan, N. W., & Hirshman, E. (1997). Measure the bases of recognition memory: an
investigation of the process-dissociation framework. Journal of Experimental Psychol-
ogy: Learning, Memory, and Cognition, 23, 280–304. http://dx.doi.org/10.1037/0278-
7393.23.2.280.
Neisser, U. (1967). Cognitive psychology. New York: Appleton-Century-Crofts.
Otero, S. C., Weekes, B. S., & Hutton, S. B. (2011). Pupil size changes during recogni-
tion memory. Psychophysiology, 48, 1346–1353. http://dx.doi.org/10.1111/j.1469-
8986.2011.01217.x.
Papesh, M. H. (2012). Source memory revealed through eye movements and pupil dilation. Doctoral
dissertation, Arizona State University.
Papesh, M. H., & Goldinger, S. D. (2011).Your effort is showing! Pupil dilation reveals mem-
ory heuristics. In P. Higham & J. Leboe (Eds.), Constructions of remembering and metacogni-
tion (pp. 215–224). : Palgrave Macmillan.
Papesh, M. H., & Goldinger, S. D. (2012). Pupil-blah-metry: cognitive effort in speech plan-
ning reflected by pupil dilation. Attention, Perception, & Psychophysics, 74, 754–765. http://
dx.doi.org/10.3758/s13414-011-0263-y.
Papesh, M. H., Goldinger, S. D., & Hout, M. C. (2012). Memory strength and specific-
ity revealed by pupillometry. International Journal of Psychophysiology, 83, 56–64. http://
dx.doi.org/10.1016/j.ijpsycho.2011.10.002.
Partala, T., & Surakka, V. (2003). Pupil size variation as an indication of affective pro-
cessing. International Journal of Human-computer Studies, 59, 185–198. http://dx.doi.
org/10.1016/S1071-5819(03)00017-X.
Porter, G., & Troscianko, T. (2003). Pupillary response to grating stimuli. Perception, 32, 156.
Porter, G., Troscianko, T., & Gilchrist, I. D. (2007). Effort during visual search and counting:
insights from pupillometry. Quarterly Journal of Experimental Psychology, 60, 211–229.
Rajaram, S. (1993). Remembering and knowing: two means of access to the personal past.
Memory & Cognition, 21, 89–102.
Rajkowski, J., Kubiak, P., & Aston-Jones, G. (1993). Correlations between locus coeruleus
(LC) neural activity, pupil diameter and behavior in monkey support a role of LC in
attention. Society of Neuroscience Abstracts, 19, 974.
Rajkowski, J., Majczynski, H., Clayton, E., & Aston-Jones, G. (2004). Activation of monkey
locus coeruleus neurons varies with difficulty and performance in a target detection task.
Journal of Neurophysiology, 92, 361–371.
Ranganath, C., Yonelinas, A. P., Cohen, M. X., Dy, C. J., Tom, S. M., & D’Esposito, M. D.
(2004). Dissociable correlates of recollection and familiarity within the medial tem-
poral lobes. Neuropsychologia, 42, 2–13. http://dx.doi.org/10.1016/j.neuropsycholo-
gia.2003.07.006.
Ratcliff, R., Sheu, C., & Gronlund, S. D. (1992). Testing global memory models using
ROC curves. Psychological Review, 99, 518–535. http://dx.doi.org/10.1037/0033-
295X.99.3.518.
Richardson-Klavehn, A., Gardiner, J. M., & Java, R. I. (1996). Memory: task dissociations,
process dissociations, and dissociations of awareness. In G. Underwood (Ed.), Implicit
cognition (pp. 85–158). Oxford: Oxford University Press.
Rotello, C. M., & Zeng, M. (2008). Analysis of RT distributions in the remember-know

paradigm. Psychonomic Bulletin & Review, 15, 825–832. http://dx.doi.org/10.3758/
PBR.15.4.825.
Sauvage, M. M., Fortin, N. J., Owens, C. B.,Yonelinas, A. P., & Eichenbaum, H. (2008). Rec-
ognition memory: opposite effects of hippocampal damage on recollection and familiar-
ity. Nature Neuroscience, 11, 16–18.
Squire, L. R., & Wixted, J. T. (2011). The cognitive neuroscience of human memory since
H.M. Annual Review of Neuroscience, 34, 259–288.
Stanners, R. F., Coulter, M., Sweet, A. W., & Murphy, P. (1979). The pupillary response as an
indicator of arousal and cognition. Motivation and Emotion, 3, 319–340. http://dx.doi.
org/10.1007/BF00994048.
Steinhauer, S. R., & Hakerem, G. (1992). The pupillary response in cognitive psychophysiol-
ogy and schizophrenia. Annals of the New York Academy of Sciences, 658, 182–204.
Steinhauer, S. R., Siegle, G. J., Condray, R., & Pless, M. (2004). Sympathetic and parasympa-
thetic innervations of pupillary dilation during sustained processing. International Journal
of Psychophysiology, 52, 77–86. http://dx.doi.org/10.1016/j.ijpsycho.2003.12.005.
Sterpenich,V., D’Argembeau, A., Desseilles, M., Balteau, E., Albouy, G.,Vandewalle, G., et al.
(2006). The locus ceruleus is involved in the successful retrieval of emotional memories
in humans. Journal of Neuroscience, 26, 7416–7423. http://dx.doi.org/10.1523/JNEU-
ROSCI.1001-06.2006.
Tulving, E. (1985). Memory and consciousness. Canadian Psychologist, 25, 1–12.
Uncapher, M. R., & Rugg, M. D. (2005). Effects of divided attention on fMRI correlates
of memory encoding. Journal of Cognitive Neuroscience, 17, 1923–1935. http://dx.doi.
org/10.1162/089892905775008616.
Van Gerven, P. W.M., Paas, F., Van Merriënboer, J. J. G., & Schmidt, H. G. (2004). Memory
load and the cognitive pupillary response in aging. Psychophysiology, 41, 167–174. http://
dx.doi.org/10.1111/j.1469-8986.2003.00148.x.
Verfaellie, M., Rajaram, S., Fossum, K., & Williams, L. (2008). Not all repetition is alike: dif-
ferent benefits of repetition in amnesia and normal memory. Journal of the International
Neuropsychological Society, 14, 365–372.
Võ, M. L.-H., Jacobs,A. M., Kuchinke, L., Hofmann, M., Conrad, M., Schacht,A., et al. (2008).
The coupling of emotion and cognition in the eye: introducing the pupil old/new effect.
Psychophysiology, 45, 130–140. http://dx.doi.org/10.1111/j.1469-8986.2008.00745.x.
Wais, P. E. (2008). fMRI signals associated with memory strength in the medial temporal
lobes: a meta-analysis. Neuropsychologia, 46, 3185–3196. http://dx.doi.org/10.1016/j.
neuropsychologia.2008.08.025.
Wais, P. E., Squire, L. R., & Wixted, J. T. (2009). In search of recollection and familiarity
signals in the hippocampus. Journal of Cognitive Neuroscience, 22, 109–123. http://dx.doi.
org/10.1162/jocn.2009.21190.
Wais, P. E., Wixted, J. T., Hopkins, R. O., & Squire, L. R. (2006). The hippocampus supports
both the recollection and the familiarity components of recognition memory. Neuron,
49, 459–466. http://dx.doi.org/10.1016/j.neuron.2005.12.020.
Wheeler, M. E., & Buckner, R. L. (2004). Functional-anatomic correlates of remembering
and knowing. NeuroImage, 21, 1337–1349.
Whittlesea, B. W. A. (1997). Production, evaluation and preservation of experiences: con-
structive processing in remembering and performance tasks. In D. Medin (Ed.), The
psychology of learning and motivation. (Vol. 37). New York: Academic Press.
Whittlesea, B. W. A., & Leboe, J. (2000). The heuristic basis of remembering and classifica-
tion: fluency, generation and resemblance. Journal of Experimental Psychology: General, 129,
84–106.
Whittlesea, B. W. A., & Williams, L. D. (1998). Why do strangers feel familiar, but friends
don’t? A discrepancy-attribution account of feelings of familiarity. ACTA Psychologica,
98, 141–165.
Whittlesea, B. W. A., & Williams, L. D. (2001). The discrepancy-attribution hypothesis: I. The

heuristic basis of feelings of familiarity. Journal of Experimental Psychology: Learning, Mem-
ory, and Cognition, 27, 3–13.
Wixted, J.T. (2007). Dual-process theory and signal detection theory of recognition memory.
Psychological Review, 114, 152–176. http://dx.doi.org/10.1037/0033-295X.114.1.152.
Wixted, J. T. (2009). Remember/know judgments in cognitive neuroscience: an illustration
of the underrepresented point of view. Learning & Memory, 16, 406–412. http://dx.doi.
org/10.1101/lm.1312809.
Wixted, J. T., & Mickes, L. (2010). A continuous dual-process model of remember/know
judgments. Psychological Review, 117, 1025–1054. http://dx.doi.org/10.1037/a0030874.
Wixted, J.T., & Stretch,V. (2004). In defense of the signal detection interpretation of remem-
ber/know judgments. Psychonomic Bulletin & Review, 11, 616–641.
Yonelinas, A. P. (1994). Receiver-operating characteristics in recognition memory: evidence
for a dual-process model. Journal of Experimental Psychology: Learning, Memory, and Cogni-
tion, 20, 1341–1354. http://dx.doi.org/10.1037//0278-7393.20.6.1341.
Yonelinas, A. P. (2001). Components of episodic memory: the contribution of recollection
and familiarity. Philosophical Transactions of the Royal Society of London. Series B, Biological
Sciences, 356, 1363–1374.
Yonelinas, A. P. (2002). The nature of recollection and familiarity: a review of 30 years of
research. Journal of Memory and Language, 46, 441–517. http://dx.doi.org/10.1006/
jmla.2002.2864.
Yonelinas, A. P., & Jacoby, L. L. (1996). Response bias and the process-dissociation pro-
cedure. Journal of Experimental Psychology: General, 125, 422–434. http://dx.doi.
org/10.1037/0096-3445.125.4.422.
Yonelinas, A. P., Kroll, N. E. A., Quamme, J. R., Lazzara, M. M., Sauve, M. J., Widaman, K. F.,
et al. (2002). Effects of extensive temporal lobe damage or mild hypoxia on recollection
and familiarity. Nature Neuroscience, 5, 1236–1241. http://dx.doi.org/10.1038/nn961.
Yonelinas, A. P., & Parks, C. M. (2007). Receiver operating characteristics (ROCs) in
recognition memory: a review. Psychological Bulletin, 133, 800–832. http://dx.doi.
org/10.1037/0033-2909.133.5.800.
CHAPTER SIX
A Mechanistic Approach to
Individual Differences in
Spatial Learning, Memory, and
Navigation
Amy L. Shelton1, Steven A. Marchette, Andrew J. Furman
Department of Psychological and Brain Sciences, Johns Hopkins University, Baltimore, MD, USA
Contents
1. Introduction224
2. What Does It Mean to Measure Spatial Learning and Navigational Ability? 225
2.1. Differences in Navigational Success Rates 225
2.2. Differences in Strategies/Styles: Landmark–Route–Survey Framework 226
2.3. Challenges to the Landmark–Route–Survey Framework 227
3. Dual Systems for Spatial Learning in Rodents 229
4. Place and Response Learning in Humans 232
4.1. Virtual Water Maze 232
4.2. Human Radial Arm Mazes 235
4.3. Ecological Paradigms 236
5. The Place/Response Framework for Individual Differences 239
5.1. Balance of Systems as the Foundation for Individual Differences 240
5.2. Classification vs Predicting Solutions 241
5.3. Advantages and Disadvantages for Different Solutions 242
5.4. Separating Preference from Prowess 244
6. Connections to Other Sources of Variability 246
6.1. Aging247
6.2. Sex Differences and Hormonal Influences 248
7. Competition or Interaction of Systems 250
8. Conclusions252
References255
Abstract
Navigation is a complex task that depends on the processes of perception, learning,
memory, and reasoning to be successful. Given this complexity, it is not surprising that
humans (and other species) vary dramatically in their approach and success at naviga-
tion. This wide range of abilities has been of great interest to the field of human spatial
cognition. In addition, spatial navigation is a cross-species phenomenon that can speak

ISSN 0079-7421, http://dx.doi.org/10.1016/B978-0-12-407187-2.00006-X All rights reserved. 223
224 Amy L. Shelton et al.
to a variety of learning and memory processes. Therefore, understanding individual dif-

ferences in this domain can offer a wide range of insights that affect many behaviors in
the real world. A cognitive framework that gives precedent to the flexible use of spatial
information and explicit or declarative learning processes has driven much of the work
on individual differences in navigation in humans. However, animal models of basic
learning mechanisms may also offer substantial insight into individual differences in
both how well people navigate their surroundings and in the strategies or styles that
they bring to bear on the navigational problems. This mechanistic approach may offer
a stronger foundation for not only how individual differences might emerge but also
how they interact with differences in the environments and goals that drive our need
to learn, remember, and navigate in the world.
1. INTRODUCTION
Whether considering how ancestral humans learned to hunt for food
and shelter in the wilderness or the more mundane issues of how you got
to work today, there is little question that navigation is an important aspect
of the human experience. Many of our everyday tasks—finding the car in a
parking lot, getting to the checkout line at the grocery store—involve spa-
tial navigation. As humans, we accomplish this navigation in various ways,
and everyone has an opinion about his or her own navigational abilities.
Ask any random group of people, and you will get answers ranging from
“I am great with directions” to “I get lost all the time”. This massive range
of individual differences in perceived spatial prowess can also be seen in
behavior. Most people can identify those friends or family members who
they prefer to ride with on trips or those who give good or bad direc-
tions. We also know people who are dependent on their Global Positioning
System (GPS) units for even the most basic of travel tasks. This variability
in navigational ability likely reflects the fact that navigation is a complex
task that requires individuals to encode the immediate surroundings, inter-
rogate memory (working memory and/or long-term memory), determine
how to proceed, and execute the relevant movements through space. As
such, understanding individual differences in navigation requires a com-
prehensive approach to understanding how individuals might differ in the
way they learn and remember information about the spatial organization of
their environments.
Individual differences in spatial learning and navigation have enjoyed a
long history in the literature. Here, we review that literature in light of the
extensive work on nonhuman animal models of spatial learning mecha-
nisms. In particular, we explore how using these learning mechanisms to
A Mechanistic Approach to Individual Differences 225
ground individual differences in human spatial learning and navigation

might offer a framework for understanding how the wide range of differ-
ences emerge. In addition, navigational tasks have been used to understand
more general processes of learning and memory in human and nonhuman
animals (for brief review, see Bird & Burgess, 2008). In humans, much of
the focus has been on the role of explicit or declarative kinds of processes,
consistent with the description of navigation as a cognitively demand-
ing task of perceiving, planning, and executing. By couching learning and
navigational differences in basic learning mechanisms, we can more readily
investigate the contributions of other implicit or habit-based mechanisms
to complex behaviors. Moreover, this will allow extension to other sources
of variability such as aging and hormones. Taken together with the extant
literature, this framework has substantial potential to enrich a wide range
of domains.
The following sections provide a context for thinking about individual
differences in spatial learning and navigation beginning with ideas about
what it means to measure navigational ability. Subsequent sections describe
the dominant framework for thinking about individual differences in strate-
gies and elaborate on a neurobiological framework coming out of rodent
learning that might offer a broader and more comprehensive approach. The
final sections highlight the advantages of this new framework for under-
standing many sources of variability.
2. WHAT DOES IT MEAN TO MEASURE SPATIAL

LEARNING AND NAVIGATIONAL ABILITY?
Despite the remarkable variability in reports of navigational ability,
humans (and other animals) are quite successful on the daily task of navigat-
ing.That is, few of us fail to reach our office or make it home from the gro-
cery store. What, then, characterizes individual differences in this ability? In
this section, we discuss the two broad facets of performance that comprise
individual differences, success rate and strategies/style.
2.1. Differences in Navigational Success Rates

The first dimension on which we tend to classify people is on the dimen-
sion of success—or good/bad navigation ability. Interestingly, when we
consider whether people are good or bad navigators, we often empha-
size how well they perform when faced with a new learning or a novel
navigational situation. For example, if we are traveling with friends and
want to figure out how to get to a new restaurant, we often appeal to

the person or people in the group who we think will be most likely to
find this new location. Similarly, we consider who is going to be best at
giving directions. This anecdotal emphasis on these novel or flexible uses
of spatial information suggests that good/bad navigators may actually be
good/bad explorers.
This emphasis on new learning and flexibility carries through to empiri-
cal work on individual differences in navigational abilities. In most experi-
mental studies of spatial memory and navigation, participants learn novel
environments (for exceptions, see Marchette,Yerramsetti, Burns, & Shelton,
2011; Yerramsetti, Marchette, & Shelton, 2012). Individual differences are
largely characterized in terms of differences in accuracy and success rate at
navigation to make inferences about the quality of spatial learning and the
factors associated with those abilities (e.g. Allen, Kirasic, Dobson, Long, &
Beck, 1996; Fields & Shelton, 2006; Ishikawa & Montello, 2006). In many
cases, the assessment emphasizes the explicit knowledge of spatial relation-
ships. For example, Ishikawa and Montello (2006) used a test of straight-
line distances to assess how well people learned under various conditions.
Similarly, Fields and Shelton used judgments of relative direction after dif-
ferent encoding conditions. In both these cases, the measure of learning/
navigation was how accurately people judged the distances or directions.
Critically, these tests require that a person understand the global structure
and use the space flexibly. This emphasis on flexibility in learning is also
reflected in the Santa Barbara Sense of Direction questionnaire (SBSOD;
Hegarty, Richardson, Montello, Lovelace, & Subbiah, 2002), one of the most
commonly used measures of spatial navigational ability. In this self-report
measure, participants report on their general ability and comfort level with
different kinds of spatial scenarios. Like many of the measures, the SBSOD
emphasizes the ability to use spatial information flexibly, marking flexibility
in perceptual and cognitive processing as a primary factor in navigational
success (see Wolbers & Hegarty, 2010).
2.2. Differences in Strategies/Styles: Landmark–

Route–Survey Framework
Whether couched in terms of successful navigation or successful explora-
tion, success rate is not the only dimension along which humans differ in
navigation. There is also substantial anecdotal and self-report data to sup-
port the claim that humans differ dramatically in the way they successfully
navigate. For example, imagine asking a room full of healthy adults how
they navigate to work each day. Some individuals will report using the
same set route over and over again, whereas others will report navigat-
ing by just keeping a desired direction in mind. Still others might report
changing their route as a function of time of day, observed patterns of
traffic, or even weather conditions. Any comprehensive understanding of
individual differences will require an appreciation of these different styles
or strategies for navigating and how those different approaches might be
more or less effective for different individuals and under different naviga-
tional conditions.
Like success rate, navigational styles and strategies have a history in the
research literature. A common framework for thinking about such indi-
vidual differences has stemmed from ideas about the progression of knowl-
edge (Siegel & White, 1975). In this popular framework, one first learns the
landmarks in an environment. Landmarks are defined as features or objects
that can demarcate locations or directions. As landmarks become linked,
one develops route knowledge, which is generally defined as the knowledge
of a set of ordered paths. Finally, this information begins to form a repre-
sentation of the global context as survey knowledge, generally defined as
knowledge of the organization of locations in a more global framework. In
its strongest instantiation, this framework suggests that learning an environ-
ment goes through each of these stages, each building upon the previously
acquired knowledge, with survey knowledge as the ultimate representation
of an environment.
The landmark–route–survey framework has influenced the direction of
a wide range of studies in spatial cognition, including approaches to indi-
vidual differences. For example, one tool used to assess spatial navigational
strategies is the Questionnaire on Spatial Representation (QSR; Pazzaglia &
De Beni, 2001). In this self-report survey, questions are directed at assessing
the degree to which people use different types of information, and scoring
allows assessment of landmark use, route-based strategies, and survey-based
strategies. Similar efforts to classify navigational behaviors have taken a simi-
lar approach, with a strong emphasis on the route-based and survey-based
distinctions (e.g. Lawton, 1994, 1996).
2.3. Challenges to the Landmark–Route–Survey Framework

This classification scheme has been useful in advancing the ideas that navi-
gational ability includes both success rate and strategic differences, but
grounding individual differences in a stage-based theory promotes the idea
that some forms of knowledge or types of strategies are inherently better
than others. Most notably, this framework has given top status to survey
knowledge or the use of spatial information in a flexible manner based on
knowledge of the global organization. This is reflected in many empirical
studies that assess ability based on flexibility and use of detours. However,
there are important counterpoints to consider when considering how to
classify navigational style and its relationship to ability.
The first counterpoint is the evidence challenging the successive nature
of the stages of knowledge. Although there is little doubt that the distinc-
tion between route-based and survey-based information is important to
understanding the kinds of strategies people use and the cues that might
be involved in spatial learning, it is not clear that the progression from one
type of knowledge to another is a critical feature (e.g. Ishikawa & Montello,
2006; Taylor, Naylor, & Chechile, 1999). For example, Taylor, Naylor, and
Chechile gave people specific goals that emphasized the need for either
route-based knowledge or survey-based knowledge. They then gave tests
that tapped each type of knowledge. By the stage theory, individuals who
show survey-based knowledge should also have the route-based knowl-
edge from which the survey knowledge was built. By contrast, their results
showed that the specific goals drove the type of representation such that
people who developed survey-based representations did not necessarily
develop the route-based representations as well. This suggests that different
types of representations can be built up somewhat independently (or at least
without a strict dependence on each other’s presence).
Second, the sense that survey-based strategies convey a clear advantage
over other strategies does not account for the potential interaction between
the navigational situation and strategies. Emerging frameworks have started
to acknowledge that the success of navigational wayfinding likely depends
on the combination of strategies (e.g. Ishikawa & Montello, 2006) and/or
the interplay of skills or strategies and the specific environment or naviga-
tional challenges (e.g. Carlson, Hölscher, Shipley, & Dalton, 2010; Wolbers
& Hegarty, 2010). For example, Carlson et al. put forth a conceptual frame-
work for indoor navigation where navigational success depends on the
combination of the building features, the cognitive representation of the
space that has been formed, and the strategies and skills that are brought
to bear. The latter distinction between the mental representation and the
strategies and skills suggests that there is not a one-to-one link between
a particular navigational strategy and the underlying representation. This
opens the door for thinking about strategies (or styles of navigation) as a
more flexible tool that interacts with the information that is represented.
Finally, both anecdotal and empirical evidence suggests that the route-
based and survey-based dichotomy (or landmark–route–survey trichotomy)
may be insufficient to account for the variety of behaviors people actually
utilize in navigation. There have been recent proposals for additional cat-
egories of processing such as “graph knowledge” (for review, see Chrastil,
2012). However, even within a category, there may be additional distinc-
tions that need to be considered (e.g. Allen et al., 1996; Chan, Baumann,
Bellgrove, & Mattingly, 2012; Waller & Lippa, 2007). For example, Allen
et al. tested environmental learning with tasks that tapped both the for-
ward and the backward sequences of the routes. Participants could vary in
both of these depending on the degree to which they had strong explicit
knowledge of the route or more associative knowledge of the sequence in
a stimulus–response fashion. Similarly, people who report using landmarks
could be using them as cues to a specific action (e.g. turn left at the church)
or simply as beacons detached from the specific action required (e.g. look
for the church and head in that direction). These two cases would likely
result in similar classification as some combination of landmark and route-
based strategies, but they are serving different purposes and require differ-
ent levels of explicit knowledge in the act of navigating. Waller and Lippa
demonstrated how differences in performance depended on which way
the landmarks were used, suggesting that these distinctions may be impor-
tant for understanding what people can or cannot do. This represents just
one case where the landmark–route–survey framework may be incomplete.
Once we break away from this framework, the potential to identify and
characterize these additional navigational strategies will enhance our appre-
ciation for individual differences.
3. DUAL SYSTEMS FOR SPATIAL LEARNING IN

RODENTS
One framework that has been emerging as a major candidate for
explaining aspects of human navigational behavior is based on the dual-
system model in rodents. This model stems from several decades of work
on the learning mechanisms or systems used by rodents to learn an envi-
ronment. In the basic navigational paradigm, rodents are placed in a sim-
ple cross maze with one arm blocked (i.e. dual-solution T-maze; Figure
6.1(A)). The rat is placed in the same start arm relative to both the envi-
ronment and the location of a food reward over many trials. After learning,
the rat is given a probe trial on which the start location is rotated with
Figure 6.1 Diagram of the dual-solution T-maze used to explore place and response
learning in rodents (e.g. Packard & McGaugh, 1996; Restle, 1957; Tolman et al., 1947). (A)
Encoding environment with stable cues and path; (B) probe trial from novel start arm.
Shapes represent cues visible in the surrounding environment. (For color version of this
figure, the reader is referred to the online version of this book.)
respect to the cues in the larger environment (Figure 6.1(B)). On probe

trials, rats can exhibit one of two behaviors. If the rat turns down the arm
that is the same location as the reward relative to the cues in the environ-
ment (right turn in Figure 6.1(B)), this is taken as evidence of place learning.
Alternatively, if the rat makes the same physical turn as during the learn-
ing phase (left turn in Figure 6.1(B)), this is taken as evidence of response
learning. This distinction is intended to reflect the fact that during repeated
learning in a stable environment, the rat has the potential to learn the loca-
tion in (at least two ways), relative to the environmental cues or as a series
of required responses (e.g. Restle, 1957; Tolman, 1948; Tolman, Ritchie, &
Kalish, 1947).
Place and response learning have been distinguished in the rodent based
on a number of different properties in both the dual-solution T-maze and
other preparations. For example, place learning tends to emerge earlier in
learning before sufficient repetition has occurred to establish the response
(e.g. Packard & McGaugh, 1996). Similarly, conditions that require place
learning tend to be more rapidly learned than conditions that require
response learning (e.g. Ritchie, Aeschlimm, & Pierce, 1950; Tolman, Ritchie,
& Kalish, 1946). Place learning has also been viewed as more cognitively
demanding than response learning as rats placed under conditions of
distraction or stressors tend to show more response-learning behaviors
(e.g. Elliot & Packard, 2008; Schwabe, Schächinger, de Kloe, & Oitzl, 2010).
In addition to behavioral differences, these two learning mechanisms

have been associated with different underlying brain systems in the rat.
Using temporary deactivation of brain regions via lidocaine injection, Packard
and McGaugh (1996) had rats learning a T-maze preparation over time.
In different groups of rats, they temporarily deactivated either the caudate
nucleus or the hippocampus. These deactivations had both temporally and
behaviorally specific results. In control rats (saline administration), place-
learning behavior (using the environmental cues) was observed in earlier
training blocks and response-learning behavior (engaging the same left or
right turn) became more prevalent in later training blocks. This pattern was
unaffected by the site or timing of saline injection. By contrast, lidocaine
injections to the hippocampus had pronounced effects on performance
early in training with a reduction in place-learning behaviors, whereas lido-
caine injections to the caudate nucleus affected later performance with a
reduction in response-learning behaviors. This and subsequent work has
established the hippocampus and caudate nucleus as markers for the place-
and response-learning systems, respectively, in the rodent brain.
Notably, the results not only supported the use of two systems but also
suggested that they are simultaneously acquired. In particular, rats that
received lidocaine injections to the caudate late in learning not only showed
a reduction in their ability to manifest response-learning behaviors but also
shifted to place-learning behaviors. This unveiling of place-learning behav-
ior suggests that those animals had the necessary place-learned knowledge
to compensate when the response-learning system was deactivated. That is,
although their typical behavior was the expression of response learning, the
capacity to show the place-learning behavior was intact.There was no sym-
metrical boost in response-learning behavior during early inactivation of
the hippocampus because there had presumably been insufficient repetition
to acquire the response. Taken together with the behavioral characteriza-
tion, this work offers both behavioral and neurobiological markers of place
and response learning that can be useful for thinking about how humans
develop their capacity to remember and navigate in the world.
An added advantage to incorporating nonhuman animal models is the
potential to address a wider range of factors.The dual-system model has been
used to investigate the influences of hormones, neurotransmitter systems, and
aging on rodent learning and memory. Ultimately, the use of this framework
as a foundation for individual differences will invite these critical extensions
as well.We return to these ideas in the latter sections. First, we consider how
place and response learning have been characterized in humans.
4. PLACE AND RESPONSE LEARNING IN HUMANS

The growing popularity of both desktop and immersive virtual real-
ity has broadened the range of tools and types of questions about spatial
learning and navigation that can be addressed in humans. One important
consequence of these new tools has been the motivation to adapt rodent
paradigms for use in humans. Although this adaptation comes with a variety
of challenges (Box 6.1), there have been variations on most of the stan-
dard paradigms. Studies using variations on the virtual water maze (e.g.
Cornwell, Johnson, Holroyd, Carver, & Grillon, 2008; Hamilton, Driscoll,
& Sutherland, 2002; Newman & Kraszniak, 2000) and the virtual radial
arm maze (Bohbot, Iaria, & Petrides, 2004; Iaria, Petrides, Dagher, Pike, &
Bohbot, 2003) employ close approximations of the rodent preparations,
whereas other more ecological tasks such as virtual towns (e.g. Hartley,
Maguire, Spiers, & Burgess, 2003), mazes (e.g. Brown, Ross,Tobyne, & Stern,
2012), and dual-solution mazes (e.g. Marchette, Bakker, & Shelton, 2011)
offer more conceptual adaptations.Together, these tools have begun to offer
both general distinctions in human navigational behavior and insights into
the potential mechanisms of individual differences.
4.1. Virtual Water Maze

The Morris water maze (MWM; Morris, 1981, 1984) is a circular pool of
cloudy water with a hidden platform at a goal location (Figure 6.2). Rats
are placed in the maze and must swim to the platform. Early in learning,
rats must explore the environment to locate the platform, but most animals
become quite adept at finding the platform. Manipulations of the environ-
ment have demonstrated that rats tend to use stable environmental cues
to guide navigation (place learning), and performance on the mnemonic
aspects of this task breaks down with damage to the hippocampus (e.g.
Morris, 1984; Morris, Garrud, Rawlins, & O’Keefe, 1982; Stewart & Morris,
1993). As such, the MWM has been used in a variety of rodent studies as a
classic test of spatial memory associated with the hippocampal place-learning
system.
Variations of the MWM for humans (Cornwell et al., 2008; Driscoll,
Hamilton, Yeo, Brooks, & Sutherland, 2005; Hamilton et al., 2002;
Newman & Kaszniak, 2000; Skelton, Ross, Nerad, & Livingstone, 2006) typ-
ically involve placing individuals inside a virtual drum or multisided room
(for an actual physical hidden sensor version, see Bohbot et al., 1998).As in the
Box 6.1 Challenges to Adapting Rodent Paradigms

for Humans
One key to making connections between human navigation and the dual-system
model of place and response learning in rodents has been the adaptation of the
rodent paradigms for use in humans. This adaptation brings with it some obvious
and not-so-obvious challenges. Here, we summarize some of those challenges and
the key questions that one might need to address in developing new paradigms.
Scalability
Virtual reality has made it easy to create environments of nearly any size and shape,
but one can ask whether a physically scalable environment is still appropriate for
asking the same questions in humans. The virtual radial arm maze (e.g. Bohbot et al.,
2004; Iaria et al., 2003) may represent the most direct analog to a rodent paradigm
in terms of scalability because both the human versions can essentially replicate the
physical structure of the rodent mazes. The virtual variants on the MWM for humans
(Cornwell et al., 2008; Hamilton et al., 2002; Newman & Kaszniak, 2000; Skelton et al.,
2006) are often similar in structure to the rodent paradigm, in that they tend to
use circular (or multisided) drums scaled up for humans; however, humans are not
dropped into murky water to hide the locations and instead navigate in "dry" ver-
sions, which may have additional implications for other factors noted below.
In contrast to the radial arm maze and MWM, some rodent paradigms are phys-
ically scalable in virtual space but likely operate very differently for humans. For
example, the dual-solution T-maze has been a long-standing tool for eliciting place-
and response-learned behaviors in rodents (e.g. Packard & McGaugh, 1996; Restle,
1957; Tolman, 1948), but they are not particularly useful in (adult) humans because
humans reduce them to single-trial learning. In pilot work, healthy adult humans
very quickly learned that they had two things to remember and often asked after
just one training trial whether they were supposed to just make a specific turn or go
toward a specific arm in the room. As such, these mazes did not offer us the same
window into place and response systems in humans that they did in rodents.
Motivation
A second important issue is to consider whether the motivation for navigation
matters. For example, when a rat is dropped into the MWM, it is arguably swim-
ming for survival, whereas a human placed in a virtual water maze has little or
no threat to survival if he/she decides to stand still. At worst, the human par-
ticipant faces boredom (or perhaps reduced compensation in incentive-based
paradigms), and it would be unethical to introduce truly comparable manipu-
lations. The cases that use reward in rodents may allow for more comparable
motivational states if "getting out", and/or monetary compensation in humans is
sufficiently rewarding. However, given that navigation and memory systems are
affected by such things as stress (e.g. Kim & Yoon, 1998), it is important to at least
consider how the motivational factors contribute.
(Continued)
Box 6.1 Challenges to Adapting Rodent Paradigms

for Humans—(cont’d)
Comparable Ecological Validity
Finally, it is useful to consider whether the ecological validity is similar across
species. For both rats and humans, the water maze-type tasks may be equally
unusual navigational situations, but what about T-mazes and radial arm mazes?
Rats outside the lab tend to gravitate to tight places—underground, in walls and
foundations, etc., whereas humans largely evolved in larger open spaces and
have extensive experience in highly structure buildings and rich developed envi-
ronments. One could imagine that something like the radial arm maze is more
“natural” for rats than for humans. Critically, the ecological validity in general is
important for making claims about real-world navigational ability, but it is equally
important to make sure that the ecological validity is at least comparable in the
rodents and humans if the goal is to allow for cross-species insights. Although
this is not evidence based, it represents an issue worth considering when adapt-
ing paradigms across species and understanding the limitations of their interpre-
tation. Given these challenges, paradigm development should begin with careful
consideration of what conceptual elements one is trying to capture to offer the
strongest analogs between humans and rodents. For example, when creating an
analog to the dual-solution T-maze (see Marchette, Bakker, et al., 2011), it became
abundantly clear that we needed an alternative environment, and we considered
what elements of the paradigm were critical. We identified several, including
(1) encoding conditions had to afford learning a stable environment and learn-
ing a repeated behavior; (2) retrieval had to offer a way of distinguishing which
sources of learning was driving behavior; and (3) the retrieval had to be suffi-
ciently ambiguous so that individuals would be free to use either behavior. We
then considered additional elements such as whether to use incentives for suc-
cessful completions, whether to impose stressors or time limits, or whether the
stability of the environment had to be based solely on extra-maze distal cues. The
intended result was a paradigm that looked very different on the surface but (we
hope) shared a deeper conceptual structure with its rodent analog. Moreover,
this careful analysis also offered a large number of potential manipulations that
could be useful directions for future research.
rodent MWM, there is a predefined goal location that is typically hidden

from view (or shares the same visual features as other items in the envi-
ronment). Humans are given repeated opportunities to learn the location
of the hidden goal using the information in and around the virtual maze.
Much of this work has shown that, as in rodents, performance depends on
the integrity of the hippocampus and surrounding medial temporal lobe
(e.g. Astur, Taylor, Mamelak, Philpott, & Sutherland, 2002; Bohbot et al.,
1998), supporting the parallel between the rodent and human work.
Figure 6.2 Diagram of a typical water maze preparation (e.g. Morris, 1981). Shapes rep-
resent cues visible in the surrounding environment or on the walls of the drum. The
platform is not visible to the animal during testing. (For color version of this figure, the
4.2. Human Radial Arm Mazes

Radial arm mazes in rodents (e.g. Olton, Collison, & Werz, 1977) require
rats to explore a series of visually identical arms arranged in around a central
start location (Figure 6.3). In the simplest version, each arm is baited with
a reward, and the rat is required to remember which arms he has visited.
Returning to a previously visited arm is recorded as an error, and one can
infer how the rat is using various cues based on the pattern of navigation
and errors. The human virtual radial arm maze has been used in healthy
controls and patients with medial temporal lobe damage to investigate dif-
ferences between the use of “spatial” strategies that depended on flexible
understanding of the spatial relations akin to place learning and “nonspatial”
strategies that depend on more associative behavioral responses (e.g. Bohbot
et al., 2004; Iaria et al., 2003). Iaria et al. observed that healthy individuals
were evenly split on their spontaneous use of the spatial (place-like) and
nonspatial (response-like) strategies early in the experiment. Over time,
many of those who began with a spatial strategy shifted to the nonspatial
strategy after many repetitions. Bohbot et al. used this same preparation to
demonstrate that activation in the hippocampus and caudate was associated
with the spatial and nonspatial strategies, respectively, suggesting that the
putative place and response systems were supporting the different strategies,
in a manner consistent with what has been observed in rats.
Figure 6.3 Diagram of a typical eight-arm radial arm maze, similar to those used in rats
(e.g. Olton et al., 1977) and humans (e.g. Bohbot et al., 2004; Iaria et al., 2003). Shapes
represent cues visible in the surrounding environment. (For color version of this figure,
the reader is referred to the online version of this book.)
4.3. Ecological Paradigms

An alternative to using paradigms identical to those used in rats has been
to develop paradigms for humans that capture the critical features of the
rodent paradigms in tasks more suited to human spatial learning and navi-
gation. These tasks offer ecologically grounded tasks and environments but
also include elements that mirror or approximate the features of rodent
paradigms known to evoke the place/response distinction.
Maguire and colleagues have examined both real and virtual navigation
ability (e.g. Hartley et al. 2003; Maguire, Frith, Burgess, Donnett, & O’Keefe,
1998; Spiers & Maguire, 2007). Maguire et al. showed that activation in the
right hippocampus was associated with learning and flexibly navigating in
a virtual environment, consistent with building up and utilizing a place-
learned representation (see also Shelton & Gabrieli, 2002, 2004). Con-
versely, the degree to which participants demonstrated rapid travel through
the environment was associated with activation in the right caudate nucleus.
Similarly, Hartley et al. varied the type of learning task and observed a simi-
lar distinction. Activation in the hippocampus and caudate was more closely
associated with exploratory wayfinding and route following, respectively.
Figure 6.4 The human DSP: (A) schematic of an environment showing the familiar path
learned during encoding and locations of the 12 target objects. (Adapted from Figure 1
of Marchette, Bakker, et al. (2011)). (B) Image of the actual environment as seen by par-
ticipants during initial encoding and subsequent navigation. (For color version of this
figure, the reader is referred to the online version of this book.)
Exploratory wayfinding offers more flexible exploration, promoting place-

like learning, whereas route following offers the repetition required for
response-like learning. Activation in these regions was also associated with
measures of performance on the respective tasks, suggesting that they might
support individual difference in strategies.
More recently, we have introduced the dual-solution paradigm for
humans (DSP; Marchette, Bakker, et al., 2011) as a human analog to the
dual-solution T-maze. In this paradigm, participants learn the locations of 12
objects in a virtual maze by viewing the same circuitous path through the
environment repeatedly. The path is constructed such that one experiences
the entire environment (all possible pathways are visible at some point in the
path), but the direct experience also allows learning a single specific path. As
such, this is an environment that should allow both place-like and response-
like learning of the locations (Figure 6.4). At test, participants are placed
along the learned path and asked to navigate to a given target. On any given
trial, the participant is free to use whatever path he/she prefers; on critical
trials, there is always a novel path that is shorter than the familiar learned
path. If people are trying to navigate efficiently, as they are instructed to do,
then they should opt for shortcuts if they are able to recognize them.That is,
when there are two solutions available, they should arguably take the shorter
option. The primary behavioral measure is the propensity to use either the
shortcut or the familiar path (i.e. solution index).
Figure 6.5 Histogram of individual differences in the solution index in the DSP. (Adapted
from Marchette, Bakker, et al. (2011), Figure 2). (For color version of this figure, the reader
is referred to the online version of this book.)
The DSP has been useful in characterizing individual differences

because the solution index has revealed substantial variability (Figure 6.5;
Marchette, Bakker, et al., 2011) and appears to have good stability in the
basic paradigm (Furman, Marchette, & Shelton, 2013). More importantly,
people are not strictly using one type of solution or the other. Instead, par-
ticipants fall along a continuum of performance from people who always
use familiar paths to people who always use shortcuts when available, with
many individuals using a mixture of both familiar paths and novel short-
cuts. In addition, brain activation in the markers for the putative place
and response systems is associated with the solution index. In Marchette,
Bakker, and Shelton, 20 participants learned the DSP environment during
functional magnetic resonance imaging (fMRI) scanning and then per-
formed the subsequent retrieval. Brain activation in the hippocampus and
caudate nucleus was measured, and a normed ratio of activation was cal-
culated. This normed ratio of activation during encoding was significantly
correlated with the subsequent solution index (Figure 6.6), indicating that
the balance of these systems during encoding predicts how one chooses to
interrogate and utilize the memory during navigation.This predictive value
of encoding activation suggests that the interplay of these two systems may
offer important clues to why individual differ in their preferred solutions
to navigational challenges.
Figure 6.6 Regions of interest analysis for linking brain activation during encoding to
navigational solutions. Left panel shows the regions identified in an overall contrast of
learning to control. Right panel shows the correlation between the normed ratio of acti-
vation in these regions and the subsequent solution index that was observed. (Adapted
from Marchette, Bakker, et al. (2011), Figure 4). (For color version of this figure, the reader
is referred to the online version of this book.)
Taken together, these human analogs to rodent studies provide a clear

case both for the existence of place-like and response-like mechanisms and
for individual differences in the engagement of these systems. In the follow-
ing section, we take this a step further to consider how framing individual
differences in terms of the contributions of these well-described learn-
ing mechanisms might strengthen the current framework and offer novel
insights into both the existence and the emergence of different strategies
and styles both within and across individuals.
5. THE PLACE/RESPONSE FRAMEWORK FOR

INDIVIDUAL DIFFERENCES
Given the wealth of evidence for a distinction akin to place and
response learning in humans, it seems plausible that this framework could
be considered in parallel with the classic landmark–route–survey framework
that has dominated the literature on individual differences in navigational
strategies/styles. In this section, we describe how this framework might be
utilized and what questions still need to be addressed to fully appreciate
a new classification scheme that begins with underlying mechanisms and
builds up to how people talk about their own navigational strategies in their
daily lives.
5.1. Balance of Systems as the Foundation for Individual

Differences
At the core of this framework is the idea that any given individual has
both place-like and response-like learning mechanisms available and that
these mechanisms underlie most, if not all, human spatial learning. How-
ever, even at this mechanistic level, individuals differ in the engagement
of these systems during spatial learning. This framework proposes that it
is these low-level differences that lay the foundation for the wide-ranging
variability in navigational success and strategy across individuals. The clas-
sic landmark–route–survey framework and the associated measures start at
the end state of what people have learned or prefer to learn and at least
implicitly assumes that this classification will generalize to a broad range of
navigational challenges. By incorporating a place/response framework, we
can begin to address what motivates an individual to report or express a
certain end-state strategy. That is, preferences and strategies emerge as skills
and experiences that are built upon differences in the bias associated with
the relative engagement of the place and response systems. Those experi-
ences and skills will then feed back on the place and response systems to
either maintain or shift that relative engagement, allowing both stabilization
of preferences and the potential for developing a range of strategies.
The framework does not posit a one-to-one mapping from place and
response to specific strategies (e.g. place = survey) but rather offers a more
basic level of analysis for thinking about how strategies and preferences
emerge. Place learning, as described above, has been associated with terms
like cognitive map and explicit learning of space. It is generally defined as
developing a mental representation of locations with respect to the struc-
ture and cues in the environment such that it allows flexible interrogation
and inferences about the space. Arguably, most people exhibit some form
of this type of learning (although many people would claim otherwise),
and this is consistent with the extensive observation of activation in the
“navigation network” in the brain (e.g. Maguire et al., 1998; Marchette,
Bakker, et al., 2011; Shelton & Gabrieli, 2002). However, stronger engage-
ment of this system likely offers access to more survey-like strategies (e.g.
Chocolatea is north of campus) as well as more explicit uses of route
information (e.g. the Starbuck’s is coming up on the right).That is, it is the
system that allows one to learn and represent relationships among loca-
tions. By contrast, response learning has largely been defined in terms of
habits or triggers for specific actions in space. A system that offers ready
access to associated actions would be well suited for serving route-based

strategies either by priming particular reactions at decision points in the
environment or in coordination with place learning to support more
explicit routes.
An important difference between this framework and the more classic
classification is the ability to consider how different strategies and prefer-
ences might have different profiles of advantages and disadvantages rather
than whether those strategies are generally good or bad. When making
broad classifications, the interaction of ones classification and the context of
navigation is largely limited to whether one is likely to be successful or not
given the conditions. For example, an individual classified as a route learner
will be expected to succeed or fail as a function of whether the environ-
ment affords learning clear paths that can be recreated later. However, if
individual differences emerge from biases in lower level learning mecha-
nisms, we can begin to consider what different strategies might bring to any
given navigational situation. An additional consequence of this framework
is that it affords thinking about a given individuals range of available strategies
and how that range might interact with the specific environment or con-
text. These advantages are elaborated in the following sections.
5.2. Classification vs Predicting Solutions

Although one important goal in studying individual differences is to offer
clear classification, as noted above, any general classification of navigational
strategy has the potential to ignore the role of strategy shifts in response to
changing navigational challenges. One advantage of the learning mecha-
nisms framework is that it does not reduce individual differences to single
categories but offers ideas about how having different biases might alter
the set of available solutions and affect an individual’s ability to select from
within that set. When measuring strategies and solutions with the DSP, we
find that most participants varied not in whether they used shortcuts or
familiar paths but in how often they used each solution (Marchette, Bakker,
et al., 2011).This variability in degree of use suggests that an important part
of characterizing individual differences needs to address when and why a
given individual might select one solution in one case and a different solu-
tion in another case.
By considering learning mechanisms, we can characterize individuals
according to their bias toward place- or response-learning mechanisms as
a starting point for developing strategies and predicting which solution(s)
might be employed. For example, if an individual shows a strong caudate/

response-learning bias (in brain activation and/or in tendency to use familiar
paths), he/she may be primed to use that type of strategy in many situations
and may have difficulty engaging alternative solutions. Alternatively, if an
individual has a more balanced initial bias, he/she may be more readily able
to shift between solutions. By introducing a variety of navigational chal-
lenges that favor one solution or another, we can begin to develop a model
that not only considers one’s initial bias but also explores the ability to shift.
To make some early strides toward this effort, in a recent series of experi-
ments, participants in the DSP were first classified with a solution index. In a
subsequent series of trials, participants were told when shortcuts were avail-
able. This manipulation was intended to motivate people to look for a short-
cut.The results suggested that people who were already using some shortcuts
often shifted to using them more frequently once they were cued, whereas
some people who were using very limited shortcuts ignored the cues (Furman
et al., 2013). Even in this simple case of reduced ambiguity, an individual’s
initial score on the solution index was a reasonable indicator of how he/she
responded to the cues. Ultimately, this will still offer ways to classify individu-
als, but the classification will be more continuous and offer insights into the
range of strategies a given individual might be able to employ.
5.3. Advantages and Disadvantages for Different Solutions

Inherent in the above discussion is the notion that different solutions/strate-
gies may be more or less appropriate for different navigational conditions
and challenges. This is in direct contrast to the landmark–route–survey pro-
gression, which suggests that flexible survey knowledge is the gold standard.
Given that the learning mechanisms framework actually stresses the coexis-
tence of (at least) two systems, appealing to learning mechanisms raises the
important question of why these two systems coexist for navigation. One
obvious answer is that each system may have an important part to play in
the wide range of spatial learning and navigational challenges that humans
(and other animals) face. Bringing this to bear on individual differences
opens the door for thinking about how different solutions and strategies
might offer distinct advantages and disadvantages to allow successful naviga-
tion under different conditions.
First, there are known advantages to the place- and response-learning
systems in rodents. Whereas place learning affords flexible use of the envi-
ronmental cues in the face of detours or changes in start location, response
learning offers a form of “automaticity” that can be engaged under
conditions of stress or distraction (e.g. Kim, Lee, Han, & Packard, 2001;
Packard & Wingard, 2004). Human navigators also face a wide range of
conditions, including both unexpected detours and concurrent distractions
while navigating. Individual differences in the bias to engage the place and
response systems likely affect the reaction to these changing conditions. In
a study of distraction, we asked participants to complete the DSP naviga-
tional task with and without verbal shadowing (Clark, Marchette, Furman,
& Shelton, in preparation).We measured their propensity to use shortcuts or
familiar paths (solution index) and observed that people shifted to using the
familiar path more often when distracted by verbal shadowing than when
not distracted.We also found that performance was worse under distraction.
In addition, those who maintained the use of shortcuts in the face of distrac-
tion also appeared to take the largest hit on accuracy, suggesting that these
individuals were not switching to the more optimal familiar route solutions.
To consider more broadly how different solutions and different learning
mechanisms might convey different advantages, it is useful to think in terms
of the interaction of (1) the environmental structure and available cues, (2)
the skills that the observer brings to bear, and (3) the goals and demands
of the navigational challenge. This interaction of factors is consistent with
recent models (e.g. Carlson et al., 2010), and the learning mechanisms
framework offers a starting point for thinking about how some aspects of
this coordination might occur.
In the simplest form, the goal of navigation is to get to one’s destination.To
that simple goal, we can add different features (one-way streets, hills, multiple
stoplights) or different constraints (running late, avoiding bad neighborhoods)
that may affect which solutions or strategies might be optimal. For example,
in a city with a lot of one-way streets, establishing known routes may be the
safest and most efficient way to be successful. This might convey an advan-
tage for someone traditionally classified as a route learner or someone in the
place/response framework who has a stronger tendency toward familiar paths.
However, this advantage may disappear in that same environment when walk-
ing instead of driving. Similarly, we can imagine an environment rich with
salient landmarks. In such an environment, encoding locations with respect to
landmark proximity might offer an advantage over established sets of routes.1
1 This
example reminds me of an anecdote from a visit to a colleague’s campus. A graduate student
was guiding us to lunch and seemed to be taking a somewhat circuitous route. Upon our arrival at the
building, he admitted that he was nervous about getting there correctly because he had been using
a crane to beacon to that location for the past 2 years. The crane had recently been removed, and his
previously efficient strategy was no longer available.
In addition, there is the internal motivation of an individual that might

drive what he/she is willing to do. For example, if the immediate goal is to
get somewhere as quickly as possible without error, then it is critical that
the individual uses a strategy that will be error free. In this case, any uncer-
tainty about a potential path would reduce the likelihood of using that path.
In our work on cuing shortcuts (Furman et al., 2013), we observed this
phenomenon in several people. Individuals who opted for the familiar path
when uncued would sometimes switch to using a shortcut once they had
the very basic knowledge that there was a shortcut available. Many of these
individuals reported that they often knew that there might be a shortcut but
indicated that they were only willing to try it when the cue offered them
certainty. Notably, we also had individuals who found the cues frustrat-
ing because they simply had no idea which paths might be the shortcuts;
these individuals did not shift their behavior in response to the cues. In the
landmark–route–survey framework, both types of individuals would likely
be labeled as route learners, but the latter group clearly has a different set of
available solutions and must employ them differently.
Finally, by thinking in terms of advantages and disadvantages, we can
also begin to think about a broader range of strategies that might emerge
from the interaction of these learning mechanisms with experiences, envi-
ronments, and goals. Even within the place- and response-learning net-
works, there have been suggestions for additional distinctions. For example,
within the striatal system that is believed to support response-like learning,
different portions of the neuroanatomy have been associated with differ-
ent kinds of landmark use: beaconing has been associated with activation
in the ventral and medial aspects of the striatum (Devan & White, 1999),
whereas the use of landmarks as associative cues has been associated with
more dorsal and lateral regions (Featherstone & McDonald, 2004, 2005;
see also Chang & Gold, 2004). Understanding how such distinctions in the
underlying neural systems manifest themselves in complex behaviors will
offer richer ideas about how individual differences emerge. Taken together,
thinking about different solutions and strategies as having distinct advan-
tages also raises important questions about what it means to be a good or
bad navigator.
5.4. Separating Preference from Prowess

As noted earlier, the classification of spatial ability in terms of success has
strong anecdotal and empirical bases, and there has been an emphasis on
certain strategies being better or worse than others. However, there has also
been clear recognition that “success” has been defined in specific contexts
that may favor one solution or strategy over another. One of the important
features of the DSP, driven by the place/response framework, is that different
solutions are equally successful.That is, whether someone sticks to the famil-
iar route or opts for a novel path, the trial will be successful as long as they
arrive within the time limit. This ability to use different solutions to solve
the same problem is consistent with the real-world situation of navigating
familiar environments. Although we do encounter the occasional roadblock,
our environments generally afford finding and using a particular route or
knowing the environmental structure well enough to use multiple routes.
The first important question then is whether there is a clear relationship
between success rate and the types of solutions people use. In earlier stud-
ies, it is clear that sense of direction as measured by SBSOD was related to
an individual’s ability to complete certain navigational tasks (for review, see
Wolbers & Hegarty, 2010); however, as we have suggested, both the SBSOD
and the empirical tests of navigation tend to emphasize the flexible use of
space and should therefore by related. In the DSP, where success is deter-
mined by taking a clean path to the destination within a reasonable time
limit, we can compare solution index (rate of use of shortcuts or familiar
paths) to the success rate (percent trials completed). As shown in Figure 6.7,
in a large sample, the ability to successfully complete the trials was not
dependent on which solution the individual preferred. Critically, the rela-
tionship trends in the opposite direction of that predicted from the idea that
flexible use of space is best; that is, people using more shortcuts (more place
learning) appear more variable and slightly less successful on average than
those using more familiar paths. More recent work (as yet unpublished) has
replicated this work and substantiated it by showing that self-report mea-
sures of strategic preferences are also unrelated to success rate on the DSP,
even with more stringent time limits.
Predicting success also carries with it the important question of general-
ization. Taking the individual’s most frequent solution or most pronounced
solution may miscalculate their actual ability to respond to certain situations.
In other words, just because someone prefers to use familiar routes does not
necessarily mean that he/she does not have the knowledge to use shortcuts.
The use of a more continuous measure of preference (as in the solution index
and its variants) may offer more insight into how to measure success rates. Sub-
stantial efforts are underway to try to make use of more quantitative approaches
to classification, and the place/response framework may offer one of the clearest
starting points from which to work.
Figure 6.7 Data from the Marchette, Bakker, et al. (2011) studies showing no relation-
ship between solution index and rate of success. (For color version of this figure, the
Finally, returning to the question of what it means to be a good or

bad navigator, we offer a more adaptive approach. In the simplest form,
being good means getting where you need to go. In this sense, the ability
to apply an appropriate strategy or solution, given the constraints of the
environment, the skills of the observer, and the current goals, should be the
mark of good navigation. Our work on shifting strategies under instruction
(Furman et al., 2013) and conditions of distraction (Clark et al., in prepara-
tion) begin to weigh in on this issue by assessing who can shift solutions
and how much they shift as a function of a wide variety of spatial skills and
preferences. Ongoing work in this domain is likely to offer remarkable new
insights and new ways to think about individual differences in navigational
ability.
6. CONNECTIONS TO OTHER SOURCES OF

VARIABILITY
In addition to providing a general framework for thinking about indi-
vidual differences, the appeal to a place/response model has implications for
other sources of variability. A large body of research has identified factors,
such as aging, sex differences, and hormonal balance that can modulate nav-
igational success and strategy. Although post hoc descriptions of how these
factors impact navigation have been marshaled, there is currently no frame-
work for explaining how these modulators may shape differences among
individuals However, a direct benefit of considering neurobiologically
grounded learning mechanisms is that it allows us to build a connection
between these factors and how they may shape individual differences in the
way people use strategies and solutions.
6.1. Aging
Aging is associated with decline in a number of cognitive domains; among
them are the processes of spatial learning and navigation. When compared
with their younger counterparts, older adults are impaired in their ability
to learn and remember ways of finding their way to important destinations
(for review, see Moffat, 2009), the physical layout of the environment (Iaria,
Palermo, Committeri, & Barton, 2009; Kirasic & Mathes, 1990; Kirasic, 2000),
as well as their position within it (Iaria et al., 2009; Mahmood, Adamo,
Briceno, & Moffat, 2009).This suggests one important area where an appeal
to learning mechanisms might offer insights into the effects of aging.
In rodents, there has been substantial effort to understand how such
changes in behavior with aging might be related to differential changes in
the underlying spatial learning mechanisms (e.g. Gallagher & Pelleymounter,
1988; Rapp, Rosenberg, & Gallagher, 1987; Rosenzweig & Barnes, 2003).
For example, when allowed to solve a task using place- or response-related
information, aged rats rely on more response-based solutions than younger
animals (Barnes, Nadel, & Honig, 1980). In humans, a similar case appears to
be building. Notably, the kinds of tasks that demonstrate the spatial memory
decline focus on explicit route learning or the ability to make flexible infer-
ence using spatial representations of the environment. For example, in vari-
ants of the MWM adapted for humans, the common observation is that older
adults take consistently longer to locate the platform, spend less time near its
former location on probe trials, and are less able to explicitly indicate where
the platform is located compared with younger adults (e.g. Driscoll et al.,
2005; Moffat, 2009; Moffat & Resnick, 2002). In addition, these age-related
deficits are accompanied by functional impairment within the medial tem-
poral lobe (e.g. De Leon et al., 1997; Jack et al., 1998) and often the broader
navigation network (parahippocampal gyrus and retrosplenial cortex; see
e.g. Meulenbroek, Petersson, Voermans, Weber, & Fernández, 2004; Moffat,
Elkins, & Resnick, 2006).Taken together, these results in rodents and humans
are consistent with a relatively selective impairment in the place-learning
system.
In the framework we have outlined, these apparent deficits in place
learning with aging would suggest the loss or impairment with some strat-
egies and only in some situations, allowing specific predictions about pat-
terns of impairment and sparing. First, individuals who favor solutions and
strategies that depend more heavily on a response-like system might show

less degradation in their overall navigational experiences. Similarly, individ-
uals who can readily shift strategies should have the capacity to compensate
the loss of place-learning proficiency by utilizing other solutions. Indeed,
on versions of the virtual MWM that allow more response-based strategies,
caudate nucleus volume has been positively correlated with successful per-
formance, suggesting that increased dependence on the response-learning
system might counter at least some of the deficits observed with aging
(Moffat, Kennedy, Rodrigue, & Raz, 2007). These predictions suggest that
the way we test for spatial learning and memory deficits in aging may need
to include tests for alternative solutions and styles.
A final key fact about aging is that the impairments are graded in nature,
with some individual rats even seeming to escape impairment (Gallagher,
Burwell, & Burchinal, 1993), and similar sparing in human aging is associ-
ated with a reduced loss of hippocampal volume (Driscoll et al., 2003).
These results suggest that age-related impairments in spatial learning may
be tied to the degree place-learning impairment, an observation in line
with our results showing that use of strategies was associated with place
learning falls along a continuum. Although it currently remains unexplored,
an expectation of our model is that these graded deficits in place learning
should be met with similarly graded increases in reliance on response learn-
ing. In addition, our work on shifting strategies suggests that some individu-
als may fail to show this shift toward response learning. These predictions
offer another potential avenue for examining why some individuals show
more marked decline then others, which may feed ideas for methods for
helping aging individuals to adapt their strategies more readily.2
6.2. Sex Differences and Hormonal Influences

The domain of navigation has traditionally been a fertile ground for study-
ing sex differences because there have been many reports of differences
between males and females in their ability to learn and use particular sorts
of spatial information. Most commonly, males tend to prefer using knowl-
edge of their position with respect to locations within the environment, or
survey knowledge, whereas females tend to prefer more route-based forms
of information (e.g. Lawton, 1994, 1996). In line with these self-reports,
2 Much of this discussion is also relevant to some of the navigational deficits observed in such
neurological disorders as mild cognitive impairment and Alzheimer’s disease, where impairments have
been associated with hippocampal damage or dysfunction (e.g. Bellassen, Iglói, de SouzaDubois, &
Rondi-Reig, 2012).
sex differences tend to be revealed in tasks that require participants to make

judgments over the larger scale global structure (e.g. Lawton, Charleston,
& Zieles, 1996) and in situations that require navigation using cardinal
directions rather than landmark information (Saucier et al., 2002). These
results suggest that men and women may rely on and learn about differ-
ent sources of information in the environment (see also Kelly, McNamara,
Bodenheimer, Carr, & Rieser, 2009; Sandstrom, Kaufman & Huettel, 1998).
Sex differences can also be linked to different hormonal profiles, another
potential sources of variability that may affect navigational behaviors. More-
over, place- and response-learning mechanisms have already been associ-
ated with different hormonal profiles in both rodents (e.g. Davis, Jacobson,
Aliakbari, & Mizumori, 2005; Gold & Korol, 2010; Korol, 2004; Naghdi,
Majlessi, & Bozorgmehr, 2005; Roof & Havens, 1992; Warren & Juraska,
1997) and humans (Bell & Saucier, 2004; Driscoll et al., 2005). For example,
work by Korol and colleagues has shown that experimentally manipulated
levels of estrogen in female rats modulate the preferential use of place and
response learning. In particular, they found that estrogen infusions into the
hippocampus, but not striatum, enhanced place learning, whereas estro-
gen infusions into the striatum, but not hippocampus, impaired response
learning. Although results with testosterone have been more varied
(e.g. Clark, Mitre, & Brinck-Johnsen, 1995; Naghdi et al., 2005; Roof &
Havens, 1992; Sandstrom, Kim, & Wasserman, 2006), there have been
suggestions that testosterone has a relationship to MWM performance

that may be associated with sex difference. For example, Roof and Havens
observed that the hippocampal formation in female rats given neonatal
testosterone was more similar to male rats than to other females that did
not receive testosterone; the treated females also showed more male-like
performance than nontreated rats. These hormonal influences in rodents
offer a potential explanation for sex differences as a function of differential
hormonal effects on place and response learning.
In humans, circulating testosterone levels appear to be a predictor of both
spatial reasoning and the ability to acquire spatial knowledge (Bell & Saucier,
2004; Driscoll et al., 2005). Critically, lower levels of testosterone were asso-
ciated with better performance in men, whereas higher levels of testosterone
are associated with better performance in women, suggesting that too much
or too little circulating hormone affected performance on explicit kinds of
spatial learning tasks. Our framework, in conjunction with the observations
in rodents, would suggest that these associations with apparent place learn-
ing might predict changes in response-learning rates as well. In particular, it
predicts that conditions that reduce place learning might enhance response
learning, which would, in turn, change the balance of strategies that indi-
viduals utilize. Again, as in other areas, this would take our thinking away
from ideas about how hormones are good or bad for navigation and toward
ideas about what people might engage under different hormonal influences.
Hormones are a particularly interesting factor to consider because, in
addition to capturing features about the specific individual (e.g. baseline
levels of circulating testosterone), hormones are sensitive to acute state
changes. For example, stress-related hormones may play a similar role in
changing the bias between the memory systems associated with place and
response learning (for review, see Packard, 2009; Brain Research). In par-
ticular, chronic stress appears to reduce the hippocampal plasticity that is
thought to be a core component of the place-learning system (e.g. Kim &
Yoon, 1998). What’s more, systemic administration of corticosteroids, a class
of hormones released in response to stress, biases rodents away from place
learning and toward response learning in a maze that allowed both solu-
tions (Sadowski, Jackson, Wieczorek, & Gold, 2009; Schwabe et al., 2010).
This may offer interesting insights into human navigational styles because
high spatial anxiety tends to coincide with more frequent use or reliance on
familiar routes (Hund & Minarik, 2006; Lawton, 1994). Thinking in terms
of biasing systems allows us to make specific predictions about the interplay
of stress hormones, acute or chronic stress/anxiety, and navigational styles.
This is another example of the theoretical richness a neurobiologically
grounded framework of individual differences might provide.
7. COMPETITION OR INTERACTION OF SYSTEMS

Before drawing some final conclusions about the role of a place/
response framework in thinking about individual differences, it is important
to address the relationship between the putative place and response sys-
tems. In the literature on rodents and humans, there has long been a debate
about the degree to which these systems are competing with each other
for resources or interacting with each other to give rise to behavior
(e.g. Poldrack & Packard, 2003). Although a thorough discussion of this
issue is beyond the scope of this chapter, there are some important points
to make with respect to how the debate affects the proposed framework
for individual differences.
In some sense, the basic place and response characterization based on
rodent behavior evokes a sense of competition by necessity. That is, only
one type of behavior can be produced, and therefore, only one system can
be manifested at any given time. For example, when Tolman and colleagues
placed rats in a T-maze from a novel start position, the rat had to either turn
toward the arm consistent with the target location relative to the environ-
mental cues or turn toward the arm consistent with the repeated action
during the acquisition (e.g. Packard & McGaugh, 1996; Restle, 1957;Tolman
et al., 1946, 1947). Therefore, in terms of being able to drive behavior, there
must at least be some form of competition that allows one system to dominate
at any given time.
Beyond the competition to drive behavior, one can ask whether there
is a necessary trade-off in resources between the two neural systems that
support the different learning mechanisms.The first important point is that
the two systems are operating in parallel during encoding, as evidenced
by the extensive work showing that rodents can successfully learn under
conditions that require either a place or a response approach. Moreover,
shutting down one system at the time of retrieval allows the other system
to control behavior (Packard & McGaugh, 1996), further supporting the
ideas that learning was occurring simultaneously. This coincidental learning
is consistent with an account that allows for either independence or coor-
dination of these systems, with the “competition” largely at the behavioral
output end.
In humans, the story with respect to competition or cooperation is
less clear. Marchette, Bakker, et al. (2011) found that the activation in
the hippocampus and caudate nucleus during encoding predicted the
solutions that were employed. Critically, although the individual regions
predicted performance as expected, it was the relative activation of the
two regions that offered the strongest correlation. This appeal to the bal-
ance of activation suggests that both systems may be operating at higher
or lower levels such that when they are in balance, people are likely
to use a mixture of solutions and that mixture shifts as the balance of
engagement shifts.
In an alternative approach to the distinction between hippocampal and
striatal contributions, Brown et al. (2012) investigated the concurrent acti-
vation in hippocampus and caudate, as well as prefrontal regions, in par-
ticipants learning and navigating in virtual mazes. Participants learned to
navigate in a series of small mazes and were then asked to navigate those
learned environments during fMRI scans. The activation was then interro-
gated for mazes that shared overlapping features compared with those that
did not. The results revealed greater functional connectivity between the
hippocampus and the caudate in conditions that required making behav-

ioral distinctions in environments with overlapping features compared with
conditions in which the environments did not overlap.This finding suggests
that the underlying place-like and response-like systems may operate in the
disambiguation of different navigational context.
Overall, the question of whether these two systems are competing, inde-
pendent, or interactive may depend on the particular situation in which
an individual is acting. In the framework proposed here, the critical con-
cept is that these two systems are present and show individual variability in
their engagement. As both the questions of competition vs coordination are
addressed and the issues surrounding individual differences become clari-
fied, the proposed place/response framework will become elaborated and
enriched.
8. CONCLUSIONS
Our understanding of individual differences in navigation has long
been shaped by models and approaches that emphasize how people engage
novel spaces. The standard landmark–route–survey framework proposes
that, in an attempt to maximize their later success, people try to focus on
learning the information they think will be important to remember later,
such as the location of a particularly salient landmark, a sequence of direc-
tions along a route, or the spatial relationships between important places.
People then ultimately meet with different amounts of success that reflect
how much their strategy has allowed them to accurately learn about the
environment, with good navigators identified by their ability to form more
holistic spatial representations. This framework stresses what people explic-
itly try to do, and much of our understanding of these complex strate-
gies necessarily comes from people’s self-report.These descriptions of what
people report using are then in turn used to classify navigators into differ-
ent discrete categories.
The work summarized here on human place- and response-learning
mechanisms paints a complementary picture to the landmark–route–survey
model but may also challenge some of its norms for defining individual
differences. The principle challenge stems from the observation that when
we tested participants on the DSP, we did not see two clear clusters of
people who preferred place-like or response-like solutions. Instead, par-
ticipants ranged continuously across the spectrum of possible preferences.
Consistent with the rodent models, this continuous range suggests that both
learning mechanisms are available to any individual and that each individual
has some bias of variable strength toward one learning mechanism or the
other. In practice, this also means that an individual is using fundamentally
different solutions on different trials. This observation may push us out of
old norms for thinking about distinct categories for classifying individuals
based on what they report to be their dominant strategy.
Once we allow the possibility that the use of navigational strategy,
at least at the level of the deployment of learning mechanisms, is con-
tinuous rather than discrete, we have to develop ideas about why strate-
gies are deployed continuously. One possibility is that the strategy of the
moment may be shaped by the particular circumstances of navigation,
rather than being determined by what participants try to do in general.
This is a concept that heralds a change from thinking about there being
strategies that are “better” and “worse” overall: normatively, under a bet-
ter/worse scheme, an individual should principally use the best strategy
that he or she can!
An alternative to the standard viewpoint, and one that we subscribe to,
is that different strategies may convey advantages and disadvantages, which
might be suitable to particular situations or optimize different navigational
criteria, such as weighing speed of travel against certainty of arrival. Having
access to both learning mechanisms is then adaptive because it allows an
individual to select the strategy that will best match their own navigational
goals and constraints. This framing may redirect our attention to how strat-
egy selection may reflect the constraints that an individual navigator places
on defining a successful trip.
Taken together, these different profiles of strengths and weaknesses mean
that some situations will call for place learning whereas others may call for
response learning. Navigational success, then, is not simply associated with
which solution one may prefer in the DSP. However, a fine point in this
argument is that we are not arguing that one’s predisposition for place and
response learning is unrelated to navigational success in the real world. The
DSP was designed such that both place and response learning can support
successful navigation, but in the real world, this may not always be the case.
Our suggestion is that, while prowess with a chosen strategy matters, navi-
gational success on a particular trip will be based primarily on how well the
individual’s strategy fits the particular situation. Our traditional predisposi-
tion as researchers to call place-like strategies “good” consequently results
from the fact that they are widely applicable, rather than universally more
successful than response-like strategies. Under this view, a truly successful
navigator is not someone with a commitment to a particularly “good” strat-

egy, but rather one who is able to recognize the right strategy at the right
time and have the flexibility, and ability, to apply it.
The learning mechanisms framework addresses a contradiction that
arises in the study of individual differences in navigation. A common trope
that virtually every paper on navigation uses to motivate their study is
that navigation is a core skill for mobile beings and is crucial to survival.
Despite this claim, a notable portion of the population reports being quite
bad at navigation without experiencing daily threats to their survival. In
our view, the resolution of this contradiction must be to accept that most
healthy individuals have adapted to the navigational needs of their time
and environment and found solutions that work. We do not question the
value of research into how people learn and adapt to novel environments.
Indeed, it is precisely these kinds of studies that will enlighten our under-
standing of individual differences. However, we suggest that studying
learning mechanisms will open up a rich largely unstudied territory in
navigational research: what are the challenges and criteria of daily naviga-
tion in the real world and how might an individual’s strategies and apti-
tudes be mustered to meet them? How flexibly can these be shifted as
circumstances change?
In conclusion, the distinction between place and response learning is
both historically important and incredibly promising for enlightening our
understanding of spatial learning mechanisms. However, it is not immedi-
ately obvious how place- and response-learning mechanisms translate to
the many complex strategies that people report for navigating the world.
There do seem to be some analogies between place learning and survey
or “map-like” strategies and response learning and route-based navigation;
however, we suggest here that this superficial similarity masks more inter-
esting relationships. Rather than simple one-to-one correspondences, the
concepts of place and response learning form the foundations for under-
standing how different learning mechanisms with different intrinsic biases
can cooperate or compete to give rise to complex strategies that might be
appropriate for the challenges of navigation. Here, we have argued that the
study of these learning mechanisms provides a framework for understand-
ing individual differences in navigation—not a fully specified model. This
is not an oversight on our part, but instead reflects the multidimensional
nature of navigation and the need to expand our focus to the range of
navigational circumstances, criteria, and constraints that individuals face
on a daily basis.
REFERENCES
Allen, G. L., Kirasic, K. C., Dobson, S. H., Long, R. G., & Beck, S. (1996). Predicting
environmental learning from spatial abilities: an indirect route. Intelligence, 22(3),
327–355.
Astur, R. S., Taylor, L. B., Mamelak, A. N., Philpott, L., & Sutherland, R. J. (2002). Humans
with hippocampus damage display severe spatial memory impairments in a virtual Mor-
ris water task. Behavioural Brain Research, 132(1), 77–84.
Barnes, C. A., Nadel, L., & Honig, W. K. (1980). Spatial memory deficit in senescent rats.
Canadian Journal of Psychology, 34(1), 29.
Bellassen,V., Iglói, K., de Souza, L. C., Dubois, B., & Rondi-Reig, L. (2012). Temporal order
memory assessed during spatiotemporal navigation as a behavioral cognitive marker for
differential Alzheimer’s disease diagnosis. Journal of Neuroscience, 32(6), 1942–1952.
Bell, S., & Saucier, D. (2004). Relationship among environmental pointing accuracy, mental
rotation, sex, and hormones. Environment and Behavior, 36(2), 251–265.
Bird, C. M., & Burgess, N. (2008). The hippocampus and memory: insights from spatial pro-
cessing. Nature Reviews Neuroscience, 9(3), 182–194.
Bohbot, V. D., Iaria, G., & Petrides, M. (2004). Hippocampal function and spatial memory:
evidence from functional neuroimaging in healthy participants and performance of
patients with medial temporal lobe resections. Neuropsychology, 18(3), 418.
Bohbot, V. D., Kalina, M., Stepankova, K., Spackova, N., Petrides, M., & Nadel, L. Y. N. N.
(1998). Spatial memory deficits in patients with lesions to the right hippocampus and to
the right parahippocampal cortex. Neuropsychologia, 36(11), 1217–1238.
Brown,T. I., Ross, R. S.,Tobyne, S. M., & Stern, C. E. (2012). Cooperative interactions between
hippocampal and striatal systems support flexible navigation. NeuroImage, 60(2), 1316.
Carlson, L. A., Hölscher, C., Shipley, T. F., & Dalton, R. C. (2010). Getting lost in buildings.
Current Directions in Psychological Science, 19(5), 284–289.
Chan, E., Baumann, O., Bellgrove, M. A., & Mattingley, J. B. (2012). From objects to land-
marks: the function of visual location information in spatial navigation. Frontiers in Psy-
chology, 3, 304.
Chang, Q., & Gold, P. E. (2004). Inactivation of dorsolateral striatum impairs acquisition of
response learning in cue-deficient, but not cue-available, conditions. Behavioral Neurosci-
ence, 118(2), 383.
Chrastil, E. R. (2012). Neural evidence supports a novel framework for spatial navigation.
Psychonomic Bulletin & Review.
Clark, S., Marchette, S. A., Furman, A. J., & Shelton, A. L. Talking while walking: investigating the
effects of verbal distraction on human navigation performance and strategy selection, in preparation.
Clark, A. S., Mitre, M. C., & Brinck-Johnsen, T. (1995). Anabolic-androgenic steroid and
adrenal steroid effects on hippocampal plasticity. Brain Research, 679(1), 64–71.
Cornwell, B. R., Johnson, L. L., Holroyd, T., Carver, F. W., & Grillon, C. (2008). Human hip-
pocampal and parahippocampal theta during goal-directed spatial navigation predicts
performance on a virtual Morris water maze. Journal of Neuroscience, 28(23), 5983–5990.
Davis, D. M., Jacobson, T. K., Aliakbari, S., & Mizumori, S. J.Y. (2005). Differential effects of
estrogen on hippocampal-and striatal-dependent learning. Neurobiology of Learning and
Memory, 84(2), 132–137.
De Leon, M. J., George, A. E., Golomb, J., Tarshish, C., Convit, A., Kluger, A., et al. (1997).
Frequency of hippocampal formation atrophy in normal aging and Alzheimer’s disease.
Neurobiology of Aging, 18(1), 1–11.
Devan, B. D., & White, N. M. (1999). Parallel information processing in the dorsal striatum:
relation to hippocampal function. Journal of Neuroscience, 19(7), 2789–2798.
Driscoll, I., Hamilton, D. A., Petropoulos, H.,Yeo, R. A., Brooks, W. M., Baumgartner, R. N.,
et al. (2003). The aging hippocampus: cognitive, biochemical and structural findings.
Cerebral Cortex, 13(12), 1344–1351.
Driscoll, I., Hamilton, D. A., Yeo, R. A., Brooks, W. M., & Sutherland, R. J. (2005). Virtual
navigation in humans: the impact of age, sex, and hormones on place learning. Hormones
and Behavior, 47(3), 326–335.
Elliott, A. E., & Packard, M. G. (2008). Intra-amygdala anxiogenic drug infusion prior to
retrieval biases rats towards the use of habit memory. Neurobiology of Learning and Memory,
90(4), 616–623.
Featherstone, R. E., & McDonald, R. J. (2004). Dorsal striatum and stimulus–response learn-
ing: lesions of the dorsolateral, but not dorsomedial, striatum impair acquisition of a
simple discrimination task. Behavioural Brain Research, 150(1), 15–23.
Featherstone, R. E., & McDonald, R. J. (2005). Lesions of the dorsolateral or dorsomedial
striatum impair performance of a previously acquired simple discrimination task. Neuro-
biology of Learning and Memory, 84(3), 159–167.
Fields, A.W., & Shelton, A. L. (2006). Individual skill differences and large-scale environmen-
tal learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 32(3), 506.
Furman, A. J., , Marchette, S. A., & Shelton, A. L. (2013). Individual differences in navigational
style and its relationship to volitional strategy shifting. Manuscript submitted for publication.
Gallagher, M., Burwell, R., & Burchinal, M. R. (1993). Severity of spatial learning impair-
ment in aging: development of a learning index for performance in the Morris water
maze. Behavioral Neuroscience, 107(4), 618.
Gallagher, M., & Pelleymounter, M. A. (1988). Spatial learning deficits in old rats: a model for
memory decline in the aged. Neurobiology of Aging, 9, 549–556.
Gold, P. E., & Korol, D. L. (2010). Hormones and memory. In G. Koob, M. Le Moal &
R. F. Thompson (Eds.), Encyclopedia of behavioral neuroscience (Vol. 2, pp. 57–64). Oxford:
Academic Press.
Hamilton, D. A., Driscoll, I., & Sutherland, R. J. (2002). Human place learning in a virtual
Morris water task: some important constraints on the flexibility of place navigation.
Behavioural Brain Research, 129(1), 159–170.
Hartley, T., Maguire, E. A., Spiers, H. J., & Burgess, N. (2003). The well-worn route and the
path less traveled: distinct neural bases of route following and wayfinding in humans.
Neuron, 37(5), 877–888.
Hegarty, M., Richardson, A. E., Montello, D. R., Lovelace, K., & Subbiah, I. (2002). Develop-
ment of a self-report measure of environmental spatial ability. Intelligence, 30(5), 425–447.
Hund, A. M., & Minarik, J. L. (2006). Getting from here to there: spatial anxiety, wayfinding
strategies, direction type, and wayfinding efficiency. Spatial Cognition and Computation,
6(3), 179–201.
Iaria, G., Palermo, L., Committeri, G., & Barton, J. J. (2009). Age differences in the formation
and use of cognitive maps. Behavioural Brain Research, 196(2), 187–191.
Iaria, G., Petrides, M., Dagher, A., Pike, B., & Bohbot, V. D. (2003). Cognitive strategies
dependent on the hippocampus and caudate nucleus in human navigation: variability
and change with practice. Journal of Neuroscience, 23(13), 5945–5952.
Ishikawa,T., & Montello, D. R. (2006). Spatial knowledge acquisition from direct experience
in the environment: individual differences in the development of metric knowledge and
the integration of separately learned places. Cognitive Psychology, 52(2), 93–129.
Jack, C. R., Petersen, R. C., Xu, Y., O’Brien, P. C., Smith, G. E., Ivnik, R. J., et al. (1998).
Rate of medial temporal lobe atrophy in typical aging and Alzheimer’s disease. Neurology,
51(4), 993–999.
Kelly, J. W., McNamara, T. P., Bodenheimer, B., Carr, T. H., & Rieser, J. J. (2009). Individual
differences in using geometric and featural cues to maintain spatial orientation: cue
quantity and cue ambiguity are more important than cue type. Psychonomic Bulletin &
Review, 16(1), 176–181.
Kim, J. J., Lee, H. J., Han, J. S., & Packard, M. G. (2001). Amygdala is critical for stress-induced
modulation of hippocampal long-term potentiation and learning. Journal of Neuroscience,
21(14), 5222–5228.
Kim, J. J., & Yoon, K. S. (1998). Stress: metaplastic effects in the hippocampus. Trends in Neu-
rosciences, 21(12), 505–509.
Kirasic, K. C. (2000). Age differences in adults’ spatial abilities, learning environmental layout,
and wayfinding behavior. Spatial Cognition and Computation, 2(2), 117–134.
Kirasic, K. C., & Mathes, E. A. (1990). Effects of different means for conveying environmental
information on elderly adults’ spatial cognition and behavior. Environment and Behavior,
22(5), 591–607.
Korol, D. L. (2004). Role of estrogen in balancing contributions from multiple memory
systems. Neurobiology of Learning and Memory, 82(3), 309–323.
Lawton, C. A. (1994). Gender differences in way-finding strategies: relationship to spatial
ability and spatial anxiety. Sex Roles, 30(11), 765–779.
Lawton, C. A. (1996). Strategies for indoor wayfinding: the role of orientation. Journal of
Environmental Psychology, 16(2), 137–145.
Lawton, C. A., Charleston, S. I., & Zieles, A. S. (1996). Individual-and gender-related differ-
ences in indoor wayfinding. Environment and Behavior, 28(2), 204–219.
Maguire, E. A., Burgess, N., Donnett, J. G., Frackowiak, R. S., Frith, C. D., & O’Keefe,
J. (1998). Knowing where and getting there: a human navigation network. Science,
280(5365), 921–924.
Mahmood, O., Adamo, D., Briceno, E., & Moffat, S. D. (2009). Age differences in visual path
integration. Behavioural Brain Research, 205(1), 88–95.
Marchette, S. A., Bakker, A., & Shelton, A. L. (2011). Cognitive mappers to creatures of habit:
differential engagement of place and response learning mechanisms predicts human
navigational behavior. Journal of Neuroscience, 31(43), 15264–15268.
Marchette, S. A.,Yerramsetti, A., Burns, T. J., & Shelton, A. L. (2011). Spatial memory in the
real world: long-term representations of everyday environments. Memory & Cognition,
39(8), 1401–1408.
Meulenbroek, O., Petersson, K. M., Voermans, N., Weber, B., & Fernández, G. (2004). Age
differences in neural correlates of route encoding and route recognition. NeuroImage,
22(4), 1503–1514.
Moffat, S. D. (2009). Aging and spatial navigation: what do we know and where do we go?
Neuropsychology Review, 19(4), 478–489.
Moffat, S. D., Elkins, W., & Resnick, S. M. (2006). Age differences in the neural systems sup-
porting human allocentric spatial navigation. Neurobiology of Aging, 27(7), 965–972.
Moffat, S. D., Kennedy, K. M., Rodrigue, K. M., & Raz, N. (2007). Extrahippocampal
contributions to age differences in human spatial navigation. Cerebral Cortex, 17(6),
1274–1282.
Moffat, S. D., & Resnick, S. M. (2002). Effects of age on virtual environment place navigation
and allocentric cognitive mapping. Behavioral Neuroscience, 116(5), 851.
Morris, R. G. (1981). Spatial localization does not require the presence of local cues. Learning
and Motivation, 12(2), 239–260.
Morris, R. (1984). Developments of a water-maze procedure for studying spatial learning in
the rat. Journal of Neuroscience Methods, 11(1), 47–60.
Morris, R. G. M., Garrud, P., Rawlins, J. N.P., & O’Keefe, J. (1982). Place navigation impaired
in rats with hippocampal lesions. Nature, 297(5868), 681–683.
Naghdi, N., Majlessi, N., & Bozorgmehr, T. (2005). The effect of intrahippocampal injec-
tion of testosterone enanthate (an androgen receptor agonist) and anisomycin (protein
synthesis inhibitor) on spatial learning and memory in adult, male rats. Behavioural Brain
Research, 156(2), 263–268.
Newman, M. C., & Kaszniak, A. W. (2000). Spatial memory and aging: performance on a
human analog of the Morris water maze. Aging, Neuropsychology, and Cognition, 7(2),
86–93.
Olton, D. S., Collison, C., & Werz, M. A. (1977). Spatial memory and radial arm maze per-
formance of rats. Learning and Motivation, 8(3), 289–314.
Packard, M. G. (2009). Anxiety, cognition, and habit: a multiple memory systems perspective.
Brain Research, 1293, 121–128.
Packard, M. G., & McGaugh, J. L. (1996). Inactivation of hippocampus or caudate nucleus
with lidocaine differentially affects expression of place and response learning. Neurobiol-
ogy of Learning and Memory, 65(1), 65–72.
Packard, M. G., & Wingard, J. C. (2004). Amygdala and “emotional” modulation of the relative
use of multiple memory systems. Neurobiology of Learning and Memory, 82(3), 243–252.
Pazzaglia, F., & De Beni, R. (2001). Strategies of processing spatial information in survey and
landmark-centred individuals. European Journal of Cognitive Psychology, 13(4), 493–508.
Poldrack, R. A., & Packard, M. G. (2003). Competition among multiple memory systems:
converging evidence from animal and human brain studies. Neuropsychologia, 41(3),
245–251.
Rapp, P. R., Rosenberg, R. A., & Gallagher, M. (1987). An evaluation of spatial information
processing in aged rats. Behavioral Neuroscience, 101(1), 3.
Restle, F. (1957). Discrimination of cues in mazes: a resolution of the “place-vs.-response”
question. Psychological Review, 64(4), 217.
Ritchie, B. F., Aeschliman, B., & Pierce, P. (1950). Studies in spatial learning.VIII. Place per-
formance and the acquisition of place dispositions. Journal of Comparative and Physiological
Psychology, 43(2), 73.
Roof, R. L., & Havens, M. D. (1992). Testosterone improves maze performance and induces
development of a male hippocampus in females. Brain Research, 572(1–2), 310–313.
Rosenzweig, E. S., & Barnes, C. A. (2003). Impact of aging on hippocampal function: plastic-
ity, network dynamics, and cognition. Progress in Neurobiology, 69(3), 143–179.
Sadowski, R. N., Jackson, G. R., Wieczorek, L., & Gold, P. E. (2009). Effects of stress, cor-
ticosterone, and epinephrine administration on learning in place and response tasks.
Behavioural Brain Research, 205(1), 19–25.
Sandstrom, N. J., Kaufman, J., & Huettel, S. A. (1998). Males and females use different distal
cues in a virtual environment navigation task. Cognitive Brain Research, 6(4), 351–360.
Sandstrom, N. J., Kim, J. H., & Wasserman, M. A. (2006).Testosterone modulates performance
on a spatial working memory task in male rats. Hormones and Behavior, 50(1), 18–26.
Saucier, D. M., Green, S. M., Leason, J., MacFadden, A., Bell, S., & Elias, L. J. (2002). Are sex
differences in navigation caused by sexually dimorphic strategies or by differences in the
ability to use the strategies? Behavioral Neuroscience, 116(3), 403.
Schwabe, L., Schächinger, H., de Kloet, E. R., & Oitzl, M. S. (2010). Corticosteroids operate
as a switch between memory systems. Journal of Cognitive Neuroscience, 22(7), 1362–1372.
Shelton, A. L., & Gabrieli, J. D. (2002). Neural correlates of encoding space from route and
survey perspectives. Journal of Neuroscience, 22(7), 2711–2717.
Shelton, A. L., & Gabrieli, J. D. (2004). Neural correlates of individual differences in spatial
learning strategies. Neuropsychology, 18(3), 442.
Siegel, A. W., & White, S. H. (1975). The development of spatial representations of large-scale
environments. Advances in Child Development and Behavior, 10, 9–55.
Skelton, R. W., Ross, S. P., Nerad, L., & Livingstone, S. A. (2006). Human spatial navigation
deficits after traumatic brain injury shown in the arena maze, a virtual Morris water
maze. Brain Injury, 20(2), 189–203.
Spiers, H. J., & Maguire, E. A. (2007). A navigational guidance system in the human brain.
Hippocampus, 17(8), 618–626.
Stewart, C. A., & Morris, R. G. M. (1993). The watermaze. In A. Saghal (Ed.), Behavioral neu-
roscience. A practical approach (Vol. 1, pp. 107–122). Oxford: IRL Press.
Taylor, H. A., Naylor, S. J., & Chechile, N. A. (1999). Goal-specific influences on the repre-
sentation of spatial perspective. Memory & Cognition, 27(2), 309–319.
Tolman, E. C. (1948). Cognitive maps in rats and men. Psychological Review, 55(4), 189.
Tolman, E. C., Ritchie, B. F., & Kalish, D. (1946). Studies in spatial learning. II. Place learning
versus response learning. Journal of Experimental Psychology, 36(3), 221.
Tolman, E. C., Ritchie, B. F., & Kalish, D. (1947). Studies in spatial learning. V. Response
learning vs. place learning by the non-correction method. Journal of Experimental Psychol-
ogy, 37(4), 285.
Waller, D., & Lippa,Y. (2007). Landmarks as beacons and associative cues: their role in route
learning. Memory & Cognition, 35(5), 910–924.
Warren, S. G., & Juraska, J. M. (1997). Spatial and nonspatial learning across the rat estrous
cycle. Behavioral Neuroscience, 111(2), 259.
Wolbers, T., & Hegarty, M. (2010). What determines our navigational abilities? Trends in
Cognitive Sciences, 14(3), 138–146.
Yerramsetti, A., Marchette, S. A., & Shelton, A. L. (2012). Accessibility versus accuracy in
retrieving spatial memory: evidence for suboptimal assumed headings. Journal of
Experimental Psychology: Learning, Memory, and Cognition. http://dx.doi.org/10.1037/
a0030905. Advance online publication.

CHAPTER SEVEN
When Do the Effects

of Distractors Provide a Measure
of Distractibility?
Alejandro Lleras*,1, Simona Buetti*, J. Toby Mordkoff†
*Department of Psychology, University of Illinois at Urbana-Champaign, Champaign, IL, USA
†Department of Psychology, University of Iowa, Iowa City, IA, USA
Contents
1. Introduction262
2. When Do “Distractors” Cause Distraction? 264
2.1. Visual Search 264
2.1.1. Effect of Distractors on Search Performance 264
2.1.2. When Distractors Do Not Impact Search Performance 265
2.1.3. The End of Preattentive Vision 267
2.1.4. Nothing about Distraction Can be Learned from Visual Search Experiments 269
2.2. Divided Attention 271
2.3. Flanker Effect 272
2.3.1. Traditional View: The Flanker Effect Implies Late Selection 272
2.3.2. The Flanker Effect is Not a Measure of Distraction 273
2.3.3. The Information Processing Tradition and the Flanker Task 274
2.3.4. Why are Flankers Referred to as “Task-Irrelevant”? 276
2.4. A Different Form of Distractor: The Inattentional Blindness’ “Critical Stimulus” 278
2.4.1. The Unexpected Event Paradigms 278
2.4.2. Recruitment of “Central” Resources and the Ensuing Blindness 280
2.5. Empirical Study: Comparing the Salience vs the Relevance of a Distractor 281
2.6. Distraction or Distractor Interference? 285
2.6.1. The Current State of Confusion 285
2.6.2. Distractor Interference as a Measure of Attentional Success 288
2.6.3. Connecting Flanker Experiments to Real-World Situations 289
3. A Brief Case Study on Distraction 291
3.1. Limitations of Unexpected Event Paradigms 291
3.2. A New Paradigm 293
3.3. Discussion296
4. A Theory of Attention and Distractibility 300
4.1. The Need for Inner Focus 300
4.2. Predicting Inattentional Blindness 301
4.3. A Look Back at Visual Attention 303
5. Conclusions307
References310

262 Alejandro Lleras et al.
Abstract
We discuss how, at the present time, there is a large deal of confusion in the attention
literature regarding the use of the label “distractor” and what may be inferred from
experiments using distractors. In particular, investigators seem to use the concepts of
distractor interference and distractibility almost interchangeably. In contrast, we argue
at both the theoretical and empirical levels that these two concepts are not only differ-
ent, but in fact mutually exclusive. To that end, a brief review of several subliteratures is
presented, in which we identify some examples of the misuse of these terms. We also
propose a new paradigm for the study of distraction, as well as present a contemporary
general theory of visual attention that provides a better framework for understanding
distractor-interference effects, as well as instances of true distraction.
1. INTRODUCTION
What is “distraction”? To borrow some from James’ famous definition
of attention, one could say that distraction is the taking possession of the
mind in clear and vivid form by a thought or stimulus that one never intended
to process in the first place. It is a thought or stimulus that takes us “out” of
our intended task and drives our attention onto new thoughts or sensations.
This definition seems quite close to our intuitions of what being distracted
is like, and matches the vernacular definition of distraction (e.g. “the draw-
ing away [of the mind or thoughts] from one point or course to another;
diversion of the mind or attention”, Oxford English Dictionary). It is also a
definition that is implied by work in applied psychology, such as the recent
studies on distracted driving (e.g. Cooper, Vladisavljevic, Medeiros-Ward,
Martin, & Strayer, 2009; Strayer, Watson, & Drews, 2011). It certainly is an
important concept in our everyday life, so it is to no surprise that we, in
the attention field, may want to draw inferences about this concept from
our investigations. One of the primary goals of this chapter is to point out
where this effort has faltered and identify one reason why this has happened:
the inappropriate assumption that the effects of distractors are always a mea-
sure of distraction.
The history of the term “distractor” is as old as the discipline of psy-
chology itself. One can easily find papers using this term as far back as
the nineteenth century (e.g. Darlington & Talbot, 1898). In the early days,
distractors were understood to be that type of stimulus that can distract
us from our main tasks and, thereby, disturb our performance. Darlington
& Talbot (1898) were interested, for example, in testing whether music
could be considered a distractor and used a weight-lifting task to assess
the impact of music on performance. (Results suggested that, no, music
Distractors and Distractibility 263
was not a “distractor”.) Another example was Tinker (1925) who studied
whether the presence of bells ringing intermittently during an intelligence
test would affect performance of the test, a marker that distraction hinders
performance. (On average, it did not.) Yet, from the start, the early intuition
that distraction ought to hinder performance was there and a significant
debate arose as to what ought to be called a distractor (see Dulsky, 1932,
Psychological Review). In fact, early efforts to study attention using distractors
often failed “because it was impossible to obtain a satisfactory distractor”
(Dulsky, 1932, p. 590).That said, what did come from those early years was a
new way of talking about distractors: “distractor” is not a label to be used in
relation to attention (or as a means to study that abstract concept), but rather
in relation to performance. The rationale is simple: given that we don’t
know what attention is, but we do know how to measure performance, it
is best to simply define distractors as those stimuli that hinder performance
(Dulsky, 1932).
The goal of this chapter is to challenge this conception of “distrac-
tor,” which is closely related to the modern use of this term. Our hope is
to highlight the problems with this term, the confusion that it engenders
and, hopefully, provide a way out of this state of affairs by creating a more
modern context in which we can (a) redefine the study of distraction; (b)
clarify the terminology that is used to study attention and distraction; and
(c) identify current sources of confusion in the literature that are very much
related to the confusability between the labels “distractor” and “distrac-
tion”. We will begin by presenting a literature where the term distractor
is heavily used (viz, visual search) in spite of the fact that the goal of this
work is not to study distraction at all (but in fact attention). Along the
way, we will argue that not all “distractors” are created equal and, thus,
one must be careful when trying to extend the findings from laboratory
visual search to real-world search tasks. Next, we will quickly discuss (and
dismiss) an example of the use of the term “distractor” within the literature
on redundancy gains. Then we will describe a relatively more nuanced case
study regarding the perils of the use of the label “distractor”: the Flanker
Effect literature, a domain in which this term has gained a very promi-
nent role. From this section, we will conclude that it is in fact inappropriate
to make claims about the phenomenon of “distraction” from experiments
using modern-day “distractors,” in spite of our field’s tendency to want to
make such claims. One of our goals, too, is to differentiate the concept of
“distraction” from that of “distractor interference,” which are actually quite
different phenomena.
We will next discuss the inattentional blindness paradigm, which might

be a better way of studying and conceptualizing distraction than the previ-
ous paradigms. We will also describe data from an experiment using this
methodology to highlight why it is so critical to acknowledge the difference
between distractor interference and actual distraction. We will argue that
distractor-interference effects reflect, at a basic level, an important degree of
success by attentional selection (rather than a failure, as it frequently claimed).
Even more, we will argue that the presence of a “Flanker Effect” should be
taken as evidence in favor of early selection models of attention, rather than
as evidence for late selection models, as it is commonly done.
The next section of this chapter will focus on a proposed new task to
study distraction and distractibility using an eye-tracking methodology. We
will present data from our laboratory showing that heightened cognitive
focus reduces distractibility. And we will conclude by presenting, in rough
strokes, the skeleton of a model of attention that incorporates the properties
of attention highlighted in this chapter, as well as other factors like motiva-
tion and the need for inner focus when completing complex cognitive tasks.
2. WHEN DO “DISTRACTORS” CAUSE DISTRACTION?

2.1. Visual Search
2.1.1. Effect of Distractors on Search Performance
To answer the question of when do distractors cause distraction? we will begin
with the most popular paradigm that uses the first of these terms: namely,
visual search. Another reason for doing this is that attention is often thought
of as being the opposite or the antidote to distraction, and the concept of
attention has played a major role in our understanding of visual search from
the very beginning.
In the context of visual search, the term “distractor” is used to refer to
a majority of the elements within the display, and was first used by Shiffrin
and Gardner (1972). The collection of all distractors (plus the target item,
if there is one) is referred to as set size. Visual search is typically studied by
measuring response time (RT and/or proportion correct) as a function of set
size. In other words, distractors are a set of a priori possible targets that one
must inspect in order to find the target in the display. We draw inferences
about the workings of visual attention by inspecting the reaction time × set
size function (e.g. Duncan & Humphreys, 1989; Treisman & Gelade, 1980;
Wolfe, 1994). Notice that the distractors are stimuli created by the inves-
tigator to probe how attention works. Therefore, in most laboratory tasks
concerning search, there is a certain degree of (purposeful) overlap between

the distractors and target stimuli (e.g. all items share a common shape, or vary
within a small set of colors, orientations, etc.), because the goal is to induce
a measurable effect on performance, and performance, itself, is defined as a
function of set size (i.e. the number of distractors). So, if one defines
“distraction” as either “attention is being allocated to a nontarget stimulus”
or “an effect on performance that is due to distractors,” then, yes, visual
search would provide data that are relevant to the study of distraction. But,
as we will argue, what people really mean by “attention” in the context of
visual search is very different from the kind of attention that is needed to
avoid being distracted. And what people really mean by “distraction” is very
different from the kinds of effects that distractors have on response time in
visual search. But before discussing what can be learned about distraction
from cases of visual search where distractors do have an impact on per-
formance, we will address a related, but separate issue, regarding how the
absence of such effects have been interpreted, and how we believe that they
ought to be interpreted, in their relation to the concept of attention.
2.1.2. When Distractors Do Not Impact Search Performance

Most theories on attention that rely on visual search experiments indeed
focus most of their explanatory efforts on accounting for performance when
set size matters (e.g. Bundesen, 1990; Duncan & Humphreys, 1989; T reisman
& Gelade, 1980; Wolfe, 1994). In fact, whenever the number of distractors is
found to have no effect on RT (i.e. so-called flat search slopes), the experi-
mental results are brushed under the attention rug: clearly, if performance
was unaffected by set size, then attention must have not been involved in
the processing of the display and the detection of the target. In fact, when-
ever we find a case where performance is not affected by set size, we refer
to it as being the result of “preattentive” processing. Color pop-out is a
fine example of this: if asked to look for a red apple among green ones, the
number of green apples does not influence RT to find the red target: the
red apple “pops out” from the green ones and we introspectively believe that
our attention automatically moved to the red target apple. But did attention
really have nothing to do with us finding that red apple? Evidence from
at least two smaller literature within the attention field would argue that
this conclusion is misguided: the contingent-capture literature (e.g. Folk,
Leber, & Egeth, 2008; Folk, Remington, & Wright, 1994) and the prim-
ing of pop-out literature (e.g. Fecteau, 2007; Maljkovic & Nakayama, 1994,
1996). For now, simply consider the following data. In Wan and Lleras
(2010), subjects were asked to detect the presence of a color oddball item
(i.e. the one item that differed from all others in color). Two different sets of
color pairs were used. In one block of trials, participants were tested with red
and green items, while, in a different block, the items were pink and fuchsia
(two colors closely matched in luminance). Both color pairs produced visual
pop-out: in both cases, RTs were unaffected by set size (3, 8, or 12), with
search slopes of −0.09 ms/item for red among green and 1.19 ms/item
for pink among fuchsia, neither of which was different from zero. So far so
good: the standard interpretation would be that these pop-out discrimina-
tions were computed in preattentive fashion by the visual system.Yet, there
is a kink with this interpretation. Overall, responses were much slower with
the pink and fuchsia stimuli (mean RT = 698 ms) than with the red and
green stimuli (565 ms). How are we to interpret this 133 ms increase in
mean RT? At what point(s) in processing was this rather large delay added?
And what were observers doing during this time? W ere they waiting for pre-
attentive vision to finish processing, without attending to the display? This
seems implausible given other performance benchmarks of the attention
system. For instance, saccades to sudden onsets that match our attentional
template can be triggered extremely quickly, in the range of 130–150 ms
(Hollingworth, Matsukura, & Luck, in press; Kirchner & Thorpe, 2006). Not
only it is difficult to believe that participants would have an added processing
difficulty in “preattentive” vision of a magnitude of 133 ms (during which
attention would not be involved), but the overall magnitude of the RTs is
also equally difficult to interpret. Preattentive vision is supposed to be “fast
and automatic;” the sort of stuff that gets done “before we know it.” How
can this type of processing produce detection RTs on the order of 700 ms?
Are we to infer that preattentive vision can sometimes be that slow? No, we
would argue, even though this answer is at odds with the prevailing view
in the field on how to interpret results from visual search experiments (e.g.
Duncan & Humphreys, 1989). Attention is involved in finding a pop-out
target, and some pop-outs are, in fact, rather hard to find.
Going back to the central issue of what distractors are and how they
are used, it is clear that we (as vision scientists) have a very poor model
for what goes on with attention when distractors fail to have a measurable
effect on performance. This area of potential research has gone unrecog-
nized for the most part, and is still in its infancy (with some exceptions,
Purcell et al., 2010; Schall, Purcell, Heitz, Logan, & Palmeri, 2011; Townsend,
1972; Tseng, Glaser, Caddigan, & Lleras, submitted for publication; Wenger
and Townsend (2000), who earlier recognized the perceived need to use
different tools to analyze presumably parallel forms of processing). Yet, this

is not surprising given that one of our principal tools to index attention at
work has been the manipulation of set size. For instance, a very influential
model of visual search, Similarity Theory (Duncan & Humphreys, 1989),
is predicated on the idea that preattentive vision precedes attentive vision
in time. Preattentive vision is parallel and capable of producing null effects
of set size, but, almost by definition, is uninteresting for studying attention
per se. The main thrust of the theory is in interpreting performance when
distractors do affect RTs, because one can infer attentional processing times:
for instance, the more similar targets are to distractors, the more complex
the attentional involvement is in the search task (e.g. larger inspection times
per item). So, the logic goes, the more that distractors impact RTs, the
more attention is involved in determining performance. This assumption
regarding the normal processing steps in vision (encoding, in some parallel,
preattentive form, followed by identification, in a more serial, attentive
form) is common to most theories of attention, both formal (e.g. Treisman
& Gelade, 1980; Wolfe, 1994; Wolfe, Vo, Evans, & Greene, 2011) and com-
putational ones (e.g. Itti & Koch, 2001; Navalpakkam & Itti, 2007; Peters,
Iyer, Itti, & Koch, 2005; Townsend, 1972; Wenger & Townsend, 2000). But
the lack of a set-size effect does not imply a lack of attentional processing.
If a set-size effect implies attentional processing (A ergo B), it is an error
in logic to argue that the absence of a set-size effect implies an absence of
attentional processing (not A ergo not B).Those are not logically equivalent
propositions.What we can be sure of is that, if attention is not involved, then
one can never expect a set-size effect to emerge (not B ergo not A). That
would be correct, but it is not the argument being made in these cases.
2.1.3. The End of Preattentive Vision

How can we understand attention when our principal tool of inquiry pro-
duces no results? Understanding attentional processing in pop-out situations
will require newer methodologies (new computational models and new
brain imaging techniques). But it is important to realize that it is a logical
flaw to argue that because our tool (set-size manipulation) failed to influence
performance, that attention was not involved in producing that performance.
Thus, we cannot brush aside all forms of “parallel” searches under the preat-
tentive rug. In fact, there is overwhelming evidence that preattentive vision is
dead. And we are not the first ones to say so (e.g. Di Lollo, Kawahara, Zuvic,
& Visser, 2001, “The pre-attentive emperor has no clothes”, JEP:General).
The general idea that there is a temporal order to visual processing whereby
some processing always occurs prior to the allocation of attention is likely

wrong. Biased competition theory (Desimone & Duncan, 1995), for exam-
ple, suggests that attention can bias processing to favor one type of stimulus
(or property) over another at an extremely early point. And, thus far, the evi-
dence in favor of this proposal is quite strong. In fact, evidence suggests that
top–down attention can bias neural processing before the onset of a visual
stimulus, which can result in extremely fast eye movements (Hollingworth
et al., in press; Kirchner & Thorpe, 2006), as well as produce parallel feature-
based selection over the entire visual field (Bichot, Rossi, & Desimone, 2005;
Serences & Boynton, 2007). Even more, thanks to fMRI, we now know
that every possible layer of visual processing in the brain can be modulated
by attention, not only at the cortical level (reference V1 and V2, which are
now routinely shown to be modulated by attention), but even subcortical
structures like the LGNshow modulations by attention (e.g. Kastner et al.,
2004; O’Connor, Fukui, Pinsk, & Kastner, 2002), that entail as complex a
computation as increasing the precision encoding of attended objects while
completely filtering unattended objects (Fischer & Whitney, 2012).
Thus, given our understanding of attention, not only within the frame-
work of biased competition, it stands to reason that (other than the process-
ing in the retina) there is likely no strict temporal order to visual processing
such that some degree of visual processing always occurs prior to the alloca-
tion of attention. Even the initial levels of visual processing are likely modu-
lated by attention, under the right circumstances (e.g. Hillyard & Munte,
1984). That is not to say that there is not a lot of processing that goes on
outside of the focus of attention. Jeremy Wolfe’s most recent model of vision
now includes a “nonselective” pathway in vision: a parallel route of visual
processing that proceeds independently from attention and that is respon-
sible for things like computing the sensation that the visual world is more
than the object being attended, as well as extracting visual regularities from
the world (Chong & Treisman, 2003, 2005a, 2005b; Choo & Franconeri,
2010; Haberman & Whitney, 2011). Whether there is such a pathway is a
matter for debate, but what is clear is that there is a lot of unattended vision
that goes on at any given point in time, and proposing that this form of pro-
cessing is taking place simultaneously to attended vision is also likely correct.
The distinction between unattended and attended vision is, actually, criti-
cal for us to begin to form an understanding of “distraction”. We can use a
definition of distraction that follows our common intuition: distraction is
when our current train of thought gets interrupted by a secondary train of
thought, against our wish. One can study cognitive distractions: the occurrence
of distracting thoughts, which can happen in a benign manner (as in the task-
unrelated thoughts literature, Giambra, 1989; McVay & Kane, 2009, 2012;
Smallwood, Obonsawin, & Heim, 2003) or even pathological manner
(as in Obsessive compulsive disorder, Depression and Anxiety). However, we
are here interested in perceptual distraction: when the occurrence of an event
in the world comes to derail our train of thought or interfere with our task at
hand. More specifically, we will use the domain of vision to understand this
concept. In our terminology, distraction can be defined as times when a visual
event that was initially processed by unattended vision forces the attention
system to orient to it. This reorienting must occur at the cost of disengaging
attention from whatever other object (in vision or thought) that was previ-
ously being attended.We will return to this definition after presenting theory
and methodology that is currently used to assess “distractibility” (the study of
distraction) in our field. But first we must dispense with visual search.
2.1.4. Nothing about Distraction Can be Learned from

Visual Search Experiments
If we return to the question of what the use of distractors can tell us about
distractibility in the domain of visual search, the answer is clearly nothing.
This conclusion is, hopefully, not very controversial. Distractors are a group
of stimuli that are, by design, meant to be inspected by attention because
of their resemblance to the target stimulus. That is, it would be awkward
to argue that seeing a red hat in my closet is a distraction from my search
of a red scarf in said closet. If one has set a priority for detecting red
items in the closet, seeing and evaluating a red item is not an episode of
distraction, but is an integral step in achieving my goal. I want to inspect all
red items with the hope that one of them is my scarf. Thus, visual search,
in the traditional form, is not a methodology suited for studying or under-
standing distraction. In fact, it has been very difficult to translate what we
know about visual search from our vision laboratories to visual search in
the real world, precisely because in the real world, not all objects in a scene
are a priori meant to be possible search targets. If we are looking for a fork
in a kitchen, it is irrelevant how many appliances are in that kitchen. No
one placed the refrigerator, oven or dishwasher in the room with the hope
that people looking for forks would look at them. These items are not tar-
gets for the attention system, thus cannot be counted as distractors in the
laboratory sense; however, if asked to identify all objects in a kitchen, most
observers would likely count all appliances therein as being objects in the
kitchen scene. Labeling all objects in a real-world scene as a priori targets
of the search (i.e. as distractors) will always produce an overestimate of set

size and may lead to the awkward conclusion that real-world search is more
efficient than laboratory search. If one has a black-and-white styled kitchen,
with all black-and-white utensils, appliances, and counters, searching for a
red kitchen towel would produce a very efficient search (unaffected by the
number of elements in the scene), whereas searching for a specific black
utensil will likely be highly dependent on the number of objects in the
room. And what we have learned about guidance in the laboratory is abso-
lutely applicable to this real-world scene. It is merely complicated because
every type of search target in a real-world scene determines the set size in
that scene. In other words, the set size in a scene cannot be (and should
never be) determined in the absence of a definition about what the search
target in that scene will be. Once such a definition is provided (e.g. “look
for a couch”), one can determine the subset of elements in the scene that
will be a priori interesting to the attention system (e.g. all furniture items).
We propose that we should use the term “candidates” to refer to this specific
variety of distractors: the subset of items in a scene that might be a target.
Again, candidates can only be defined once the target of search is known.
Once a target of the type furniture has been defined, the number of win-
dows, paintings, and toys on the floor are immediately discounted by that
definition. Those objects could never be targets and will not be inspected
in the search, so they are not what we here called “candidates”.Very prom-
ising work in this arena has been recently conducted by Zelinsky et al.
(Alexander & Zelinsky, 2011; Neider & Zelinsky, 2011) as well as Wolfe
et al. (Wolfe, Alvarez, Rosenholtz, Kuzmova, & Sherman, 2011). Note that,
by design, most of laboratory visual search tasks have situations where the
set of candidates is the number of objects in the scene (what people have
referred to as set size), whereas in the real world, the number of candidates
is often substantially smaller than the number of objects in the visual scene.1
1 A notable exception is the literature on the phenomenon known as Visual Marking (Watson &
Humphreys, 1997). In this phenomenon, it is typically found that previewing half of the search items
(say all green stimuli) for some time before the second set of search items (say, the blue items) is
presented produces search efficiencies identical to those that are obtained when participants only see
and search through the second set of search items (i.e. only blue items). Interestingly, this improved
search efficiency in the “preview” condition is always accompanied by a substantial main effect, in the
order of 50–150 ms. This visual marking effect is critically dependent on attention being allocated to
the previewed items (Watson & Humphreys, 1997), and we would argue that it reflects the process of
candidate selection (selecting all blue items as search candidates, while ignoring green items) and is
central to this paradigm. In spite of its importance, most of the research on visual marking has focused
overwhelmingly in the analysis of slopes, ignoring all intercept effects.
Definition
Candidates: A candidate is a stimulus that contains at least some of the
defining attributes of a target. In both visual search and many selective-
attention tasks, a candidate is an object that closely resembles the target
in terms of its event-like descriptive characteristics. Does it belong to the
same visual category of items than the target? Is it visually similar to the
target? Does it have the same temporal characteristics as the target (i.e. it
appears/disappears at the same time as a target)? Does it appear nearby or
at a possible target location? As a result, which elements of a display are
candidates can only be determined once a task is defined and the target of
the attention system is determined. In visual search, the number of candi-
dates is what determines the functional set size for a scene. Candidates are
always highly relevant to the search task because of their status as “poten-
tial targets”. Candidates do not have stimulus–response associations in a
task, but they can nonetheless inform responses. For example, in a present/
absent visual search task, every inspected and discarded distractor is further
evidence that the response on the trial is more likely to be “absent”. See
Bundesen (1990) for a similar definition regarding relevant distractors: those
that are similar to the target along the defining characteristics of the target.
2.2. Divided Attention

A very similar point can be made for the task that is often used to study
divided attention—viz, the redundant-targets detection task (e.g. Miller,
1982; Mordkoff & Yantis, 1991). In these experiments, subjects are asked
to respond to prespecified targets (usually letters), just as they are when
doing visual search, but displays are always quite small (e.g. two letters at
most) and some displays contain multiple instances of the same target. The
key result from these tasks is the response-time advantage for trials with
two targets, instead of only one, which is known as a “redundancy gain”
(Raab, 1962).
What makes this task interesting here is how some researchers have
interpreted the finding that a single target accompanied by a distractor can
produce slower responses than a single target presented alone (e.g. Grice,
Canham, & Boroughs, 1984; Grice & Gwynne, 1987). The slowing in the
former case was said to be due to a “distraction decrement;” that is, the
nontarget in the second display location was said to distract the observer by
pulling resources away from the target. As it turned out, however, the slow-
ing was actually caused by the nontarget being correlated with the absence
of targets (Mordkoff & Yantis, 1991); when these correlations were elimi-
nated, single-target trials produced exactly the same RTs whether or not a
nontarget was also included within the display. In other words, there was no
actual decrement in performance due to the nontarget item; the slowing of
RT was caused by the information carried by this item, instead. Even more,
we would here argue that this nontarget should not really be thought of as
being a distractor, because it was a candidate, due to being in one of the very
few locations in the entire display at which targets ever appear.
The second way in which this work might be relevant to the study of
distraction is in how a distinction was made between nontargets that some-
times appeared with targets, which were called “distractors,” and nontargets
that never appeared with targets, which were called “noise”. The main rea-
son for keeping these two types of nontarget stimuli separate was to be able
to calculate separate correlations between their presence and the correct
response (Mordkoff & Yantis, 1991); these calculations were crucial to the
demonstration that the so-called “distraction decrement” does not actually
exist. But the use of two different labels for items that, on the surface, play
identical (non-)roles in the task also shows that the idea that some nontar-
gets are more equal than others is not completely new. When one adds in
that these nontargets can carry different amounts of information—in the
technical sense of the word—a few doubts should be raised about the use of
the label “task-irrelevant” to refer to nontargets. But this is probably better
explained in the context of the task to be discussed next.
2.3. Flanker Effect

2.3.1. Traditional View: The Flanker Effect Implies Late Selection
One can also study attention in the absence of location uncertainty, that is,
when one knows where the target is going to be (e.g. Eriksen & Eriksen,
1974), yet there is uncertainty about the identity of the upcoming target.
This experimental procedure has come to be known as the Flanker Para-
digm and is typically a variation of the following set up: target position is
defined a priori and participants are asked to report as quickly as possible
the identity of the target, almost always a letter. Just as in visual search, the
target is presented among “noise” elements that may be related to the target
in some way. For example, in the original experiment, the noise elements
could resemble the target in terms of their low-level features (target and
noise letters looked alike) or in terms of response associations (noise letters
were letters picked from the possible set of target letters). As in the visual
search literature (Estes & Taylor, 1966), over the years, in this paradigm,
the term “noise” element was replaced by “distractor” (or sometimes just
“flanker” because the distractors tend to be presented on both sides of the
target letter, flanking it). In other words, in this context, one uses the word
“distractor” to refer to target-like stimuli (most often, exact copies of the
possible targets themselves) that are presented at locations that are near to
the target. The goal of this paradigm is to measure the degree to which
attention can focus solely on the target item itself (a test for what is referred
to as “early selection”, Neisser & Becklen, 1975; Sperling, 1960;Treisman &
Geffen, 1967). If performance is unaffected by the presence or identity of
the distractors, then one can conclude that selective attention successfully
“filtered out” the “irrelevant” information from the display. In most cases,
however, it is found that response times are strongly modulated by the iden-
tity of these flanking distractors (e.g. evidence for so-called “late selection”,
Deutsch & Deutsch, 1963; Norman, 1968), an effect known as the Flanker
Effect.This result is interpreted as evidence that the distractors have been pro-
cessed to a sufficiently deep level of processing (at least to the level of stimulus
identification) such that they, too, can activate response selection processes. In
fact, the Flanker Effect is most often measured as the difference in RTs between
two conditions: one in which the distractors activate the same response as the
target for the trial (compatible or congruent condition) and one in which the
distractors activate a different (incorrect) response from the target (incompatible
or incongruent condition), with responses on incongruent trials typically being
anywhere between 30 and 100 ms slower than on congruent trials. When a
Flanker Effect is obtained, it is generally concluded that selective attention has
failed in some way, allowing task-irrelevant items to be processed.2
2.3.2. The Flanker Effect is Not a Measure of Distraction

The Flanker Effect is an example of distractor interference, and it is not the
only one. The Simon effect (Simon & Rudell, 1967) and the Stroop effect
( Jaensch, 1929; Stroop, 1935) are also well-known examples of distractor
interference. What these effects have in common is that one can consis-
tently find evidence that some aspect of the distractor stimulus (or the
distractor attribute) influences behavior. But is distractor interference a
measure of distractibility? Hardly so. If we follow the logic presented above
2 It should be noted that, with regard to the presence or absence of a Flanker Effect, the same logical
error is being made as is for visual search. While the presence of a (significant) Flanker Effect is
evidence that the flanking distractors were, indeed, processed, the absence of a Flanker Effect does not
necessarily imply that the flankers were not processed. This point was made rather forcefully by Driver
& Tipper (1989).
in the context of visual search, the Flanker Effect is no different. Distrac-

tors in this task are selected by the experimenter with the hope that, if
they are processed, they can have a measureable effect on performance.
To achieve a measureable effect on performance, the critical distractors in
the flankers task are, in the vast majority of studies, exact replicas of one of
the possible target stimuli. The rationale is simple: given that participants
have a response associated with every target, we can hijack those stimulus–
response associations to measure distractor processing. So, all we need to
do is use targets (or target-like stimuli) as the distractors. Participants will
know that they are not the target on any given trial because targets and
distractors are always presented at different locations. Yet, if participants
process those distractors, they will activate those stimulus–response associa-
tions and compete (with the information coming from the actual target)
to determine the response on the trial. In sum, congruent and incongruent
distractors in a flankers task are by design stimuli that fit the participants’
task set. For example, if the participants’ task is to identify whether a central
letter is an X or an O, the critical distractors in most designs of a flankers
task would be either Xs or Os. Thus, if one assumes that participants have
a task set that stresses the importance of Xs and Os, it would be difficult to
come up with more task-relevant stimuli than congruent and incongruent
distractors (i.e. Xs and Os). Therefore, given that the critical distractors in
this task are actually very much task-relevant, they cannot tell us anything
about distractibility.
2.3.3. The Information Processing Tradition and the Flanker Task

Overall, this simple analysis of the flankers task reveals that most of the
distractors used in these studies are by design task-relevant and they should
be understood as such. However, there has been a tradition in this litera-
ture, going back more than 25 years, to describe the Flanker distractors as
being task-irrelevant, instead (e.g. Lavie, Hirst, de Fockert, & Viding, 2004;
Miller, 1987; see also Figure 7.4). To understand why, one must under-
stand the theoretical tradition within which the initial studies on this topic
were conducted.The dominant methodology in the mid-twentieth century
experimental psychology was the detection of target signals embed-
ded in noise (often interpreted in the terms of signal-detection theory).
In fact, early studies on visual search refer to the distractor elements
in those displays as “noise” elements (e.g. Estes & Taylor, 1966), just as
Eriksen and Eriksen described them in their first report of the Flanker
Effect. Furthermore, information theory was the theoretical backdrop to
cognitive psychology. One viewed and studied the mind as an informa-

tion processing problem, with an input signal that underwent several dis-
crete transformations (or processing stages) inside the black box that was
the mind before producing an overt response. Within this worldview, the
“noise” stimuli used in tasks like the Flankers paradigm were irrelevant
to the observer because they carried no information regarding which
response should be made. Thus the term was born of distractors being
task-irrelevant: the word “task” being a short-hand for “which response
should be made” and the word “irrelevant” indicating that the distractor
carried no information with regard to the task. In sum, the identity (or
even presence) of distractors in a visual search or flanker task changed
nothing to the fact that only the target signal determined the appropri-
ate response on each trial. If, for example, the target on a given trial was
the letter X, the response on that trial was to press the button associated
with that specific letter, irrespective of any and all other noise letters in
the display. The noise elements were irrelevant to the actions of the par-
ticipant because they carried no information about the required response on any
given trial. In that sense, perhaps a better label to qualify the distractors in
a visual search or flankers task would be to say that distractors are response
uninformative.3
A great deal of confusion would be avoided if a more precise term like
this one was used, instead of the more misleading label task-irrelevant that
has now come to be the dominant term for describing flankers. Another
alternative would be to simply drop this qualifying term altogether from
this literature. Distractors would simply be referred to as distractors, with
no mention of their task relevancy, given that it is clear that they are
task-relevant, in the way that we now use these terms. Here, however,
we propose the use of the term foil to refer to congruent and incongru-
ent distractors. It implies, as it should, that they were put in place in the
experiment with the goal of somehow tricking or interfering with the
participants’ responses: in a Flanker task, foils are a close match to what
participants are looking for. An exact copy, but just off in some minor way
(like their location).
3 Note that not all stimuli need be response informative to be considered task-relevant. Candidates
are one such example: in visual search tasks where participants must report the identity of the target,
candidates themselves are uninformative to the trial’s response. A spatial cue in a Posner-type cueing
paradigm is also an example of a task-relevant but response-uninformative stimulus: it tells observers
where they might find items that contain response-related information. But the cue itself is not
informative to the response.
Definition
Foil: A foil is a type of candidate that can cause a participant to produce
the wrong response in a trial because foils have links to potentiated responses
in a task. They often contain the same response-defining attributes as the
set of targets. In a selective attention task, a foil is often an object that is
identical to one of the possible targets but is not the target itself because
of a priori reasons: it differs from the target along one (or several) defining
characteristics. For instance, the target might be determined by its unique
location, or the foil itself might have a unique prespecified location. A foil
may also appear at a different point in time in a trial than the target (as
in priming experiments). Foils are automatically selected by the attention
system, to the extent that they are perceptually well represented. Selecting
and rejecting foils is a critical part of what attention must do in order to
properly respond on a given trial.
2.3.4. Why are Flankers Referred to as “Task-Irrelevant”?

But, why did the authors in this literature feel that they had to specify that,
in their design, distractors were indeed task-irrelevant? That is because
not all distractors in a flanker task need to be response-uninformative,
though most are. Several papers have documented the “correlated Flanker
Effect” (e.g. Miller, 1987; Mordkoff & Halterman, 2008): that is, one
can, by design, pick a combination of target–distractor co-occurrence
frequencies such that the identity of distractors can indeed tell the

participant something about the response. For instance, say that the task
is to identify a target letter (an A, B, Y, or Z, with A and B mapped to
one response and Y and Z mapped to the opposite response), and that on
75% of the trials when the flankers are Ms, an A or B is the target, and
on 75% of the trials when the flankers are Ns, a Y or Z is the target. The
distractors in this task have no direct association with either response
and they do not resemble the targets. Yet, their identity carries informa-
tion about the likely response. As one can imagine, responding A/B on
an A trial is faster and more accurate when the distractors are Ms than
when they are Ns. This is the correlated Flanker Effect and it can also be
observed in more subtle ways whenever there are contingencies in the
design of the flanker task such that the identity of the distractors pro-
vides useful information of any sort (Mordkoff, 1996). In such cases, the
distractors are response-informative and, as such, can be characterized as
“task-relevant” by the older definition of task relevancy. So, it is easy to
see why there was a need in the literature to differentiate between these
two forms of Flanker Effects: the one that arises when distractors carry
some information about the response (by design or by accident), which is
the correlated-flankers effect, and the one that arises when the distractors
carry no information about the response, but have an effect because they
are the same stimulus as one of the actual targets, which is the traditional
Flanker Effect.4
In a somewhat unfortunate turn of events, contemporary cognitive
psychology has continued to use the original “task-irrelevant” modifier to
describe distractors in a flanker task. It is unfortunate because most of us
have moved beyond that use of the term “task relevancy” and use a more
contemporary definition of relevance that implies a very different under-
standing of the role of distractors in the flankers task and their relationship
to the study of distraction. The newer use of the term “task-relevant” arises
from research on phenomena like attentional capture, task switching, and
executive control. In attentional capture, for instance, people study what
aspects of a stimulus make the stimulus more or less “attractive” to the atten-
tion system. Folk et al. (1994) demonstrated the phenomenon of contingent
attentional capture: that a visually salient element captures attention to the
extent that some of its defining characteristics match those characteristics
that define the target stimulus. A moving stimulus draws attention to itself
to the extent that the attention system is looking for moving-like things
(e.g. targets that are moving or have sudden onsets, or some other stimulus
property that also characterizes motion). These attentional control settings are
determined by the task demands, and, some would argue also by the behav-
ioral urgency of certain stimuli (e.g. Franconeri, Alvarez, & Enns, 2007; Lin,
Franconeri, & Enns, 2008; von Mühlenen & Lleras, 2007).
In sum, we now use the term “task-relevant” to describe the level of
overlap between distractors and targets, either in terms of their visual char-
acteristics or their functional description with respect to the task character-
istics (Are they both sudden onsets? Do they appear in tandem? Does one
precede the other? At the same or different locations? etc. See also Yantis &
Egeth, 1999).
4 It is worth noting that, by the information theory-based definition of task relevance, all the nontargets
in nearly every published visual search study have actually been highly relevant, since each item that
is identified as not being a target increases the likelihood that any given not-yet-identified item is a
target. Likewise, the total number of items in the display is equally relevant, because, as more items are
included, the odds of any particular item being a target are reduced. In other words, reverting to the
old definition of task relevance is not going to change our earlier argument that the nontargets in a
visual search display are not a source of distraction, even when they cause responses to be slower.
To sum up, in our terminology (and in agreement with the attention

capture literature), candidates are always task-relevant. So are foils. In a later
section, we will expand on the theoretical and applied consequences of
using the term “task-irrelevant distractors” in a flankers task, but for now,
let us finish on the following point. Congruent and incongruent distrac-
tors in a flanker task are not task-irrelevant by the current definition of task
relevancy because they are exact replicas of the target and resemble the
target in every possible functional way, save for the fact that they appear at a
different location than the target. Thus, as with visual search, nothing about
distraction and distractibility can be learned from studies on the Flanker
Effect. Of course, by the same logic, one can also argue that little about dis-
tractibility can be learned from studies on attentional capture, because, in that
paradigm, distractor stimuli most often fall clearly within the “task-relevant”
category of stimuli. Therefore, we are going to have to turn to a very different
kind of task if we hope to learn anything about actual distraction.
2.4. A Different Form of Distractor: The Inattentional

Blindness’ “Critical Stimulus”
2.4.1. The Unexpected Event Paradigms
Most often, the phenomenon known as “inattentional blindness (IB)” is
studied with a paradigm that is a variant of the original experiments by
Mack and Rock (1998): participants are asked to do a fairly engaging visual
discrimination (e.g. deciding which of the two arms of a cross is longer,
when the two arms are actually almost identical in length) in a brief period
of time. After the participants have completed a few trials of this task, exper-
imenters present an unexpected “critical stimulus” within the display: an
unexpected visual object that is completely unrelated to the task at hand. It
can be, for instance, a small square near fixation. After participants complete
the response on this critical trial (also known as the “inattention trial”),
experimenters ask the subjects whether they saw anything unexpected on
that last trial. Henceforth, we will refer to this overall experimental frame-
work as the unexpected event paradigm. The term “inattentional blindness”
refers to the finding that quite often, people fail to report the unexpected
stimulus.5 Importantly, after the inattention trial, participants complete a few
additional trials, including a final trial in which they are told to completely
5 The label “inattentional blindness” was applied prematurely and is rather unfortunate, in that it
strongly suggests that the unexpected stimulus is not seen. Subsequent work (e.g. Moore & Egeth,
1997) has indicated that the unexpected stimulus is, indeed, perceived, but never reaches awareness
and/or is quickly forgotten. But we are stuck with the name for the paradigm.
ignore the stimulus for their primary task (i.e. the cross in our example) and
just focus on trying to detect any other objects in the display. In this last
trial (referred to as the full attention trial), participants are typically at ceil-
ing at detecting the critical stimulus.This is important because it guarantees
that the critical stimulus is actually perfectly visible. Furthermore, it allows
the experimenters to conclude that the reason why participants failed to be
aware of the critical stimulus was that they were completely immersed in
processing a different object (the cross) at the time it appeared.
This is not the only paradigm in which inattentional blindness has been
studied. Neisser had earlier pioneered the use of video to study the same
phenomenon (Neisser, 1979; Neisser & Becklen, 1975) asking participants
to attend to one of the two superimposed movies. Participants were, as
one would suspect, quite unaware of unexpected events in the unattended
film. The video paradigm was also popularized recently by Simons et al.
who demonstrated that one does not need two superimposed videos to
create inattentional blindness: grabbing a cue from Mack and Rock (1998)
and Simons and Chabris (1999) showed that simply asking participants to
be intensely engaged in some attentive task (like counting the number of
bounces and passes of a basketball) would trigger inattentional blindness to
unexpected events occurring in that same video. The added benefit of this
technique is that unexpected events can be prolonged in time (the time it
takes a gorilla to walk across a scene, and pump its chest a few times), as long
as they start and finish while participants are engaged in the attentive task.
Follow-up experiments also showed that the likelihood participants will
become aware of the critical stimulus increases the more the critical stimu-
lus resembles the events in the attentive task. For instance, people counting
bounces and passes in a team of people wearing black t-shirts will be more
likely to detect the appearance of a (black) gorilla, than those counting
passes in a team wearing white t-shirts (for a similar demonstration, see
Most et al., 2001).
In this sense, unexpected event paradigms do hold promise for the study
of distractibility. Why? Because they allow one to ask the question: will an
observer’s train of thought be interrupted by a secondary event, against the
observer’s own goals to concentrate on a primary task. In fact, the unex-
pected event paradigm has inspired a number of applied investigations
into issues relating to distractibility. In the driving domain, Strayer et al.
(e.g. Strayer & Drews, 2007) have proposed that talking on a cell phone
produces a form of inattentional blindness (to the extent that people overly
concentrate on the cell phone conversation) such that participants will be
less likely to remember or identify objects that they actually fixated during
the driving task, if they were talking on the cell phone when those objects
appeared. In other words, the work of Strayer et al. suggests that the addi-
tional cognitive load involved in having a cell phone conversation creates
a situation where participants are less likely to attentively experience the
objects that their eyes fixate during driving. This is perfectly in line with
Mack and Rock’s initial account of inattentional blindness but extends that
account to attentionally engaging events outside of the realm of vision.That
is to say, whereas one could have interpreted the initial findings of inatten-
tional blindness as reflecting a limitation of visual attention, the findings in
the applied literature suggest that they are more general than that. We do
not fail to see the critical stimulus in the unexpected event paradigm
(or the gorilla in Simons & Chabris’ video) because our visual attention
cannot be focused on two visual aspects of the scene at once. Rather, we
fail to see the critical stimulus because our attention system (irrespective
of modality) has a difficult time doing two things at once: processing the
information required in the demanding primary task, as well as encoding
and responding to the unexpected stimulus.
2.4.2. Recruitment of “Central” Resources and the Ensuing Blindness

Elegant evidence in support of the idea that a “central” limited resource is
responsible for inattentional blindness was found by Fougnie and Marois
(2007). They asked participants to do a task where they had to either
rehearse in memory a set of letters (the maintain condition) or reorganize
them in alphabetical order (the manipulate condition). Shortly after the
letters were presented to the participants, an unexpected critical stimulus
was presented in the display. Only 35% of participants failed to detect the
unexpected stimulus in the maintain condition, whereas 68% of participants
in the manipulate condition missed it. There was no difference between
groups in the full attention trial and participants were near ceiling at detect-
ing the critical stimulus on that trial. This evidence suggests that the degree
of involvement of central executive resources in a primary task determines
the degree to which participants will become aware of unexpected and a
priori unattended events in the world. In a clever follow-up experiment,
Fougnie and Marois presented the critical stimulus toward the end of the
retention interval, at a time where participants in the manipulate condi-
tion would have already finished reorganizing the letters in memory. In this
experiment, there were no differences in the degree of inattentional blind-
ness across groups, presumably because the central executive was no longer
actively engaged by the reorganization task (i.e. by the time the Critical
Stimulus appeared, participants were just rehearsing the ordered letters in
memory).
Overall, this brief (and admittedly selective) glance at some of the inat-
tentional blindness literature suggests that distractibility may depend on the
degree to which we are currently engaged in a centrally demanding task: the
more engaged we are in that task, the less distractible we are by unexpected
events (see also Macdonald & Lavie, 2008, 2011). To touch back briefly
on an issue raised in the section on visual search, all evidence suggests that
visual processing proceeds normally even in the absence of attention.Visual
processes like perceptual grouping and surface completion (and others) are
all performed even in the absence of attention, as operationalized by the
occurrence of inattentional blindness (e.g. Moore & Egeth, 1997; Moore,
Grosjean, & Lleras, 2003).This is important because it underscores the need
for revisiting some visual attention theories. We can be fairly confident that
(1) vision unfolds with or without attention, which is the reason we have
a rich phenomenal experience of the visual world even though we cannot
attend to it in its entirety at any given moment; (2) attention can access any
level of visual processing to bias processing in a particular direction; and
(3) therefore, there seems to be no need to assume or propose some form
of uniquely preattentive processing that must precede attention in time.
There are no proto-objects (Rensink, 2000); there is no massive parallel
preattentive vision (Treisman & Gelade, 1980). We believe that any model
of attention that assumes so is likely misguided.
Returning to the issue of distraction, the question is: why do we expe-
rience inattentional blindness in the first place? What factors determine
whether we do? Is inattentional blindness a failure of some sort or might it
reflect an efficient, calculated tradeoff, instead? These are complex question
and we will attempt to offer a possible answer to them in Section 3 of this
chapter. But first, let us reexamine the issue of task relevance in the stimulus,
using a simple inattentional blindness paradigm.
2.5. Empirical Study: Comparing the Salience vs the

Relevance of a Distractor
There has already been evidence in the inattentional blindness literature that
the degree of visual similarity between the critical stimulus and the stimulus
in the attentionally engaging task determines the degree of inattentional
blindness to the critical stimulus (Most et al., 2001; Most, Scholl, Clifford,
& Simons, 2005). Here, we just want to reiterate this point in a less subtle
variation. Rather than comparing stimuli that vary along their color, we
will simply compare two different critical stimuli: one will be clearly task-
relevant (it will be a foil), the second one will not. However, the irrelevant
stimuli will be significantly more luminant than the first. Here then, we are
pitting the salience of a stimulus (in terms of its overall luminance) against
the relevance of a stimulus (in terms of the task set).The question is: can we
create inattentional blindness when the critical stimulus is a foil (i.e. task-
relevant)? If not, then this is evidence that task-relevant stimuli (like the
distractors used in the flankers paradigm) are not good stimuli to use when
one wants to draw conclusions about distractibility. That is because such
evidence would show that task-relevant stimuli are never unattended in the
first place. To answer this question, we will use two possible critical stimuli
(CS) a bright-white square and a white letter. The main task of the partici-
pants will be (as in a flankers task) to identify the letter in the center of the
display.The task-relevant critical stimulus will be simply a traditional flanker
(i.e. a foil, one of the possible target letters, presented slightly above, below,
to the right or left of the target).We will compare performance in this con-
dition to performance in a condition where, rather than a letter, we present
a white square that is 1.5 times larger than the letter. Because the square is
completely filled-in with white, and the letter is only drawn with a small
portion of the pixels used to fill up the square, we can be certain that the
square is substantially more luminant than the letter. Thus, if we are correct
that stimuli which are highly relevant to the task are almost always selected
by our attention system (as something that attention should select from the
display), one would expect very small levels of inattentional blindness to
the foil version of the CS. In contrast, given that the square is completely
irrelevant to the letter-identification task, one would expect a much larger
degree of inattentional blindness to the square, in spite of its much larger
visual presence in the display.
Eighty four subjects were asked to complete a brief five-trial study, using
a traditional unexpected event paradigm. Participants were told that their
task was to identify a briefly presented letter and report its identity. The
letter could be an A or an E, in lower or upper case, although the letter
case was irrelevant to the response. To make matters challenging, the let-
ter (about 1° of visual angle in extent) was preceded by a ∼100 ms long
random-dot mask (about 2° of visual angle in extent), which was imme-
diately followed by the letter (presented only for ∼70 ms), which was in
turn immediately followed by a new random-dot pattern mask that stayed
on the screen until participants had recorded a response. In the first three
trials of the experiment, those were the only visual events in the display, and
participants completed the task with a good degree of accuracy. Average
accuracy in the letter identification task in the first trial was 74%, and rose to
90% and 92% on the second and third trials, respectively. By the third trial,
participants had a mean response time of 1470 ms (st. dev. 316 ms), which
is not unusually long given that response time was not stressed in the task.
Then came the unexpected event trial: the critical stimulus was pre-
sented on the screen at the same time as the first random-dot mask and
remained on the display until the target letter was replaced by the sec-
ond random-dot mask. In all, the critical stimulus stayed on the screen for
160 ms, more than twice the duration of the target.The critical stimulus was
presented 2° of visual angle away from the mask and could appear above,
below, to the right or left of it.
Let us begin by establishing whether or not our main task and stimuli
can, indeed, produce inattentional blindness by looking at performance
in the group of subjects that was assigned to the square critical stimulus
(which was approximately 0.75° of visual angle), a stimulus that was very
much like the critical stimulus in many of Mack & Rock’s original stud-
ies. How many participants reported seeing the unexpected square? Only
18 out of 42 participants (43%). When forced to guess the location of
the unexpected stimulus, 17 out of 42 correctly reported its location. In
other words, we produced a substantial degree of inattentional blindness
to the square. Did the onset of the unexpected square interfere with the
participants’ effort in the primary task? Trial 4 accuracy was 97%, so, at
first sight, one would argue that it did not. A post hoc RT analysis (limited
by the variability of a one-observation/participant measure) suggested
that it did not: subjects who did not report seeing the square responded
to the letter target with an average RT of 1450 ms (st. dev. 600 ms), well
in line with their performance on Trial 3. In other words, even though
an unseen stimuli could theoretically have affected their RT (e.g. Moore
et al., 2003), it failed to do so. In contrast, there was some suggestion in
the data that those who did see the square actually slowed down in their
response to the letter: the RT for the 18 participants that did see the
square was 1661 ms (st. dev. 510 ms).The difference failed to reach signifi-
cance, mostly because of the inherent variability in the measure. Finally, it
should be noted that in the full attention trial, 40 out of 42 observers saw
the square, and 38 out of 42 correctly reported its location. So, about 90%
of subjects had no trouble seeing the square when not attentively engaged
in the central letter task.
Overall, we can conclude that with our task parameters, and using a task-
irrelevant stimulus, we can create substantial levels of inattentional blindness
(almost 50%, the difference in detection between full attention and inatten-
tion trials). What happens when we replace the white square with a much
smaller foil (i.e. a task-relevant letter)? In this task-relevant group, the critical
stimulus was either the capital letter A or E. It should be noted that the two
experimental groups were ran simultaneously and that our research assis-
tants were blind to the experimental condition. This is important because
it guarantees that all participants received the same instructions and that
research assistants could not knowingly or unknowingly affect the outcome
of the experiment. So, was this smaller “flanking” letter treated differently
by the attention system of our subjects than the square? Absolutely: 34 out
of 42 (81%) participants reported seeing the distractor letter, almost twice
as many subjects who saw the square. Did the onset of this letter affect their
performance in the central letter identification task before they even knew we
would ask questions about this unexpected letter? Yes, again. Remember, when
the critical stimulus was a square, participants in Trial 4 identified the target
letter almost perfectly (97% correct). When the critical stimulus was a foil
(one of the possible targets), target identification performance dropped to a
dismal 69% (chance being 50%). RTs also substantially increased to 2022 ms
and were highly variable (st. dev. 1408 ms). Importantly, it was only after
participants had completed this response that we asked whether they had
seen an unexpected event in the display. By the time we asked, all perfor-
mance indicators suggested that yes, indeed, the foil had by and large not
gone unnoticed. How visible was the foil? In the full attention trial, 40 out
of 42 participants reported seeing the foil and 38 out of 42 (90%) correctly
reported the foil’s location. So, the a priori visibility of the square and letter
foil was well matched across conditions, both were at 90%. Given that 81%
of participants reported seeing the letter in the inattention trial, our results
suggest that for the most part, there was little-to-no inattentional blindness
in this condition. In sum, the attention system spontaneously detected and
selected the foil (letter distractor), even though participants knew nothing
about the possibility that a letter might appear in the periphery, and the
act of selecting this letter substantially impacted their performance in the
primary task.
Knowing that participants were not blind to the foil, one can do a
second pass at the results and ask the question: was there a Flanker-like
congruency effect in our experiment? Was the main letter identification
task modulated by the congruence between target and foil? Yes again.
The average RT in Trial 4 for participants who had a congruent target–foil

relation was 1743 ms (st. error = 175), whereas RT for participants who had
an incongruent relation was 600 ms slower: 2359 ms (st. error = 425).When
accuracy is analyzed, we found that 100% of participants in the congruent
condition reported the correct letter. In contrast, only 31% of responses
were correct in the incongruent condition. Finding an accuracy rate below
chance suggests that participants were actually reporting the identity of the
foil, not that of the target letter. If one recodes performance as a function
of foil identity, participants’ accuracy on Trial 4 becomes 90.5%, perfectly
in line with their performance on Trial 3. This is interesting. The evidence
from this experiment suggests not that the foil interfered with the tar-
get, but that, in fact, the target may have interfered with the foil (in terms
of producing elevated response times). The results further suggest that the
Flanker Effect may very well be a failure to select the appropriate of several
potential targets (Lachter, Forster, & Ruthruff, 2004; Yigit-Elliott, Palmer, &
Moore, 2011).
From this simple experiment we can conclude that, as we had hinted,
using one of the possible target letters as a distractor in a selective attention
task makes that distractor task-relevant, by our contemporary standards of
what “task relevancy” means, but more importantly, by empirical standards
as well. In fact, in Section 3 we will propose that when we ask the attention
system to identify the presence of As or Es on a display, it will likely do so
over the entire display, provided that there is sufficient signal to represent
letters in the periphery (i.e. that there is no crowding and that the letters
are sufficiently large to be accurately represented). So, the so-called “task-
irrelevant” distractors in flankers task are actually always selected by our
attention, precisely because they are actually very much task-relevant. Given
that the presence and identity of such distractors alters our performance on
the primary task (identifying the central target), does that mean that these
distractors are distracting us in a Flanker Task?
2.6. Distraction or Distractor Interference?

2.6.1. The Current State of Confusion
In 2010, Lavie wrote a brief review on “Attention, Distraction and Cognitive
Control under Load” (Lavie, 2010). The title is important because it identifies
the concept of distraction as one of the central topics of the article.The conclu-
sions of the article are as follows. If one wants to reduce the likelihood of dis-
traction (at work, for example), it is advisable to work in a perceptually loaded
environment. This means we should increase clutter in our desks. Further, we
might be tempted to infer that adding clutter (or otherwise increasing percep-
tual load) would be advised for situations where attentional focus is required,
such as driving, but – as even Lavie (2010) admits – this advice would best be
not taken. At gut level, this proposal sounds just wrong and ill-advised. Another
important finding to come out of the review is that increased cognitive load creates
more distractibility.We have not yet defined cognitive load here, but it can intui-
tively be thought as reflecting the cognitive difficulty of a task or the additional
cognitive difficulty brought on by a secondary task. So, one should try to work
under conditions of low cognitive load to avoid distractibility.While at first this
sounds reasonable, let us translate the technical jargon in this conclusion: Lavie
proposes that when we are mentally busy, we are at our most distractible, and the
more we try to concentrate, the more distractible we will be. This is why she
proposes that we should try to work under conditions of low mental load. But,
what would be the point of concentrating on a task if not to shut out the world
and protect our thoughts from worldly distractions? Lavie’s “cognitive load”
proposal again is not only counterintuitive, but it is a proposal that flies in the
face of the entire inattentional blindness literature (in addition to being at odds
with data from her own lab, Macdonald & Lavie, 2008, 2011): if nothing else,
the theoretical and applied literature on this phenomenon has taught us that we
are insensitive to otherwise easily visible events when we are highly focused on
a cognitive task, it be judging the lengths of the arms of a cross (Mack & Rock,
1998), getting ready to produce a list of complicated navigation instructions
(Simons & Levin, 1998), identifying a briefly presented letter (above), alpha-
betizing a list of letters (Fougnie & Marois, 2007) or talking on the cell phone
(Strayer & Drews, 2007).
So, what gives? We propose that these conclusions (all based on soundly
conducted experiments and robust effects) arise from a fundamental misun-
derstanding of the term “task-irrelevant distractor” and confusion between
the concepts of distraction and distractor interference. Lavie (2010) calls the Flanker
Effect: “the most conventional laboratory index of distraction” (p. 144). Lavie’s
Load Theory was entirely inspired by results from Flankers experiments and
most of her work on attention uses variations on “task-irrelevant distractor”
paradigms in which the distractors are not really task irrelevant.We have already
suggested these are not appropriate paradigms for studying distraction. Dis-
traction is the concept about how events or thoughts that we have no a priori
intention of attending to or thinking about, come nonetheless to be attended
or thought of. To reiterate, foils in a flanker task are task-relevant by design,
so they fail the basic litmus test to be informative regarding the phenome-
non of distraction. Our attention is actively set to select and scrutinize foils.
Thus, (1) we should not be surprised if they end up influencing behavior, and
(2) they do not really distract us, as much as they interfere with our stated goal of
reporting the identity of only one letter in the display (the target). We propose
that there is a fundamental difference between these two ways in which our
thoughts can be affected by an event (or thought): distraction, which we already
defined, and distractor interference. Unlike distraction, distractor interference refers
to the impact on performance of stimuli that are a priori relevant to our behavior
but not the target of our task.
Let’s return to the example of a visual search task: if the goal is to find
a red circle and we bias our visual attention to select red items, the degree
to which the number of distractors will interfere with our task of finding
the red target will be determined by the level of functional overlap (or task
relevance) between the distractors in the display and the target. If all distrac-
tor shapes are blue, there will be no distractor interference in this task and
participants will find the target irrespective of set size. If, on the other hand,
there are also red square distractors, these red distractors (because they are
candidates) will interfere with our goal of finding the red circle, and RTs
will increase with set size. Unless of course, we can bias our attention s ystem
to find the target in terms of its shape (rather than its color), in which
case, there will be no functional overlap between our a priori intentions
(of finding circles) and the distractors (squares). In Section 4 of this chapter
we will review how we believe selection works in these simple examples.
For the time being, it is important to recognize that the set-size effect in
visual search is an example of distractor interference, not one of distraction.
We are not distracted by the distractors, we are interfered with by the distractors,
and in particular, by those distractors that are relevant to our task of finding
the target. In our new terms, we inspect any and all and only candidates.
The same is true with the Flanker Effect and this follows because the flanker
task is functionally equivalent to a visual search task without spatial uncertainty
(Eriksen & Eriksen, 1974). In other words, it is incorrect to say that we are dis-
tracted by foils in a flanker task, we are simply interfered with by them. From
this perspective, it is also easy to understand why the actual number of objects
in a real-world scene will always be larger than the functional set size for any
given visual search in that scene (the number of candidates).The functional set
size will be determined by the number of objects that are a priori of interest
to us in the scene. Looking for a person in a room filled with various furniture
items will be accomplished without any effect of the number of pieces of
furniture because the furniture cannot exert any distractor interference with
our search task since they bear no functional overlap with our task settings
(for finding a person). But, if we are looking for a specific furniture item, then
furniture items will be a priori items of interest to attention and will produce
distractor interference effects (a significant set-size effect), whereas the number
of persons in the room will be irrelevant to performance in that case.
Returning to Lavie’s arguments, as explained earlier, we believe that
extrapolating from Flanker tasks to the domain of distraction is inappropri-
ate. The problem is that, while the flankers might not be providing infor-
mation in the sense of information theory, they are highly task-relevant
because they are foils—i.e. they are very much like the targets. This greatly
contrasts with the truly irrelevant stimuli that are used in most experiments
on inattentional blindness, such as the presentation of an unexpected square.
Even more, the data from our small experiment confirmed that attention
does treat those two situations extremely differently: when we are looking
for As and Es, the presentation of an A or E near the target location will not
go unnoticed, whereas the presentation of a bigger, brighter white square
can be completely ignored. Thus, the foil in a flanker task will be processed,
which in a way is the point of the Flanker Effect. But, as our results suggest,
one cannot generalize between situations with flanking foils (which are
actually task-relevant) to those obtained with flanking squares (which
are substantially less task-relevant). The two types of distractors are treated
in a profoundly different manner by the attention system.
In sum, one cannot, as Lavie does, extend the findings from traditional
congruent and incongruent flanking stimuli (the sort that produce Flanker
Effects) to task-irrelevant stimuli (the sort that cannot produce distractor
interference), like those in most real-life situations pertaining to actual dis-
traction. In fact, we want to argue that finding a Flanker Effect in an experi-
ment is, to some important degree, a measure of the success of our attention
system (not of its failures) because it demonstrates the degree to which our
attention and information-processing system is able to simultaneously per-
form the task (with accuracies close to 100%) while also picking up several
other candidate stimuli from the display. So, whereas Lavie (and many others
before her) has characterized the presence of a Flanker Effect as a limitation
of the selective attention system, we propose that it might index its efficiency.
The Flanker Effect is not to be interpreted as evidence that attention works
as a late selection filter. Just the opposite: it is clear evidence of attention
working as an early filter, a filter whose setting is determined by our goals.
2.6.2. Distractor Interference as a Measure of Attentional Success

Case in point: a beautiful study by Torralbo and Beck (2008) demonstrated
that larger Flanker Effects are found when there is less cortical competition
for representation among items in the display. In other words, when the
brain can (without the need for biased competition, i.e. constrained spatial
attention) represent all letters in the display, there will be larger Flanker
Effects, precisely because the brain can successfully represent and select
the relevant items in the display, and these representations have, in turn, an
opportunity to independently activate response selection processes. The
fact that this opportunity exists is what gives rise to distractor interfer-
ence. Well-represented/identified distractors can activate stimuli–response
associations and thereby interfere with our goal of producing a response to
the target only. In contrast, when display items are poorly represented, our
attention system ends up having to bias the competition to each stimu-
lus separately (via spatial attention, which is effortful and works in serial
manner) and thus there are fewer opportunities for distractor interference
to arise. In other words, we believe that the presence of a Flanker Effect
reflects to some degree a level of attentional success: it reflects the success
of selecting multiple stimuli from a display. Which stimuli? Those that are
relevant to our task, those that we a priori set our system to select from the
display. Certainly, the magnitude of the Flanker Effect is likely determined
by a number of factors (e.g. how small can we make our “spotlight” of
spatial attention; the spatial arrangement of targets and distractors; is the
target at fixation and distractor in periphery or vice versa; as well as a
host of grouping and segregation cues that may further help attention
to discard distractors, like color, etc.). The point here, though, is that the
fact that there is an opportunity for distractor interference should be inter-
preted as evidence that attention is efficiently selecting from the display
the things it cares about.
2.6.3. Connecting Flanker Experiments to Real-World Situations

This is why the recommendations from Lavie are, ultimately, exactly the
opposite of what they should be. When she proposes that we should all
work with highly cluttered displays (i.e. displays with high perceptual load)
so that we can avoid Flanker Effects (i.e. in her worldview, avoid distrac-
tion), her recommendation is exactly opposite to what her data suggest we
should do. We should actively seek display arrangements that are conducive
to Flanker Effects precisely because the conditions that produce Flanker
Effects are those in which we can effortlessly select several task-relevant
objects from the display in parallel, whenever we need to find them. We
will not select task-irrelevant objects, just like we do not select blue circles
when looking for a red one, or just like we do not tend to select a square
when looking for a letter.
With regards to her second recommendation, Lavie proposes that we

should work under conditions of low cognitive load to prevent distractibil-
ity by “task-irrelevant” stimuli that is observed under high cognitive load.
Again, this is exactly opposite to what one should conclude. She is again
confusing distractor interference with distractibility. Distractibility is what
has been measured in the inattentional blindness literature: when cognitive
engagement is high, we are less sensitive to truly task-irrelevant distractors.
And, if cognitive engagement is on the primary task (i.e. on identifying
a letter), this will make it more likely that our system will pick up infor-
mation from the world that is relevant to our task, which is exactly what
one should hope it would do. This is also why there was almost no inat-
tentional blindness to letters in our experiment. And what the literature
on inattentional blindness has taught us is that a high level of cognitive
engagement in the main task comes at the “cost” of less processing of truly
task-irrelevant events in the world, so much so, that we can even become
completely unaware of otherwise easily detectable events. And this is why
our participants failed to see the square. In sum, when we concentrate on a
task, (1) we are more likely to select the information that we are looking for
in order to complete our task; and (2) we are less likely to select completely
irrelevant information—i.e. less likely to be distracted. Countless real-life
examples come to mind. When we are engrossed in a TV show, we may fail
to hear the teapot whistling (task-irrelevant stimulus).When we are focused
on writing a manuscript, we may fail to hear the clock in our office ticking.
And when we are focused on a computer project and we want to look for
folders on our desktop, we will not be distracted by the background image
on our monitors, yet we will likely quickly know where all the folders on
our desktop are.
Before we move on to the next brief section of this chapter, we want to
point out that Lavie is not the only investigator to have misused the term
“task-irrelevant” to qualify distractors or to have confused the concepts of
distractor interference with that of distraction. Another visible example is
Kim, Kim, and Chun (2005), Proceedings of the National Academy of Sciences
who write “Concurrent working memory load can reduce distraction”.
Instead of using a Flanker Task to draw this conclusion, they used several
variations of a Stroop task. In other words, they were measuring effects of
task-relevant stimuli (foils) on target processing. This is not a measure of
distraction; it is a measure of distractor interference. To be clear, we are not
questioning the replicability of the results in any of the several dozen papers
on perceptual load, or any paper on the Stroop effect, or Simon Effect.
These are all tasks that ask the same question: to what degree aspects of
a task-relevant stimulus impact target processing. They measure distractor
interference.Yet, in our haste to extend our results to the real world, to claim
that our experiments can be applied to important real-life challenges, like
the understanding of distraction and distractibility, we might be tempted to
substitute one term for the other. This is wrong. This is not to say that the
results from these experiments do not inform important real-world deci-
sions like, for example, how to design a high-efficiency workplace. They
clearly do, as we reviewed in the paragraph above. That said, we must be
more careful about what the conclusions that we can draw about our
experiments are and avoid overreaching in our explanatory power. To avoid
making such mistakes, we believe it is important that we start being more
mindful in the use of terms in our discipline, like distraction, distractibility
and task (ir)relevance. Or we risk drawing the exact opposite conclusions
from our data, as Lavie (2010) has done, or perhaps more subtle, use the
results of our experiments to make inferences about phenomena that they
cannot inform (as Kim et al., 2005).To this end, we have proposed the use of
new terms, candidates and foils, to identifying different types of distractors.
These labels are useful because they also help us to determine the explana-
tory bounds of the effects observed with each type of distractor.
3. A BRIEF CASE STUDY ON DISTRACTION

3.1. Limitations of Unexpected Event Paradigms
We have reviewed one set of experimental paradigms that has been useful in
bringing about an initial understanding of distraction: the unexpected event
paradigm that produces inattentional blindness. Unfortunately, though
powerful, there are two main problems with this particular paradigm if one
wants to further explore the issue of distraction. The first concern is simply
a question of methodology: the typical unexpected event paradigm pro-
duces only one observation per subject.Though one can draw some impor-
tant theoretical inferences from these experiments (as we have done above),
it remains a fundamentally limiting paradigm. The main reason is that, once
participants are made aware of the fact that the experimenters will be asking
questions about the critical stimulus, they start attending it on every trial.
Because of this, the critical stimulus becomes extremely relevant to the sub-
ject, so we can no longer argue that future instances of similar stimuli will
be treated in the same “unattended” fashion as the first instance was. Good
evidence for the “incorporation” of the critical stimulus into the main
attentive task can be seen in Moore, Lleras, Grosjean, and Marrara (2004). In
that experiment, participants were asked to report the identity of a centrally
presented letter (forward and backward masked). In addition to the target
letter, a small dot could appear either to the right or to the left of the target.
Participants completed a first block of 64 trials with this setup, followed by
a short block of trials where we tested participants’ awareness for the small
lateralized square, as in an inattentional blindness experiment. Only 16% of
participants reported having seen the square in the inattentional blindness
trial. Critically, participants were asked to complete a second full set of 64
trials, identical in every way to the first block of trials.We used the accessory
Simon effect (e.g. Simon, Acosta, Mewaldt, & Speidel, 1976) as a measure
of distractor processing (via distractor interference). Whereas the square did
not produce any interference on the letter identification task before the inat-
tentional blindness trial, it produced robust levels of distractor interference
in the second block of trials. In other words, participants in the first block
of trials were unaware of the square (as assessed by the inattentional blind-
ness results) and therefore it failed to impact performance. In contrast, once
we asked questions about the square, they not only became aware of it, but
incorporated it into their task. From then on, the presence of this square
interfered with the responses to the central target on every trial.
A second problem with the unexpected event paradigm is one of gener-
alizability. It is true that the paradigm does reflect a setting that one may find
often in the real world (when one is effortfully attending to one event and a
completely unexpected event takes place).Yet, there is the perhaps far more
common situation of being in a busy environment (and knowing some of
the characteristics of the environment) while trying to effortfully engage in
some mental activity. For instance, one may be trying to read an important
email from work while children are playing around. Or one may be talk-
ing on the cell phone while driving. In both cases, one is well aware of the
characteristics of the environment. And one may decide to only pay atten-
tion to one thing (i.e. ignore the children in the first example) or pay mostly
attention to one thing (i.e. momentarily decreasing attention to the driv-
ing). How will we react to events along these unattended (or significantly
underattended) channels of information? This is an important question that
has been asked as far back as Treisman (1960). And one that deserves being
asked. With respect to the domain of driving, Strayer et al. have shown, for
example, that drivers talking on the phone are 10 times more likely to fail
to come to a stop at a four-way intersection, and they are significantly less
likely to remember traffic-related events that occurred during their driving
(in a simulator) than when the events appeared and they were not talking
on the phone (Strayer & Drews, 2007). What Strayer et al. argue is that this
is because of inattentional blindness: drivers actually are less aware of the
world while on the phone, and this decreased level of awareness does not
appear to be mitigated by the traffic relevance of worldly events. That is,
once drivers start underattending the driving environment, they seem to
lose sensitivity to the driving relevance of events in the world. But one must
highlight a certain incompatibility between Strayer’s findings and those in
the inattentional blindness’ laboratory: whereas Strayer and Drews’ findings
suggest that the relevance of events is not associated with drivers’ awareness
of those events, the inattentional blindness literature shows that relevance
does impact awareness. Clearly, more research needs to be done to under-
stand underattended perception in a paradigm where we can systematically
manipulate such things as the degree of cognitive difficulty in a task (cogni-
tive load), the cognitive engagement in a task (motivation), as well as the
relevance of events in the unattended/underattended channels of informa-
tion. Below, we propose one such paradigm.
3.2. A New Paradigm

We here propose a simple paradigm to study actual distraction—i.e. the
ability of information in unattended or underattended channels to draw
attention onto itself while the observer is actively engaged on a primary
task (Lleras & Buetti, 2012a,b). In this task, we ask participants to perform
a sustained cognitive task: each trial lasted about 1 min. In the experiments
below, we asked participants to perform a 1-back, 2-back, or running arith-
metic task. The task-relevant stimuli for this task were presented auditory
once every 3 s and each stimulus itself lasted somewhere between 500 and
800 ms. In other words, the participant had between 2200 and 2500 ms
of time between task-relevant stimuli. They used this time to perform the
required cognitive operation in their minds. The critical manipulation is
what happened on the display while participants were focused on per-
forming this ongoing task. We informed participants that while they were
performing this cognitive task, completely irrelevant photographs would
appear on the computer monitor. They were told that they could ignore
those pictures and that they would not be tested on them at any point.
There were 360 photos, all taken from the International Affective Picture
System (Lang, Bradley, & Cuthbert, 2008). They were selected as being
somewhat neutral on valence and not arousing; they were mildly interesting
photos. The images appeared 1800–2300 ms after the offset of the previous
task-relevant stimulus and 700–1200 ms before the onset of the next task-
relevant stimulus. Images could appear at one of the four possible locations
(above, below, to the right or left of the center of the display). Only one
image was present at a time.When a new image appeared, it always appeared
at one of the three unoccupied locations. An image disappeared exactly at
the time the next image in the sequence appeared.
To summarize the experimental task, participants engaged in a com-
plex cognitive task for a period of 1 min at a time. While they did this,
images appeared every now and then on the display. Importantly, when a
new image appeared, it was the only event in the world, the only “interest-
ing” location to look at on the monitor and its presentation was isolated
in time from the presentation of the task-relevant stimuli (that is, it did not
compete with other stimuli for sensory attention). A schematic of the trial
events is shown in Figure 7.1(a). Figure 7.1(b) and (c) show results from
this task in three versions of the task. The experimental stimuli and events
were identical in all three experiments. The only thing that differed was
the task we asked participants to perform on the task-relevant stimuli. The
task-relevant stimuli were a set of audio files with someone’s voice recit-
ing a simple mental operation (“plus one”, “minus one”, “plus two”,… up
to “minus five”). First group of participants performed a 1-back task on
the list of operations, that is, they counted the number of times the same
math instruction was presented twice in a row. This was a “low-cognitive
load” group. A second group of participants performed a 2-back task on the
operations, that is, they counted the number of times in the 1 min trial, the
current operation was the same as the operation before the last one. This
was a “high-cognitive load” group. Finally, a third group of participants was
asked to actually compute all the math operations starting from a three-digit
number that was presented to them at the start of the 1 min trial (a running
arithmetic task). This was also a “high-cognitive load” group.
As can be seen from Figure 7.1, the data we collect is simply a measure
of where people look when a new and very salient photograph suddenly
appears in the display.The results speak for themselves. As one would expect
(and as would be predicted by all theories of visual attention), when par-
ticipants are only doing a very easy cognitive task (one that only requires
rehearsal of auditory information), participants spontaneously look at our
photos: 75% of the time, within 200 ms of a new image appearing, their
eyes land on the photo. No surprise there. In fact, if there is a surprise is
that the percentage is this low: one might have expected something closer to
100%. Be that as it may, it provides us with a baseline level of performance:
(a) +1 -2
123 …
1500 ms 1500 ms 700 + 200 ms ~ 650 ms 1650 + 200 ms 700 + 200 ms ~ 650 ms
3000 ms
(b) Regards to new image (c) Regards to blank regions of the screen
before and after onset of image before and after onset of image
1.0 1.0
0.9 0.9
Proportion of regards
0.8 0.8
to area of interest
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0.0 0.0
–200 –100 0 +100 +200 +300 +400 +500 –200 –100 0 +100 +200 +300 +400 +500
Time (ms) Time (ms)
1-back task (low load) 2-back task (high load) math task (high load)
Figure 7.1 Oculomotor capture during mental engagement. (a) Schematic of events in
a trial. Every 3 s, participants heard a new operation. The sound file lasted on average
650 ms. A new image always appeared at an unoccupied location, on average 1650 ms
after the end of the sound file. (b) Proportion of regards to the new image within 500 ms
after image onset (time = 0). Participants reflexively oriented to about 75% of images
in the 1-back task, whereas they oriented to fewer than 50% of new image onsets both
in the 2-back and math tasks. (c) Proportion of regards directed to empty regions of
the display as a function of time across the three cognitive load conditions. All three
experimental groups showed a gross insensitivity to the image onset (time 0). If a par-
ticipant had been looking at an empty region of the display before the image onset, he/
she continued to do so after the onset. Furthermore, the degree of regards to “nothing”
was about twice as likely in the high-cognitive load conditions as it was in the low-load
condition. (For color version of this figure, the reader is referred to the online version of
this book.)
these irrelevant (though admittedly expected) images are inspected on a

vast majority of instances. However, this is not what happens when we
increase the degree of cognitive effort on the main task. The data also show
that when people have to actively engage executive resources and apply
them to the auditory information, their spontaneous regards of the new
images drop to <50%.This is remarkable. Not only did their regards to new
images significantly decrease (Figure 7.1(b)), but there is a large percentage
of regards to “nothing” (completely blank regions of the monitor, Figure

7.1(c)), and these regards are completely unaffected by the appearance of
the new image: when participants were looking at nothing, they continued
to do so after the onset of the image. The conclusion is simple. In identi-
cal environments, when participants engage in cognitive tasks that require
some degree of executive control, they spontaneously withdraw their atten-
tiveness from irrelevant events.
3.3. Discussion
This new task to study distraction is interesting because it reproduces a
real-life situation in which one is trying to complete a mental task, and is
focused on doing so, while there are ongoing events in the world. Interest-
ingly, one has a general idea about the structure of the environment: here,
that images appear with a certain frequency and that one can ignore them,
inasmuch as one wants to ignore them, since they do not inform our deci-
sion making on the primary task. The results suggest that we are sensitive
to some degree to the cognitive demands of a task and that we tradeoff
attention, giving more to our thoughts and less to the world, when cogni-
tive demands are high. Interestingly, this does not seem to be the end of
the story. It is difficult to compare performance and mental effort across
the two high-cognitive load tasks (perhaps separate measures of effort and
mental workload like hear-rate variability and skin conductance would be
informative). So, to further investigate the effect of cognitive demands on
attention, we performed a follow-up analysis on the math task, because we
had a level of control over the actual cognitive difficulty of specific math
operations. We measured on a separate, self-timed task, how long it took
participants to complete each of the math operations in the experiment.
For instance, it took our participants 1.7 s to complete a +1 operation on a
three-digit number, whereas it took twice as long to complete a −5 opera-
tion (3.2 s).These data suggest a higher degree of cognitive effort is required
to complete +5 operations compared to +1 operations. When we analyzed
regards as a function of the difficulty of the math operation, we found no
measurable effect of the difficulty of the operation on the likelihood that
participants will look at the next new image when it appears (Figure 7.2).
Thus, even though for our participants it is substantially easier to add 1 to
a three-digit number than to add 5, the complexity of this operation does
not determine whether they will look or not at the next new image on the
screen.This suggests that the “tradeoff ” between attentiveness to the outside
world (images) and attentiveness to the mental world (cognitive operations)
(a) (b)
4 1 Proportion looks at new image 500 ms after onset
Self paced time to complete math operation
0.9
Completion time (sec)
3.5
Proportion looks
0.8
3
0.7
2.5 0.6
2 0.5
1.5 0.4
0.3
1
0.2
0.5 0.1
0 0
+1 –1 +2 –2 +4 –4 +5 –5 +1 –1 +2 –2 +4 –4 +5 –5
Operation Operation
Figure 7.2 Analysis of the math task. (a) Participants completed at the end of the exper-
iment a self-paced version of the main experiment, where they had a chance to advance
the operations at their own pace. No images were presented on the screen. Operation
times varied greatly as a function of the operation. (b) Proportion regards on the new
image, 500 ms after its onset. Comparing panels (a) and (b) demonstrates that in spite of
large differences in the degree of difficulty of specific math operations, the type of math
operation participants were performing when the image appeared had no impact on
their likelihood to look at the new image.
is not driven by a moment-by-moment need for cognitive resources, but

by an a priori and general investment of resources. That is, it appears that
the actual cognitive resources required to complete a task do not deter-
mine the level of distractibility of an individual, rather, it seems that the
degree of anticipated effort is what determines it. In other words, we believe
it is the amount of resources that participants voluntarily set aside to com-
mit to their primary task that determines the attentional tradeoff between
thoughts and outside world. The present cognitive difficulty of a task may
only play an indirect role in this tradeoff; initially, it may help to calibrate the
degree of concentration (effort) to invest in the primary task (“this is going
to be tough, I better concentrate”), and it may interact with the degree of
motivation of participants (“this is too hard”).
Evidence in favor of “effort calibration” in a cognitive task has been
documented in the abbreviated vigilance task literature (e.g. Helton &
Russell, 2012;Temple et al., 2000) in which participants are asked to engage in
brief, but cognitively demanding tasks for periods of about 2 min (each with
about 110 events of interest) (Figure 7.3). Performance in the first 2 min trial
or “period of watch” (as it is called) is typically superior than performance
at all subsequent 2 min trials, as if participants had initially engaged in the
task with great motivation, but after the first 2 min trial had decided to settle
at a more comfortable, less arduous level of engagement. Similarly, Silvia,
Jones, Kelly & Zibaie (2011) produced good evidence that motivation and
cognitive difficulty interact to determine performance (see also Gendolla &
Figure 7.3 (a) Figure 3 from Helton & Russell (2012) showing performance (as indexed
by A’) in a demanding sustained attention task, as a function of periods of watch.
Richter, 2010; Silvia, McCord, & Gendolla, 2010). In their paper, Silvia et al.
assessed motivation in terms of a personality trait (“self-focused attention”),
which has been shown to reflects participants’ willingness to exert effort in
a task, as well as their own level of trait motivation (e.g. Carver & Scheier,
2001; Duval & Silvia, 2001).The authors manipulated the difficulty of a task
by changing the response window (long, medium and short = easy, medium
and difficult, respectively), using a perceptual discrimination that was dif-
ficult to begin with. They measured systolic blood pressure (SBP) reactivity,
which is a good physiological marker of effort (e.g. Bongard, 1995; Smith,
Ruiz, & Uchino, 2000). Participants high in self-focused attention showed
increased SBP reactivity as the difficulty of the task increased. In contrast, for
participants low in self-focused attention, SBP reactivity increased from low
to medium task difficulty levels, and then it dropped when task difficulty
became high. This is evidence that some people can motivate themselves to
exert larger effort in difficult tasks than others. Here we propose that this
decision of how much effort to put into a task is as critical a determinant
of their distractibility as is the cognitive requirements of the attentive task.
Note that we believe motivation can also be directly manipulated by incen-
tives or pressure, and we believe those manipulations would result in the
same effects on attention. There is a need for research along these lines.
Periods of watch are 2 min intervals during which performance is evaluated. The entire
experiment lasts 12 min. (b) Figure 4 from Helton & Russell (2012) showing mean reaction
time in the same task. As can be seen from these two figures, participants initially engage
strongly in the task (first 2 min), both in terms of their initial effort to be accurate (higher
A′) and fast (slower RT). However, performance drops quickly after that initial period
and there is little-to-no change in the remaining five periods of watch, either in terms of
accuracy or RT. Notice that this is the opposite of a speed-accuracy tradeoff: participants
become both less accurate and slower, even though nothing has changed about the task.
This is evidence of voluntary changes in effort (or engagement) in the task, as partici-
pants calibrate their effort, and settle into the task. (c) Figure 1 from Silvia et al. (2011).
Systolic blood pressure reactivity (an index of incurred effort) as a function of task dif-
ficulty (low, medium and high) for participants either low or high in self-focus attention,
which is a measure of how people evaluate their performance with respect to their own
standards and how they motivate themselves to achieve (or not) those standards. People
high in self-focused attention will try harder than people low in self-focused attention as
success is more significant to them and they will motivate themselves endogenously to
achieve better performance. As can be seen, people high in self-focus increase the degree
of effort they put in the task as task difficulty increases from low to medium and from
medium to high. In contrast, participants low in self-focused attention try harder when
difficulty goes from low to medium, but when difficulty further increases to high levels,
they stop trying, and SBP reactivity simply drops. (All figures reprinted with permission).
4. A THEORY OF ATTENTION AND DISTRACTIBILITY

4.1. The Need for Inner Focus
Glenberg, Schroeder, and Robertson (1998) presented a wonderful dem-
onstration of the spontaneous tradeoff between attention to the world and
attention to our thoughts. In their experiments, participants were asked to
do a series of remembering tasks that varied in difficulty. The specific ques-
tions were about autobiographical events or general knowledge. In Experi-
ment 3, participants viewed such questions on a computer monitor and the
investigators rated the number of times people spontaneously averted their
gaze from the monitor (e.g. closed their eyes or looked away). Their results
showed that the more people averted their gaze, the better their perfor-
mance in the memory task. In Experiment 4, they converted this measure
to a manipulation: in some conditions, they asked participants to close their
eyes while thinking of the answer to the question. They found that people
performed better at both the memory task and on a set of simple math
problems, as well, when they closed their eyes than when they kept their
eyes open.Together, these results are good indicators of two main character-
istics of our attention system. First, when we have a need to concentrate on
some internal cognitive process, we tend to actively prevent the encoding of
potentially distracting information. Second, the prevention of encoding of
potentially distracting information is, in fact, functional: we are better think-
ers when we shut our eyes. This explains why we tend to avert our gaze
from a listener when we are about to start speaking (e.g. Beattie, 1978a,b;
Doherty-Sneddon, Bruce, Bonner, Longbotham, & Doyle, 2002; Doherty-
Sneddon & Phelps, 2005; Ehrlichman, 1981; Goldman-Eisler, 1967): we
are actively trying to shut down a large source of potential distraction (our
listener’s face) while we are engaged in trying to plan new sentences.
Thus, our instinct to move our eyes away from potential sources of
distractions is both a reflection of our intuitive understanding of the hazards
of the environment (where distracting information may be found) as well as
an effective strategy for controlling the allocation of cognitive resources in
times of need. Interestingly, Glenberg et al.’s results also showed that when
overly difficult questions were used (such as questions to which people knew
they did not know the answer for example), people no longer averted their
gaze. They had stopped trying to do the task. They knew that a momen-
tary boost of cognitive focus would not change the fact that they did not
know the answer to the question posed. So, they no longer engaged in gaze
aversion. This makes it clear how gaze aversion is not merely a reaction to
heightened cognitive demands; it is strategic in nature. We propose that the
attentional system has an executive component that is sensitive to anticipated
cognitive demands (and whether said demands will exceed capacity), as well
as motivational states to bring about a tradeoff between sensory attention
(i.e. attention to world events) and mental attention (i.e. attention to inner
cognitive processing). And we propose that it is this tradeoff, when pushed
to extremes, that leads to the more blatant failures of awareness observed
in inattentional blindness and when driving while talking on the phone. In
a momentary need for extreme inner focus, attention to the world suffers.
The need to reorient attention inward may also help explain other psy-
chological phenomena, such as the attentional blink (Raymond, Shapiro,
& Arnell, 1992). This phenomenon refers to the finding that people often
miss the second of two successive targets, at least when those targets are
presented embedded in a rapid visual stream of distractors and the second
target appears 200–500 ms after the first one. There are several theories as
to why the attentional blink occurs, but most models agree that the blink
itself seems to be tightly linked to the need to “save” information from the
first target in working memory, so that it can be accessible for response later
on. Functionally, this seems like a brief but acute need for inner focus: upon
detection of the first target, and given the presence of so many distractors
(the other letters in the stream), the participant may need to effortfully pick
and store the detected target in working memory. If it is true that an acute
need for inner processing triggers a disengagement from potential sources
of distraction, then the need to store the first target, once it is detected
(“Oh, that’s a target! I need to remember it.”), in the face of rapid steam of
stimuli, may be what causes the blink to occur. This account is not novel
per se. What is new is the explicit presentation of a framework from which
to understand why consolidation into working memory (or any other
demanding and inwardly directed process) would create a blink in the first
place (as several theories have argued, Bowman & Wyble, 2007; Chun
& Potter, 1995; Craston, Wyble, Chennu, & Bowman, 2009; Di Lollo,
Kawahara, Ghorashi, & Enns, 2005; Raymond et al., 1992; Shapiro, Caldwell,
& Sorensen, 1997).
4.2. Predicting Inattentional Blindness

We believe our proposed framework for attention allows us to reframe one
of the principal challenges in the inattentional blindness literature: how to
predict when a given participant will fail to report an unexpected stimulus.
Several studies have shown that more difficult (primary) tasks lead to more
blindness (e.g. Cartwright-Finch & Lavie, 2007; Fougnie & Marois, 2007;
Simons & Jensen, 2009). But performance in the primary task alone is not a
good predictor of inattentional blindness (Cartwright-Finch & Lavie, 2007;
but see Simons & Jensen, 2009). In fact, Simons & Jensen (2009) recently
demonstrated that someone’s skill at a task also fails to predict the degree
of blindness. Why is that? The attentional tradeoff we propose is sensitive to
the anticipated need for executive control, as well as the motivational state
of the participant. The anticipated need for executive control will depend
on a number of factors.The self-perceived skill of a participant at a task will
play a big role in determining need: cigar-rolling experts will require less-
executive oversight of the activity than a beginner, thus, one would predict
that the expert will be less likely to suffer from inattentional blindness while
engaged in this activity. Once an anticipated need is identified, one must ask
the question: how well can the participant fulfill that request? Does he have
the ability to precisely match the need with the corresponding resources?
It is possible and more research is needed to answer this question, but it is
also likely that participants spontaneously assign resources in gross fashion:
“I will focus on this task” vs “I will focus on this task a lot,” rather than
“I will focus on this task with 47% of my resources” vs “I will focus on this
task with 66% of my resources.” This is not to say that participants are inca-
pable of allocating their attention in fine quanta, but that it is more likely
that they spontaneously allocate in relatively large chunks. This may also be
a reason why experimental manipulations of the degree of difficulty of the
attentive task (aimed at capturing incremental levels of attentiveness in the
participants) may not lead to a nicely titrated relationship with performance.
It may also explain why the actual difficulty of the mental arithmetic task
failed to influence gaze aversion in the task presented above: participants are
either attentive to the math task or they are not, as they anticipate a need
for focusing on it to succeed. But once that determination has been made,
whether a specific operation is easy (+1) or hard (−5) is unlikely to change
their overall attentive engagement to the task (Figure 7.2).
Finally, the amount of attentiveness to the main task (and ergo the
degree of unattentiveness to the rest of the world) will also be determined
by motivational and personality characteristics of each individual. As shown
by Silvia et al. (2011), once a task becomes too difficult, some people will
continue to increase their effort to match the demands of the task, whereas
others will simply stop trying. Factors like the self-focused attention trait
(as suggested by Silvia et al.), as well as conscientiousness (e.g. Trautwein,
Lüdtke, Roberts, Schnyder, & Niggli, 2009), will probably be important

factors to determine engagement in the attentive task. If one thinks back to
experts performing a task that does not require (of them) lots of executive
control, one might still see differences in inattentional blindness as deter-
mined by their conscientiousness: in spite of their expertise, conscientious
experts may deliberately pay more attention to the task, even if they do not
need to. And of course, this is likely to change as a function of both intrinsic
and extrinsic motivation. If there is a high reward to perform well in the
attentive task, one would expect that individuals more responsive to rewards
will be more attentive and thereby exhibit larger degrees of inattentional
blindness.
In sum, if one wants to predict performance in any situation that would
seem to be parallel to inattentional blindness, there are a number of factors
that will need to be considered and measured. Failure to take these factors
into account will likely produce just noise. Take the Simons and Jensen
result, for example, that skill at a motion tracking task failed to predict inat-
tentional blindness. Skill was rated as the speed at which a participant could
accurately track to a level of 75% accuracy. Importantly, all subjects in the
inattentional blindness task were tested with stimuli moving at a speed that
fell more or less at the half-way point of skills in the participant pool. At
one end of the spectrum, one is measuring people low in skill at the task,
performing at a level beyond their skill. These subjects will either try hard
(if conscientious or high in self-focused attention) or simply give up
(“it’s too hard for me!”). Thus, at the bottom end of the skill distribution,
one cannot predict inattentional blindness; it can go either way. The same is
true at the top end of the distribution: all these participants can perform the
task, but if the difficulty is sufficiently lower than their skill, then (1) they
may not see a need for engaging high levels of attention in the task or (2) if
they are conscientious, they will likely focus a lot on the task, to make sure
that they excel in accordance to their own standards. So, neither at the low
nor high end of the skill distribution will skill alone predict inattentional
blindness. Admittedly, our proposal remains to be tested, but we hope that it
at least provides some guiding parameters for what considerations should be
taken when one is trying to predict IB in a large group of people.
4.3. A Look Back at Visual Attention

We will not elaborate much on how visual attention works, except to
emphasize some general properties that we believe guide how visual selec-
tion takes place. Our starting points are theories like Bundesen’s theory
of visual attention as well as Desimone and Duncan’s biased competition

theory. First, we propose that the visual system uses its “nodes” of specializa-
tion to select in parallel and across the visual field stimuli that are processed
by those nodes. For instance, if the visual system has a level of specialization
that focuses on processing motion, this node can be leveraged to select the
presence of particular motions in the world (e.g. feature-based attention,
Serences & Boynton, 2007). The same ought to be true for higher or lower
nodes, so that one can select on the basis of orientations (V1), shapes (V4),
colors (V1, V4a) but also alphanumeric characters (visual word form area,
Nestor, Plaut, & Behrmann, 2012), and other objects of expertise (fusiform
face area, like faces or cars, if one is a car expert, see McGugin, Gatenby,
Gore, & Gauthier, 2012). In other words, one can easily set a filter to detect
all faces in a scene. To the extent that those faces are sufficiently well repre-
sented at the neural level (i.e. this probably means there is no competition
at the level of that nodes’ receptive fields from nonpreferred stimuli), they
will likely be selected in parallel. Thus, effortlessly, one can know where all
the persons in a room are, and likely too, process in parallel properties of the
group of selected faces (e.g. Chong & Treisman, 2003, 2005a,b; Haberman
& Whitney, 2009).
Note that this “parallel” function of visual attention is both powerful
and intuitive. We know that visual information about the entire scene flows
in parallel throughout the visual system (i.e. is being analyzed by these dif-
ferent nodes), thus for example, the FFA is analyzing face-like inputs across
the visual field all at once, or MT motion information across the visual field.
What we are arguing is that attention leverages this massively parallel pro-
cessing toward biasing a subset of elements in the scene.This allows for effi-
cient selection (or adoption) of candidates across the visual field. Note, too,
that there are natural limits to this parallel form of processing. For instance,
in addition to competition for representation, cortical magnification will
make it so that information farther from fixation will be represented with
less fidelity than information at or near fixation, thereby imposing a con-
straint on the quality of input that is available to this parallel function of
attention.
This initial form of parallel selection is a function of relevance (Bundesen,
1990; Fischer & Whitney, 2012). And it is evident in visual search, where
observers are extremely efficient at massively rejecting in parallel large sets
of items that are not relevant to the search task (Cunningham & Wolfe,
2012). Furthermore, to the extent that distractors are similar to each
other and group well, observers can reject these in large groups, as well
(e.g. Anderson, Vogel, & Awh, in press; Duncan & Humphreys, 1989). So,
we can both put in place large “adoption” as well as large “rejection” filters
to process information in parallel. The output of the “adoption” filter is the
set of likely targets for the current task (i.e. candidates). Again, assuming that
the items in the display can be accurately represented (i.e. no crowding),
how fine the “adoption” filter will be in part determined by the ability of a
particular “node” to process specific sets of features that discriminate targets
from other stimuli in the set. The more diagnostic target features there are
(relative to the specific vision node responsible for computing said features),
the smaller the set of candidates or likely targets will be (i.e. the more
effective the first sweep of selection will be). The less diagnostic the target
features are, the larger the set of items is that will be selected in that initial
sweep of selection.
What happens next? It depends on the task. In visual search, the task is
to select the “one” target among the “selected few” possible candidates at
various locations. To perform this operation, a comparison must be made
between the candidates and the target template, and this comparison will
be made in parallel in working memory and will be determined by the
capacity of working memory (Anderson et al., in press). The larger the
capacity is, the more of these comparisons can be made simultaneously.
The smaller it is, the fewer of these comparisons can be made. So, in
fact, the “serial” aspect of visual search observed in most experiments is
actually the result of two massively parallel operations (see also Young &
Hulleman, 2012). An initial sweep that determines candidates (i.e. a priori
targets), that is likely of unlimited capacity (not counting visual acuity
limitations or representational crowding). This initial sweep is followed by
a second parallel sweep that compares candidates to the target template,
also in parallel. However, this second sweep is capacity-limited (Anderson
et al., in press; Pashler, 1987). For the current purposes, let us say that the
maximum number of items is four. Search then will proceed in serial
manner, in groups of four, inspecting the set of candidates, until a match
is found (or until acuity and/or crowding require refixations to new loca-
tions in the search space). Note that the comparison process will also be
limited by the precision of the target representations in memory as well
as the confusability between candidates and target templates (e.g. Alvarez
& Cavanagh, 2004; Awh, Barton, & Vogel, 2007). It is also likely that when
the target is fixed, the comparison process will become more and more
efficient over time (e.g. Carlisle, Arita, Pardo, & Woodman, 2011; Logan,
1988; Woodman & Arita, 2011).
From this overview, a second important function of attention can be

determined: attention must act to decrease the noise in the representation
of visual items (as proposed by biased competition theory), either covertly
or overtly. That is, when items compete for representation, attention must
act on each item individually to improve the representation of that item,
at the momentary expense of surrounding items. Indeed, Scalf & Beck
(2010) nicely demonstrated that attention to multiple items simultane-
ously does not result on improvements in representation of those items, if
those are competing for representation because of their spatial arrangement.
Eye movements, guided by attention, then work to solve this problem: by
directly gazing at potential items of interest, the representation of those
items is greatly improved. Alternatively, attention can be directed serially to
each of the competing items to improve, one at a time, the representation
of each of those items.
If the task is a Flanker task, the candidates selected by attention in paral-
lel can actually be foils. That is, not only do candidates look like a target,
congruent and incongruent flankers have links to potentiated responses
(press right/left, for example). Thus, the Flanker Task is actually a task in
which response selection processes must select which of the already selected
foils they want to respond to. The problem comes from the fact that by
virtue of having been selected in the first sweep, the foils are already activat-
ing potentiated responses, which is why the congruency between foils and
target ends up producing a Flanker Effect. Studies in the Flanker Effect lit-
erature are, therefore, centered on the processes that allow for more efficient
suppression of foil interference, or alternatively, better identification of the
proper target among the foils (Lachter et al., 2004; Yigit-Elliott et al., 2011).
And there are studies showing that the executive control system is actively
trying to learn as much as it can from the task ecology to try to be as effec-
tive as possible in that regard (Max & Tsal, 2011, 2012). Yet, the critical
message that should be remembered is that the presence itself of a Flanker
Effect represents a specific sign of attentional success: the foil was success-
fully selected by our “adoption” filter as a candidate. The relative reduction
(or elimination of a Flanker Effect when one was expected) reflects success
of a different kind: success in the efficient suppression of foil interference.
But note that the simple Flanker Effect can be modulated by both factors. In
displays with “high perceptual load,” a Flanker Effect may not be observed
for two categorically different reasons: because the flanker was not selected
as a candidate (perhaps because of crowding or competition for represen-
tation, which requires slow allocation of spatial attention to disambiguate
visual representations) or because foil interference was entirely suppressed

(perhaps because the executive system found ways to more efficiently sup-
press response-related foil processing).
5. CONCLUSIONS
A Flanker Effect experiment is an experiment that studies the
response interference that is created by selected foils on target processing—
what people may call a form of “distractor interference”. Another example
of distractor interference is seen in the color-word Stroop task, where foils
are words that semantically activate the same response tokens as the colors
of the ink in which the word is written (which is what determines the
response). So, there are various forms of distractor interference effects, but,
in general, it is important to recognize that the effects arise because the dis-
tractors are in fact “foils” of some sort (foils because of their visual similarity
to the target as in the traditional flanker or their semantic overlap with the
target information as in the classic Stroop). The attention system efficiently
selects foils from a display (as it does with candidates). As we have argued
before then, paradigms like the Flanker and Stroop cannot measure distract-
ibility and cannot be used to study distractibility because the stimulus caus-
ing the interference was selected in the first place. Inspecting candidates and
foils is a critical part of the task. It is what is intended from the start. Given
that foils are present in the display, inspecting them is in an important way
part of what the participant must do to successfully complete the task.
We have reviewed a number of issues regarding various selective atten-
tion paradigms and more specifically, we have identified the problematic
use of certain labels within this literature (Figure 7.4). We have proposed
a new labeling system for qualifying “distractors” across various tasks, so
as to minimize issues with the interpretation of results from these para-
digms. Importantly, the labels themselves make it clear what the explana-
tory power and the explanatory domain are of a given stimulus and task
set. One cannot and should not try to generalize results from one type of
distractors to another set of distractors without proper evidence that such
generalization is warranted. Conducting a Flanker Effect and conclud-
ing something about distraction is therefore problematic. Having defined
this new set of labels is important because it allows investigators who are
interested in using our paradigms to understand other aspects of cogni-
tion (such as cognitive problems in clinical populations) to know which
tasks and stimuli are most appropriate for studying their phenomenon of
200
Selective attention and distraction
180 Task-irrelevant distractors
160
140
120
100
80
60
40
20
0
1960s 1970s 1980s 1990s 2000s 2010–2012
Figure 7.4 Citation counts each decade for two different searches in PsychInfo.
In white are the citation counts for the keyword “distraction” in all fields within the
specific subject “selective attention”. In black are the citation counts for the search
“task-irrelevant distractors” or “task-irrelevant distractor”. This citation count simply
illustrates the sudden explosion of articles using the terms “task-irrelevant distractors”
in the last dozen years (2000–2012). The citation count for the conjunction selective
attention and distraction is presented here as a baseline. This is instructive as it docu-
ments the research interest in understanding distraction, within the selective atten-
tion literature. That said, most selective attention paradigms are not appropriate for
drawing conclusions about distraction, if they use candidates or foils as stimuli. With
respect to the use of the term “task-irrelevant distractor,” its rise in prominence in the
literature is clearly worrisome as it is likely people are misusing and/or misunderstand-
ing the meaning of the term: 75 of those papers (almost 40%) also include the word
“distraction” as a descriptor of their paper.
interest. As an example, if one is interested in understanding distractibility

in aging or in Alzheimer’s disease, it would be simply inappropriate to use
a Flanker or a Stroop task.
With respect to the critical difference between distractor interference
and distractibility, one final comment is warranted. It is important to under-
stand that distractor interference effects are all “within-task” effects. The
subject never feels like he or she has “changed” what he/she is doing while
performing a Flanker or a Stroop task. The instructions for the task are set
and the participant feels in compliance throughout the trial. In terms of
executive control, the goal structure to complete the task does not change,
even when one is suffering from severe levels of distractor interference, as
in the traditional Stroop task. A person trying to name the ink of a color
word throughout the trial is struggling with completing one specific goal:
naming the ink (not saying the word) of the colored word. That much is
clear. A corollary to this is that no matter the type of task-relevant distractor,
a task-relevant distractor can never alter the boundaries of the task space. In
contrast, the study of distraction is itself a study of changes to the task space.
Distraction is the taking possession of the mind by a stimulus or thought
that the subject never intended to process in the first place. It is a thought or
event that interferes with the “goal stack,” therefore changes the task space.
Participants are suddenly doing (or thinking about) something other than
their stated intention. If prompted, participants can answer the question “are
you currently doing the task you are supposed to be doing”. If distracted,
the answer to that question will be no (Giambra, 1989; McVay & Kane,
2009, 2012; Smallwood et al., 2003), because they can recognize that their
goal stack has changed.
What is particularly pernicious about distraction is that we are not always
aware that a change in the goal stack has occurred.That is troublesome.That
is why understanding distraction is a matter of crucial importance to us as
we try to understand problems like “distracted driving” or rumination in
anxiety and depression. If we are not doing what we are meant to be doing
and are not aware of it, that is a problem. The example that jumps to mind
is distracted reading (Smilek, Carriere, & Cheyne, 2010). Often we realize
we have been “reading” a book but have no idea of what it is that we just
read, because our goal of reading was pushed down our goal stack to a lower
priority than, for example, our preoccupation with Mom’s dinner plans
for thanksgiving. In the visual world, understanding sensory distractibility
is also of crucial importance to understanding issues like performance at
work or human response to alert and alarm systems. A number of interest-
ing questions remain completely open and thus far are largely unanswered:
What events in our intended task are capable of bringing us back to our
main task? The death of the main character in our book? In the case of dis-
tracted driving, will the change of a traffic light to red bring us back from
our cell phone conversation and into the driving task again? This issue of
“competition for goal status” is fascinating and will likely be the source of
much fruitful research in the years to come.
REFERENCES
Alexander, R. G., & Zelinsky, G. J. (2011).Visual similarity effects in categorical search. J ournal
of Vision, 11(8).
Alvarez, G. A., & Cavanagh, P. (2004). The capacity of visual short term memory is set
both by visual information load and by number of objects. Psychological Science, 15,
106–111.
Anersdon, D. E., Vogel, E. K., & Awh, E. A common discrete resource for visual working
memory and visual search. Psychological Science, in press.
Awh, E., Barton, B., & Vogel, E. K. (2007).Visual working memory represents a fixed number
of items, regardless of complexity. Psychological Science, 18(7), 622–628.
Beattie, G. W. (1978a). Floor apportionment and gaze in conversational dyads. British Journal
of Social and Clinical Psychology, 17(1), 1–9.
Beattie, G.W. (1978b). Sequential temporal patterns of speech and gaze in dialogue. Semiotica,
23, 1–24.
Bichot, N., Rossi, A., & Desimone, R. (2005). Parallel and serial neural mechanisms for
visual search in macaque area V4. Science, 308(5721), 529–534.
Bongard, S. (1995). Mental effort during active and passive coping: a dual-task analysis.
Psychophysiology, 32(3), 242–248.
Bowman, H., & Wyble, B. (2007). The simultaneous type, serial token model of temporal
attention and working memory. Psychological Review, 114(1), 38–70.
Bundesen, C. (1990). A theory of visual attention. Psychological Review, 97(4), 523–547.
Carlisle, N. B., Arita, J. T., Pardo, D., & Woodman, G. F. (2011). Attentional templates in visual
working memory. Journal of Neuroscience, 31(25), 9315–9322.
Cartwright-Finch, U., & Lavie, N. (2007). The role of perceptual load in inattentional
blindness. Cognition, 102(3), 1–20.
Carver, C. S., & Scheier, M. F. (2001). On the self-regulation of behavior. New York: Cambridge
University Press.
Chong, S., & Treisman, A. (2003). Representation of statistical properties. Vision Research,
43(4), 1–12.
Chong, S., & Treisman, A. (2005a). Attentional spread in the statistical processing of visual
displays. Perception & Psychophysics, 67(1), 1–13.
Chong, S., & Treisman, A. (2005b). Statistical processing: computing the average size in
perceptual groups. Vision Research, 45(7), 1–10.
Choo, H., & Franconeri, S. (2010). Objects with reduced visibility still contribute to size
averaging. Attention, Perception & Psychophysics, 72(1), 86–99.
Chun, M., & Potter, M. (1995). A two-stage model for multiple target detection in rapid
serial visual presentation. Journal of Experimental Psychology: Human Perception and
Performance, 21(1), 109–127.
Cooper, J. M., Vladisavljevic, I., Medeiros-Ward, N., Martin, P. T., & Strayer, D. L. (2009).
An investigation of driver distraction near the tipping point of traffic flow stability.
Human Factors, 51(2), 261–268.
Craston, P., Wyble, B., Chennu, S., & Bowman, H. (2009). The attentional blink reveals serial
working memory encoding: evidence from virtual and human event-related potentials.
Journal of Cognitive Neuroscience, 21(3), 550–566.
Cunningham, C. A., & Wolfe, J. M. (2012). Extending “Hybrid” visual X memory search to
very large memory sets and to category search. Paper presented at the 53th Annual Meeting
of the Psychonomic Society.
Darlington,T., & Talbot, E. B. (1898). A study of certain methods of distracting the attention.
III. Distraction by musical sounds; the effect of pitch upon attention. American Journal of
Psychology, 9(3), 332–345.
Desimone, R., & Duncan, J. (1995). Neural mechanisms of selective visual attention. Annual
Review of Neuroscience, 18, 1–30.
Deutsch, J., & Deutsch, D. (1963). Some theoretical considerations. Psychological Review, 70,
80–90.
Di Lollo,V., Kawahara, J., Zuvic, S., & Visser, T. A.W. (2001). The preattentive emperor has no
clothes: a dynamic redressing. Journal of Experimental Psychology: General, 130(3), 479–492.
Di Lollo, V., Kawahara, J., Ghorashi, S., & Enns, J. (2005). The attentional blink: resource
depletion or temporary loss of control? Psychological Research, 69(3), 191–200.
Doherty-Sneddon, G., & Phelps, F. (2005). Gaze aversion: a response to cognitive or social
difficulty? Memory & Cognition, 33(4), 727–733.
Doherty-Sneddon, G., Bruce, V., Bonner, L., Longbotham, S., & Doyle, C. (2002).
Development of gaze aversion as disengagement from visual information. Developmental
Psychology, 38(3), 438–445.
Driver, J., & Tipper, S. P. (1989). On the nonselectivity of ‘selective’ seeing: contrasts between
interference and priming in selective attention. Journal of Experimental Psychology: Human
Perception and Performance, 15, 304–314.
Dulsky, S. G. (1932). Discussion: what is a distractor? Psychological Review, 39(6), 590–592.
Duncan, J., & Humphreys, G. (1989). Visual search and stimulus similarity. Psychological
Review, 96(3), 433–458.
Duval, T., & Silvia, P. (2001). Self-awareness & causal attribution: A dual systems theory. Boston:
Kluwer Academic.
Ehrlichman, H. (1981). From gaze aversion to eye-movement suppression: an investigation
of the cognitive interference explanation of gaze patterns during conversation. British
Journal of Social Psychology, 20(4), 1–9.
Eriksen, B. A., & Eriksen, C. W. (1974). Effects of noise letters upon identification of a target
letter in a non-search task. Perception & Psychophysics, 16, 143–149.
Estes,W. K., & Taylor, H. A. (1966).Visual detection in relation to display size and redundancy
of critical elements. Perception & Psychophysics, 1, 9–16.
Fecteau, J. (2007). Priming of pop-out depends upon the current goals of observers. Journal
of Vision, 7(6), 1.
Fischer, J.T., & Whitney, D. (2012). Distractor suppression: attention gates visual coding in the
human pulvinar. Nature Communications.
Folk, C. L., Remington, R. W., & Wright, J. H. (1994). The structure of attentional control:
contingent attentional capture by apparent motion, abrupt onset, and color. Journal of
Experimental Psychology: Human Perception and Performance, 20(2), 317–329.
Folk, C. L., Leber, A. B., & Egeth, H. E. (2008).Top-down control settings and the attentional
blink: evidence for nonspatial contingent capture. Visual Cognition, 16(5), 616–642.
Fougnie, D., & Marois, R. (2007). Executive working memory load induces inattentional
blindness. Psychonomic Bulletin and Review, 14(1), 142–147.
Franconeri, S., Alvarez, G., & Enns, J. (2007). How many locations can be selected at once?
Journal of Experimental Psychology: Human Perception and Performance, 33(5), 1003–1012.
Gendolla, G. H., & Richter, M. (2010). Effort mobilization when the self is involved: some
lessons from the cardiovascular system. Review of General Psychology, 14(3), 212–226.
Giambra, L. M. (1989). Task-unrelated thought frequency as a function of age: a laboratory
study. Psychology and Aging, 4(2), 136–143.
Glenberg, A., Schroeder, J., & Robertson, D. (1998). Averting the gaze disengages the envi-
ronment and facilitates remembering. Memory & Cognition, 26(4), 651–658.
Goldman-Eisler, F. (1967). Sequential temporal patterns and cognitive processes in speech.
Language and Speech, 10(2), 122–132.
Grice, G. R., & Gwynne, J. W. (1987). Dependence of target redundancy effects on noise
conditions and number of targets. Perception & Psychophysics, 42, 29–36.
Grice, G. R., Canham, L., & Boroughs, J. M. (1984). Combination rule for redundant
information in reaction time tasks with divided attention. Perception & Psychophysics, 35,
451–463.
Haberman, J., & Whitney, D. (2009). Seeing the mean: ensemble coding for sets of faces.
Journal of Experimental Psychology: Human Perception & Performance, 35(3), 718–734.
Haberman, J., & Whitney, D. (2011). Efficient summary statistical representation when
change localization fails. Psychonomic Bulletin and Review, 18(5), 855–859.
Helton, W., & Russell, P. N. (2012). Brief mental breaks and content-free cues may not keep
you focused. Experimental Brain Research, 219(1), 37–46.
Hillyard, S. A., & Münte, T. F. (1984). Selective attention to color and location: An analysis
with event-related brain potentials. Perception and Psychophysics, 36, 185–198.
Hollingworth, A., Matsukura, M., & Luck, S. J.Visual working memory modulates rapid eye
movements to simple onset targets. Psychological Science, in press.
Itti, L., & Koch, C. (2001). Computational modelling of visual attention. Nature Reviews
Neuroscience, 2(3), 194–203.
Jaensch, E. R. (1929). Grundformen menschlichen Seins. Berlin: Otto Elsner.
Kastner, S., O’Connor, D. H., Fukui, M. M., Fehd, H. M., Herwig, U., & Pinsk, M. A. (2004).
Functional imaging of the human lateral geniculate nucleus and pulvinar. Journal of
Neurophysiology, 91(1), 438–448.
Kim, S., Kim, M., & Chun, M. (2005). Concurrent working memory load can reduce
distraction. Proceedings of the National Academy of Sciences of the United States of America,
102(45), 16524–16529.
Kirchner, H., & Thorpe, S. (2006). Ultra-rapid object detection with saccadic eye movements:
visual processing speed revisited. Vision Research, 46(11), 1762–1776.
Lachter, J., Forster, K. I., & Ruthruff, E. (2004). Forty years after Broadbent: still no
identification without attention. Psychological Review, 111, 880–913.
Lang, P., Bradley, M., & Cuthbert, B. (2008). International affective picture system (IAPS): Affective
ratings of pictures and instruction manual. Technical Report A-8, Gainesville, FL: University
of Florida.
Lavie, N., Hirst, A., de Fockert, J. W., & Viding, E. (2004). Load theory of selective attention
and cognitive control. Journal of Experimental Psychology: General, 133(3), 339–354.
Lavie, N. (2010). Attention, distraction, and cognitive control under load. Current Directions in
Psychological Science, 19(3), 143–148.
Lin, J., Franconeri, S., & Enns, J. (2008). Objects on a collision path with the observer
demand attention. Psychological Science, 19(7), 686–692.
Lleras, A., & Buetti, S. (2012a). Ocular capture by abrupt onsets is severely reduced during
mental arithmetic: evidence against cognitive load theory. Paper Presented at the 53rd
Annual Meeting of the Psychonomic Society.
Lleras, A., & Buetti, S. (2012b). Where do the eyes go when you think? Away from visually
salient information. Paper presented at the 12th Annual Meeting of the Vision Sciences Society,
Naples, Florida. 12(9), 1261, http://dx.doi.org/10.1167/12.9.1261.
Logan, G. D. (1988). Toward an instance theory of automatization. Psychological Review, 95,
492–527.
Macdonald, J., & Lavie, N. (2008). Load induced blindness. Journal of Experimental Psychology:
Human Perception and Performance, 34(5), 1078–1091.
Macdonald, J., & Lavie, N. (2011). Visual perceptual load induces inattentional deafness.
Attention, Perception & Psychophysics, 73(6), 1780–1789.
Mack, A., & Rock, I. (1998). Inattentional blindness. Cambridge, MA: MIT Press.
Maljkovic, V., & Nakayama, K. (1994). Priming of pop-out: I. Role of features. Memory &
Cognition, 22(6), 657–672.
Maljkovic,V., & Nakayama, K. (1996). Priming of pop-out: II. The role of position. Perception
& Psychophysics, 58(7), 977–991.
Max, R., & Tsal, Y. (2011). Efficient selection is modulated by target-distractor discrim-
inability and not by perceptual load. Poster presented at the 52nd Annual Meeting of the
Psychonomic Society, Seattle, WA, November 4, 2011.
Max, R., & Tsal,Y. (2012). Exact temporal locus of visual distraction. Poster presented at the 53rd
Object Perception Attention and Memory Workshop, Minneapolis, MN, November 15, 2012.
McGugin, R.W., Gatenby, J. C., Gore, J. C., & Gauthier, I. (2012). High-resolution i maging of
expertise reveals reliable object selectivity in the FFA related to perceptual performance.
Proceedings of the National Academy of Sciences of the United States of America, 109(42),
17063–17068.
McVay, J. C., & Kane, M. J. (2009). Conducting the train of thought: working memory
capacity, goal neglect, and mind wandering in an executive-control task. Journal of
Experimental Psychology: Learning Memory and Cognition, 35(1), 196–204.
McVay, J. C., & Kane, M. J. (2012). Drifting from slow to “D’oh!”: working memory capacity
and mind wandering predict extreme reaction times and executive control errors. Journal
of Experimental Psychology: Learning Memory and Cognition, 38(3), 525–549.
Miller, J. (1982). Divided attention: evidence for coactivation with redundant signals. Cognitive
Miller, J. (1987). Priming is not necessary for selective-attention failures: semantic effects of
unattended, unprimed letters. Perception & Psychophysics, 41, 419–434.
Moore, C., & Egeth, H. (1997). Perception without attention: evidence of grouping under con-
ditions of inattention. Journal of Experimental Psychology: Human Perception and Performance,
23(2), 339–352.
Moore, C., Grosjean, M., & Lleras, A. (2003). Using inattentional blindness as an opera-
tional definition of unattended: the case of surface completion. Visual Cognition, 10,
299–318.
Moore, C., Lleras, A., Grosjean, M., & Marrara, M. (2004). Using inattentional blindness as an
operational definition of unattended: a response-end effect. Visual Cognition, 11, 705–719.
Mordkoff, J.T., & Halterman, R. (2008). Feature integration without visual attention: evidence
from the correlated flankers task. Psychonomic Bulletin and Review, 15, 385–389.
Mordkoff, J. T., & Yantis, S. (1991). An interactive race model of divided attention. Journal of
Experimental Psychology: Human Perception and Performance, 17(2), 520–538.
Mordkoff, J.T. (1996). Selective attention and internal constraints: there is more to the flanker
effect than biased contingencies. In A. F. Kramer, M. G. H. Coles & G. Logan (Eds.),
Converging operations in the study of visual selective attention (pp. 483–502). Washington,
DC: APA.
Most, S., Simons, J., Scholl, B., Jimenez, R., Clifford, E., & Chabris, C. (2001). How not to
be seen: the contribution of similarity and selective ignoring to sustained inattentional
blindness. Psychological Science, 12(1), 9–17.
Most, S., Scholl, B. J., Clifford, E. R., & Simons, D. J. (2005). What you see is what you
set: sustained inattentional blindness and the capture of awareness. Psychological Review,
112(1), 217–242.
Navalpakkam,V., & Itti, L. (2007). Search goal tunes visual features optimally. Neuron, 53(4),
605–617.
Neider, M., & Zelinsky, G. (2011). Cutting through the clutter: searching for targets in evolving
complex scenes. Journal of Vision, 11(14).
Neisser, U., & Becklen, R. (1975). Selective looking: attending to visually specified events.
Cognitive Psychology, 7(4), 480–494.
Neisser, U. (1979). The control of information pickup in selective looking. In A. D. Pick
(Ed.), Perception and its development: A tribute to Eleanor J. Gibson (pp. 201–219). Hillsdale,
NJ: Lawrence Erlbaum.
Nestor, A., Plaut, D., & Behrmann, M. (2012). Orthographic form processing—a multi-
variate investigation of its neural basis. Cerebral Cortex, Cereb. Cortex firxt published
online June 12, 2012, http://doi.10.1093/cercor/bhs158.
Norman, D. A. (1968). Towards a theory of memory and attention. Psychological Review, 75,
522–536.
O’Connor, D., Fukui, M., Pinsk, M., & Kastner, S. (2002). Attention modulates responses in
the human lateral geniculate nucleus. Nature Neuroscience, 5(11), 1203–1209.
Pashler, H. (1987). Detecting conjunctions of color and form: reassessing the serial search
hypothesis. Perception & Psychophysics, 41(3), 191–201.
Peters, R., Iyer, A., Itti, L., & Koch, C. (2005). Components of bottom-up gaze allocation in
natural images. Vision Research, 45(18), 1–20.
Purcell, B., Heitz, R., Cohen, J., Schall, J., Logan, G., & Palmeri, T. (2010). Neurally
constrained modeling of perceptual decision making. Psychological Review, 117(4),
1113–1143.
Raab, D. H. (1962). Statistical facilitation of simple reaction times. Transactions of the New York
Academy of Sciences, 24, 574–590.
Raymond, J., Shapiro, K., & Arnell, K. (1992). Temporary suppression of visual processing in
an RSVP task: an attentional blink? Journal of Experimental Psychology: Human Perception
and Performance, 18(3), 849–860.
Rensink, R. (2000). Seeing, sensing, and scrutinizing. Vision Research, 40(10–12), 1–19.
Scalf, P. E., & Beck, D. M. (2010). Competition in visual cortex impedes attention to multiple
items. Journal of Neuroscience, 30(1), 161–169.
Schall, J., Purcell, B., Heitz, R., Logan, G., & Palmeri, T. (2011). Neural mechanisms of sac-
cade target selection: gated accumulator model of the visual-motor cascade. European
Journal of Neuroscience, 33(11), 1991–2002.
Serences, J., & Boynton, G. (2007). Feature-based attentional modulations in the absence of
direct visual stimulation. Neuron, 55(2), 1–12.
Shapiro, K., Caldwell, J., & Sorensen, R. (1997). Personal names and the attentional blink:
a visual "cocktail party" effect. Journal of Experimental Psychology: Human Perception and
Performance, 23(2), 504–514.
Shiffrin, R. M., & Gardner, G. T. (1972). Visual processing capacity and attentional
control. Journal of Experimental Psychology, 93(1), 72–82.
Silvia, P., McCord, D., & Gendolla, G. (2010). Self-focused attention, performance expectan-
cies, and the intensity of effort: do people try harder for harder goals? Motivation and
Emotion, 34, 363–370.
Silvia, P., Jones, H., Kelly, C., & Zibaie, A. (2011). Trait self-focused attention, task difficulty,
and effort-related cardiovascular reactivity. International Journal of Psychophysiology: Official
Journal of the International Organization of Psychophysiology, 79(3), 1–6.
Simon, J., & Rudell, A. (1967). Auditory S-R compatibility: the effect of an irrelevant cue on
information processing. Journal of Applied Psychology, 51(3), 300–304.
Simon, J., Acosta, E., Mewaldt, S., & Speidel, C. (1976).The effect of an irrelevant directional
cue on choice reaction time: duration of the phenomenon and its relation to stages of
processing. Perception & Psychophysics, 19, 1–7.
Simons, D., & Chabris, C. (1999). Gorillas in our midst: sustained inattentional blindness for
dynamic events. Perception, 28(9), 1059–1074.
Simons, D., & Jensen, M. (2009). The effects of individual differences and task difficulty on
inattentional blindness. Psychonomic Bulletin and Review, 16(2), 398–403.
Simons, D., & Levin, D. T. (1998). Failure to detect changes to people during a real-world
interaction. Psychonomic Bulletin and Review, 5, 644–649.
Smallwood, J., Obonsawin, M., & Heim, D. (2003). Task unrelated thought: the role of
distributed processing. Consciousness and Cognition, 12(2), 169–189.
Smilek, D., Carriere, J. S.A., & Cheyne, J. A. (2010). Out of mind, out of sight: eye blink-
ing as indicator and embodiment of mind wandering. Psychological Science, 21(6),
786–789.
Smith,T., Ruiz, J., & Uchino, B. (2000).Vigilance, active coping, and cardiovascular reactivity
during social interaction in young men. Health psychology: Official Journal of the Division of
Health Psychology, American Psychological Association, 19(4), 382–392.
Sperling, G. (1960). The information available in brief visual presentations. Psychological
Monographs: General and Applied.
Strayer, D. L., & Drews, F. (2007). Cell-phone induced inattention blindness. Current
Directions in Psychological Science, 16, 128–131.
Strayer, D. L., Watson, J. M., & Drews, F. A. (2011). Cognitive distraction while multitasking
in the automobile. In B. Ross (Ed.), The psychology of learning and motivation (Vol. 54,
pp. 29–58). Burlington: Academic Press.
Stroop, J. (1935). Studies of interference in serial verbal reactions. Journal of Experimental
Psychology, 18(6), 643–662.
Temple, J., Warm, J., Dember, W., Jones, K., LaGrange, C., & Matthews, G. (2000). The effects
of signal salience and caffeine on performance, workload, and stress in an abbreviated
vigilance task. Human Factors, 42(2), 183–194.
Tinker, M. A. (1925). Intelligence in an intelligence test with an auditory distractor. American
Journal of Psychology, 36(3), 467–468.
Torralbo, A., & Beck, D. (2008). Perceptual-load-induced selection as a result of local com-
petitive interactions in visual cortex. Psychological Science, 19(10), 1045–1050.
Townsend, J. T. (1972). Some results concerning the identifiability of parallel and serial
processes. British Journal of Mathematical and Statistical Psychology, 25, 168–199.
Trautwein, U., Lüdtke, O., Roberts, B., Schnyder, I., & Niggli, A. (2009). Different forces,
same consequence: conscientiousness and competence beliefs are independent predic-
tors of academic effort and achievement. Journal of Personality and Social Psychology, 97(6),
1115–1128.
Treisman, A., & Geffen, G. (1967). Selective attention: perception or response? Quarterly
Journal of Experimental Psychology, 19(1), 1–17.
Treisman, A., & Gelade, G. (1980). A feature-integration theory of attention. Cognitive
Psychology, 12(1), 97–136.
Treisman, A. (1960). Contextual cues in selective listening. Quarterly Journal of Experimental
Tseng,Y. C., Glaser, J. I., Caddigan, E., & Lleras, A. Attentional decision making in pop-out visual
search, Submitted for publication.
von Mühlenen, A., & Lleras, A. (2007). No-onset looming motion guides spatial attention.
Journal of Experimental Psychology: Human Perception and Performance, 33, 1297–1310.
Wan, X., & Lleras, A. (2010).The effect of feature discriminability on the inter-trial i nhibition
of focused attention. Visual Cognition, 18, 920–944.
Watson, D. G., & Humphreys, G. W. (1997). Visual marking: prioritizing selection for new
objects by top-down attention inhibition of old objects. Psychological Review, 104,
90–122.
Wenger, M. J., & Townsend, J. T. (2000). Basic tools for attention and general processing
capacity in perception and cognition. Journal of General Psychology: Visual Attention,
127(1), 67–99.
Wolfe, J., Alvarez, G., Rosenholtz, R., Kuzmova, Y., & Sherman, A. (2011). Visual search
for arbitrary objects in real scenes. Attention, Perception & Psychophysics, 73(6),
1650–1671.
Wolfe, J.,Vo, M., Evans, K., & Greene, M. (2011).Visual search in scenes involves selective and
nonselective pathways. Trends in Cognitive Sciences, 15(2), 77–84.
Wolfe, J. (1994). Guided search 2.0: a revised model of visual search. Psychonomic Bulletin and
Review, 1(2), 202–238.
Woodman, G. F., & Arita, J. T. (2011). Direct electrophysiological measurement of attentional
templates in visual working memory. Psychological Science, 22(2), 212–215.
Yantis, S., & Egeth, H. (1999). On the distinction between visual salience and stimulus-driven
attentional capture. Journal of Experimental Psychology: Human Perception and Performance,
25(3), 661–676.
Yigit-Elliott, S., Palmer, J., & Moore, C. M. (2011). Distinguishing blocking from attenuation
in visual selective attention. Psychological Science, 22(6), 771–780.
Young, A. H., & Hulleman, J. (2012). Eye movements reveal how task difficulty moulds visual
search. Journal of Experimental Psychology: Human Perception and Performance.

INDEX
Note: Page numbers followed by “f ” and “t” indicate figures and tables respectively.
A Attentional control, 162

Abduction, 5, 22. See also Induction Augmented Multiparty Interaction (AMI),
causes and effects, 21 81–83
inconsistency, 22–23 Automata theory, 27–28
participants task, 22 Automaticity form, 242–243
putative explanations, 22
study of, 21 B
James’s remark, 22–23 Baseline effects. See Scaling effects
ADHD. See Attention-deficit hyperactivity Beck Depression Inventory (BDI),
disorder 175–176
Aging, 247 Behavioral mimicry
impairments, 248 interactive linguistic alignment, 70–71
individuals, 247–248 interlocutors, 70
in rodents, 247 mechanism of alignment, 71–72
Alternative interpretations, and relational perception–action link, 71
competition Bernstein’s analysis, of chisel and hammer,
competition among, 114 54–55
innovative interpretation, 114 Bernstein’s problem, 52
non-impaired individuals, 114–115 Biased competition theory, 267–268
AMI. See Augmented Multiparty
Interaction C
Analysis of variance (ANOVA), 105–106, 180 Candidates, 271b
Aristotelian–Thomistic view, features, CARIN theory. See Competition among
123–124 relations in nominals theory
Association hypothesis, 58–59 Carryover effects, 206
Attention theory CDP. See Continuous dual-process
inattentional blindness prediction Central nervous system (CNS), 204
performance prediction, 303 Chameleon effect, 70
principal challenges, 301–302 Classical cognitive approach, 50–51
self-focused attention trait, 302–303 Clinical disorders, 172
inner focus, need for Clinical populations
attention reorientation, 301 cognitive task, 172
characteristics, 300 context processing, 173
potential sources, 300–301 stop-signal paradigm, 172–173
visual attention, 303–304 CNS. See Central nervous system
flanker task, 306–307 Cognition building blocks, 98
parallel function, 304 Cognitive neuroscience, 47
serial aspect, 305 Cognitive processes, 83
visual search, 304–305 Color pop-out, 265–266
Attention-deficit hyperactivity disorder Competition among relations in nominals
(ADHD), 173–174 theory (CARIN theory), 101–102
317
318 Index
Complementarity Coordination process, 56. See also

interlocutors, 73 Complementarity
structured conversation, 72–73 high-level cognition, 69
Complex system linguistic coordination, 70
emerging models, 52 social coordination, 69
intersection system, 51–52 CV. See Context variability
naturalistic language performance, CVC trigrams. See Consonant–vowel–
52–53 consonant trigrams
Compound processing
constituent representations, 116 D
construction process, 115 Deduction, 4
lexical representation, 117 Delay effects
opaque compounds, 115–116 DF costs and benefits, 161
relational interpretations diversity, testing effect, 161
118–119 Depression, 175–176
semantic representation, 116–117 depressive rumination, 176–177
transparent compounds, 117–118 dysphorics, 175–176
Conceptual combination Depressive rumination, 176–177
CARIN theory, 101–102 Different-relation prime conditions,
evaluation process, 103 113–114
relational gist interpretation, 104 vs. same-relation prime condition, 113
relational structures, 102–103 Directed forgetting (DF), 133
RICE theory, 102, 102f impairment, 148–149
Conceptual composition, 98 changes in mental context, 148
modifier-noun phrases, 98–99 encoding strategy, 148
relational information process, 99 manipulations, 133
relational interpretations, 99–100 broad theory, 133–134
Consonant–vowel–consonant trigrams list-method, 133
(CVC trigrams), 206–207 Disjunction
Contemporary threshold theories, 196 double, 8
Context change, 139–140, 164–165 exclusive, 7
Context cues, 148 inclusive, 7
category cues vs., 154–155 Distractibility, 290
methodological differences, 155 Distraction, 262
Context variability (CV), 156 attention task, 298f–299f
forgetting impairment, 157f citation counts, 308f
low and high CV items, 156 effort calibration, 297–299
word frequency, 156–157 using eye-tracking methodology, 264
Continuous dual-process (CDP), 196 flanker effect, 263
Continuous models, 195–196, 195f. See also inattentional blindness paradigm, 264
Hybrid models math task analysis, 297f
Control parameter, 64–66 new paradigm
Controlled strategies cognitive task, 293–294
DF magnitude, 141 experimental task, 294
forget cue, 141–142 oculomotor capture, 295f
forgetting impairment, 142f participants, 294–296
importance, 141 performance and mental effort, 296–297
Index 319
unexpected event paradigms inattentional blindness

generalizability, 292–293 recruitment, 280–281
limitations, 291–292 unexpected event paradigms, 278–280
Distractor interference, 263, 273–274 visual search, 264
connecting flanker experiments attention, 264–265
cluttered displays, 289 set size, 264–265
distractibility, 290 DPSD theory. See Dual-process signal
using flanker task, 290–291 detection theory
high-efficiency workplace, 290–291 DSP. See Dual-solution paradigm
current state of confusion Dual-process account, 167
attention system, 288 L2 benefits, 168–169
flanker effect, 287–288 switch encoding strategies, 167
Lavie’s arguments, 288 two-factor account, 168
Lavie’s cognitive load proposal, Dual-process
285–286 models, 192–193
task-irrelevant distractor, 286–287 theories, 25, 196
visual search task, 287 Dual-process signal detection theory
via spatial attention, 288–289 (DPSD theory), 196
Distractors, 152, 262–263, 271 Dual-solution paradigm (DSP), 237, 244–245
using eye-tracking methodology, 264 Dysexecutive disorder, 173
flanker effect literature, 263 Dysphorics, 176
inattentional blindness paradigm, 264
Distractors causing distraction E
distraction interference Ecological paradigms, 236
attentional success measure, 288–289 behavioral measure, 237
confusion, 285–288 hippocampus and caudate, 236–237
connecting flanker experiments, human analogs, 239
289–291 human DSP, 237f
divided attention individual differences, 238, 238f
distraction decrement, 271–272 interest analysis regions, 239f
noise and targets, 272 Ecological validity, comparable, 233b–234b
redundancy gain, 271 Egocentric processes, 56
empirical study Encoding strategy, 148
critical stimulus, 281–282 Encoding strength
flanker-like congruency effect, context strength vs., 155
284–285 strengthening manipulations, 155–156
participants, 282–283 Event-related potential (ERP), 197–198
RT analysis, 283 waveforms, 206–207
task relevancy, 285 Exponential function, 111–112
unexpected event trial, 283 Extensional probability. See also Intensional
flanker effect probability
distraction measurement, 273–274 clause relation, 24, 24f
information processing tradition and equiprobability, 24
flanker task, 274–276 assumption of, 24
task-irrelevant, 276–278 principle, 24
traditional view, 272–273 extensional reasoning, 23
flat search slopes, 265–266 mental models, 24, 24f
320 Index
Extensional probability (Continued ) “just for practice”, 134

nonextensional reasoning, 23 list-method DF studies, meta-analysis of,
principle of indifference, 23 135, 146
probabilistic reasoning, 23 Forget-group participants, 141
red and blue marbles, 24, 24f Forgetting, 133
Extralist cued recall controlled strategies
associative strength, 158 DF magnitude, 141
stronger test cues, 158 forget cue, 141–142
targets, 157 forgetting impairment, 142f
importance, 141
F directed (DF). See Directed forgetting
Facilitation. See Inhibition (DF)
Familiar compound, 114–115 intentional, preexisting beliefs about,
Familiarity, 196, 198 144–146. See also Intentional
estimation, 198–199 forgetting
PDP, 199–201 remember group, 142–143
remember/know paradigm, 201–203 strategic decision, 140–141
First response functions, 137 controlled, 141
Flanker effect, 272–273 strategies, 146
distractor interference, 273–274 categories, 146
information processing tradition controlled behaviors, 146–148
distractors, 275 forgetting costs and benefits, 147f
task-irrelevant, 274–275 Four-list design, 136–137
task-irrelevancy, 276–277 Fragment completion test, 163
characteristics, 277 Full attention trial, 278–279
distractor stimuli, 278 Functional magnetic resonance imaging
distractors and targets, 277 (fMRI), 238
Flanker task, 275, 306–307
foils in, 286–288 G
information processing tradition and, GAD. See Generalized anxiety disorder
274–276 Generalized anxiety disorder (GAD), 175
congruent and incongruent distractors Graph knowledge, 229
in, 278
“noise” elements, 274–275 H
“task-irrelevant” modifier, 277 Haken–Kelso–Bunz model, 65
Flat search slopes, 265–266 Head-based relational information, 120
Floor effects High-frequency relation, 106–107
measurement error, 180 High-low threshold model, 194–195,
types of item, 180–181 194f
fMRI. See Functional magnetic resonance High-threshold model, 193–194
imaging Hormones, 250
Foil, 275, 276b Human interaction
For-loops, 29 centipede’s dilemma, 44
for computing minimal solutions, cognitive mechanisms, 44
31t–32t computational models, 46
Forget group consensus definitions, 44–45
DF, costs and benefits, 147f dynamical systems framework, 45
dual-process account, 146–148 generic expectations, 45
Index 321
linguistic interaction, 45–46 theories of reasoning, 6–7

multimodal coordination, 45 social exchange, 2–3
self-organization stringent experiments, 3
Bernstein’s analysis, 54–55 symbols
central controller, 53 affirmative possibilities, 10–11
degrees of freedom, 54 compound assertions, 11
process threading, 53–54 concept of negation, 11
synergies, 54 denial of assertions, 11
Human navigators, 242–243 denial of conjunction, 11
Human radial arm mazes, 235 denial of inclusive disjunction, 11f
eight-arm radial arm maze, 236f mental model of preceding assertion,
Human reasoning. See also Mental model 11f
theory; Reasoning for negation, 10–11
computationally intractable, 2 scope of negation, 10–11
icons theories of deduction, 2
alternative model, 10 verbal rules, 3
iconic representations, 9 Hybrid models, 196. See also Threshold-
mental model of assertion, 9 based models
order of mention, 10 dual-process model, 196–197
role, 10 recollection and familiarity processes,
transcend iconicity, 11 197–198
transitive inference, 9 temporal properties, 198
typical situation, 10 two-criterion model, 196–197
inferential machinery, 2–3 Hypothetical research scenario, 79–81
informal programs
informal algorithms, 33 I
Kolmogorov complexity, 30 Icons
loops for computing minimal alternative model, 10
solutions, 31t–32t iconic representations, 9
loops of operations, 33 mental model of assertion, 9
palindromes, 30 order of mention, 10
sorts of rearrangement, 30 role, 10
paradigm shift, 2–3 transcend iconicity, 11
possibilities of models, 6 transitive inference, 9
assertion, 6 Inattentional blindness, 264, 278–279
categorical assertion, 7 prediction
deliberations, 6–7 performance prediction, 303
double disjunctions, 8 principal challenges, 301–302
dual process, 6–7 self-focused attention trait, 302–303
exclusive disjunction, 7 recruitments
inclusive disjunction, 7 central limited resource, 280–281
indefinite number of possibilities, 6 distraction issue, 281
intuitions, 6–7 ensuing blindness, 281
mental models, 6f–7f unexpected event paradigms
modal errors, 8 distractibility, 279–280
side effect of, 8–9 participants, 278–279
spatial relations, 8 using superimposed videos, 279
systems, 6–7 Inclusion recognition, 152
322 Index
Inclusion test, 199 Intrusion errors

Induction, 4–5, 20, 23 source discrimination, 166
mariner’s mental model and reality, 21 source errors, 166
monotonic, 20–21 ITI. See Intertrial interval
nonmonotonic, 20–21
retractions, 20–21 J
Informal programs Joint probability distribution (JPD), 25
informal algorithms, 33 JOL. See Judgment of learning
Kolmogorov complexity, 30 JPD. See Joint probability distribution
loops Judgment of learning ( JOL), 143
for computing minimal solutions,
31t–32t K
of operations, 33 Knowledge, use of, 16–20
palindromes, 30 Kolmogorov complexity, 30
sorts of rearrangement, 30
Inhibition, 113, 138 L
modifier’s relational distribution, 113 L1 item retrieval
relational priming results, 113 fragment completion test, 163
Intensional probability part-set cuing, 162–163
arithmetical operations, 25 L2 learning
conditional probabilities, 26 DF impairment, 158–159
deliberative procedure, 27 first-order paradigms, 159
disjunctive probabilities, 26 inhibitory processes, 159
dual process theory, 25 retrieving
intuitive and deliberative systems, meta-analysis, 160
26–27 multiple retrieval trials, 160–161
JPD, 25 second-order paradigms, 159
loops of operations, 26 Landmark–route–survey framework, 239
model theory of probabilities, 27 adaptive approach, 246
negative probability in third conjunction, advantages and disadvantages
26–27, 26–27 automaticity form, 242–243
nonextensional reasoning, 25 constraints and features, 243
probabilities, 25 flexible survey knowledge, 242
values of probabilities in JPD, 25–26 interaction, 243
Intentional forgetting internal motivation, 244
forget cue, 145–146 using landmark, 244
forgetting impairment, 145f anecdotal and self-report data, 226–227
preexisting beliefs, 144–145 challenges
Interactional patterns graph knowledge, 229
speech turns, 74 knowledge stages, 228
task-related interactions, 74–75 navigational ability, 227–228
Interpersonal synergies, 72, 77 survey-based strategies, 228
complementarity, 72–73 classification vs. predicting solutions
interactional patterns, 74–75 DSP, 242
Interrogate memory, 224 navigational challenges, 241
Intertrial interval (ITI), 206 place- or response-learning
Intriguing, 28–29 mechanisms, 241–242
Index 323
data, 246f test orders, 134–135

foundation for individual differences two-list design, 134
classic classification, 241 GAD, 176–177
cognitive map and explicit learning, forgetting neutral words, 177
240–241 perseverative difficulties, 177
low-level differences, 240 multilist designs
one-to-one mapping, 240–241 four-list design, 136–137
navigational styles and strategies, 227 three-list design, 136
SBSOD, 245 nonclinical research, 181
spatial ability, 244–245 research designs, 178
in spatial cognition, 227 forgetting designs, 178t
substantial efforts, 245 remember control groups,
Landmarks, 227 178–179
Language processing, 61 separate remember condition, 178
Large-scale replication study Locus coeruleus–norepinephrine system
forgetting strategies, 169 (LC–NE system), 203–204
switch strategy, 169–170
LC–NE system. See Locus coeruleus– M
norepinephrine system Means-ends analysis, 28
Lexicalized compounds, 107–108 Memory capacity
Linguistic coordination, 70, 76f attentional control, 162
linguistic repetitions levels, 76 context encoding, 162
speech production, 77 Memory development beliefs during
synergy, 75–76 experiment
task-oriented conversations, 75 DF impairment, 144, 144f
testing models, 75 participants’ predictions, 143
Linguistic expressions, 98–99 Memory failures, 192
Linguistic interaction, 45–46 Mental context-change paradigm
cognition in, 49–50 behavioral effects, 149–150
dynamical systems framework, 50 forgetting, 149
Linguistic repetition levels, 76 list-length factor, 150
Linguistic representation, 100 Mental reinstatement
List-method directed forgetting context account, 153
manipulations (List-method DF episodic context, 154
manipulations), 133 memory formal theories, 154
causes Mental simulations, 29
episode, 139 automata theory, 27–28, 27f
inhibition, 138 intriguing, 28–29
retrieval inhibition account, 138 loop halts, 30
RIF, 139 means-ends analysis, 28
selective rehearsal account, 138 primitive recursion, 29
dependent measures, 137 protocol, 29
design issues sorts of rearrangement, 29
control group, 135–136 palindromes, 29
costs and benefits effects, 134 parity sorts, 29
output interference, 135 reversals, 29
R–F measure, 135–136 Metacognitive beliefs, 143
324 Index
Model theory, 3 Morris water maze (MWM), 232–234

as counterexamples, 14–15 Motivation, 233b–234b
arithmetical formula, 14–15 Multilist designs
consistent premises, 14 four-list design, 136–137
invalid conclusion, 14 three-list design, 136
possibility, 15, 15f Multimodal coordination, 45
putative conclusion, 13–14 MWM. See Morris water maze
right frontal pole, 15
set of premises, 13–14 N
essentials of, 3–4 Navigation, 224
principle of truth dominant framework, 225
biconditional assertion, 12f individual differences in, 224–225
cognitive illusions, 13 Navigational ability measurement
compound assertion, 12 challenges to landmark–route–survey
exclusive disjunction, 11 framework, 227–229
explicit models, 12–13, 13f differences
fallacies, 13 in navigational success rates, 225–226
first disjunction, 12, 12f in strategies/styles, 226–227
fully explicit, 12 Navigational success rates, differences in
second disjunction, 12, 12f–13f good/bad navigation ability, 225–226
Modification effect paradigm, 122–123 learning and flexibility, 226
Modifier-noun phrases, 98–99 Network analysis approach, 81, 82f
computation, 100–101 Neutral context, 106–107
construction process, 101
linguistic expressions, 100 O
Modifier’s relational distribution, 113 Observer’s sensitivity, 195–196
Modulation, 16–17. See also Human Opaque compounds, 115–116
reasoning Oscillators, 72–73
conclusion Output interference, 135
conditional, 18, 18f Output order, 179–180
disjunctive, 18 Own-race bias (ORB), 210–211
as possibility, 18, 18f
conditionals, 16, 16f P
construction of models, 17 Palindromes, 29
contrasting inference, 17 Parasympathetic nervous system, 204
different patterns of inference, 17 Parity sorts, 29
false clause, 16, 16f PDP. See Process-dissociation procedure
inference, 17–18 Perception–action link, 71
formal rule of, 19 Perspective-taking social modulation
interpretation, 17 cognitive demands, 68
effect of modulation, 18 communication goal, 67–68
premises, 19 data timing and response characteristics,
responses, 17 66–67, 67f
tacit modulation, 19 egocentric accounts, 62
temporal inferences, 19 ground information, 61
temporal modulation effect, 19 high-dimensional neural process, 65–66
transparent in logic, 20 interactions, 63
Index 325
interactive perceptual detection task, 68f face race and memory performance, 213f
low-dimensional dynamic process, 65f key criticisms, 200
moment-by-moment linguistic ORB, 211–212
processing, 62 participants, 199–200
social interaction, 62–63 recognition memory, 200–201
social variables, 60–61 recollection, 199
Phasic pupillary reflexes, 205 Property mapping. See Relation linking
Place learning, 230. See also Spatial learning Psychophysiological correlation of memory
in humans, 232 for faces, 210
ecological paradigms, 236–239 ORB, 210–211
human radial arm mazes, 235 PDP, 211
virtual water maze, 232–234 anecdotal experiences, 213–214
for individual differences, 239 Asian and White functions, 212
advantages and disadvantages, 242–244 baseline-corrected pupil diameter, 212f
classification vs. predicting solutions, cognitive resources, 212–213
241–242 face race and memory performance,
separating preference, 244–246 213f
systems balance as foundation, ORB, 211–212
240–241 RK procedure
rodent paradigms adaption, 233b–234b baseline-corrected pupil diameters,
Place systems 214f
characterization, 250–251 ORB, 214–215
competition or cooperation, 251 Pupillometry, 203–204
during encoding, 251 and memory
during fMRI scans, 251–252 CVC trigrams, 206–207
in individual differences, 250 fMRI and ERP investigations, 206
variability, 252 memory strength, 207–208
Principle of truth memory-based effects, 209–210
biconditional assertion, 12f phasic pupillary changes, 210
cognitive illusions, 13 production process, 208
compound assertion, 12 pupil diameter function, 209f
exclusive disjunction, 11 pupillary old/new effect, 208–209
explicit models, 12–13, 13f recollection, 207
fallacies, 13 TEPR investigations, 206
first disjunction, 12, 12f pupillary reflex
fully explicit, 12 LC–NE system, 203–204
second disjunction, 12, 12f–13f memory-based effects, 209–210
Process-based account neurophysiological evidence, 204
context change, 139–140 TEPRs
two-factor, 140f cognitive processing, 204–205
Process-dissociation procedure (PDP), phasic pupillary reflexes, 205
198–199, 211
anecdotal experiences, 213–214 Q
Asian and White functions, 212 QSR. See Questionnaire on Spatial
automatic processes, 199 Representation
baseline-corrected pupil diameter, 212f Questionnaire on Spatial Representation
cognitive resources, 212–213 (QSR), 227
326 Index
R Relational competition
Reasoning alternative interpretations
abduction, 5 competition, 114
assertion, 6 innovative interpretation, 114
categorical, 7 non-impaired individuals, 114–115
contradiction, 5 compound processing
deduction, 4 constituent representations, 116
deliberations, 6–7 construction process, 115
disjunction lexical representation, 117
double, 8 opaque compounds, 115–116
exclusive, 7 relational interpretations diversity,
inclusive, 7 118–119
dual process, 6–7 semantic representation, 116–117
indefinite number of possibilities, 6 transparent compounds, 117–118
induction, 4–5 inhibition
intuitions, 6–7 modifier’s relational distribution, 113
mental models, 6f–7f relational priming results, 113
modal errors, 8 nature of relations, 108
parsimonious conclusions, 5 conceptual representations, 108–109
possibilities, 6 relational abstraction level, 111
side effect of, 8–9 relations representation, 109–110
spatial relations, 8 numbers of competitors, 111–112
systems, 6–7 parallel process, 112–113
theories of, 6–7 stronger competitors, 111–112
theory of deductive competence, 5–6 Relational gist interpretation, 104, 121–122
Recognition memory, 192–193 Relational information, 98, 101–102
signal-detection models, 195f head-based, 120
single-process, 195f process, 99
Recognition tests Relational interpretation competitive
associative recognition, 152–153 evaluation theory (RICE theory), 102
DF null effects, 151 clay machine, 119–120
distractors, 152 head-based relational information, 120
inhibitory account, 151 relation verification, 120–121
item recognition, 153 semantic information, 119–120
memory tests, 150 stage, 119
meta-analysis, 151 Relational interpretations, 99–100
rehearsal process, 153 Relational priming effect, 108–109
Recollection, 196, 198 lexicalized compounds, 107–108
estimation, 198–199 novel combination, 107
process-dissociation procedure, 199–201 Relations
remember/know paradigm, 201–203 filler items, 110–111
Redundancy gain, 271 frequencies, 106
Regression analysis, 105–106 linking, 99
Reinterpreting clinical list-method relational effects, 109–110
ADHD, 173 selection
context, 173 elaboration process, 121
Relational abstraction level, 111 full noun phrases, 124
Index 327
head noun’s conceptual representation, during encoding, 251

123 during fMRI scans, 251–252
mechanistic merging, 122–123 in individual differences, 250
predication relation, 123–124 variability, 252
transference process, 121–122 Response-learning behaviors, 231
semantic priming effect, 110 Retrieval-induced forgetting (RIF), 139
suggestion Reversals, 29
comprehension, 105 RICE theory. See Relational interpretation
distributions, 105 competitive evaluation theory
modifier’s role in, 104 RIF. See Retrieval-induced forgetting
novel phrase, 107 RK paradigm. See Remember/know
phrase, 105–106 paradigm
prediction, 104–105 Rodent paradigms adaption for humans,
relation frequencies, 106 233b–234b
relational effects, 106–107
subjective familiarity, 106 S
verification, 120–121 Same-relation condition, 113–114
Remember group, 142–143 Santa Barbara Sense of Direction question-
benefits of DF, 134, 147f naire (SBSOD), 226, 245
cost of DF, 134, 147f SBP. See Systolic blood pressure
and “do nothing” people, 142–143 Scalability, 233b–234b
dual-process account, 146–148 Scaling effects, 180
list-method DF studies, meta-analysis of, Schizophrenia
135, 146 control group, 174–175
Remember/know paradigm (RK hallucinations, 175
paradigm), 201, 214–215 memory and inhibitory control, 174
baseline-corrected pupil diameters, 214f Self-focused attention, 297–299
dissociations, 201 Semantic priming effect, 110
hybrid CDP model, 202 Serial position effects, 165–166
interpretation, 201–202 Signal-detection model, 208–209
neuropsychological evidence, 202–203 Single-process models, 192–193
ORB, 214–215 Social coordination, 69
strength confound, 202–203 Social modulation
Response learning, 230 of cognitive dynamics, 56
in humans, 232 low-level visual attention
ecological paradigms, 236–239 association hypothesis, 58–59
human radial arm mazes, 235 cognitive approach, 58
virtual water maze, 232–234 context experiments, 59–60, 59f
for individual differences, 239 focus group, 58
advantages and disadvantages, 242–244 form of eye movements, 57–58
classification vs. predicting solutions, friction and social context, 57
241–242 social tuning, 59
separating preference, 244–246 perspective-taking
systems balance as foundation, 240–241 cognitive demands, 68
Response systems communication goal, 67–68
characterization, 250–251 data timing and response
competition or cooperation, 251 characteristics, 66–67, 67f
328 Index
Social modulation (Continued ) of inclusive disjunction, 11f

egocentric accounts, 62 mental model of preceding assertion, 11f
ground information, 61 for negation, 10–11
high-dimensional neural process, scope of negation, 10–11
65–66 Synergies, 54
interactions, 63 alignment and, 75–77
interactive perceptual detection task, 68f interpersonal, 72
low-dimensional dynamic process, 65f complementarity, 72–73
moment-by-moment linguistic interactional patterns, 74–75
processing, 62 Systolic blood pressure (SBP), 297–299,
social interaction, 62–63 298f–299f
social variables, 60–61
social constraint T
control parameter, 64–65 Task relevancy, 285
dynamical simulation, 63–64 Task-evoked papillary response (TEPR),
mental rotation functions, 64f 204–205
Spatial learning. See also Place learning cognitive processing, 204–205
dual systems phasic pupillary reflexes, 205
dual-solution T-maze, 230f Task-irrelevant distractors, 278
nonhuman animal models, 231 Task-irrelevant modifier
place and response learning, 230 characteristics, 277
rat, 229–230 distractor stimuli, 278
response-learning behaviors, 231 distractors and targets, 277
using temporary deactivation, 231 Task-relevant critical stimulus, 281–282
individual differences in, 224–225 TEPR. See Task-evoked papillary
navigational ability, 225 response
Spatial learning measurement Test order
challenges to landmark–route–survey DF benefits, 170
framework, 227–229 encoding strategy, 170
differences forgetting costs and benefits, 171f
in navigational success rates, 225–226 Theoretical debates. See also Human
in strategies/styles, 226–227 interaction
Stronger competitors, 111–112 cognitive mechanisms, 47
Structural priming. See Perception–action cognitive neuroscience, 47
link coordination, 46–47
Supportive context. See Neutral context interaction-dominant dynamic complex
Surface network analysis, 81–83 systems, 48–49
behaviors, 79 language, 47–48
hypothetical research scenario, 79–81, 80f theoretical integration, 48
multimodal models, 78–79 two-stage process, 46
Symbols Theory of mental models. See Model
affirmative possibilities, 10–11 theory
compound assertions, 11 Three-list design, 136
concept of negation, 11 Threshold-based models, 193–195. See also
denial Continuous models
of assertions, 11 high-low threshold model, 194f
of conjunction, 11 high-threshold model, 194f
Index 329
Time for more models, in self-organization hormones, 250

interaction, 77–78 using survey knowledge, 248–249
wholesale integration, 78 testosterone levels circulation, 249–250
Trial-by-trial basis, 75 Virtual water maze, 232
Two-criterion model, 196–197 MWM, 232–234
Two-high threshold model, 194–195 water maze preparation, 235f
Two-list design, 134 Visual attention, 303–304
flanker task, 306–307
U low-level
Unexpected event paradigms association hypothesis, 58–59
distractibility, 279–280 cognitive approach, 58
generalizability, 292–293 context experiments, 59–60, 59f
limitations, 291–292 focus group, 58
participants, 278–279 form of eye movements, 57–58
using superimposed videos, 279 friction and social context, 57
Unified theory, 33, 37. See also Mental models social tuning, 59
cardinality, 35 parallel function, 304
deliberative system, 36 serial aspect, 305
extensional representations, 34 visual search, 304–305
iconic model diagram, 34 Visual processing, 59–60, 268
inferences, 36 Visual search, 264
intensions of assertions, 34 distractors effect, 264–265
intuitive system, 35 experiments, 269–270
model updates, 35 impacts
spatial relations, 33–34 central issue, 266–267
syllogistic reasoning, 36 color pop-out, 265–266
preattentive vision, 267–268
V cognitive distractions, 268–269
Variability sources, connections to visual processing, 268
aging, 247 set size, 264–265
impairments, 248
individuals, 247–248 W
in rodents, 247 Water maze preparation, 235f. See also
hormonal influences, 248–249 Morris water maze (MWM);
estrogen infusions, 249 Virtual water maze
hormones, 250 While-loops, 29
using survey knowledge, 248–249 for computing minimal solutions,
testosterone levels circulation, 249–250 31t–32t
place/response model, 246–247 Word frequency, 156–157
sex differences during recognition test, 29, 209f
estrogen infusions, 249 Words, in interpreting, 6

CONTENTS OF PREVIOUS VOLUMES
VOLUME 40 Multimedia Learning

Richard E. Mayer
Different Organization of Concepts and
Meaning Systems in the Two Cerebral Memory Systems and Perceptual Categori-
Hemispheres zation
Dahlia W. Zaidel Thomas J. Palmeri and Marci A. Flanery
The Causal Status Effect in Categorization: Conscious Intentions in the Control of
An Overview Skilled Mental Activity
Woo-kyoung Ahn and Nancy S. Kim Richard A. Carlson
Remembering as a Social Process Brain Imaging Autobiographical Memory
Mary Susan Weldon Martin A. Conway, Christopher W.
Pleydell-Pearce, Sharon Whitecross, and
Neurocognitive Foundations of Human
Helen Sharpe
Memory
Ken A. Paller The Continued Influence of Misinforma-
tion in Memory: What Makes Correc-
Structural Influences on Implicit and
tions Effective?
Explicit Sequence Learning
Colleen M. Seifert
Tim Curran, Michael D. Smith, Joseph M.
DiFranco, and Aaron T. Daggy Making Sense and Nonsense of Experi-
ence: Attributions in Memory and
Recall Processes in Recognition Memory
Judgment
Caren M. Rotello
Colleen M. Kelley and Matthew G.
Reward Learning: Reinforcement, Incen- Rhodes
tives, and Expectations
Real-World Estimation: Estimation Modes
Kent C. Berridge
and Seeding Effects
Spatial Diagrams: Key Instruments in the Norman R. Brown
Toolbox for Thought
Index
Laura R. Novick
Reinforcement and Punishment in the
Prisoner’s Dilemma Game VOLUME 42
Howard Rachlin, Jay Brown, and Forest
Memory and Learning in Figure—Ground
Baker
Perception
Index Mary A. Peterson and Emily Skow-Grant
Spatial and Visual Working Memory: A
VOLUME 41 Mental Workspace
Robert H. Logie
Categorization and Reasoning in Relation
to Culture and Expertise Scene Perception and Memory
Douglas L. Medin, Norbert Ross, Scott Marvin M. Chun
Atran, Russell C. Burnett, and Sergey Spatial Representations and Spatial Updating
V. Blok Ranxiano Frances Wang
On the Computational basis of Learning Selective Visual Attention and Visual Search:
and Cognition: Arguments from LSA Behavioral and Neural Mechanisms
Thomas K. Landauer Joy J. Geng and Marlene Behrmann
331
332 Contents of Previous Volumes
Categorizing and Perceiving Objects: Explor- The Immersed Experiencer: Toward an

ing a Continuum of Information Use Embodied Theory of Language Com-
Philippe G. Schyns prehension
From Vision to Action and Action to Vision: Rolf A. Zwaan
A Convergent Route Approach to Speech Errors and Language Production:
Vision, Action, and Attention Neuropsychological and Connectionist
Glyn W. Humphreys and M. Jane Perspectives
Riddoch Gary S. Dell and Jason M. Sullivan
Eye Movements and Visual Cognitive Psycholinguistically Speaking: Some Matters
Suppression of Meaning, Marking, and Morphing
David E. Irwin Kathryn Bock
What Makes Change Blindness Interesting? Executive Attention, Working Memory
Daniel J. Simons and Daniel T. Levin Capacity, and a Two-Factor Theory of
Index Cognitive Control
Randall W. Engle and Michael J. Kane
Relational Perception and Cognition:
VOLUME 43 Implications for Cognitive Architecture
Ecological Validity and the Study of Concepts and the Perceptual-Cognitive Interface
Gregory L. Murphy Collin Green and John E. Hummel
Social Embodiment An Exemplar Model for Perceptual Catego-
Lawrence W. Barsalou, Paula M. Nied- rization of Events
inthal, Aron K. Barbey, and Jennifer A. Koen Lamberts
Ruppert On the Perception of Consistency
The Body’s Contribution to Language Yaakov Kareev
Arthur M. Glenberg and Michael P. Kas- Causal Invariance in Reasoning and Learning
chak Steven Sloman and David A. Lagnado
Using Spatial Language Index
Laura A. Carlson
In Opposition to Inhibition
Colin M. MacLeod, Michael D. Dodd, VOLUME 45
Erin D. Sheard, Daryl E. Wilson, and Exemplar Models in the Study of Natural
Uri Bibi Language Concepts
Evolution of Human Cognitive Architec- Gert Storms
ture Semantic Memory: Some Insights From
John Sweller Feature-Based Connectionist Attractor
Cognitive Plasticity and Aging Networks
Arthur F. Kramer and Sherry L. Willis Ken McRae
Index On the Continuity of Mind: Toward a
Dynamical Account of Cognition
Michael J. Spivey and Rick Dale
VOLUME 44 Action and Memory
Goal-Based Accessibility of Entities within Peter Dixon and Scott Glover
Situation Models Self-Generation and Memory
Mike Rinck and Gordon H. Bower Neil W. Mulligan and Jeffrey P. Lozito
Aging, Metacognition, and Cognitive Control Reconsidering the Role of Structure in

Christopher Hertzog and John Dunlosky Vision
The Psychopharmacology of Memory and Elan Barenholtz and Michael J. Tarr
Cognition: Promises, Pitfalls, and a Conversation as a Site of Category Learning
Methodological Framework and Category Use
Elliot Hirshman Dale J. Barr and Edmundo Kronmuller
Index Using Classification to Understand the
Motivation-Learning Interface
W. Todd Maddox, Arthur B. Markman,
VOLUME 46 and Grant C. Baldwin
The Role of the Basal Ganglia in Category Index
Learning
F. Gregory Ashby and John M. Ennis
Knowledge, Development, and Category VOLUME 48
Learning The Strategic Regulation of Memory Accu-
Brett K. Hayes racy and Informativeness
Concepts as Prototypes Morris Goldsmith and Asher Koriat
James A. Hampton Response Bias in Recognition Memory
An Analysis of Prospective Memory Caren M. Rotello and Neil A. Macmillan
Richard L. Marsh, Gabriel I. Cook, and What Constitutes a Model of Item-Based
Jason L. Hicks Memory Decisions?
Accessing Recent Events Ian G. Dobbins and Sanghoon Han
Brian McElree Prospective Memory and Metamemory:
SIMPLE: Further Applications of a Local The Skilled Use of Basic Attentional
Distinctiveness Model of Memory and Memory Processes
Ian Neath and Gordon D. A. Brown Gilles O. Einstein and Mark A. McDaniel
What is Musical Prosody? Memory is More Than Just Remembering:
Caroline Palmer and Sean Hutchins Strategic Control of Encoding, Access-
ing Memory, and Making Decisions
Index
Aaron S. Benjamin
The Adaptive and Strategic Use of Memory
by Older Adults: Evaluative Processing
VOLUME 47 and Value-Directed Remembering
Relations and Categories Alan D. Castel
Viviana A. Zelizer and Charles Tilly
Experience is a Double-Edged Sword: A
Learning Linguistic Patterns Computational Model of the Encoding/
Adele E. Goldberg Retrieval Trade-Off With Familiarity
Understanding the Art of Design: Tools for Lynne M. Reder, Christopher Paynter,
the Next Edisonian Innovators Rachel A. Diana, Jiquan Ngiam, and
Kristin L. Wood and Julie S. Linsey Daniel Dickison
Categorizing the Social World: Affect, Moti- Toward an Understanding of Individual Dif-
vation, and Self-Regulation ferences In Episodic Memory: Modeling
Galen V. Bodenhausen, Andrew R. Todd, The Dynamics of Recognition Memory
and Andrew P. Becker Kenneth J. Malmberg
Memory as a Fully Integrated Aspect of Protected Values and Omission Bias as

Skilled and Expert Performance Deontological Judgments
K. Anders Ericsson and Roy W. Roring Jonathan Baron and Ilana Ritov
Index Attending to Moral Values
Rumen Iliev, Sonya Sachdeva, Daniel M.
Bartels, Craig Joseph, Satoru Suzuki,
VOLUME 49 and Douglas L. Medin
Short-term Memory: New Data and a Noninstrumental Reasoning over Sacred
Model Values: An Indonesian Case Study
Stephan Lewandowsky and Simon Farrell Jeremy Ginges and Scott Atran
Theory and Measurement of Working Development and Dual Processes in Moral
Memory Capacity Limits Reasoning: A Fuzzy-trace Theory
Nelson Cowan, Candice C. Morey, Approach
Zhijian Chen, Amanda L. Gilchrist, and Valerie F. Reyna and Wanda Casillas
J. Scott Saults Moral Identity, Moral Functioning, and the
What Goes with What? Development of Development of Moral Character
Perceptual Grouping in Infancy Darcia Narvaez and Daniel K. Lapsley
Paul C. Quinn, Ramesh S. Bhatt, and “Fools Rush In”: AJDM Perspective on the
Angela Hayden Role of Emotions in Decisions, Moral
Co-Constructing Conceptual Domains and Otherwise
Through Family Conversations and Terry Connolly and David Hardman
Activities Motivated Moral Reasoning
Maureen Callanan and Araceli Valle Peter H. Ditto, David A. Pizarro, and
The Concrete Substrates of Abstract Rule Use David Tannenbaum
Bradley C. Love, Marc Tomlinson, and In the Mind of the Perceiver: Psychological
Todd M. Gureckis Implications of Moral Conviction
Ambiguity, Accessibility, and a Division of Christopher W. Bauman and Linda J. Skitka
Labor for Communicative Success Index
Victor S. Ferreira
Lexical Expertise and Reading Skill
Sally Andrews VOLUME 51
Index Time for Meaning: Electrophysiology
Provides Insights into the Dynamics
of Representation and Processing in
VOLUME 50 Semantic Memory
Causal Models: The Representational Infra- Kara D. Federmeier and Sarah Laszlo
structure for Moral Judgment Design for a Working Memory
Steven A. Sloman, Philip M. Fernbach, Klaus Oberauer
and Scott Ewing When Emotion Intensifies Memory Inter-
Moral Grammar and Intuitive Jurispru- ference
dence: A Formal Model of Unconscious Mara Mather
Moral and Legal Knowledge Mathematical Cognition and the Problem
John Mikhail Size Effect
Law, Psychology, and Morality Mark H. Ashcraft and Michelle M.
Kenworthey Bilz and Janice Nadler Guillaume
Highlighting: A Canonical Experiment The Self-Organization of Cognitive Structure

John K. Kruschke James A. Dixon, Damian G. Stephen,
The Emergence of Intention Attribution in Rebecca Boncoddo, and Jason Anastas
Infancy Index
Amanda L. Woodward, Jessica A. S ommer-
ville, Sarah Gerson, Annette M. E.
Henderson, and Jennifer Buresh
VOLUME 53
Adaptive Memory: Evolutionary Con-
Reader Participation in the Experience of
straints on Remembering
Narrative
James S. Nairne
Richard J. Gerrig and Matthew E.
Jacovina Digging into Déà Vu: Recent Research on
Possible Mechanisms
Aging, Self-Regulation, and Learning from
Alan S. Brown and Elizabeth J. Marsh
Text
Elizabeth A. L. Stine-Morrow and Lisa M. Spacing and Testing Effects: A Deeply Criti-
S. Miller cal, Lengthy, and At Times Discursive
Review of the Literature
Toward a Comprehensive Model of Com-
Peter F. Delaney, Peter P. J. L. Verkoeijen,
prehension
and Arie Spirgel
Danielle S. McNamara and Joe Magliano
How One’s Hook Is Baited Matters for
Index
Catching an Analogy
Jeffrey Loewenstein
VOLUME 52 Generating Inductive Inferences: Premise
Naming Artifacts: Patterns and Processes Relations and Property Effects
Barbara C. Malt John D. Coley and Nadya Y.Vasilyeva
Causal-Based Categorization: A Review From Uncertainly Exact to Certainly Vague:
Bob Rehder Epistemic Uncertainty and Approxi-
mation in Science and Engineering
The Influence of Verbal and Nonverbal Pro-
Problem Solving
cessing on Category Learning
Christian D. Schunn
John Paul Minda and Sarah J. Miles
Event Perception: ATheory and Its Applica-
The Many Roads to Prominence: Under-
tion to Clinical Neuroscience
standing Emphasis in Conversation
Jeffrey M. Zacks and Jesse Q. Sargent
Duane G. Watson
Two Minds, One Dialog: Coordinating
Defining and Investigating Automaticity in
Speaking and Understanding
Reading Comprehension
Susan E. Brennan, Alexia Galati, and Anna
Katherine A. Rawson
K. Kuhlen
Rethinking Scene Perception: A Multi-
Retrieving Personal Names, Referring
source Model
Expressions, and Terms of Address
Helene Intraub
Zenzi M. Griffin
Components of Spatial Intelligence
Index
Mary Hegarty
Toward an Integrative Theory of Hypothesis
Generation, Probability Judgment, and VOLUME 54
Hypothesis Testing Hierarchical Control of Cognitive Processes:
Michael Dougherty, Rick Thomas, and The Case for Skilled Typewriting
Nicholas Lange Gordon D. Logan and Matthew J. C. Crump
Cognitive Distraction While Multitasking in There Is Nothing So Practical as a Good

the Automobile Theory
David L. Strayer, Jason M. Watson, and Robert S. Siegler, Lisa K. Fazio, and Aryn
Frank A. Drews Pyke
Psychological Research on Joint Action: The Power of Comparison in Learning
Theory and Data and Instruction: Learning Outcomes
Günther Knoblich, Stephen Butterfill, and Supported by Different Types of
Natalie Sebanz Comparisons
Self-Regulated Learning and the Allocation Bethany Rittle-Johnson and Jon R. Star
of Study Time The Role of Automatic, Bottom-Up Pro-
John Dunlosky and Robert Ariel cesses: In the Ubiquitous Patterns of
The Development of Categorization Incorrect Answers to Science Questions
Vladimir M. Sloutsky and Anna V. Fisher Andrew F. Heckler
Systems of Category Learning: Fact or Fan- Conceptual Problem Solving in Physics
tasy? Jose P. Mestre, Jennifer L. Docktor, Natalie
Ben R. Newell, John C. Dunn, and E. Strand, and Brian H. Ross
Michael Kalish Index
Abstract Concepts: Sensory-Motor Ground-
ing, Metaphors, and Beyond
Diane Pecher, Inge Boo, and Saskia Van VOLUME 56
Dantzig Distinctive Processing: The Co-action of
Thematic Thinking: The Apprehension and Similarity and Difference in Memory
Consequences of Thematic Relations R. Reed Hunt
Zachary Estes, Sabrina Golonka, and Lara Retrieval-Induced Forgetting and Inhibi-
L. Jones tion: A Critical Review
Index Michael F.Verde
False Recollection: Empirical Findings and
Their Theoretical Implications
VOLUME 55 Jason Arndt
Ten Benefits of Testing and Their Applica- Reconstruction from Memory in Naturalis-
tions to Educational Practice tic Environments
Henry L. Roediger III, Adam L. Putnam Mark Steyvers and Pernille Hemmer
and Megan A. Smith Categorical Discrimination in Humans and
Cognitive Load Theory Animals: All Different and Yet the Same?
John Sweller Edward A. Wasserman and Leyre Castro
Applying the Science of Learning to Multi- How Working Memory Capacity Affects
media Instruction Problem Solving
Richard E. Mayer Jennifer Wiley and Andrew F. Jarosz
Incorporating Motivation into a Theoretical Juggling Two Languages in One Mind:What
Framework for Knowledge Transfer Bilinguals Tell Us About Language Pro-
Timothy J. Nokes and Daniel M. Belenky cessing and its Consequences for Cog-
On the Interplay of Emotion and Cognitive nition
Control: Implications for Enhancing Judith F. Kroll, Paola E. Dussias, Cari A.
Academic Achievement Bogulski and Jorge R.Valdes Kroff
Sian L. Beilock and Gerardo Ramirez Index
VOLUME 57 VOLUME 58
Meta-Cognitive Myopia and the Dilemmas Learning Along With Others
of Inductive-Statistical Inference Robert L. Goldstone, Thomas N. Wisdom,
Klaus Fiedler Michael E. Roberts, Seth Frey
Relations Between Memory and Reasoning Space, Time, and Story
Evan Heit, Caren M. Rotello and Brett Barbara Tversky, Julie Heiser, Julie Morrison
K. Hayes The Cognition of Spatial Cognition: Domain-
The Visual World in Sight and Mind: How General within Domain-specific
Attention and Memory Interact to Holly A. Taylor, Tad T. Brunyé
Determine Visual Experience Perceptual Learning, Cognition, and Expertise
James R. Brockmole, Christopher C. Philip J. Kellman, Christine M. Massey
Davoli and Deborah A. Cronin
Causation, Touch, and the Perception of
Spatial Thinking and STEM Education: Force
When, Why, and How? Phillip Wolff, Jason Shepard
David H. Uttal and Cheryl A. Cohen
Categorization as Causal Explanation: Dis-
Emotions During the Learning of Difficult counting and Augmenting in a Bayesian
Material Framework
Arthur C. Graesser and Sidney D’Mello Daniel M. Oppenheimer, Joshua B.
Specificity and Transfer of Learning Tenenbaum, Tevye R. Krynski
Alice F. Healy and Erica L. Wohldmann Individual Differences in Intelligence and
What Do Words Do? Toward a Theory of Working Memory: A Review of Latent
Language-Augmented Thought Variable Models
Gary Lupyan Andrew R. A. Conway, Kristof Kovacs
Index Index

PLM 59

Uploaded by

Copyright:

Available Formats

PLM 59

Uploaded by

Document Information

Original Title

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

PLM 59

Uploaded by

Copyright:

Available Formats

VOLUME FIFTY NINE

AMSTERDAM • BOSTON • HEIDELBERG • LONDON

Copyright © 2013 Elsevier Inc. All rights reserved.

No part of this publication may be reproduced, stored in a retrieval system or transmitted

For information on all Academic Press publications

Printed and bound in USA

1. Toward a Unified Theory of Reasoning  1

2. The Self-Organization of Human Interaction  43

3. Conceptual Composition: The Role of Relational Competition in

3.  Theoretical Framework: A Three-Stage Theory of Conceptual Combination  101

4. List-Method Directed Forgetting in Cognitive and Clinical

5. Recollection is Fast and Easy: Pupillometric Studies of Face

6. A Mechanistic Approach to Individual Differences in Spatial

7. When Do the Effects of Distractors Provide a Measure of

Toward a Unif ied Theory of

Psychology of Learning and Motivation, Volume 59 © 2013 Elsevier Inc.

abandon norms in favor of descriptions (Evans, 2012). One revenant is that

increase information. But semantics should not be confused with epistemol-

alternative to a premise. So, a theory of deductive competence—of what the

The left-to-right axis of the model corresponds to the left-to-right axis

cannot carry out recursive processes, including arithmetical operations such

conclusions that throw semantic information away by adding disjunctive

4. ICONS AND SYMBOLS

5. THE PRINCIPLE OF TRUTH

It has two mental models:

The participants read a statement of the problem, then three premises,

7. MODULATION AND THE USE OF KNOWLEDGE

Therefore, if Luisa played music then she didn’t play soccer.

8. INDUCTION AND ABDUCTION

at the heart of human performance is the abduction of explanations that

The pilot fell into a deep snowdrift and so wasn’t hurt.

which instead of explaining the inconsistency, they rate which assertion

9. PROBABILITIES: EXTENSIONAL AND INTENSIONAL

Hence, the equiprobability principle predicts correctly that individuals will

Turning to nonextensional reasoning, consider this question about a

10. MENTAL SIMULATIONS AND INFORMAL

on knowledge of programming languages. Hence, our studies focused

For-loops Lisp Informal English

For-loops Lisp Informal English

P. N. Johnson-Laird and Sangeet S. Khemlani

Our experiment corroborated the predictions. Individuals were able to

11. TOWARD A UNIFIED THEORY

All A are B A B (A is included in B.)

Some A are B A B (Intersection of A and

Some A are not B A−B (Set of A that are not

Most A are B |A B|> |A− B| (Cardinality of intersection >

More than half the |A B|> |A|/2 (Cardinality of intersection >

A corollary is that determiners, such as “most”, have parameters specifying

Actor Baker Colleague

Actor Baker Colleague

The implementation of the unified theory has so far focused on unique

The Self-Organization of Human

Psychology of Learning and Motivation, Volume 59 © 2013 Elsevier Inc.

1. INTRODUCTION: THE “CENTIPEDE’S DILEMMA” OF

or even take on similar bodily postures and movements. Through explora-

many behaviors involved in interaction, at several levels of analysis. We look

2. AN EXAMPLE THEORETICAL DEBATE AND THE

integration, adaptation, and process flexibility. In the study of basic cognitive

3. SELF-ORGANIZATION AND HUMAN INTERACTION

1. Toward a Unified Theory of Reasoning 1

2. The Self-Organization of Human Interaction 43

3. Conceptual Composition: The Role of Relational Competition in

3. Theoretical Framework: A Three-Stage Theory of Conceptual Combination 101

4. List-Method Directed Forgetting in Cognitive and Clinical

5. Recollection is Fast and Easy: Pupillometric Studies of Face

6. A Mechanistic Approach to Individual Differences in Spatial

7. When Do the Effects of Distractors Provide a Measure of