The Psychology of Counterfactual Thinking (International Series in Social Psychology)

416.qxd

9/29/2006

2:33 PM

Page 1

Batch number: 1

CIRCULATED

Date:

SEEN BY DESK EDITOR: REVISE NEEDED

Initial: Date:

APPROVED FOR PRESS BY DESK EDITOR

Initial: Date:

Back board: ❑ ISBN ❑ Barcode

THE PSYCHOLOGY OF COUNTERFACTUAL THINKING Edited by David R. Mandel, Denis J. Hilton, and Patrizia Catellani

CHECKLIST (must be completed before press) (Please cross through any items that are not applicable) Front board: Spine: ❑ Title ❑ Title ❑ Subtitle ❑ Subtitle ❑ Author/edited by ❑ Author/edited by ❑ Series title ❑ Extra logo if required ❑ Extra logo if required General: ❑ Book size ❑ Type fit on spine

The Psychology of Counterfactual Thinking

Edited by David R. Mandel, Denis J. Hilton, and Patrizia Catellani

ISBN 978-0-415-32241-6

,!7IA4B

Routledge research international series in social psychology www.routledge.com ï an informa business

PC4

Royal

Demy

B-format

Spine back edge

The Psychology of Counterfactual Thinking

People have a penchant for thinking about how things that matter to them might have turned out differently – either for the better or for the worse. For the past two decades psychologists have been intrigued by this phenomenon, which they call counterfactual thinking. Specifically, researchers have sought to answer “big” questions like: Why do people have such a strong propensity to generate counterfactuals, and what functions does counterfactual thinking serve? What are the determinants of counterfactual thinking, and what are its adaptive and psychological consequences? The Psychology of Counterfactual Thinking brings together a collection of thought-provoking papers by social and cognitive psychologists who have made important theoretical and empirical contributions to our understanding of this topic. The essays in this volume contain novel theoretical insights and, in many cases, descriptions of previously unpublished empirical studies. The Psychology of Counterfactual Thinking will provide an excellent overview of this fascinating topic for researchers, as well as advanced undergraduates and graduates in psychology, particularly those with an interest in social cognition, social judgment, decision making, thinking, and reasoning. David R. Mandel is a Defence Scientist with the Department of National Defence in Canada and an adjunct Associate Professor of Psychology at the University of Toronto. His areas of research expertise include thinking and reasoning, judgment and decision making, and social cognition. Denis J. Hilton is Professor of Social Psychology at the University of Toulouse-II. His research interests include social cognition, reasoning, judgment, and experimental economics. Patrizia Catellani is Professor of Social Psychology at the Catholic University of Milan, Italy. Her research is focused on the area of cognitive social psychology, with particular emphasis on applications to the political and judicial contexts.

Routledge research international series in social psychology Edited by W. Peter Robinson University of Bristol, UK

This series represents a showcase for both the latest cutting-edge research in the field, and important critiques of existing theory. International in scope, and directed at an international audience, applied topics are well represented. Social psychology is defined broadly to include related areas from social development to the social psychology of abnormal behaviour. The series is a rich source of information for advanced students and researchers alike. Routledge is pleased to invite proposals for new books in the series. In the first instance, any interested authors should contact Professor W. Peter Robinson, Department of Experimental Psychology, University of Bristol, 8 Woodland Road, Bristol BS8 1TN. E-mail: [email protected]. Routledge research international series in social psychology 1 Cooperation in Modern Society Promoting the welfare of communities, states and organizations Edited by Mark van Vugt, Mark Snyder, Tom R. Tyler and Anders Biel 2 Youth and Coping in Twelve Nations Surveys of 18–20-year-old young people Edited by Janice Gibson-Cline 3 Responsibility The many faces of a social phenomenon Hans-Werner Bierhoff and Ann Elisabeth Auhagen 4 The Psychological Origins of Institutionalized Torture Mika Haritos-Fatouros 5 A Sociocognitive Approach to Social Norms Edited by Nicole Dubois 6 Human Rights as Social Representations Willem Doise

7 The Microanalysis of Political Communication Claptrap and ambiguity Peter Bull 8 The Justice Motive in Adolescence and Young Adulthood Origins and consequences Edited by Claudia Dalbert and Hedvig Sallay 9 The Psychology of Counterfactual Thinking Edited by David R. Mandel, Denis J. Hilton, and Patrizia Catellani Also available in International Series in Social Psychology, now published by Psychology Press Children as Consumers A psychological analysis of the young people’s market Barrie Gunter and Adrian Furnham Adjustment of Adolescents Cross-cultural similarities and differences Ruth Scott and William Scott Social Psychology and Education Pam Maras Making Sense of Television The psychology of audience interpretation Sonia Livingstone Stereotypes During the Decline and Fall of Communism Gyorgy Hunyady Understanding the Older Consumer The grey market Barrie Gunter Adolescence: From Crisis to Coping A thirteen nation study Edited by Janice Gibson-Cline Changing European Identities Social psychological analyses of social change Edited by Glynis M. Breakwell and Evanthia Lyons

Social Groups and Identities Developing the legacy of Henri Tajfel Edited by Peter W. Robinson Assertion and its Social Context Keithia Wilson and Cynthia Gallois Children’s Social Competence in Context The contributions of family, school and culture Barry H. Schneider Emotion and Social Judgements Edited by Joseph P. Forgas Game Theory and its Applications In the social and biological sciences Andrew M. Colman Genius and Eminence Edited by Robert S. Albert The Psychology of Gambling Michael Walker Social Dilemmas Theoretical issues and research findings Edited by Wim Liebrand, David Messick and Henk Wilke The Theory of Reasoned Action Its application to AIDS-preventive behaviour Edited by Deborah Terry, Cynthia Gallois and Malcolm McCamish The Economic Psychology of Everyday Life Paul Webley, Carole B. Burgoyne, Stephen E.G. Lea and Brian M. Young Personal Relationships Across the Lifespan Patricia Noller, Judith Feeney and Candida Peterson Language in Action: Psychological Models of Conversation William Turnbull Rival Truths: Common Sense and Social Psychological Explanations in Health and Illness Lindsay St Claire

The Psychology of Counterfactual Thinking

Edited by David R. Mandel, Denis J. Hilton, and Patrizia Catellani

First published 2005 by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN Simultaneously published in the USA and Canada by Routledge 270 Madison Ave, New York, NY 10016 Routledge is an imprint of the Taylor & Francis Group © 2005 David R. Mandel, Denis J. Hilton and Patrizia Catellani selection and editorial matter; the contributors their contributions Typeset in Garamond by Wearset Ltd, Boldon, Tyne and Wear Printed and bound in Great Britain by MPG Books Ltd, Bodmin All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Library of Congress Cataloging in Publication Data A catalog record for this book has been requested ISBN 0-415-32241-3

Contents

List of figures List of tables List of contributors

ix x xi

Introduction

1

PART I

Counterfactuals, causality, and mental representation 1 Counterfactual and causal explanation: from early theoretical views to new frontiers

9 11

DAVID R. MANDEL

2 The relation between counterfactual and causal reasoning

28

BARBARA A. SPELLMAN, ALEXANDRA P. KINCANNON, AND STEPHEN J. STOSE

3 The course of events: counterfactuals, causal sequences, and explanation

44

DENIS J. HILTON, JOHN L. McCLURE, AND BEN R. SLUGOSKI

4 The mental representation of what might have been

61

CLARE R. WALSH AND RUTH M. J. BYRNE

PART II

Functional bases of counterfactual thinking 5 Reflective and evaluative modes of mental simulation KEITH D. MARKMAN AND MATTHEW N. MCMULLEN

75 77

viii Contents 6 Scenario simulations in learning: forms and functions at the individual and organizational levels

94

SUSANA SEGURA AND MICHAEL W. MORRIS

7 Finding meaning from mutability: making sense and deriving significance through counterfactual thinking

110

ADAM D. GALINSKY, KATIE A. LILJENQUIST, LAURA J. KRAY, AND NEAL J. ROESE

PART III

Counterfactual thinking and emotion 8 When a small difference makes a big difference: counterfactual thinking and luck

127 129

KARL HALVOR TEIGEN

9 On the comparative nature of regret

147

MARCEL ZEELENBERG AND ERIC VAN DIJK

PART IV

Counterfactual thinking in the context of crime, justice, and political history

163

10 Escape from reality: prisoners’ counterfactual thinking about crime, justice, and punishment

165

MANDEEP K. DHAMI, DAVID R. MANDEL, AND KAREN A. SOUZA

11 When the social context frames the case: counterfactuals in the courtroom

183

PATRIZIA CATELLANI AND PATRIZIA MILESI

12 Theory- versus imagination-driven thinking about historical counterfactuals: are we prisoners of our preconceptions?

199

PHILIP E. TETLOCK AND ERIKA HENIK

References Author index Subject index

217 245 247

Figures

1.1 1.2 2.1 5.1 5.2 6.1 8.1 10.1 10.2 10.3 10.4 11.1 12.1

Confusion matrix representing implications of factual and counterfactual cases for a hypothesis about an actual generative cause Confusion matrix representing implications of factual and counterfactual cases for a hypothesis about a foregone inhibitory cause Illustration of the relation between counterfactual and causal reasoning The interaction between simulation direction and mode Mediator model of simulation mode, affect, and task persistence Experiential learning framework Fortune wheels presented in a study of luck and closeness to failure Mean self-blame assigned as a function of thinking focus and stage Mean emotional intensity as a function of emotion type and thinking focus Mean intensity of anger as a function of upward counterfactuals and stage Mean perception of fairness as a function of upward counterfactuals about trial and sentence Mean proportion of counterfactuals focused on the rape victim’s nonconforming behaviors as a function of source trustworthiness Inevitability and impossibility curves from the “rise of the West” scenario

18 19 30 78 84 97 131 171 173 176 177 196 214

Tables

2.1 3.1 8.1 8.2 8.3 8.4 10.1 11.1

Results from the causality, counterfactual reasoning, and regret study Five kinds of causal chain Peer ratings of outcome attractiveness and counterfactual closeness of thirty “good luck” and thirty “bad luck” stories Peer ratings of luck, outcome attractiveness, and counterfactual closeness of thirty-eight “dangerous” stories Peer ratings of luck, outcome attractiveness, and counterfactual closeness of thirty-eight “careless” behaviors Percentage of respondents assigning statements to “lucky” and “unlucky” stories Prisoners’ counterfactuals about how things could have turned out better and fairer Counterfactuals focused on the rape victim’s nonconforming behaviors as a function of stereotype endorsement

40 50 134 137 138 142 179 191

Contributors

Ruth M.J. Byrne, Professor, Department of Psychology, University of Dublin, Trinity College, Ireland. Patrizia Catellani, Professor, Department of Psychology, Catholic University of Milan, Italy. Mandeep K. Dhami, Lecturer, Institute of Criminology, University of Cambridge, UK. Adam D. Galinsky, Assistant Professor, Kellogg School of Management, Northwestern University, Evanston, Illinois, USA. Erika Henik, Doctoral candidate, Walter A. Haas School of Business, University of California, Berkeley, California, USA. Denis J. Hilton, Professor, Department of Psychology, University of Toulouse II, France. Alexandra P. Kincannon, Ph.D., Department of Psychology, University of Virginia, Charlottesville, Virginia, USA. Laura J. Kray, Assistant Professor, Haas School of Business, University of California, Berkeley, California, USA. Katie A. Liljenquist, Doctoral candidate, Kellogg School of Management, Northwestern University, Evanston, Illinois, USA. David R. Mandel, Defence Scientist, Judgment and Decision Making Group, Command Effectiveness and Behaviour Section, Defence Research and Development Canada, Toronto, Canada. Keith D. Markman, Assistant Professor, Department of Psychology, Ohio University, Athens, Ohio, USA. John L. McClure, Associate Professor, Department of Psychology, Victoria University of Wellington, New Zealand. Matthew N. McMullen, Associate Professor, Department of Psychology, Montana State University, Billings, Montana, USA.

xii Contributors Patrizia Milesi, Assistant Professor, Department of Psychology, Catholic University of Milan, Italy. Michael W. Morris, Professor, Graduate School of Business, Columbia University, New York, USA. Neal J. Roese, Associate Professor, Department of Psychology, University of Illinois, Urbana, Illinois, USA. Susana Segura, Assistant Professor, Department of Psychology, University of Malaga, Spain. Ben R. Slugoski, Associate Professor, Department of Psychology, James Cook University, Townsville, Queensland, Australia. Karen A. Souza, Master’s candidate, Department of Psychology, University of Victoria, British Columbia, Canada. Barbara A. Spellman, Associate Professor, Department of Psychology, University of Virginia, Charlottesville, Virginia, USA. Stephen J. Stose, Doctoral candidate, Department of Psychology, University of Virginia, Charlottesville, Virginia, USA. Karl Halvor Teigen, Professor, Department of Psychology, University of Oslo, Norway. Philip E. Tetlock, Lorraine Tyson Mitchell Professor of Leadership, Walter A. Haas School of Business, University of California, Berkeley, California, USA. Eric van Dijk, Professor of Social Psychology, Department of Social and Organizational Psychology, Leiden University, Netherlands. Clare R. Walsh, Postdoctoral Research Associate, Department of Cognitive and Linguistic Sciences, Brown University, Providence, Rhode Island, USA. Marcel Zeelenberg, Professor of Social Psychology, Department of Economic and Social Psychology, Tilburg University, Netherlands.

Introduction

One of the great debates in the history of science concerned the reversibility of time. Classical dynamicists such as Isaac Newton viewed time as essentially reversible. Simultaneously invert the velocities of all the parts of a system and the system goes backwards in time, much like playing a movie in reverse creates the visual impression of backward time travel. The science of heat, thermodynamics, told a different story. According to the second law, the entropy of the universe was always increasing. The universe as a whole was aging irreversibly, and much like the constituents of a burnt piece of paper could never be reassembled into a clean, white sheet, time cannot go backwards either in practice or in principle. In many ways, people’s subjective experience of time reconciles these disparate perspectives. Most individuals (at least in Western cultures) are aware of a forward direction to “real” time, moving from the past through the present to the future. In this type of real time, people are always literally “in the present.” However, in the present, people can do a remarkable thing. They can travel forwards or backwards in subjective time. Indeed, even at the present moment, one can imagine oneself in the past thinking about how some event in the further past might have been different! Likewise, one can imagine oneself in the future replaying a “past” episode that, in fact, has never happened yet. According to Endel Tulving, this ability for “mental time travel” is largely due to the fact that humans have episodic memory, a capacity to recollect events experienced in one’s past. Without this cognitive capacity, Tulving has argued, humans would be unable to form a stable concept of “the self” over time and, even more significantly, the myriad of human cultures that have existed and that exist today would never have been able to evolve. The story of human cognition is all the more remarkable because not only do humans have the ability to travel in subjective time, they also have the ability to effect changes to the historical record as they time-travel – to imagine actions with consequences in these (at least seemingly) possible worlds. The capacity for humans to explore and be influenced by the counterfactual worlds they construct is a truly outstanding evolutionary feat – one that has propelled our species far beyond even the most formidable

2

Introduction

powers of retrospection. The present volume explores the psychological bases of this remarkable feat of human cognition: counterfactual thinking. Strictly speaking, counterfactuals refer to thoughts or statements that include at least some premises believed to be contrary to fact. According to this broad definition, counterfactuals do not require a temporal reference. For instance, one might say (counterfactually), “If all circles were squares, then all spheres would be cubes.” The “contrariness to fact” aspect of counterfactuals has long been of interest to logicians, such as Nelson Goodman, David Lewis, and Richard Stalnaker, who have sought to explain how knowledge could be derived from false conditional premises. Psychologists, on the other hand, have paid greater attention to counterfactual thinking that focuses specifically on how the past might, could, would, or even should have turned out differently. These researchers have been intrigued by the psychologically compelling nature of what if and if only thoughts, and the propensity for people to mentally time-travel. They have directed their attention to two broad questions. First, what are the affective, motivational, cognitive, and social determinants of counterfactual thinking? Second, what are the functional and psychological consequences of counterfactual thinking? The contributions to this volume bring together a collection of contemporary theoretical insights and descriptions of recent empirical research that bears on these two overarching questions. Traveling backward in time, the sustained attempt to address these questions can be traced back to 1982 when Danny Kahneman and Amos Tversky published a brief, but thought-provoking, chapter on what they called the “simulation heuristic.” There they proposed that, in addition to using the availability heuristic to arrive at judgments (namely, a strategy by which the frequency or likelihood of an event is judged by the ease with which similar instances could be recalled from memory), people often rely on a more constructive process in which they mentally simulate a model of a temporally extended episode and then examine its contents and implications. They proposed that such models are often constructed with the aim of examining how a past outcome might have been “undone” or, more generally, what the consequences of a slight change to the historical record would likely have been (i.e., an intuitive “what if?” analysis). Kahneman and Tversky offered a number of key proposals that set the research agenda for many years to come. One idea was that counterfactual simulations are normality-restoring. People tend to undo outcomes that they perceive as abnormal. Moreover, they tend to undo abnormal outcomes by mentally mutating antecedents that they similarly perceive as being abnormal under the circumstances. Rarely do people undo events by making what Kahneman and Tversky called “uphill changes” – mutations that involve mentally deleting normal antecedents or mentally inserting abnormal ones. Another important idea that Kahneman and Tversky advanced was that the ease of undoing had important consequences for how people responded emotionally and judgmentally to the actual events. Thus, a flight missed by five

Introduction

3

minutes will likely evoke more disappointment and perhaps more selfrecrimination than a flight missed by thirty minutes because it is easier to imagine having been on the flight in the former case than in the latter. The possibility of “making the flight” seems “closer” to reality when the flight was “just missed” than when it was missed a half hour ago. Both of these key proposals, among others, were elaborated in Danny Kahneman and Dale Miller’s 1986 exposition of norm theory. According to norm theory, judgmental and affective reactions to events are largely influenced by the standards of comparison that they recruit. These cognitive reference points, Kahneman and Miller proposed, were not merely based on a priori assessments of an event’s likelihood, but rather were based on alternatives that were “post-computed” – that is, mentally constructed on the fly in response to feelings of surprise that often accompany expectancy violations. The notion of “norm restoration” had a profound influence on researchers who sought to explain the cognitive rules by which people mentally reconstructed the past using counterfactual thought experiments. Many of these “mutability constraints,” such as preference for mutating exceptional rather than routine antecedents, actions rather than inaction, and proximal rather than distal events, were understood in terms of this “normality principle.” Moreover, Kahneman and Miller’s emotional amplification hypothesis, which states that emotional responses to events are contrasted away from the affective direction of the counterfactual reference point, highlighted an important fact: counterfactuals tend to have a direction, either being upward (i.e., better than reality) or downward (i.e., worse than reality), but rarely are they horizontal (i.e., just “different”). Spurred by these seminal contributions, the next decade witnessed the first major wave of sustained research and theory development. Many of the core insights from this period were described in a thought-provoking 1995 volume edited by Neal Roese and Jim Olson and entitled What Might Have Been: The Social Psychology of Counterfactual Thinking. Perhaps foremost among these developments was the idea that counterfactual thinking has a functional basis. On the one hand, it was proposed that upward counterfactual thinking served a preparatory function by allowing individuals to explore the causal bases of past outcomes, especially those that deviated from expectation and that had negative consequences. On the other hand, downward counterfactual thinking was proposed to regulate affective responses by making people feel better about reality upon realizing how “it could have been worse.” Also taking shape during this first wave of research was the influential idea that counterfactual thinking plays a key role in how people select the causes of past events. Again, the idea can be traced back to Kahneman and Tversky’s seminal chapter, although similar (and, indeed, more elaborate) proposals had been made by the legal philosophers H.L.A. Hart and A.M. Honoré and by the philosopher J.L. Mackie. The basic proposal was that one of the ways that people would assign causal status to an outcome would be

4

Introduction

to run a counterfactual test in which the proposed cause was negated in simulation. If the simulation ran on in such a way that the outcome was subsequently undone, it would lend support to the proposed cause. If the outcome remained intact in the simulation, then the proposed cause could be ruled out by this kind of thought experiment. In the decade since the publication of Roese and Olson’s edited volume, a second wave of counterfactual research has emerged. Although this work continued to be influenced by earlier theoretical assumptions and over a decade’s worth of empirical research, a number of important new themes were evident. First, the methodology for doing counterfactual thinking research steadily matured. The use of “vignette studies,” cleverly pioneered by Kahneman and Tversky, had been a favored method for exploring the determinants and consequences of counterfactual thinking in a substantial number of later studies. Over time, however, some researchers started to move beyond the use of scenario studies by exploring the counterfactuals that people reported in response to real events that they had either experienced or had spent a considerable amount of time thinking about. The contributions to the present volume by Dhami, Mandel and Souza (Chapter 10) and by Tetlock and Henik (Chapter 12) clearly illustrate this thrust. In the former case, the authors examine the counterfactual thoughts of hundreds of sentenced prisoners in response to their arrest, conviction and sentencing. In the later case, the authors examine the use of counterfactual arguments by experts on world politics and history. There were also important theoretical developments taking shape. First, the idea that counterfactual thinking influences the causal selection process faced serious challenges. Part I on “Counterfactuals, Causality, and Mental Representation” is largely devoted to contemporary views on this issue. In Chapter 1, Mandel presents an overview of research on the relation between causal and counterfactual explanation. He summarizes the major critiques to the “counterfactual thinking influences causal selection” accounts that held sway in the first wave of counterfactual research, and then he sketches a new “judgment dissociation” theory of the relation between counterfactual and causal thinking. In Chapter 2, Spellman, Kincannon and Stose propose an alternative theoretical account of the causal selection process. They posit that causal selection is based primarily on subjective assessments of probability, but they also importantly explain how counterfactual thinking, in turn, can influence those probability judgments. In Chapter 3, Hilton, McClure and Slugoski contrast two main approaches to the study of causal judgment: an ahistorical approach that describes how people discover the causal relations between different types of events (as in the natural sciences), and a historical approach that describes how people discover the causes of particular effect occurring within an unfolding chain of events (as in the legal case approach). Contributing to the latter approach, they offer a typology of causal chains that expands on the “causal–temporal” distinction invoked in earlier counterfactual research.

Introduction

5

A second theoretical development came from cognitive psychologists who had long been interested in the process of mental representation. Notably, Ruth Byrne and her colleagues initiated a program of research in the mid1990s that offered a representational account of counterfactual thinking in terms of mental models theory she and Phil Johnson-Laird had developed (building on earlier ideas by Kenneth Craik). According to this account, by default, people represent only the true possibilities in a given case. However, because counterfactuals prompt individuals to represent false possibilities alongside a model of what is believed to be true, these thoughts can trigger different patterns of inference than their factual counterparts. The final chapter of Part I, Chapter 4 by Walsh and Byrne, reviews the key contributions that mental models theorists have made to the counterfactual thinking literature. As noted earlier, one of the key objectives of the second wave of counterfactual thinking research was to develop an account of the functional (and possibly dysfunctional) bases of counterfactual thinking. The consensus that emerged from earlier work was, essentially, that upward counterfactuals help people learn and better prepare for the future but make them feel bad in the process by pointing to ways that things could have been better, whereas downward counterfactuals make people feel good by showing them how bad it could have been, but don’t teach them very much. Since then, however, there have been important revisions and expansions of this functionalist account. Accordingly, we have devoted Part II on “Functional Bases of Counterfactual Thinking” to three contributions that explore some of the more recent and intriguing functionalist themes. In Chapter 5, Markman and McMullen propose that the functions of upward and downward counterfactual thinking described in earlier accounts are characteristic of an evaluative mode of thinking in which counterfactual representations are contrasted against their factual counterparts. According to their account, however, people also generate counterfactuals in a more experiential, reflective mode of thinking in which little attention is devoted to what actually happened. In this mode, they propose, upward counterfactuals can improve affect and downward counterfactuals can function as “wake-up calls” that prompt preparatory responses. In Chapter 6, Segura and Morris examine the role of counterfactual thinking in experiential learning at individual and organizational levels. Segura and Morris unpack the learning cycle into three stages – evaluating outcomes, inducing rules, and implementing actions – and they examine the role that counterfactual simulations can play at each of these distinctive stages. In Chapter 7, Galinsky, Liljenquist, Kray, and Roese strike out into new functionalist territory by proposing not only that counterfactual thinking facilitates learning, but that it also plays a crucial role in how people formulate meaning in their everyday lives. They suggest that by thinking about the many ways in which events could have happened differently, people often conclude that personally significant events must have happened as they did for “good reason.”

6

Introduction

A fourth notable area of theoretical development during the second wave of counterfactual research has been on the relation between counterfactual thinking and emotion. Part III on “Counterfactual Thinking and Emotion” examines how counterfactual thinking contributes both to the experience of both luck and regret. In Chapter 8, Teigen argues that people typically consider themselves “lucky” when avoiding a disaster and “unlucky” when a small mistake or accident turns out to have disproportionate consequences. Using a number of intriguing examples, Teigen examines three different mechanisms that elicit these two feelings, and related ones such as gratitude. In Chapter 9, Zeelenberg and van Dijk review their work on regret that, for example, identifies circumstances in which people can feel more regret after inaction than after actions. They then turn to the comparative nature of regret, and show that better forgone outcomes that are similar to the actual outcome are more likely to cause regret than those that are not. This effect is principally observed in people with low social comparison orientations, as those with strong tendencies to make social comparisons with others tend to feel regret regardless of whether the forgone outcome is similar to the target outcome or not. Part IV on “Counterfactual Thinking in the Context of Crime, Justice, and Political History” offers three examples of how the extension of counterfactual research to applied domains may not only add to the external validity and methodological rigor of this research area, but also contribute to important theoretical developments. Focusing attention on counterfactual thinking in sentenced prisoners, Dhami, Mandel and Souza (Chapter 10) expand our knowledge of the influence of counterfactuals on emotions and attributional judgments. They show that the influence of counterfactual thinking on guilt is not only due to affective contrast, but that it is also mediated by attributions of self-blame. Moreover, they show that the situational context to which counterfactual thinking refers (e.g., committing a crime, being arrested, convicted, or sentenced) has an important moderating effect on the relation between upward counterfactual availability and anger. In Chapter 11, Catellani and Milesi examine counterfactual thinking in the context of rape cases and they show how social contextual factors may influence counterfactuals and, as a consequence, social judgments. Whereas norm theory focused primarily on intrapersonal norms, Catellani and Milesi show that social norms triggered by the actors in a given case can also constrain the content of counterfactuals. They also show how communication-related goals and the mutual expectancies of people involved in interpreting the event (e.g., during a trial) may constrain the generation, expression, and evaluation of counterfactuals. In Chapter 12, attention shifts to the historical and political domain, where Tetlock and Henik show that well established views of reality, such as political ideologies, also constrain which counterfactuals will be entertained as plausible. Demonstrating a key function of counterfactual thinking – belief-system defense – Tetlock and Henik show that, in reasoning about a wide range of historical events, experts selectively invoke

Introduction

7

second-order “even if” counterfactuals and they are more likely to challenge connecting principles when they are presented with counterfactual arguments that fly in the face of their own preferred theories than when presented with belief-confirming counterfactuals. Over little more than two decades, research on the psychology of counterfactual thinking has made important strides. The contributions to our book highlight in different ways the methodological, empirical, and theoretical advances that have been made in recent years. Projecting forward in subjective time, we have little doubt that the next few years will continue to reveal important new insights into the psychology of counterfactual thinking. Finally, several thanks are in order. The idea for this book emerged following a stimulating conference on counterfactual thinking that we organized in May 2001, in Aix-en-Provence, France. If it were not for the financial support of the European Association for Experimental Social Psychology (EAESP), we probably would not have been able to organize the meeting and, in turn, this book might never have been. Thus, we thank EAESP for its generous support. We should also note that most of the contributors to our book also attended the conference. We thank Peter Robinson, the editor of the Research Monographs in Social Psychology series, who encouraged us early on to submit a proposal for this book. Finally, we sincerely thank Yeliz Ali, Katherine Carpenter, Terry Clague, Joe Whiting, and the rest of the staff at Routledge and Taylor & Francis, who have done a superb job of assisting us from start to finish. David Mandel Denis Hilton Patrizia Catellani

Part I

Counterfactuals, causality, and mental representation

1

Counterfactual and causal explanation From early theoretical views to new frontiers David R. Mandel

In everyday and not-so-everyday life, we encounter situations that seem to demand an explanation of why something happened, how it happened, or how it could have been prevented. For example, following the 9/11 terrorist attacks, many people sought explanations for each of these questions. Explanations of why it happened have focused on Islamic fundamentalism and US hegemony in world politics. Explanations of how it happened, by contrast, have focused on the actions of the terrorists and their accomplices who were involved in instigating or directly carrying out the attacks. Differently still, explanations of how the attacks might have been prevented have focused on errors of judgment and ineffectual policies of US government agencies such as the CIA and the FBI that bore responsibility for preventing such attacks. As the example illustrates, explanations are “tuned” by the type of question they are meant to address. They are meant to be relevant, not “merely” true or probable (Hilton and Erb 1996). That causal thinking plays a key role in the explanation process may seem obvious. After all, how and why questions are causal questions. As Kelley (1973: 107) put it, “Attribution theory is a theory about how people make causal explanations, about how they answer questions beginning with ‘why?’” Less obvious, perhaps, is the role that counterfactual thinking may play in that process. Yet, over the past two decades psychologists have proposed that counterfactual thinking does indeed play a key role. In this chapter, I examine some of the theoretical claims, critiques, and reconciliation attempts that have emerged from this literature. Although I draw on evidence from various pertinent sources, my focus in this chapter is on reasoning directed at explaining an effect in a specific case. Readers interested in the role of counterfactual reasoning about causal laws might consult Tetlock and Belkin (1996a).

Early theoretical views Early claims regarding the effect of counterfactual thinking on causal explanation have focused on two interrelated routes of influence: (1) the selection of contrast cases used to define the effect to be explained and

12

David R. Mandel

(2) counterfactual conditional simulations used to test the plausibility of particular hypothesized causes (for reviews, see Spellman and Mandel 1999, 2003). I discuss these proposed routes in the subsections below. Contrastive counterfactual thinking The desire to explain is roughly proportional to the perceived discrepancy between expectancies and outcomes. When our expectancies are confirmed, there is little need for explanation. But, when expectancies are disconfirmed, they are likely to trigger spontaneous searches for causal explanations (Hastie 1984; Kanazawa 1992; Weiner 1985). The persistence of attention to disconfirmed expectancies is, by definition, then, a form of counterfactual thinking. Contrastive in nature, counterfactuals of this sort recapitulate expectancies that are juxtaposed against the reality of surprising outcomes. Although theorists tend to associate close counterfactuals with the term almost (Kahneman and Varey 1990; see also Chapter 8, Teigen) and counterfactual conditionals with the term if only and, more recently, even if, contrastive counterfactuals have not been provided with a natural-language marker. I propose that the term rather than might be appropriate in this regard. That is, contrastive counterfactuals often convey if not in form, then in gist, the following: “This unexpected event X occurred rather than Y, which I expected to occur instead.” As many theorists have proposed, contrastive counterfactuals play an important role in causal selection by defining the nature of the effect to be explained (e.g., Einhorn and Hogarth 1986; Gorovitz 1965; Hesslow 1983; Hilton 1990; Mackie 1974). As Kahneman and Miller (1986) noted, in situations in which an outcome is viewed as normal in the circumstances – and hence expected – it is reasonable to answer the question “Why?” with the reply “Why not?” In such cases, no effect is meaningfully defined. The why question therefore presupposes a deviation between occurrence and what had been expected to occur by an explainee: “Why did this happen rather than what I expected would happen?” Theorists have proposed that, as a general rule, counterfactual thinking will recruit contrast cases that restore normality because people expect normal events to occur (Hart and Honoré 1985; Hilton and Slugoski 1986; Kahneman and Miller 1986). In the present context, normal not only means what was likely or is frequent, but also what is normative in the circumstances (McGill and Tenbrunsel 2000; Chapter 11, Catellani and Milesi). These norm-restoring “downhill climbs” (Kahneman and Tversky 1982a) assist in defining a backgrounded set of factors – or causal field in Mackie’s (1974) terms – that are assumed to be common to both the factual case and the contrastive counterfactual set of cases and that are ruled out as causal candidates. Therefore, to the extent that contrastive counterfactual thinking plays a role in defining a causal field, it can be said to play an important role in determining what would not normally be deemed the cause. As we shall

Counterfactual and causal explanation

13

see next, counterfactual thinking has also been ascribed a more positive role in the process of causal explanation. Counterfactual conditional simulations While disconfirmed expectancies may be deemed counterfactual, not all counterfactuals merely recapitulate disconfirmed expectancies. An important function of counterfactual thinking, which Kahneman and Tversky (1982a) brought to psychologists’ research attention, is that it allows people to run “if–then” simulations in working memory, which can allow us to explore our own intuitions about how manipulations to aspects of a case might have influenced how the case would have subsequently unfolded. Accordingly, the second proposed route of counterfactual influence on causal explanation involves the idea that counterfactual “if–then” or “even if–then” simulations can be used to identify – perhaps even verify – various causal contingencies that may later be deemed “the cause” (e.g., Lipe 1991; Mackie 1974; McGill and Klein 1993; Roese and Olson 1995a; Wells and Gavanski 1989). These conditional representations may be regarded as generic response forms to the counterfactual question “Would Y have happened if X had not?” For instance, according to this “counterfactual simulation account,” a student who learns that she failed an exam and who thinks “If I had studied harder, I would have passed the exam” will be more likely to view her lack of preparation as a cause (if not “the” cause) of her performance than a student in a comparable position who instead thinks “Even if I had studied harder, I still would have failed” (McCloy and Byrne 2002). According to Mackie (1974), the counterfactual test in which the cause is negated is crucial because the very meaning of the expression “X caused Y” is, first, that X and Y in fact happened and, second, that had X not happened, Y also would not have happened. This idea has been recurrent in psychological literature linking counterfactual thinking and causal explanation. For instance, Kahneman and Tversky (1982a: 202) proposed that “to test whether event A caused event B, we may undo A in our mind, and observe whether B still occurs in the simulation.” More forcefully, Wells and Gavanski (1989: 161) proposed that “an event will be judged as causal of an outcome to the extent that mutations to that event would undo the outcome” (italics mine). Roese and Olson (1995a: 11) took the argument even further, claiming that although “not all conditionals are causal . . . counterfactuals, by virtue of the falsity of their antecedents, represent one class of conditional propositions that are always causal” (italics mine). These authors explain that “[t]he reason for this is that with its assertion of a false antecedent, the counterfactual sets up an inherent relation to a factual state of affairs” (1995a: 11). Mental model theorists (Byrne and Tasso 1999; Thompson and Byrne 2002) have similarly proposed that, whereas counterfactual conditionals automatically recover factual models, thus establishing a salient contrast case, factual

14

David R. Mandel

conditionals do not automatically recover counterfactual models because people do not spontaneously represent false events. Thus, no contrast case would be evoked unless the implicit models were deliberatively unpacked. These later accounts suggest not only that counterfactual thinking may influence causal explanation, but that counterfactual thinking will have a stronger influence on causal explanation than factual thinking.

Empirical and theoretical challenges Although some evidence supporting the idea that causal judgments (and related attributions) are influenced by counterfactual thinking has accrued (e.g., see Branscombe et al. 1996; Roese and Olson 1997; Wells and Gavanski 1989), most of it is based on studies that have manipulated the mutability of antecedents or outcomes in a given case (but see Chapter 10, Dhami et al.). Outcome mutability has been manipulated by constructing different versions of scenarios in which alternatives to a chosen option would have led either to the same outcome or to a better outcome (Wells and Gavanski 1989). Antecedent mutability has been manipulated, for instance, by varying the abnormality (Kahneman and Tversky 1982a) or controllability (Mandel and Lehman 1996) of antecedents in scenarios. The core assumption underlying such research has been that if the relevant manipulation influenced judgment, then it must have been mediated by counterfactual thinking. Other research (e.g., Davis et al. 1995; N’gbala and Branscombe 1995), however, suggests that the effect of “mutability manipulations” on causal judgments may have had more to do with the particular hypothetical scenarios that had been used in previous research than with a robust effect of counterfactual thinking on causal judgment. For example, Mandel and Lehman (1996: Experiment 3) demonstrated that, when participants read about a hypothetical case that afforded the opportunity to make different counterfactual and causal selections, antecedent mutability manipulations influenced participants’ counterfactual listings, but these same manipulations did not influence participants’ causal judgments. Moreover, the antecedent that was perceived as most causal differed from that which was most frequently mutated as a way of undoing the outcome. Trabasso and Bartolone (2003) pointed out that manipulations of antecedent normality in past studies are confounded with the extent to which such antecedents were themselves explained in the relevant scenarios. These authors independently manipulated level of explanation and normality in Kahneman and Tversky’s (1982a) “Mr Jones” car-accident scenario and asked participants to rank the likely availability of four counterfactual “if only” statements that mutated either the route Jones took, the time he left work, his decision to brake at the yellow light, or the other vehicle charging into the intersection. Calling into question the relation between normality and counterfactual availability, Trabasso and Bartolone found that

Counterfactual and causal explanation

15

counterfactual rankings were influenced by level of explanation only. Antecedent normality had no effect on participants’ judgments of how likely it would be that a given counterfactual statement would be generated. Another set of studies (Mandel 2003b) examined the idea that counterfactual thinking about what could have been has a stronger effect on attribution than factual thinking about what was. For example, participants in Experiment 2 first recalled an interpersonal event they had recently experienced and were then instructed either to think counterfactually about something they (or someone else) might have done that would have altered the outcome or to think factually about something they (or someone else) did that contributed to how the outcome actually occurred. Participants rated their level of agreement with causality, preventability, controllability, and blame attributions, each of which implicated the actor specified in the thinking manipulation. Compared to participants who received no thinking directive, participants in the factual and counterfactual conditions reported more extreme attributions. However, mean agreement did not differ between the factual and counterfactual conditions – a finding that was replicated in two other experiments (cf. Mandel and Dhami in press; Tetlock and Lebow 2001). Counterfactual tests of necessary causes Another problem faced by counterfactual simulation accounts is that they imply that causal reasoners assign greater weight to causes that are necessary rather than sufficient to explain the relevant effect or yield the focal outcome. Strictly speaking, X is a necessary cause of Y, if X is implied by Y and, in contrapositive form, if ¬X implies ¬Y (the symbol ¬ is read as “the negation of”).1 By contrast, X is a sufficient cause of Y, if X implies Y and, in contrapositive form, if ¬X is implied by ¬Y (Cummins 1995; Fairley et al. 1999). However, as Mackie (1974) explained, the concept of a causal field requires a qualified interpretation of these relations, such that necessity and sufficiency are interpreted as meaning necessary or sufficient in the circumstances, where the latter qualification may be interpreted as meaning “given the presence of all the events in the case that are backgrounded.” On either interpretation, counterfactual conditionals that proceed by negating a hypothesized cause provide information relevant to the assessment of whether that factor was necessary to bring about the effect. I regard this as one of the principal limitations of counterfactual simulation accounts because there is mounting evidence that what is meant by the term cause in everyday discourse tends to reflect sufficiency in the circumstances rather than necessity in the circumstances. Indeed, even Mackie (1974: 38), a key proponent of the “necessary cause” view, wrote: There is, however, something surprising in our suggestion that “X caused Y” means, even mainly, that X was necessary in the circumstances

16

David R. Mandel for Y. Would it not be at least as plausible to suggest that it means that X was sufficient in the circumstances for Y? . . . After all, it is tempting to paraphrase “X caused Y” with “X necessitated Y” or “X ensured Y,” and this would chime in with some of the thought behind the phrase “necessary connection”. But if “X necessitated Y” is taken literally, it says that Y was made necessary by X or became necessary in view of X, and this would mean that X was sufficient rather than necessary for Y.

One of the key problems with treating “X caused Y” as meaning, primarily, that “had X been absent, Y would not have occurred” is that the definition is too inclusive (Lombard 1990). As Hilton et al. (Chapter 3 below) put it, “Th[e] plethora of necessary conditions brings in its train the problem of causal selection, as normally we only mention one or two factors in a conversationally given explanation . . .” Contrary to this reasonable proposal, according to the “negate-X, verify-Y” counterfactual criterion, oxygen would be the cause of all fires, and birth would be the cause of all deaths. Clearly, statements such as these violate our basic understanding of what causality means. Although we can all agree that birth is necessary for death, few would say the latter is brought about by the former. It is the quality of being instrumental in “bringing about,” even if not without the assistance of a background set of enabling conditions, which suggests that “X caused Y” means X was sufficient in the circumstances for Y to occur. Since Mackie posed the preceding question, psychologists have conducted considerable research to empirically address it. Studies of naïve causal understanding indicate that people define causality primarily in terms of sufficiency. For example, Mandel and Lehman (1998: Experiment 1) asked participants to provide open-ended definitions of the words cause and preventor. They found that a majority of participants defined cause (71 percent) and preventor (76 percent) in terms of sufficiency (e.g., “if the cause is present, the effect will occur”). By contrast, only a minority defined these concepts in terms of necessity (22 percent and 10 percent for cause and preventor, respectively; e.g., “if the cause is absent, the effect won’t occur”). Although Mandel and Lehman (1998) did not report the cross-tabulated frequencies of response, a re-analysis of the dataset revealed an interesting result. Without exception, the minority of participants who provided a necessity definition also provided a sufficiency definition for the same term. That is, not a single participant in their study provided a definition of causality or preventability only in terms of necessity, whereas the majority provided definitions that focused exclusively on sufficiency. Given the possibility for bias in coding of open-ended responses, Mandel (2003c: Experiment 2) attempted to replicate these findings by asking participants whether they thought the expression “X causes Y” means “When X happens, Y also will happen” (i.e., X is sufficient to cause Y) or “When X doesn’t happen, Y also won’t happen” (i.e., X is necessary to cause Y). Eighty-one percent of the sample interpreted the causal phrase in terms

Counterfactual and causal explanation

17

of sufficiency and a comparably high percentage (84.5 percent) thought that other people would do so too. Goldvarg and Johnson-Laird (2001) provide converging support for the sufficiency view. Their research examined the types of mental models that people view as being consistent with expressions like “X will cause Y.” In Experiment 1, causal expressions were associated with the three possibilities implied by the notion of a sufficient cause for roughly half of the sample (i.e., X and Y, ¬X and Y, and ¬X and ¬Y).2 The other half of the sample indicated that the causal expressions were associated with the two possibilities implied by the notion of a necessary and sufficient cause (i.e., X and Y, and ¬X and ¬Y). However, consistent with the cross-tabulation findings just reported, no participant selected the set of possibilities implied by the notion of a necessary cause alone (i.e., X and Y, X and ¬Y, and ¬X and ¬Y). Considerable support for the sufficiency view also comes from causal induction studies, which have demonstrated that people assign greater weight to the sufficiency-relevant cause-present cases than the necessity-relevant cause-absent cases (e.g., Anderson and Sheu 1995; Cheng 1997; Kao and Wasserman 1993; Mandel and Lehman 1998; McGill 1998; Schustack and Sternberg 1981). In sum, the weight of available empirical evidence favors the sufficiency interpretation of causality and, in the single-case context, this assertion must be qualified by adopting the “in the circumstances” interpretation.

Theoretical reconciliations Mandel and Lehman’s (1996) prevention-focus account To reconcile the idea that counterfactual conditionals do tell us something of a causal nature with the substantial body of evidence indicating that people place greater weight on sufficient rather than necessary conditions, Mandel and Lehman (1996) proposed an alternative “prevention-focus” account. This account posits that people often treat “negate-X” counterfactual conditionals not as tests of the necessary generative cause of an effect but as explanations of sufficient but forgone ways in which the effect might have been prevented. To illustrate the account, it will be instructive to map the representations of factual and counterfactual events in a given case onto a 2  2 confusion matrix, such as researchers have used to summarize contingency information in multiple-case studies (e.g., Mandel and Lehman 1998). In the present context, however, the cells of this matrix do not index the frequencies of the relevant event conjunctions across a set of real cases but, rather, in line with Kahneman and Tversky’s (1982a) idea of the simulation heuristic, they index the cognitive ease of representing ways in which the various possibilities might have occurred in the focal case.

18

David R. Mandel

In Figure 1.1, Y represents an event to be explained (e.g., an unexpected car accident) and X represents a mutable antecedent event (e.g., the route the driver took). Further, let us assume that one is evaluating the hypothesis that “X caused Y.” Cell A represents the factual case in which X and Y cooccur. Cell B represents counterfactual cases in which X occurs but Y nevertheless does not. Cell C represents counterfactual cases in which the negation of X fails to undo Y. And, cell D represents counterfactual cases in which the negation of X successfully undoes Y. Both cells A and D provide evidence that supports the focal hypothesis. By contrast, cells B and C provide evidence against the focal hypothesis, but the nature of the detraction differs in the two cases and is best thought of in terms of the contrast between competing sources of evidence. Specifically, the contrast of cause-present information in cells A and B permits a test of whether X is sufficient to cause Y, whereas the contrast of cause-absent information in cells C and D permits a test of whether X is necessary to cause Y. As noted, theorists have focused on the counterfactual question “If X had not occurred, would Y still have occurred?” or the more restrictive probe “But for X, would Y still have occurred?” Answers to such questions will evoke representations that fall into either cells C or D (cf. Lipe 1991). If the negation of X does not undo Y (e.g., “Even if I had taken a different route, I probably still would have had an accident”), the C-cell representation will likely disconfirm the hypothesis that X was necessary in the circumstances for Y. Conversely, if the negation of X undoes Y (e.g., “If only I had taken a different route, I wouldn’t have had this accident”), then the D-cell representation will likely support the necessity claim. The prevention-focus account shifts the emphasis on counterfactual conditionals in three key respects. First, emphasis is shifted from reasoning about generative causes to reasoning about inhibitory causes. Second, emphasis is shifted from assessing necessary conditions to assessing sufficient conditions. Third, there is an epistemological shift such that counterfactuals Y A

B Confirmation by factual case

X

C ¬X

¬Y Sufficiency violation by counterfactual case D

Necessity violation by counterfactual case

Confirmation by counterfactual case

Figure 1.1 Confusion matrix representing implications of factual and counterfactual cases for a hypothesis (“X caused Y”) about an actual generative cause.

Counterfactual and causal explanation

19

are intended to refer directly to hypotheses about counterfactual possibilities (what might have been) rather than hypotheses about factual possibilities (what was). Figure 1.2 shows how the factual and counterfactual contingencies considered earlier change in terms of their implications when the focal hypothesis is reframed from “X caused Y” to “¬X could have prevented Y.”3 In this case, the contrast of cells A and B provides information relevant to assessing whether ¬X would have been necessary in the circumstances to prevent Y, whereas the contrast of cells C and D provides information relevant to assessing whether ¬X would have been sufficient in the circumstances to prevent Y. In support of the prevention-focus account, Mandel and Lehman (1996) demonstrated that participants who completed “If only . . .” sentence stems provided responses that were aligned more closely with other participants’ listings of how the effect might have been prevented than with participants’ listings of how the effect was caused. The intuition underlying the prevention-focus account is also captured by considering the types of explanations offered for the 9/11 attacks. As noted earlier, although the cause of the attacks was attributed to the terrorists and their accomplices, most counterfactual assessments have been directed at US prevention failures. For instance, one New York Times article began by stating “The director of the FBI, Robert S. Mueller III, acknowledged today for the first time that the attacks of Sept. 11 might have been preventable if officials in his agency had responded differently to all the pieces of information that were available” (Lewis 2002: § 1). Such counterfactuals are widespread and influence blame assignment, as the 9/11 commission in the United States is now proving (see, e.g., Johnston and Dwyer 2004), but one would be hard-pressed to find news stories claiming that US agencies caused 9/11. It appears that although upward counterfactual thinking is, broadly speaking, causally informative and that it does play a key role in the explanation process, the focus of counterfactual explanations differs from Y A

B Confirmation by factual case

X

C ¬X

¬Y Necessity violation by counterfactual case D

Sufficiency violation by counterfactual case

Confirmation by counterfactual case

Figure 1.2 Confusion matrix representing implications of factual and counterfactual cases for a hypothesis (“¬X could have prevented Y”) about a forgone inhibitory cause.

20

David R. Mandel

that of direct causal explanations. In particular, as Mandel and Lehman (1996) observed, counterfactual listings are more likely than generativecause listings to focus on controllable behaviors (also see Girotto et al. 1991; Mandel 2003a; McCloy and Byrne 2000). The emphasis on personal control over the outcome and on “human error” (Morris et al. 1999) suggests that counterfactual explanations tend to conform to notions described by some manipulability theories of causality. As Collingwood (1940: 296) put it, “the question ‘What is the cause of an event y?’ means in this case ‘How can we produce or prevent y at will?’” Collingwood (1940: 307) even proposed that “for the mere spectator (meaning someone who cannot produce or prevent any of the conditions of an event) there are no causes.” Although this definition of causality is too restrictive, it captures fairly well the sense in which people offer counterfactual explanations for past events. Spellman’s (1997) probability-updating account Another reconciliatory account was offered by Spellman (1997; Spellman and Kincannon 2001; Chapter 2, Spellman et al.), who proposed a probability-updating account of causal selection akin to a stepwise multiple regression analysis (cf. Hilton 1988). According to SPA (for Spellman’s probability-updating account), a reasoner first identifies the factual outcome to be explained (Y), as well as a set of causal candidates (X1, X2, . . . Xn) from the chain of events in the relevant case. Next, the reasoner assesses the posthoc probability of Y prior to X1 and then reconditionalizes the probability on X1, observing the change in probability accounted for by X1. This process is repeated for each candidate in the set, each time assessing the change in probability from the last step. For instance, the causal efficacy of X2 would be assessed by assessing P(Y | X1  X2) and subtracting from it the prior assessment, P(Y | X1). Causal selection is determined by identifying the candidate that accounts for the largest change in outcome probability. SPA is essentially a probabilistic model of causal selection adapted for single-case judgments, but it also posits that counterfactual thinking can influence causal explanation when the resultant counterfactuals affect relevant subjective probabilities. In particular, SPA predicts that if it is easy to imagine counterfactual alternatives to the actual outcome, then P(Y) will be lowered, correspondingly increasing the chance that X1 will be selected as the cause. More precisely, if Y can occur k ways from a set of n alternatives, then holding k constant, the number of counterfactuals alternatives to Y should be a positive monotonic function of n. Although this idea is plausible, at present there is a lack of direct evidence demonstrating that subjective probability mediates the effect of counterfactual thinking on causal explanation. Indeed, it has yet to be demonstrated that the factor that yields the largest change in subjective probability will be deemed the cause. In the next section, I report on a recent test of the latter prediction (see also Chapter 3, Hilton et al.).

Counterfactual and causal explanation

21

Reflections and new frontiers It is indisputable that counterfactual, causal, and for that matter, covariational statements can convey similar meanings, thereby indicating some degree of conceptual overlap. For instance, if a child becomes ill after eating a poisonous mushroom, one observer might say “He got sick because he ate the mushroom,” while another observer might say “Yes, if only he hadn’t eaten that mushroom, he wouldn’t have been sick,” while yet another observer might say “Given that he ate the mushroom, the chances of him getting sick were almost certain.” Although each observer uses a different form of expression, the reasoning in each case seems closely aligned. For this reason, many theories of causal explanation have posited a central role of counterfactual and/or covariational reasoning. Surely, counterfactual simulations can be used to probe the plausibility of causal explanations, sometimes with surprising effects (Tetlock and Belkin 1996b), but it is also true that causal knowledge influences the counterfactuals that people generate. Rather than arguing that one form of thinking precedes the other, it may be more sensible to examine the features of cognition that account for similarities and differences between causal and counterfactual explanations. Reflections on similarities: is the mind a determinist or an indeterminist? On the surface, one similarity between people’s causal and counterfactual explanations is that both reveal, as Kahneman (1995: 383) put it, that “the mind is not a determinist.” In hindsight, people tend to perceive outcomes and potential causes as having been more or less mutable. I say “on the surface,” however, because in other respects people don’t appear to be indeterminists either. To illustrate this, consider a simple thought experiment. First, imagine a chain of events culminating in a particular outcome. Now, replay the scenario keeping everything exactly the same right up to the point before the final outcome and then ask yourself whether the outcome would be identical or even slightly different. Assuming that the antecedents were exactly the same (and notwithstanding any knowledge one might have about quantum electrodynamics), and concurring with Tetlock and Henik (Chapter 12) who assert that historical observers tend to err in the direction overly deterministic thinking, I suspect most would conclude that the outcome would also be exactly the same in every possible replay. I propose that, although the mind is a determinist at heart, it behaves like an indeterminist, mutating Xs and Ys left and right. That the mind is a determinist at heart is revealed, for instance, by people’s proclivity for causal explanations and their dissatisfaction with purely statistical ones. Indeed, a key heuristic strategy for judging likelihood is representativeness, which comprises not only similarity assessments but assessments of causality too (Kahneman and Tversky 1972). Moreover, statistical information such as

22

David R. Mandel

base rates (Tversky and Kahneman 1982b) and conditional probabilities (Tversky and Kahneman 1982a) are given greater weight if they are interpreted in causal terms that directly shed light on how the relevant outcomes or criteria may have been determined. Deterministic thinking is also revealed by the examples of finding “too much meaning” discussed by Galinsky et al. (Chapter 7). What keeps the mind from experiencing considerable dissonance, I suspect, is its bounded sense of rationality and its pragmatism. Unlike a true philosophical determinist, the mind is rarely concerned with heady questions about whether the initial conditions of the universe necessitated all that has followed. It simply is not equipped to handle such complexities. First, unlike Laplace’s demon, it cannot store knowledge about everything. Second, what the mind thinks it knows must be manipulated within the confines of working memory. Hence, the mind focuses on a piece of a larger puzzle that wins out, at least momentarily, in a contest of pragmatic relevance. The piece that it focuses on will likely be organized into a scenario that has story-like properties including goals, means, and outcomes (Read 1987). Accordingly, when one searches for either a causal or counterfactual explanation, the antecedents in the scenario that serve as candidates will only be partially constrained. For instance, social perceivers often explain their own and other actors’ behaviors in terms of choices, intentions, and so on without accounting for (or apparently feeling the need to account for) the determinants of these ostensibly indeterminate explanatory constructs. If the causal determinants of those antecedents are located outside of the focal scenario or are otherwise inaccessible, attention to the constraints on causes – namely, the meta-causes – will be highly restricted. The steepness of attentional decline can foster the appearance that the mind is an indeterminist. However, within its field of attention, the mind constructs simplified explanations of how or why, which conform well to a deterministic stance. In essence, the appearance of indeterminism in causal and counterfactual reasoning is a consequence of focalism, akin to that found in other areas of research (e.g., Schkade and Kahneman 1998; Wilson et al. 2000). New frontiers: judgment dissociation theory In this subsection, I describe an account called judgment dissociation theory (JDT, Mandel 2003c), which attempts to explain how counterfactual, causal, and covariational (i.e., conditional-probability) reasoning differ from each other. Consistent with past functionalist proposals (Roese and Olson 1997; Taylor and Pham 1996; Weiner 1985), JDT posits that causal, counterfactual, and covariational reasoning can aid in prediction, control, and explanation across many situational contexts. Nevertheless, JDT draws functional distinctions between these types of reasoning and predicts dissociations in the content of causal, counterfactual, and covariational judgments under specified conditions.

Counterfactual and causal explanation

23

A key proposal of JDT is that causal, counterfactual, and covariational judgments differ in terms of how “the outcome” of a scenario is conceptualized. The theory proposes that causal selection is guided by the actuality principle – namely, causal reasoners tend to focus on antecedents that played a critical role in how the actual outcome of a case was generated. Antecedents that “merely” explain how an outcome similar to the actual one, which might have been probable or even inevitable up to a point in the case, will likely be rejected as “the” cause. That is, JDT posits that causal selections will be perceived as informative to the extent that they elucidate causal processes that in fact generated the outcome. In this regard, JDT draws conceptual connections to accounts that posit the importance of perceived causal mechanisms (Ahn and Kalish 2000; Michotte 1963), force dynamics (Wolff and Song 2003), and perceived changes of propensity that can serve as indicators of causal processes (Kahneman and Varey 1990). As the philosopher W.C. Salmon (1984: 170) put it, “causal processes are the means by which causal influence is propagated, and changes in processes are produced by causal interactions.” According to JDT, then, causal explanations will focus on causal interactions that produce salient changes in causal processes whose propagated force is perceived as having culminated in the actual outcome. By contrast, JDT proposes that counterfactual and covariational assessments are guided by the substitution principle – namely, they tend to focus on ad hoc categories of outcome in which the actual outcome represents the category norm. Ad hoc categories (e.g., “ways to prevent Mr Jones from having been injured on this day”) are constructed in response to short-term goals that, once satisfied, allow the category to be disbanded (Barsalou 1983, 1991). In JDT, the goals that generate such categories are associated in predictable ways with different types of reasoning. Specifically, category exemplars in counterfactual assessments are constrained by the goal of finding ways to undo the outcome or something like it, whereas category exemplars in covariational assessments are constrained by the goal of finding ways that would increase the probability of the outcome or something like it. Thus, whereas selected counterfactual conditionals are sensitive to factors that are sufficient in the circumstances to prevent an entire category of outcome, conditional-probability assessments are sensitive to contingencies that raise the probability of any exemplar. More generally, the substitution principle coheres with Kahneman and Tversky’s (1982c: 149) idea that “it is frequently appropriate in conversation to extend the definition of an event X to ‘X or something like it’.” However, the substitution principle generalizes this idea by proposing that a similar extension from X to “X or something like it” aptly characterizes the mental representations that people invoke in counterfactual and covariational reasoning. Mandel (2003c) tested JDT’s predictions by constructing cases with two sequential sufficient conditions for a given type of outcome. For example, in Experiment 2, the protagonist (John) works at one company division for a

24

David R. Mandel

year, after which he can keep the same job or switch to another division. He decides to switch and, soon afterwards, he is relocated. Unbeknownst to John, he is exposed to asbestos fibers at the new location and he develops a terminal case of lung cancer. One day, on a routine medical visit to the hospital, a nurse gives John the wrong medication. He immediately suffers cardiac arrest and, although there is a resuscitation attempt, he dies moments after receiving the drug. After reading the scenario, participants listed factors that caused John’s death and ways of undoing John’s death in counterbalanced order. They also rated the importance of each listing from 0 to 10. Participants then estimated the probability of John dying prematurely conditional on the points at which he: (1) was first hired, (2) switched jobs, (3) was moved to the new location with asbestos, (4) spent thirty years in the building with asbestos, and (5) was given the wrong medication. According to JDT, this scenario will produce a full dissociation in the content of causal, counterfactual, and conditional-probability assessments. Specifically, causal selections will focus primarily on the drug error because that factor played a critical role in how John actually died. By contrast, participants will judge the move to the new location and the ensuing exposure to asbestos as the joint factor that accounted for the greatest increase in the probability of his death – even though he actually died as a consequence of the drug error. Hence, contrary to the prediction of SPA, JDT predicts that participants will judge the drug error as having done little to increase the probability of John’s death even though they will select it as the primary cause. Finally, participants’ counterfactuals will focus primarily on John’s decision to switch jobs because this factor alone is sufficient in the circumstances to undo the actual death and the inevitable death by cancer. Note that this prediction is at odds with counterfactual simulation accounts that predict strong overlap in counterfactual and causal content (Wells and Gavanski 1989). As JDT predicts, participants were more likely to list the drug error as the cause (95 percent) than either the cancer (66 percent) or the decision to switch jobs (38 percent). By contrast, the most frequent counterfactual listing focused on the decision to switch jobs (79 percent), although a sizable percentage (71 percent) mentioned the nurse’s error, probably because of its highly controllable nature. These findings do not support counterfactual simulation accounts. Indeed, contrary to that view (Roese and Olson 1997), the mean correlation in rated importance across antecedent targets was not significantly greater when the counterfactual tasks preceded rather than followed the causal tasks. Analyses of participants’ probability judgments also supported JDT. Participants judged that the move to the new location and the ensuing asbestos exposure, taken together, accounted for a 45 percent increase in the probability of John’s premature death, whereas the decision to switch jobs and the drug error accounted for only 11 percent and 3 percent increases, respectively. These findings are inconsistent with SPA, which predicts on

Counterfactual and causal explanation

25

the basis of these findings that participants would be most likely to select the relocation and subsequent cancer-causing exposure to asbestos as the cause. Moreover, disconfirming the fundamental test of SPA called for earlier, the correlations between probability-change scores and causalimportance ratings were invariably nonsignificant across the targets. In summary, the findings of this experiment and others reported in Mandel (2003c) support the idea that, under certain boundary conditions, causal, counterfactual, and covariational assessments will diverge in focus. These dissociations are particularly likely in scenarios with multiple sufficient causes, as in the experiment just described. In such cases, SPA and JDT predict different forms of asymmetric causal discounting. SPA predicts that the earliest sufficient condition will be deemed the cause because of its effect on probability updating, whereas JDT predicts that the last sufficient condition is likely to be deemed the cause because of its probable role in the process by which the effect was generated. In this regard, there is strong agreement between JDT and Mackie’s analysis. Mackie (1974: 44–6) considered causal scenarios that included cases much like the one just described. He concluded that, in such cases, “What we accept as causing each result, though not necessary in the circumstances for the result described in some broad way, was necessary in the circumstances for the result as it came about” (1974: 46, italics in original).

Conclusion Our understanding of the relationship between counterfactual and causal thinking has evolved from the idea that counterfactuals provide input into the causal explanation process to the idea that counterfactuals provide a distinct form of explanation that shares similarities with causal explanation but that also has important differences. Recent developments indicate several avenues for future research. Although research has shown that causal reasoning influences categorization (Rehder and Hastie 2001), the effect of categorization on causal and counterfactual reasoning is just starting to be examined (Mandel 2003c). As JDT predicts, the explanation of an effect depends on how the effect is conceptualized. Whereas causal explanations favor an instance-based view of effects, counterfactual explanations favor a category-based view. But what types of features influence the perceived similarity between actual and onceinevitable outcomes? JDT posits that perceived functional similarity of consequences is one factor, but what people treat as similar, as well as possible asymmetries in similarity assessments, remains to be explored. For example, a counterfactual explanation might require undoing an inevitable premature death that was precluded by an unrelated severe injury. But would an inevitable injury precluded by an unrelated premature death also require undoing? And, if so, how serious would the injury have to be? There are also many intriguing questions concerning the validity and

26

David R. Mandel

normative status of counterfactual explanations. As Dawes (1996) noted, counterfactuals may be accurate yet vacuous if they are not supported by reliable statistical evidence. In hindsight, it is easy to imagine ways that one could have made an outcome turn out better, and the availability of such thoughts can obscure the fact that, in foresight, success by the counterfactual course of action may have been even less probable (Miller and Turnbull 1990; Sherman and McConnell 1995). Research by Goldinger et al. (2003) has also shown that, compared to people with relatively high working memory capacity, those with lower capacity were more likely to blame victims when their actions were highly mutable, thus raising interesting questions about whether people can make appropriate inferential corrections once counterfactuals lacking in validity have been considered (see also Koehler 1991). Dawes’ warning echoes Kahneman and Tversky’s (1982c) earlier point that forecasts and decisions based on “inside view” scenarios of a case are poor substitutes for assessments of statistical information based on the “outside view” of many cases. Counterfactual simulations are prime examples of inside-view thinking and, therefore, they are poor methods for rigorously assessing support for hypothesized causes. For instance, counterfactual simulations would seldom provide accurate frequency estimates in the four cells of the confusion matrices shown in Figures 1.1 or 1.2, although they could assist in determining whether it is possible that a given cell has a non-zero value. Such thought experiments, in fact, have proved to be very important in science. Error will arise, however, when the ease of generating such examples is mistaken for the frequency or likelihood of such examples. Note, however, that the role of counterfactual thinking in possibility probing may be far greater than that previously assigned to it. As Figure 1.1 revealed, not only can people inquire about the status of Y had X been negated (drawing attention to cells C or D depending on the simulation result), they can also probe whether it would have been possible for Y not to have happened even if X had occurred (drawing attention to cells A and B). If people do in fact assign the term cause to conditions that are perceived as sufficient in the circumstances to bring about the actual outcome, then “affirm-X, verify-Y” probes may be even more important in the causal explanation process than the oft-discussed “negate-X, verify-Y” probes. For instance, B-cell counterfactuals could signal the insufficiency of a causal candidate by drawing attention to the importance of enabling conditions that might not have been present or the possibility of disabling conditions might have intervened. Another normative issue concerns the co-tenability of counterfactual explanations. Counterfactual thinkers often simulate the consequences of a “minimal rewrite” (Tetlock and Belkin 1996b) by changing X and nothing else, but as Jervis (1996) points out, in complex systems it may be impossible to change just one thing. Counterfactual simulations may thus resemble

Counterfactual and causal explanation

27

tightly controlled experiments with high internal validity, but they may also be externally invalid, reflecting the perception that causes are separable in cases in which main effects are the exception and higher-order interactions are the rule. Despite all their real and possible shortcomings, counterfactuals and the cognitive processes that produce them are indispensable, and may be essential for mental health (Hooker et al. 2000). The ability to think counterfactually represents a truly remarkable evolutionary achievement, which probably much more than we have yet acknowledged, defines what it is to be human. Humans not only relive the past in the present, they also relive a multitude of pasts that never existed, and they use these simulated realities to map out possible futures whose truth values remain indeterminate. As Tulving (2004) argued in a keynote address to the American Psychological Society, this ability for forward “time travel,” which he calls proscopic chronesthesia, is a necessary condition for the inception and evolution of all human cultures. This is an important part of the reason why the study of counterfactual thinking continues to intrigue. If progress over the last decade of counterfactual thinking research is a reasonable indicator of what is to come, then the next decade of research examining the determinants and consequences of how people explain what was, what might have been, and what may be will surely be revealing. I recommend that we treat this as our bestguess hypothesis and proceed as if!

Notes Preparation of this chapter was facilitated by research grant 249537-02 from the Natural Sciences and Engineering Research Council of Canada. 1 The nature of the conditional implications defining necessity and sufficiency is of considerable debate and beyond the scope of this chapter. Elsewhere (Mandel 2005), I argue that these concepts and their associated conditional implications are not interpreted as the material conditional implication in everyday discourse but rather in terms of a modified Ramsey test. 2 Goldvarg and Johnson-Laird (2001) define true possibilities in terms of the material conditional implication (cf. Evans et al. 2003; Mandel 2005). 3 This framing effect may be restated as the following truth-functional question: Is ¬X to be interpreted as “X is false” or as “¬X is true?” Whereas counterfactualsimulation accounts assume the former interpretation, the prevention-focus account assumes the latter interpretation.

2

The relation between counterfactual and causal reasoning Barbara A. Spellman, Alexandra P. Kincannon, and Stephen J. Stose

Among the vast array of thinking skills that humans possess are the abilities to engage in counterfactual and causal reasoning. Counterfactual reasoning allows us to imagine something in the world being other than it actually was or is (i.e., counter-to-fact); we can then imagine, or mentally simulate, the world continuing to unfold in a direction other than the direction it has actually taken. This ability allows us to torment ourselves with regret (“If only I hadn’t gone for a drive that night . . .”) or to create our own personalized version of It’s a Wonderful Life. It also allows us to plan for the future and to learn from our mistakes. Whereas counterfactual reasoning is about possibility, causal reasoning is about reality. Finding causes is at the heart of our scientific endeavors; assessing causality is essential for meting out justice in our legal system. As many researchers are fond of noting, discovering causal relations allows us to understand, explain, predict, and control our world. In the past researchers have posited various relations between the two types of reasoning; for example, that counterfactual reasoning was at the heart of causal reasoning or that they were fundamentally identical (for history see Spellman and Mandel 1999, 2003). Currently, however, researchers seem to agree that causal reasoning and counterfactual reasoning ask different questions and serve different functions (Mandel 2003c; Mandel and Lehman 1996; also Roese 1994). In counterfactual reasoning we focus on prevention; we ruminate about counterfactuals that might have prevented an outcome (e.g., “If only I hadn’t gone for that drive”). In causal reasoning we focus on how an event actually occurred (e.g., a man ran the red light, smashing into my car). Given the history of the investigation, researchers must acknowledge that despite their different foci, counterfactual and causal reasoning are related. Our goal in this chapter is to offer answers to the questions: what is that relation and why does it exist? One suggestion, gleaned from the early literature, might be this: Events people select as “causes” must be earlier events (“antecedents”) that pass the but for test for later outcomes (“consequences”). The but for test of causality relies on counterfactual reasoning as illustrated by the philosopher Mackie (quotation from Lipe 1991: 457–8):

Counterfactual and causal reasoning

29

Mackie (1974) argued that our concept of causation is intimately tied to such counterfactual questions. He argued that when we are able to imagine or observe instances of the effect (the shattering of the goblet) without the proposed cause (the high note), causality is not inferred. However, when we are unable to imagine such alternative situations, the proposed causal link remains intact. Hart and Honoré (1959), cited in Mackie (1974), stated that the lawyer approaches causal statements in the following way, “When it is suggested that A is the cause of B he is apt to ask as the first question, would B have happened without A?” (Mackie, 1974, p. 121). Note the two steps in inferring causality: first, one must be able to imagine that A (the antecedent) can be changed or mutated; and second, such a change must “undo” B (the consequence). Thus, another way to state the proposed relation between causal and counterfactual reasoning would be to say that for an antecedent to be called a “cause” of a consequence, if the antecedent were changed the consequence would be changed. Typically, when participants are asked to select a cause of an outcome, they will select a cause but for which the outcome would not have occurred (e.g., the man running the red light is a cause; if he hadn’t done so then there would have been no accident). However, simple as the example of the but for test seems, and important as it is in the legal system (Hart and Honoré 1985; see Spellman and Kincannon 2001), that test does not work for all causes. The obvious exceptions are situations involving multiple sufficient causes: that is, when two (or more) causes act simultaneously or sequentially and either cause alone would be enough to cause the outcome. Simultaneous multiple sufficient causes are illustrated by an example from Spellman and Kincannon (2001). Participants read the following story: Reed hates Smith and wants to kill him. West also hates Smith (for an entirely different reason) and also wants to kill him. One day Reed shoots Smith in the head. At the exact same instant, West shoots Smith in the heart. Smith dies. The coroner says that either shot alone would have been enough to kill Smith. Participants realized that changing Reed or West’s actions alone would not change the outcome (death) yet they attributed full causality to both Reed and West individually and sentenced them each to maximum jail time. Sequential multiple sufficient causes occur when, for example, A gives X a lethal dose of poison, but before it can take effect, B shoots X dead (see Katz 1987 for the classic legal examples). In such cases, participants tend to attribute causality to B, even though changing B’s actions would not change the fact of X’s death (Greene and Darley 1998; Mandel 2003c). Thus, cases

30

Spellman, Kincannon, and Stose

of multiple sufficient causes show that but for causality does not capture the psychological relation between counterfactual and causal reasoning. Instead, the suggestion we offer here is that counterfactual and causal reasoning are similar (1) because they rely on the same underlying information and (2) because the counterfactuals people consider provide input into their computations of causality.

Our theory of the relation between counterfactual and causal reasoning Our theory of the relation between counterfactual and causal reasoning (or, more accurately, between mutability and causality judgments) can best be explained by reference to one figure and one equation. The figure illustrates the way that mutability and causality judgments are based on similar information; the equation details how specific causality judgments might rely on specific counterfactual information. Before describing the details of our theory, however, we should explicate what kind of judgments it makes predictions about. In experiments, mutability and causality judgments are usually elicited after participants read a scenario. When making causality judgments, participants are typically asked to either list causes of the outcome, rate events that are on a provided list of potential causes, or both. Counterfactual judgments (mutability judgments) are typically solicited in one of two ways. Participants may be asked how a character in the story would complete a sentence beginning with “if only” (as in Kahneman and Tversky 1982b; Mandel and Lehman 1996). Alternatively, they may be asked to list some number of ways in which the story External cues (e.g., description of events; framing of questions)

Internal cues (e.g., potential answers generated to earlier questions)

Availability of alternatives Pre-existing knowledge (especially causal knowledge)

p(Oafter)

p(Obefore)

Counterfactual judgments (what gets mutated)

Causal judgments p(Oafter) – p(Obefore)

Figure 2.1 Illustration of the relation between counterfactual and causal reasoning.

Counterfactual and causal reasoning

31

could be changed (i.e., “mutated”) so that the outcome would be different or “undone” (e.g., Girotto et al. 1991; Wells and Gavanski 1989; Wells et al. 1987). The information The essence of our figure is that mutability and causality judgments rely on similar underlying information but do not rely directly on each other. (Note that this does not preclude one type of judgment from influencing the other indirectly, as explained below.) The relevant information is: (1) the availability of alternatives to the cause and effect and (2) pre-existing knowledge – in particular, pre-existing causal knowledge. Pre-existing causal knowledge is necessary for both kinds of judgments. To make a mutability judgment, the reasoner must know whether changing an antecedent would change the consequence; such a judgment requires causal knowledge. To make a causality judgment, the reasoner must have either knowledge of covariation, beliefs about causal mechanisms, or both. Alternatives to the cause and effect may become more available through both external and internal cues. External availability cues are provided by the experimenter. For example, the order in which a story is presented can affect the availability of alternatives (Byrne et al. 2000; Spellman 1997, 2003) as can the way in which questions about the events are phrased (Mandel and Lehman 1996). Internal availability cues are provided by the participant. For example, when a participant judges both mutability and causality in a single experiment, answers considered during the earlier judgment are more available for the later judgment (Spellman 2003). Consequences of such order effects are described below. The equation: causality judgments for single-occurrence events Following Spellman’s (1997) SPA model (“Spellman Probability-updating Account,” coined by Mandel 2003c), we assume that a causality judgment about a person or event is a function of how much that person or event increases the probability of the outcome above its previous probability.1 Below is a simple equation depicting that function. C  p(Oafter)  p(Obefore)

(1)

Equation 1 says that causality (C) is a function of the probability of the outcome occurring estimated after the target event has occurred (“probability after”) minus the probability of the outcome occurring estimated before the target has occurred (“probability before”). Note that when there are a series of events that might be viewed as causal, each estimate can be inserted in sequence. SPA suggests that if people are asked to identify a single cause of an outcome, they will choose the cause that maximizes C in Equation 1.

32

Spellman, Kincannon, and Stose

The probabilities in Equation 1 can be “unpacked” by expressing each as a probability divided by 1 and then expanding 1 into two complementary components. (Note that this kind of unpacking is common in psychological functions, e.g., Bayes’s theorem.) p(Oafter) p(Obefore) C     p(Oafter)  p(Oafter) p(Obefore)  p(Obefore)

(2)

Given that the events in question are one-time events, how can people estimate the probability that an outcome would happen? One way would be to consider all of the ways that the world could unfold, and sum up, for each way, the product of the likelihood that that “way” would happen [p(way)] and the probability of the outcome given that “way” [p(O|way)]. ways p(way)*p(Oafter|way) C    ways p(way)*p(Oafter|way)  ways p(way)*p(Oafter|way) ways p(way)*p(Obefore|way) C   ways p(way)*p(Obefore|way)  ways p(way)*p(Obefore|way)

(3)

We can use these equations, together with the figure, to explain several phenomena involving counterfactual and causal reasoning. Defining “the outcome” An important question raised by Mandel (2003c; see also Spellman and Kincannon 2001; Strassfeld 1992) is what should count as the “outcome” in these situations. For example, which was the outcome of the Reed and West story: Smith’s death? Smith’s death by gunshot? Smith’s death by two gunshots? In Mandel’s (2003c) Experiment 1, which involved sequential multiple sufficient causes, participants read about a career criminal (Mr Wallace) who is poisoned by Assassin 1, but before the poison can take effect, his car is intentionally run off the road by Assassin 2. Wallace dies when his car explodes. Participants were asked to list and then rate up to four factors that caused Wallace’s death. They were also asked to estimate the probability of Wallace dying “given” neither, each, and both of the assassination attempts. Mandel argues that according to SPA (1) participants should view the poison as most causal (because it most increases the probability of death) and (2) causal ratings should be correlated with the change in estimated probability of death attributable to each assassination attempt. Neither of those results was found. Rather, consistent with Mandel’s Judgment Dissociation Theory, participants rated the crash as most causal of Wallace’s death but assessed the administration of poison as making the biggest change to the probability of death. Mandel proposed the “actuality principle”: when selecting the most

Counterfactual and causal reasoning

33

causal event, participants choose a sufficient condition that plays a direct role in generating the actual specific outcome, not the event that most increases the probability of that general type of outcome. We don’t have a problem with Mandel’s data and we like (and are not surprised by) his point; however, we disagree with the way he applied his results to our theory. When participants are asked what caused Wallace’s death, we agree that most will interpret the question as what caused him to die in this particular manner (i.e., car explosion). We know, in fact, that in cases of sequential multiple sufficient causes, the law (which here is designed to track human intuition; Hart and Honoré 1985) will call Assassin 2 guilty of murder whereas Assassin 1 is only guilty of attempted murder. Why? Because Wallace died from the explosion, not from the poison. We also know (Spellman 2003) that causal ratings are correlated with probability ratings only in hindsight – that is, only when the rater knows what actually caused the specific outcome. For example, “driver inattention” could be rated highly as something that might cause an accident in general, but it might not have changed the probability of a specific accident at hand. Mandel’s dependent variable question, asking for the probability of Wallace dying (generally), thus creates a problem: it calls for an interpretation. Is it asking about the probability of him ever dying? (in which case, the answer is 100 percent). About the probability of him dying in some manner having to do with the information in the story? About the probability of him dying in the exact way he actually died? We believe that different participants in different conditions interpret the question in different ways. We also believe that if the question specified the outcome, so that the interpretation was consistent across conditions (i.e., death in an off-road car explosion), the probability-change data would be correlated with the causality data, just as SPA predicts.

The phenomena Thus, with a properly specified outcome, we can use the equations, together with the figure, to explain several phenomena involving counterfactual and causal reasoning. Phenomenon 1: effects of the number of counterfactual alternatives One prediction that Equation 2 makes is that the more alternative choices there are (some of which would lead to an alternative outcome), the more causality will be attributed to an event that causes the actual outcome to occur. That prediction might be viewed as a generalization of the well known Wells and Gavanski (1989) wine experiment. In that experiment, participants read about a woman who was taken out to dinner by her boss. The boss orders for both of them, but the dish he orders contains an

34

Spellman, Kincannon, and Stose

ingredient – wine – to which the woman is allergic. She eats the dish, gets sick, and dies. In both conditions of the experiment the boss had considered ordering something else: in one condition (one-wine) the alternative dish did not contain the fatal wine, in the other condition (two-wine) the alternative dish also contained wine. Participants rated his choice as more causal of the employee’s death in the one-wine condition (i.e., when he chose between one dish with it and one dish without it). Wells and Gavanski argued that these results show that counterfactual reasoning affects causal judgments because in the one-wine condition, in which there was a counterfactual alternative to the boss’s decision that would have undone the outcome, the boss was viewed as more causal; therefore it must be that the availability of the non-fatal alternative dish affected that causal judgment.2 According to our equations, increasing the number of alternatives that would not lead to the outcome (i.e., increasing p(Obefore)) would decrease the overall “probability before” and thus should increase the causality of an event that does lead to the outcome. The equations predict this result not only for experiments in which there is a 50 percent (one-wine condition) versus a 100 percent (two-wine condition) chance of choosing a bad option, but also for conditions with varying numbers of choices. We set out to demonstrate this predicted effect with a replication and extension of Wells and Gavanski (1989).3 In our modernized variant (Spellman and Meyers 2003), participants read about a seafood restaurant in which either one-of-five, three-of-five, or all five-of-five dishes that one person considered ordering for another (while the other was busy and at the other’s request) contained a “spoiled” ingredient (mussels). A dish containing mussels is ordered and the other person gets food poisoning. Participants were asked how strongly they (dis)agreed with the statement that the decision to order the mussel dish caused the other person’s illness (5  totally disagree; 0  neutral; 5  totally agree). Following Equation 2, participants rated the decision as most causal in the condition in which only one of the five dishes had the bad ingredient (M  1.0); less causal when three dishes had the ingredient (M  0.7); and even less causal when all five dishes had the ingredient (M  0.3).4 Thus, causal ratings not only increase when one outcome-undoing alternative is available (as in Wells and Gavanski 1989), but they also change in predictable ways depending on the number and type of alternatives considered and where those alternatives fit into the equation. Phenomenon 2: order effects when making both counterfactual and causal judgments Several experiments have demonstrated that judging mutability before causality will affect that later causality judgment; a few have demonstrated the opposite (see examples below). Our proposal can explain the discrepancy

Counterfactual and causal reasoning

35

between these results. One key idea is that answers considered when making the first judgment become available for the subsequent judgment – that is, that there are now internal cues to availability. The other key idea is that the existence of outcome-undoing counterfactuals affects the probability information used to judge causality. Mutability before causality: the “if only” effect The “if only” effect refers to the finding that when participants are asked to imagine ways in which an actor in a story might have done something different so that the outcome would not have occurred, causal ratings for those actors increase (Branscombe et al. 1996; McCloy and Byrne 2002). For example, Branscombe et al. (1996) had participants read a story about a date rape and then listen to a mock lawyer’s closing argument suggesting possible mutations to the story. If the argument mutated the defendant’s actions so that the rape would be undone, the rapist was assigned more fault (cause, blame, and responsibility) than if his actions were mutated but the rape still would have occurred. Similarly, if mutating the victim’s actions would undo the rape, she was assigned more fault than if her actions were mutated but the rape still would have occurred (see also McCloy and Byrne 2002; Spellman and Kincannon 2001). In addition to hearing other people’s suggestions for undoing the outcome, generating one’s own undoing mutations can affect later causal judgments. In another experiment by Branscombe et al. (1996), some participants read the same date rape story as described above and were asked to write down a change to the victim’s actions in the story. Participants who changed the victim’s action such that the rape would be undone later rated the victim as more at fault than those whose changes would not have undone the rape. Similarly, Wells and Gavanski (1989) found overall that generating undoing mutations first increased later causal ratings but rating causes first did not affect later mutations. Why should judging mutability before causality have this effect? Imagining alternatives that undo the outcome will create in the participants’ minds more ways in which the outcome could have been avoided. The effect on Equation 2 is that such imagining will increase the denominator (by increasing p(Obefore)) but not affect the numerator for the “probability before.” Thus, the value of that fraction will decrease and the value of the “probability after” minus the “probability before” will increase, resulting in more causality being attributed. Mutability before causality: the “even if” effect The “even if” effect is the opposite of the “if only” effect: when participants are asked to imagine ways in which an actor in a story might have done something different but the outcome would remain the same, causal ratings

36

Spellman, Kincannon, and Stose

for those actors decrease (Branscombe et al. 1996; McCloy and Byrne 2002; Spellman and Kincannon 2001). Why should judging mutability before causality have this effect? Imagining alternatives that do not undo the outcome will create in the participants’ minds more ways in which the outcome would have occurred. The effect on Equation 2 is that such imagining will increase both the numerator and the denominator of the “probability before” (by increasing p(Obefore) on the top and bottom by the same amount). Thus, the value of that fraction will increase and the value of the “probability after” minus the “probability before” will decrease, resulting in less causality being attributed. Causality before mutability: when should it matter? The two examples above demonstrate that judging mutability first can affect subsequent causality judgments – but can judging causality first affect subsequent mutability judgments? Wells and Gavanski (1989) found that they did not. We, however, have created one situation in which they do and one in which they do not. Causality judgments did affect subsequent mutability judgments when we (Spellman 2003) used a variation of N’gbala and Branscombe’s (1995) Experiment 1 story to investigate order effects. (Our story was based on the description in their article; it is closest to their immoral controllable condition.) Participants read about Joe, a father, who was going to pick up his son, Jimmy, from school. On the way to his car, he stopped to talk to some people. Meanwhile, a neighbor drove by the school, waited with Jimmy for fifteen minutes, and then offered to drive the boy home. On the way home, a drunk driver came out of nowhere and struck the car, and Jimmy was seriously injured. Participants judged both mutability and causality. For the mutability judgments they were asked to complete the sentence: “The outcome of this event might have been different IF ONLY . . .” For the causality judgments they were provided with a list of people – Joe, Jimmy, Neighbor, Other Driver – and asked to rate from 0 (not at all) to 10 (very much) how much a cause of the outcome each of them was. Half of the participants judged mutability first; the other half judged causality first. When the tasks were done in the same order as in N’gbala and Branscombe (1995), that is, mutability listings before causality ratings, our results were similar to theirs: Joe was mutated most often and Other Driver was rated as most causal. However, reversing the order in which the tasks were done affected the mutability task (listing) but not the causal task (rating): participants again viewed the Other Driver as most causal but now he was most often mutated, too. Why were the mutability judgments significantly affected by the causality judgments? When doing mutability listings first, participants are influenced

Counterfactual and causal reasoning

37

by external availability from the story (primacy, focus, etc.). However, when participants do a causal rating task in which all of the names are provided by the experimenter, those names are now all highly available for mutating. The Other Driver was rated, on average, as much more causal than Joe. Thus, participants would have him available for mutating and they would notice that changing his actions would change the outcome in a big way. In contrast to the mutability task, causality ratings did not change significantly as a result of order. The magnitudes of the causal ratings for Joe and the Other Driver were related to how the participants responded to the mutability question, but not to task order. One might conclude, therefore, that order effects exist for both types of judgments. However, we believe that causality judgments do not affect subsequent mutability judgments when the effects of availability are removed. The literature has largely overlooked differences in measures. Mutability is almost exclusively measured by listings – looking at what a participant thinks up first. Causality is usually measured by ratings – looking at what a participant rates highest; importantly, sometimes the rated items are generated by the participant and sometimes they are provided by the experimenter. We believe that tasks that involve “thinking up things” are more subject to availability than those that do not. (Consider the differences between recall and recognition tests.) Therefore, to unconfound listing/rating with mutability/causality we devised a new way to measure mutability – mutability ratings on experimenter-provided alternatives (Spellman 2001, 2003; note that Mandel 2003c uses mutability ratings on participantgenerated alternatives). Participants read a story that we have used in several other experiments: A young woman was driving home from work. She had left early that day because it was a holiday weekend and traffic was very heavy. She was the first car to stop at a particular red light. Behind her was a long line of cars with a school bus at the end. As she was waiting for the light to turn green, she reached down to change the radio station. At that moment the light finally turned green, but she took an extra few seconds to find a song she liked. She then accelerated and the cars and bus accelerated behind her. Just as the school bus got into the intersection, a car driven by an upset man who had been fired that day came screaming through the red light from the other direction, hitting the bus and injuring many children. Typically, when asked to list something in the story that would change the outcome, participants are most likely to mutate something involving the woman: the time she left work, changing the radio station. When asked what caused the outcome, participants give their highest ratings to things involving the man: getting fired, running the red light (Spellman 2001, 2003).

38

Spellman, Kincannon, and Stose

In this particular experiment, however, we used ratings of experimenterprovided alternatives for both mutability and causality judgments. After reading the story, participants saw a list of twelve events from the story that had been generated by participants from previous experiments in response to either a mutability listing or causal listing question (e.g., the woman changing the radio station, the man running the red light, school being open that day). For each event, participants were asked to rate the extent to which they agreed or disagreed that the event was mutable and causal. All participants rated both mutability and causality with half the participants doing the tasks in each order. On average, mutability agreement ratings were higher than causality agreement ratings for almost all events; in fact, the only items with positive causality agreement ratings (when those ratings were made first) were things involving the man (i.e., being fired, being in a bad mood, running the red light). More important, the order in which participants judged mutability and causality affected their ratings. The most interesting finding is that rating mutability first significantly increased (most) later causal ratings whereas rating causality first had only small effects on later mutability ratings. Why did the causality judgments show task order effects? When judging mutability, one imagines alternatives to the event; such imagined alternatives should reduce the “probability before” estimates of the outcome occurring, and therefore, according to Equation 2, increase the later causal ratings of the events. Why did the mutability judgments not show task order effects? Usually (i.e., in cases not involving multiple sufficient causes), what people pick as causes are already but for causes – things that, when changed, will undo the outcome. Thus, rating causality first (1) does not make any new events more available for mutation (as long as the mutation task also has experimenter-provided alternatives) and (2) does not provide information to any mutability-computation function.5 But why did mutability judgments show order effects in our replication of N’gbala and Branscombe (1995) but not in the “Bus” experiment? In the replication, mutability was measured by listings; thus, items provided by the experimenter for the causal rating task became more available for the mutability listing task. However, when all items are provided to the participants within the mutability task, and we measure ratings rather than listings (as in the “Bus” experiment), such “availability” has no meaning or effect. Phenomenon 3: action and inaction effects in regret judgments Until now we have focused solely on mutability and causality judgments. We believe, however, that our equations can be used to explain the so-called “action” and “inaction” effects for feelings of regret. The “action effect” – the idea that people would feel more regret for actions than for inactions – was first described by Kahneman and Tversky (1982b). In their classic experi-

Counterfactual and causal reasoning

39

ment, participants read about Mr Paul and Mr George. Mr Paul owns stock A. He thinks about switching to stock B but decides against it. Later he discovers that he would have been better off by $1,200 if he had switched. Mr George owns stock C. He thinks about switching to stock D and decides to do so. Later he discovers that he would have been better off by $1,200 if he had not switched. When asked who would feel more regret for his actions, a majority of participants judged that Mr George would feel more pain. Action or inaction? Subsequent attempts to explain or reduce the action effect considered various aspects of the scenarios and the judgments, including: whether the alternative outcomes were known or unknown to the protagonist (i.e., do they know what would have happened on the road not taken), whether the outcome was good or bad, whether to consider the protagonist’s regret in the short or long run (see, e.g., Byrne and McEleney 2000; Gilovich and Medvec 1994; Landman 1987). Another breakthrough came when Zeelenberg and colleagues (Zeelenberg et al. 1998a; Zeelenberg et al. 2000c; Zeelenberg et al. 2002) suggested that regret for actions versus inactions should depend on whether it seems prudent (or not) to keep the status quo based on the outcomes of prior instances. For example, if a soccer team is having a winning season, a coach who changes his players and loses should feel more regret than a coach who keeps his players and loses (an action effect). However, if the team is having a losing season, a coach who keeps his players and loses should feel more regret than a coach who changes his players and loses (an inaction effect). Zeelenberg and colleagues argue that people experience regret when decisions appear abnormal relative to previous outcomes. These results follow from our view of how regret is related to causal and counterfactual reasoning (although we don’t believe that “abnormality” is the mechanism). We assume that regret arises as an after-the-fact emotion when, in retrospect, you believe that you had the power to choose a course of events that would have been more likely to lead to the desired outcome than the one you actually chose. Note several important factors: (1) regret requires a choice among actions, at least some of which would change the probability of the outcome (i.e., the decision is causal); (2) regret results from a comparison of what one actually did to what one could have done (i.e., it involves comparing the “factual” to one or more counterfactuals); and (3) regret requires having chosen what seems to be (in hindsight) a less promising choice. Our regret data We have replicated Zeelenberg et al.’s (2002) action and inaction effects by manipulating the prior record of a baseball pitcher and measuring the

40

Spellman, Kincannon, and Stose

coach’s post-decisional regret. Participants read about two baseball teams each with records of ten wins and ten losses. One coach decides to play his regular starting pitcher, whose record is either 7–3 or 3–7; the other coach decides not to play his regular starting pitcher, whose record is also either 7–3 or 3–7. Both teams lose. Participants were asked to estimate (1) the probability of the team winning given the coach’s actual decision and (2) the probability of winning, imagining that the coach had made the counterfactual decision. Some participants estimated those probabilities in foresight (before learning that the team lost); others estimated those probabilities in hindsight (after learning that the team lost). All participants were also asked to judge how much regret each coach would feel about his decision (on a scale from 1  no regret at all to 10  completely regrets). Table 2.1 shows some of the results from this experiment. Note that the percentages are expressed for the chances of winning, not losing (which was the actual outcome). Participants clearly believed that the decision to keep or change the starting pitcher would affect the team’s probability of winning – as can be seen by comparing the foresight factual and counterfactual probability estimates. The action effect can be seen in the top half of the table. When the pitcher had a winning record and the coach kept him, the probability of winning given the factual decision (M  58) was estimated as higher than the probability of winning given the counterfactual decision (M  40); thus, the coach would feel relatively little regret for his good decision (M  5.7) Table 2.1 Results from the causality, counterfactual reasoning, and regret study Record/action

Probability of winning judgment

Regret (1–10)

Factual

Counterfactual

Winning record Keep Starter Foresight Hindsight

53 58

43 40

F  Cf

5.7

Change starter Foresight Hindsight

43 38

53 60

F  Cf

8.5

Losing record Keep starter Foresight Hindsight

43 34

64 68

F  Cf

7.1

Change starter Foresight Hindsight

64 67

43 33

F  Cf

6.3

Counterfactual and causal reasoning

41

despite the loss. When the pitcher had a winning record but the coach changed him, the probability of winning given the counterfactual was higher than given the factual; thus, the coach felt more regret (M  8.5). The inaction effect can be seen in the bottom half of the table. When the pitcher had a losing record but the coach kept him, the probability of winning given the counterfactual was again higher than given the factual; thus, the coach felt a lot of regret (M  7.1). However, when the coach changed the pitcher with the losing record his decision was seen as better than the counterfactual decision, so his regret rating was small (M  6.3). We therefore see an action effect for winners (more regret when the pitcher is changed) and an inaction effect for losers (more regret when the pitcher is kept). Because some of these measures were collected between subjects, we are limited in which correlations we can run; further research will remedy that problem. Other possible extensions: the hindsight bias We believe that our proposal has consequences for understanding other related forms of reasoning including the hindsight bias (see Nario and Branscombe 1995; Roese and Olson 1996). In hindsight bias experiments, participants are asked to estimate the left-hand probability of Equation 1 (“probability after”); that is, estimate that the actual outcome will occur given that the target event has happened. In the control condition, participants make that estimate without knowing what outcome will occur. In the experimental condition, participants make that estimate after learning which outcome (e.g., winning the game) has occurred – but they are supposed to make the estimate as if they had not yet learned the outcome. The hindsight bias refers to the finding that after learning the outcome, experimental participants cannot put themselves back into the “unknowing” state, and their “probability after” estimates for the actual outcome are higher than those of control participants. The knowledge that an outcome has occurred fosters the belief that it was more likely to occur (compared to the control condition). Increasing the “probability after” for the actual way it happened in both the numerator and denominator will increase the value of the fraction in Equation 2. Counterfactual reasoning has often been proposed as one way to decrease the hindsight bias by attacking that change in belief in the likelihood of the actual outcome occurring. Sometimes debiasing is partially successful: participants who are asked to imagine ways in which the outcome might not have occurred often demonstrate a reduction of hindsight bias relative to those who did not imagine such alternatives (e.g., Fischhoff 1982); other times, however, attempts to debias by providing (more) alternatives seem to backfire (e.g., Roese and Olson 1996; Sanna et al. 2002).6 We are just beginning to experimentally test the predictions that follow

42

Spellman, Kincannon, and Stose

from our equations regarding the hindsight bias. For example, our equations predict that some kinds of counterfactuals should be more effective in reducing the hindsight bias than others (depending on how they affect the various components in Equation 3). The equations also suggest that estimates of the existence of unarticulated (i.e., not specifically generated) counterfactuals may explain some of the “backfiring” effects.

Conclusion We have argued that despite their differences, counterfactual and causal reasoning are related in that: (1) they rely on similar underlying knowledge and information and (2) making one type of judgment may provide information for (and thus affect) judgments of the other type. We have described how, for something to be called a cause, it need not provide an outcome-changing counterfactual (i.e., it need not be a but for cause – although it usually will); yet we still believe that counterfactual reasoning informs our causal reasoning. It does so in two ways. First, judging mutability before causality may bring to mind, and make more available, the mutated item (e.g., action or event) for consideration in the later causality judgment. Second, according to our equations, imagining ways in which an outcome could have been undone (or not) will affect our beliefs about the likelihood of that outcome occurring, which, in turn, will affect our beliefs about causation. Making causality judgments first, however, should not affect mutability judgments unless the experimenter provides items (external cues) for the participants to rate as to causality and then asks the participants to generate mutations. Generating causes before mutations will not change those latter judgments because causes, when undone, typically will change an outcome; thus, the causes considered will be a subset of the mutations considered. Our theory seems to capture several similarities and differences between counterfactual and causal reasoning. At present, we are trying to extend it to explain judgments of regret and disappointment and the hindsight bias. Indeed, because counterfactual and causal reasoning are so important for, and so ubiquitous in, thinking generally, we are optimistic that understanding them will give us insight into some of the other complex capabilities of human thought.

Notes This research was supported by a grant from the US National Institute of Mental Health (NIMH) to the first author. We would like to thank Hayley Daglis for technical assistance. 1 This view is somewhat analogous to the view in the covariation literature (Cheng 1997; Cheng and Novick 1992) that a cause is something that increases the probability of an effect above its baseline probability (see Spellman 1996, 1997, for that argument). 2 It is not clear to us why that explanation is required by the results. There was a

Counterfactual and causal reasoning

3

4

5 6

43

change made to the story and that change might have affected counterfactual and causal reasoning separately and directly without the former mediating the latter. Our lab and several others (e.g., Rachel McCloy, personal communication) have had difficulty replicating the wine experiment ten or more years later. Few participants mutate the boss’s decision or agree that his decision caused the employee’s death. These days, many participants note that people should know wine is a common ingredient in cooking; many state that the woman should have told the waiter about her allergy; some say that she should have been carrying drugs to combat her allergic reaction; others want to know why she would allow her boss to take her out to dinner, much less order for her. When all participants are included, as shown in the text, the omnibus ANOVA is significant but only the one-of-five and five-of-five conditions are different from each other. This experiment was run within subjects and when participants who gave equal ratings to the three conditions are excluded, the means are all significantly different from each other: 1.4, 0.7, and 1.6. We believe that such participants were likely to think something other than the ordering decision, for example, the restaurant, was the cause of the illness. If anything, rating causality first should lower some mutability ratings, as people have now probably considered the true causal factors and rated those items as most mutable. See Spellman (2003). Figuring out how to decrease the hindsight bias is of particular interest to lawyers whose clients have been sued for negligence (Kamin and Rachlinski 1995; Spellman and Kincannon 2001). A finding of negligence requires an afterthe-fact determination of what should have been before-the-fact foreseeable.

3

The course of events Counterfactuals, causal sequences, and explanation Denis J. Hilton, John L. McClure, and Ben R. Slugoski And every fair from fair some time declines, By chance, or nature’s changing course, untrimm’d; (Shakespeare, Sonnet 18)

Causal explanations can help us understand why events change course, and why the world turned out differently to what we might have expected. Even in understanding simple narratives of events, people form mental representations that incorporate causal inferences (e.g., Graesser et al. 1981; Hilton 1985; Read 1987; Schank and Abelson 1977; Trabasso and Sperry 1985; Trabasso and van den Broek 1985; Wilensky 1983). In addition, research on counterfactual reasoning has also used simple narratives to explore which factors are likely to be mutated (e.g., Byrne et al. 2000; Wells et al. 1987), and has demonstrated the complexity of the mental representations of these narratives through showing, among other things, that people do not always mutate causes when responding to counterfactual “if only” probes (Mandel and Lehman 1996; N’gbala and Branscombe 1995). Finally, in their analysis of close counterfactuals, Kahneman and Varey (1990) argue that people not only represent what happened, but also what almost happened. In contrast, much research on causal attribution has used simple stimuli that do not even have minimal narrative but instead present covariation information (e.g., Cheng and Novick 1990; Foersterling 1989; Hilton and Jaspars 1987; Hilton and Slugoski 1986; McArthur 1972; Sutton and McClure 2001). Indeed, some attribution research, such as that on verb effects, presents only the target event itself (e.g., Brown and Fish 1983; McArthur 1972) and examines how people may generate antecedents and consequents for those target events (e.g., Semin and Fiedler 1991). Nevertheless, even the most cursory consideration of how people interpret social life will reveal the human propensity to generate causal chains when seeking to understand events, whether it be public ones such as a space shuttle disaster (Hilton et al. 1992), the death of Princess Diana, the Concorde disaster, the attack on the Twin Towers (9/11), etc., or private ones such as an accident, a divorce, a rejection, and so on. Following Kelley’s (1983) call to pay

The course of events

45

more attention to complex causal structure, we extend earlier work (e.g., Alicke 1991; Brewer 1977; Brickman et al. 1975; Johnson et al. 1989) by considering different kinds of causal chains, and how explanations for outcomes are selected from these complex sequences of events. To do this, we need to introduce the arrow of time into models of counterfactual reasoning and the explanation of human action.

Causal explanation of events versus causal induction of characteristics As Kahneman and Varey (1990) argue, counterfactual simulations have a unique role in aiding judgments of causality. This may happen automatically, when we are surprised by a course of events turning out differently to what we had expected (Kahneman and Miller 1986), or deliberately, as when we evaluate whether a candidate factor is indeed a cause through what if reasoning about whether the target outcome would have happened if the antecedent had not (e.g., Sloman and Lagnado 2004). Whether performed automatically or deliberately, counterfactual reasoning may yield a plethora of necessary conditions for the outcome (Hesslow 1988). For example, one may reason that Princess Diana would not have died in a car accident under the Alma bridge in Paris in 1997: if her car had not been chased by paparazzi, if she had been wearing a seat-belt; if her chauffeur had not been drunk; if the tunnel under the Alma bridge had had protective barriers; if her car had run out of petrol before reaching the bridge, and so on. And so on back in time: what if she had not been seeing Dodi El-Fayed; what if she had not been estranged from Prince Charles, and so on? This plethora of necessary conditions brings in its train the problem of causal selection, as normally we only mention one or two factors in a conversationally given explanation (Hesslow 1983, 1988). In such “stories,” counterfactual simulation of what would have happened if an antecedent had not been the case produces a contrast case that is essentially historical in nature. It relies on reasoning about particulars, namely events involving particular individuals and entities that happened at particular times and places. Counterfactual reasoning enables us to wonder about how a story would have continued if a relevant antecedent had not occurred. Would Diana still be alive if she had not met Dodi? Would the patient have got better without being administered the drug? Would there still have been a Second World War if Britain and France had not signed the Treaty of Munich? A complex set of events is imagined, constituting a set of multiple necessary conditions for the outcome, organized in a temporally sequenced causal chain where each event is a precondition for the next one to act (e.g., the Munich treaty led to Hitler’s occupation of Czechoslovakia in 1938, which encouraged him to invade Poland in 1939, which led Britain and France to declare war on Germany, etc.). These support particular causal explanations about historically located events that involved identified actors and places.1

46

Hilton, McClure, and Slugoski

In contrast, causal induction is essentially ahistorical in nature, and seeks to test if general causal relations hold between types of event. If in clinical trials we administer a particular kind of drug to an experimental group, and a placebo to a control group, our aim is to go from the observation (say) that AIDS was less often observed in the experimental group than the control group (a set of particular observations) to timeless universal generalizations of the kind AZT cocktails of Type X retard AIDS. Such inductions aim to establish causal laws that do not refer to particular individuals, places and times as they hold across them. In science these universal generalizations may be thought of as covering laws (Hempel 1965), though informal versions may be thought to exist as generalizations in semantic memory of the kind people do what they want to do, pushing things causes them to fall over, gravity causes objects to fall, aspirin relieves headache, imbalance of power leads to war, etc. (Hart and Honoré 1985; Hilton 1988, 2001; Tetlock and Lebow 2001). Often, causes may act in interaction to produce an effect, as when we argue that striking a lighted match in the presence of gas produces fire. These kinds of interactions of independent factors (as manipulated in a fully crossed experimental design) are what are understood in Kelley’s (1973) ANOVA framework as a multiple necessary conditions (MNC) schema.2 However, no relation of temporal precedence between factors exists in the ANOVA framework presupposed by this notion of causal schema: it does not matter for the production of the effect whether we strike the match first (and then introduce the gas) or first introduce the gas and then strike the match. Of course, some types of causal mechanism are sequential in nature, but this is not the same as saying that they are historical in the sense of being assertions about particulars. We may believe poverty breeds discontent, and hence social strife, but this is still an ahistorical generalization that can be fitted to many peoples, socio-economic situations and periods. Universal generalizations are used to support counterfactual projections, by telling us what would normally have happened, if say, the Marshall Plan (a certain kind of economic policy) had not been adopted in post-1945 Europe, or if a certain kind of drug had not been administered to a patient last week. They can be represented probabilistically (e.g., Salmon 1998), or elliptically as “gappy” generalizations (Mackie 1974). However, establishing or supporting a generalization is not the same thing as supporting a particular counterfactual projection, as it is quite possible for something that is statistically (or generally) a cause of an outcome to have the opposite effect in a particular case. For example, miscuing shots at goal in soccer may generally lessen your chances of scoring goals (and thus be a weak or even negative covariate of scoring), though in an exceptional case a miscued shot may bounce off a defender and past the stranded goalkeeper into goal and thus be (indirectly at least) a cause of this particular goal (see Hesslow 1976 and Salmon 1998 for further discussion).3 One implication of the distinction between particular events and general

The course of events

47

laws is that it allows us to define two different classes of counterfactual world. The stories of individuals could have turned out differently than they did, but even had they done so, this counterfactual world would still be a normal one obeying familiar laws. But if the laws of the world suddenly changed – apples fall upward, medicines harm rather than heal, poverty breeds contentment, people can travel in time – the very nature of the world we know will have changed. Whereas the first kind of story might be the stuff of all kinds of biography and novels, the latter kind of counterfactual seems to be a matter for science fiction. We will concentrate our attention on the first kind of story – where events but not laws of nature could have been different – as legal theorists (Hart and Honoré 1985) have argued that the causal question everyday explanation addresses is why particular events happened when and how they did, in contrast to scientific explanation, which seeks general laws that govern relations between classes of events (see Hilton 2001). In addition, we will pay particular attention to the question of whether the mutability of an event (especially a human action) determines its selection from a causal chain as an explanation for an outcome (cf. Kahneman and Miller 1986; Wells et al. 1987).

Selecting causes from conditions in causal chains Why are intentional actions preferred as causes? All the events in a causal chain could be thought to be part of the full cause of the outcome, yet we typically only select one or two of these factors to mention in everyday explanation. So how do we select causes from such a plethora of conditions? Hart and Honoré (1985) identify two separate criteria for this task. In their classic treatise on the use of everyday concepts of causation in legal reasoning, Causation in the Law, they wrote that: Perhaps the only general observation of value is that in distinguishing between causes and conditions two contrasts are of importance. These are the contrast between what is abnormal and normal in relation to any given thing or subject-matter, and between a free deliberate human action and all other conditions. (Hart and Honoré 1985: 33) The abnormality criterion can be used to select causes from conditions, through determining which condition should be considered to have “made the difference” between the effect’s occurrence in the target case and its nonoccurrence in “normal” circumstances. In the case of Princess Diana’s accident this would eliminate “normal” conditions such as the car having petrol in its tank as candidate causes. What is considered as “normal” may be further constrained by perspective: for example, most people do not have to experience being chased at high speed by paparazzi, and for this the

48

Hilton, McClure, and Slugoski

paparazzi were soon blamed for the princess’s death. However, from the paparazzi’s point of view, such chases are part and parcel of their everyday life, and for them the factor that “made the difference” between the accident and a normal chase without incident had to be something abnormal about the chauffeur, such as his being drunk. However, Hart and Honoré (1985) argue that causality will be traced through the abnormal condition if there is an intention lying behind it. For example, if (as some conspiracy theorists seemed to believe) the driver of Diana’s car had been drugged by others wishing to bring about the accident, his lack of fitness to drive would merely be considered as a means to the assassin’s end. Both abnormality and intentionality have also been shown to be major determinants of counterfactual mutability (see Kahneman 1995 for a review) and this may in part explain why they figure so prominently as causal explanations. We have dealt with the abnormality criterion elsewhere (e.g., Hilton and Slugoski 1986), and here concentrate our attention on why the intentionality criterion is important for causal attributions in causal chains. Is it because actions are (1) highly mutable, (2) have high covariation with their effects, or (3) are “sufficient in the circumstances” for the effect in question? Or is some other criterion responsible for causal preference, for example an element’s (un)predictability, given prior occurrences in a causal chain? Support for the view that their higher mutability leads intentional actions to be preferred as explanations can be adduced from Girotto et al.’s (1991) study, which shows that controllable actions in a causal chain are frequently mutated in response to “if only” probes in order to undo an outcome. Since ease of imagining alternative actions or events that would undo an outcome should increase the perceived necessity of a target cause for an event, highly mutable actions should be judged as more necessary causes than other less mutable elements in a chain (cf. Spellman et al., Chapter 2 of this volume). This contrasts to the covariational view that not only the necessity but also the sufficiency of a cause should correlate positively with causal preference, leading to a preference for voluntary and deliberate actions as explanations due to their superior covariation with effects. The covariational model (Kelley 1967: 154) states that a cause will be “that condition which is present when the effect is present and which is absent when the effect is absent.” Cheng and Novick (1990, 1992) formalize this idea in terms of Delta-P, which is given by the proportion of times effect occurs when candidate cause is present, less the proportion of times effect occurs when candidate cause is absent. Third, intentional actions may be preferred as explanations because they are judged to be more “sufficient in the circumstances” (Heider 1958; Mackie 1974; McClure and Hilton 1997) than other factors. Finally, elements may be selected as explanations in causal chains because of their relation to other elements in the chain. For example, people may do an “intuitive multiple regression” analysis when analysing causal chains to

The course of events

49

eliminate redundant parts of the chain from the explanation (Hilton 1988). This kind of approach has found its most specific and testable expression in Spellman’s (1997) crediting causality model (but see Brickman et al. 1975; Brewer 1977; Fincham and Jaspars 1983; Fincham and Shultz 1981 for related approaches). As applied to causal chains, this implies that an item in a chain of events will be identified as causal if it increases the probability of occurrence of the effect, beyond the effect’s prior probability due to the presence of prior elements already in the chain. The crediting causality model operationalizes Salmon’s (1984) notion of successive reconditionalization using the following five steps: 1 2 3 4 5

Determine the probability of the actual outcome before the first proposed cause occurs. Determine the probability of the actual outcome after the first proposed cause occurs but before the second proposed cause occurs. Find the change in probability of the actual outcome due to the first cause (the difference between 1 and 2). Find the change in probability of the outcome due to the second cause (the difference between 2 and the outcome). Assign causality based on a comparison between 3 and 4.

A problem in assessing the role of counterfactual reasoning in causal attribution is that mutability has not been given a consistent operationalization – does it refer to the tendency to mutate an element in response to an “if only” probe, or the ease of imagining an alternative that would undo the outcome? For present purposes, we will suppose that mutability in its second sense should impact on necessity judgments; for example, Hilton and Erb (1996) have shown that making alternative causal scenarios for an outcome (e.g. a patient getting cancer) salient reduces the perceived necessity of the target cause for that outcome, and hence its quality as an explanation (see also Fischhoff et al. 1978). An additional problem rendering comparisons between studies different is that prior research on counterfactual reasoning and causal attribution from sequences of events has used a number of different kinds of causal chain. Before addressing the question of whether the preference for intentional actions as explanations in causal chains can be explained in terms of the mutability, covariation, sufficiency or crediting causality principles we therefore introduce some distinctions between kinds of causal chain (see Table 3.1). Kinds of causal chain Spellman (1997) found considerable support for the crediting causality model using stimulus stories describing noncausal temporal chains describing coin tosses or simple games with explicitly manipulated probabilities of success. A strength of the crediting causality model is that it can explain

50

Hilton, McClure, and Slugoski

Table 3.1 Five kinds of causal chain Temporal (Miller and Gunasegaram 1990) Jones and Cooper must both toss the same in order to win Distal does not constrain proximal Events are temporally switchable Coincidental (Wells et al. 1987) Speeding ticket → short cut → glass  flat tire → caught up in rush hour → being delayed by senior citizens at crossing → being late Distal does constrain proximal A causes B, B causes C, etc., in this particular case Events switchable in time in general, but not in this particular case Unfolding (Causal–generative/mechanistic) (Hart and Honoré 1985; Hilton et al. 2003) Storm(s) → ice formation → skid(s) → crash(es in general) Distal strongly constrains proximal A causes B, B causes C, etc., in this particular case, but also in general Events are not switchable in time, either in this case, or in general Opportunity chains (Hart and Honoré 1985; McClure et al. 2003) Youth sets fire to shrub  breeze fans flames → forest fire Distal enables proximal A “enables” B, B causes C Events are not switchable in time, neither in this case, nor in general Pre-emptive (Halpern and Pearl in press-a, in press-b; Mandel 2003c) Murderer poisons victim whose fate is sealed, but who is murdered by another person before he dies of poison Distal precedes proximal A is independent of B, B causes C in this case A causes C in general, B causes C in general Events are switchable in time in this case, and can be switched in general

recency effects in attributions in temporal chains, but primacy effects in coincidence chains. Thus, in a temporal chain, participants read about two men, Jones and Cooper, who have to both toss a coin and if the two coins come up both heads or both tails then they win a prize (Miller and Gunasegaram 1990). In one case where participants are told that Jones tosses heads first, then Cooper tosses heads second, and asked to judge who is more causal, they tend to say Cooper. This can be predicted from the crediting causality model as the first event (Jones’s toss) does not change the probability of winning (it is still 0.50), whereas the second event (Cooper’s toss) does makes the difference between winning and losing, thus increasing the probability of winning from 0.50 to 1 if it comes up the same as Jones’s toss.4 On the other hand, in what we call coincidence chains, where the protagonist suffers a number of normally unconnected minor misfortunes, the cred-

The course of events

51

iting causality model correctly predicts primacy effects in attribution. Spellman (1997) analyzes the following example scenario from Wells et al. (1987): The speeding ticket put William behind schedule, which made him take a short cut. Here he encountered glass, causing a flat tire, which, after changing, put him in the time frame of the rush hour, which in turn led him to contend with a group of senior citizens at a crosswalk. (Wells et al. 1987: 423) In a counterfactual “if only” reasoning task, participants tend to undo the first item, consistent with Wells et al.’s interpretation that the first event “constrains” the subsequent ones, and makes it difficult to imagine mutating the subsequent events without also mutating the prior ones. As it is the first event that makes the outcome “inevitable” in the circumstances, the crediting causality model would predict that it would be selected as the cause. Note that the explanatory preference for the first event cannot be attributed to any general covariation that it might have with the target effect, as the four kinds of event in the chain were rotated over temporal position in Wells et al.’s experiment. In what we call unfolding causal chains, the successive events follow normally and foreseeably from each other in a directed causal sequence. There is a logic of mechanism in their sequence, and they cannot be switched around and still produce the effect. They can even be thought of as sequenced “recipes” for bringing about effects, either in the sense that a saboteur may intentionally concoct a plan to bring about an aircraft crash, or simply in the sense that an unattended piece of debris on runway is a “recipe for disaster.” These chains thus involve a stronger sense of causation than that implied in the coincidence chains, in the sense that once the first cause has acted, the chain of causation only has to run normally on for the outcome (e.g. accident) to happen (e.g., the debris pierces the fuel tank, leading the fuel to escape, leading to a catastrophic fire as in the case of the Concorde crash in July 2000). Thus each prior event in the chain would not only be necessary and sufficient in the circumstances for the occurrence of the subsequent events in this particular case, but is also generally a covariate with the subsequent events and of the final effect. In what we term opportunity chains the action of a first cause creates an opportunity for the second cause to act. For example, an intentional action (throwing a cigarette butt into scrub) or a natural cause (the refraction of light through broken glass) may cause some shrub to smolder, allowing a second cause to create a bush fire; this second cause could be another intentional action (someone pours petrol on the flames) or another natural cause (e.g., wind springs up). Here the first cause will increase the probability of the outcome, but the second cause should make the outcome definite (i.e., increase the probability to 1).

52

Hilton, McClure, and Slugoski

In what we call pre-emptive causal chains an unfolding causal chain, which would have produced the outcome, is interrupted by a new causal chain, which takes over and actually produces the effect. An example would be of a fire, which, if left to itself, would inevitably burn a house down. However, another fire reaches the house first, thus pre-empting the first fire from running its course and burning down the house (Halpern and Pearl in pressa, in press-b; see also Mandel 2003c). Mandel argues that pre-emptive causal chains thus cannot be explained by the crediting causality approach, as the designated cause does not increase the probability of the outcome. Predicting causal selection of intentional actions in unfolding causal chains In unfolding causal scenarios, Hart and Honoré (1985) predict selection of proximal or distal causes as a function of whether distal cause is an intentional action or not. In the following scenario, adapted from one of their examples, Hart and Honoré would predict attribution to the proximal cause (the icy road) as the underlying distal cause is itself an abnormal condition of nature (a severe storm): During the winter, a severe storm wetted a road, thus bringing about the formation of icy patches on a country road while the temperature was below zero. Afterwards, a car came round a bend on this road, skidded on the icy patches and rolled over before colliding with a tree, resulting in considerable damage. The car driver had to be taken to hospital. (Hilton et al. 2003) However, if the distal cause is a voluntary and deliberate action, Hart and Honoré predict that causality would be traced through the proximal cause (the icy road) to the distal cause (the human action). They argue that: . . . we do not hesitate to trace the cause back through even very abnormal occurrences if the sequence is deliberately produced by some human agent: if I take advantage of the exceptional cold and plan the car accident accordingly by flooding the road with water my action is the cause of what happened . . . the icy condition of the road is now regarded as a mere means by which the effect is produced. (Hart and Honoré 1985: 43) Hart and Honoré would therefore predict attribution to the distal cause (the act of spraying the road) when the scenario was introduced with a voluntary and deliberate distal cause, such as: “During the winter, a man who wished to cause an accident sprayed a road with water, thus bringing about . . .” In addition, Hilton et al. (2003) created a condition with voluntary but nondeliberate distal causes where an action accidentally brought

The course of events

53

about the outcome in an unintended, nondeliberate fashion, introduced as follows: “During the winter, a man flooded a road without thinking about the consequences, thus bringing about . . .” Both scenarios then continued exactly as did the natural cause scenario. In our experiments the narrative would first mention a precondition, such as sub-zero temperature, which enabled the operation of the subsequent causes but which itself stayed unchanged through the period in which the action took place. The distal cause is mentioned second, which would either be a deliberate human action (spraying the road deliberately), a nondeliberate human action (spraying the road without thinking about the consequences), or an act of nature (e.g., a storm). The third link to be mentioned was the proximal cause (e.g., the formation of ice on the road), and the fourth link was the immediate cause (e.g., the driver’s loss of control of the car). Finally, the outcome itself (e.g., the car crash) would be described. Whereas the first link (the precondition) stayed unchanged through the narrative, the second, third, and fourth links described an unfolding chain of events that resulted in the final outcome. Each scenario was written to instantiate a causal chain with the following structure: Precondition → Distal → Proximal → Immediate → Outcome cause cause cause e.g., e.g., e.g., e.g., e.g., sub-zero spraying presence of loss of control crash temperature road ice on road of car The participants were first asked to evaluate how good each of the four factors in the chain was as an explanation. On a facing page, following a procedure used by Brickman et al. (1975), they were asked to evaluate the probability that the event would happen in the circumstances given the presence of each link (sufficiency), and in its absence (necessity), using an eleven-point percentage scale. Thus the study obtained judgments of four types of explanation (precondition, distal cause, proximal cause, immediate cause) across the three types of distal factor (voluntary and deliberate human action, voluntary but not deliberate human action, natural abnormal condition). Overall, our manipulations were successful in focusing participants’ attention on the parts of the chain that most interest us, in that 85 percent of the attributions were to distal (link 2) or proximal (link 3) causes, none was to preconditions (link 1) and only 15 percent to immediate (link 4) causes. These overall patterns were replicated in a second experiment that required participants to evaluate the causes on seven-point scales. Finally, in line with Hart and Honoré’s predictions, we found a preference for proximal causes when the distal cause was itself a natural event, but causality was traced through the proximal cause to the distal cause when the latter was a voluntary and deliberate action. Below we turn to the question of why these causal preferences emerged.

54

Hilton, McClure, and Slugoski

As all the necessity of all four elements (precondition, distal, proximal and immediate causes) was judged very high in all three experimental conditions (the twelve mean ratings of the likelihood that the outcome would have occurred in the absence of any one of the target factors ranged from 13 percent to 23 percent), there was unsurprisingly no correlation between necessity judgments and causal evaluations. As participants considered all four items in the chain to be highly necessary for the final outcome, this meant that all four could in principle be mutated to prevent the outcome. However, the finding that all were almost equally “mutable” in the sense of being all perceived as highly necessary obviously rules out this version of the mutability rule as a predictor of causal attribution. We also considered whether these patterns of causal selection could be predicted by various statistical rules of attribution. The perceived sufficiency of a condition predicted the explanatory quality of a cause, and regression analyses showed that this indeed generally was the case for all twelve causes examined. A measure of covariation (Delta P) created by combining the sufficiency and necessity scores (subtracting the probability that the event would occur in the absence of a target factor from the probability that it would occur in its presence) was less predictive than sufficiency alone in all comparisons. Finally, in line with Spellman’s (1997) crediting causality proposal, a conditionalized or “added value” sufficiency score was obtained for the second, third, and fourth factors by subtracting the probability that a target event would occur after a prior event (e.g., the precondition) from the probability of occurrence after the occurrence of the next link in the chain (e.g. the distal cause). Regression analyses showed that, compared to sufficiency alone, the conditionalized sufficiency score correlated less strongly with causal judgments. Nevertheless, when added into regression equations, in some cases it added additional predictive power over sufficiency alone. This was especially the case for voluntary distal causes, consistent with the idea that voluntary interventions are perceived to significantly add to the probability that the outcome will occur. In sum, causal selection from unfolding chains was not predicted by the mutability and covariational rules of attribution, but was consistent with a combination of the “sufficiency in the circumstances” view (Mackie 1974; see also Mandel, Chapter 1 of this volume, for a related view) and the crediting causality model (Spellman 1997). In addition, the finding of so few attributions to immediate causes despite their close contiguity to and high covariation with the outcome (cf. Mandel 2003c: Experiment 3) is problematic for models of causal attribution that postulate spatial and temporal contiguity as a cue to causal inference (e.g., Einhorn and Hogarth 1986). A plausible explanation is that immediate causes add no new information to what could already be inferred from prior parts of the chain. So in the case of the Concorde accident, this would explain why the proximal cause (hitting the debris on the runway) would be considered a better explanation than the immediate cause (the fuel tanks catching fire) even if the latter, as well as

The course of events

55

being more contiguous in space and time to the outcome, had higher covariation in general with aircraft crashes. This is because hitting the debris on the runway is perceived as sufficient in these particular circumstances for the crash, assuming that the world ran on normally from there, and the knowledge of the fuel tank’s catching fire therefore adds no new information, as its occurrence was already determined by the prior events in the chain. Causal selection from opportunity chains: voluntary actions versus natural causes Opportunity chains differ from unfolding chains in that the first factor does not “cause” the second factor to operate (as when a storm causes ice to form) but rather enables the second factor to act (as when the lighted cigarette in the shrub enables the wind to fan a fire). In opportunity chains the occurrence of the second event is independent of the first event, but its causal result depends on the prior occurrence of the first event. In this sense the proximal cause is “unconstrained” by the distal cause, and thus should be more mutable than is the case in unfolding causal chains, and thus be a potential candidate for causal selection according to the mutability rule. Hart and Honoré (1985) discuss what prevents an outcome from being seen as a more or less direct result of a voluntary action that is connected to it. One case they pay particular attention to is that in which there is an intervening human action: they cite the example of a person who throws a lighted cigarette into the bracken which then catches fire. Just as the flames are about to flicker out, a second person who is not acting in concert with the first deliberately pours petrol on them. Hart and Honoré consider that in this case we will consider the second person’s action to have been the cause of the fire, arguing that: “Such an intervention displaces the prior action’s title to be called the cause and, in the persistent metaphors found in the law, it ‘reduces’ the earlier action to the level of ‘mere circumstances’ or ‘part of the history’” (1985: 74). Hart and Honoré (1985) thus propose that people typically seek to trace causality back to a human action. For example, they claim that people will still trace causality back to the first person’s action (the distal cause) if the intervening event (the proximal cause) that fans the flames is a breeze that sprang up after the cigarette was thrown. The breeze springing up in that way would presumably raise the probability of the forest fire from near zero to certainty, yet not qualify as a cause because causality will be traced through to the human action preceding it. McClure et al. (2003) tested these predictions of Hart and Honoré. The design of the first experiment we performed crossed the nature of the cause (voluntary, physical) with the position of the causes in the chain of events (distal cause, proximal cause). These causes were presented in three scenarios sharing the same causal structure. We adapted the forest fire example from Hart and Honoré (1985) as a prototype for the scenarios. The scenario may

56

Hilton, McClure, and Slugoski

be rewritten so that both the distal and the natural cause are either a voluntary action or a natural event. For example, the four conditions for the forest fire scenario read: 1

2 3

4

Distal, voluntary; proximal, voluntary. “As a youth walked through a forest, he deliberately set fire to a shrub. A stranger came by, who was unaware of how the flames were caused, and intentionally fanned the flames. A forest fire followed.” Distal, voluntary; proximal, natural. “As a youth walked through a forest, he deliberately set fire to a shrub. A breeze came through, which fanned the flames. A forest fire followed.” Distal, natural; proximal, voluntary. “One bright afternoon in a forest, the sun focused by a piece of broken glass set fire to a shrub. A stranger came by, who was unaware of how the flames were caused, and intentionally fanned the flames. A forest fire followed.” Distal, natural; proximal, natural. “One bright afternoon in a forest, the sun focused by a piece of broken glass set fire to a shrub. A breeze came through, which fanned the flames. A forest fire followed.”

Consistent with Hart and Honoré’s (1985) propositions, voluntary distal causes were judged to be significantly better explanations than physical distal causes, irrespective of their position in the causal chain. However, both types were judged to be highly necessary (the perceived probability that the outcome would happen in the absence of the distal cause varied between 13 percent and 15 percent across the two types of distal cause), suggesting that the relative preference for intentional actions as explanations cannot be attributed to differences in mutability. Given that Girotto et al. (1991) found that voluntary actions are more mutable in response to “if only” probes, it may well be that there is a dissociation between the two types of mutability, as operationalized through “if only” completions and necessity judgments. Future research will do well to address this question more directly. In addition, when the distal and proximal causes were the same type (voluntary or physical), there was no difference in goodness ratings, which contradicts a second prediction by Hart and Honoré that people trace back causality to the most recent voluntary action. As with unfolding causal chains, we found no correlation between necessity judgments and causal judgments for either distal or proximal causes, suggesting that mutability had no influence on causal attribution. Importantly, this lack of correlation in opportunity chains cannot be attributed to a ceiling effect on necessity ratings, as mean necessity ratings across the four experimental conditions ranged between 57 percent and 85 percent, similar to mean sufficiency ratings, which ranged between 58 percent and 81 percent. However, we found correlations between sufficiency judgments and evaluations of the quality of explanations for both the distal cause and the proximal cause.

The course of events

57

Because sufficiency correlated with goodness judgments, we examined whether the effects of the type of cause (voluntary, physical) were mediated by perceived sufficiency. With proximal causes, type of cause predicted explanatory quality but not sufficiency, so the relation between type of cause and explanatory quality was not mediated by sufficiency. Distal causes showed partial mediation, in that the type of cause predicted explanatory quality and sufficiency, and sufficiency predicted explanatory quality but did not eliminate the variance predicted by type of cause. We also examined whether the effects of the type of cause (voluntary, natural) were mediated by Spellman’s (1997) model of crediting causality. We performed the analysis with Spellman’s (1997) formula for conditionalized sufficiency (sufficiency of the proximal cause minus sufficiency of the distal cause). The regression analysis showed partial mediation, in that the type of proximal cause predicted ratings of conditionalized sufficiency, and this in turn predicted explanatory quality but only partially reduced the variance predicted by type of cause. Finally, as type of cause had no significant effect on Delta P (sufficiency plus necessity), perceived covariation was ruled out as a potential mediator of the effect of cause type on perceived explanatory quality. The experiment shows that in these opportunity chains, voluntary causes are judged to be better explanations than physical causes, even when these causes involve an equivalent causal process and produce an identical outcome. A person fanning flames is seen as a better explanation of a forest fire than the wind fanning the flames, even though the forces that are engaged in the event are the same. A parallel effect is seen with distal causes: a person igniting the flames is seen as a better explanation of the fire than lightning igniting the flames, even though the two forces are again equivalent and have the same effect. When both causes in the chain are either voluntary or physical, participants judge them equally good as explanations. However, when one cause is voluntary and one is physical, the voluntary cause is rated better, regardless of which cause comes first in the sequence. As with unfolding causal chains, sufficiency and conditionalized sufficiency judgments predict causal preference better than necessity judgments, or sufficiency plus necessity judgments, once again suggesting that the mutability and covariational rules are insufficient to explain patterns of causal preference in the scenarios studied. However, the relation between sufficiency, conditionalized sufficiency and causality judgments is less strong in our studies on opportunity chains than in those on unfolding causal chains, and did not fully mediate the effects of cause type on causal preferences.

Conclusion: mutability and explanation in causal chains Our first conclusion is that mutability (as assessed by necessity judgments) and perceived covariation (as assessed by a combination of perceived necessity and sufficiency) do not predict selection of explanations from either

58

Hilton, McClure, and Slugoski

unfolding causal chains or opportunity chains. Preferences for voluntary and deliberate actions over natural events as causes in unfolding causal chains are more likely to be due to a combination of their greater perceived sufficiency in the circumstances, and conditionalized sufficiency (i.e., the increase in probability that the outcome will occur) that their intervention in the causal chain produces. Although sufficiency and conditionalized sufficiency also predicted explanation preferences in opportunity chains, they do not fully mediate the preference for voluntary and deliberate human actions. A second conclusion is that events can follow different kinds of courses, and that causal chains may be segmented in ways that current analyses do not yet capture. Differentiating the selection of causal segments as explanatory from selection of the explanation may be important for deciding which criterion (mutability, necessity, sufficiency, etc.) is driving causal preferences. For example, our typology allows us to predict that the first of a series of necessary and jointly sufficient conditions will be selected as the cause within an unfolding causal chain (a primacy effect) but that the second causal sequence will be considered as furnishing the explanation in an opportunity chain (a recency effect). One possibility is that highly sufficient immediate causes are omitted from an explanation in unfolding causal chains because they are predictable in the circumstances and thus uninformative to mention in an explanation (Grice 1975; Hilton 1990), whereas in opportunity chains the second cause adds predictive power to the outcome, as well as being the actual cause that “finishes off the job” by making the outcome inevitable in the circumstances. Our typology of action could also be used to predict preferences for counterfactual mutations. While Girotto et al. (1991) show that controllable actions are more likely to be mutated, less attention has been paid to foreseeability as a criterion for mutation. Thus we may ask whether voluntary (i.e., controllable) actions that deliberately produce an outcome (e.g., a car crash) are more or less mutable than those than controllable actions that produce the same effect accidentally. In addition, our typology of causal chains allows us to ask whether counterfactual mutations show the same pattern as causality judgments: following Wells et al. (1987), we would predict that the first elements will be mutated in coincidence chains as the first event constrains the later ones. However, in temporal chains such as the coin toss scenario, we would predict that losers would show a recency effect (i.e., undo or regret the last event in the chain). Similarly, in opportunity chains, because the second cause “makes the outcome definite”, we might predict that victims would undo or regret the second event more often than the first. However, in unfolding causal chains, the opposite choice is implied, as things follow on normally (inevitably) after the first cause, and we might therefore predict that victims of an accident would pick the first event over the later ones. A correspondence of patterns for counterfactual mutations and causality judgments would of course support the claim that mutability (in the sense of mental undoing) is a cue to causality.

The course of events

59

However, while Girotto et al. (1991) showed that controllable actions influence mutability judgments (as operationalized through mental undoings in response to “if only” probes), our results show that voluntary and deliberate actions had little or no impact on necessity judgments in unfolding and opportunity causal chains. Mutability – as assessed through necessity judgments – consequently had no effect on causality judgments in our studies. Sufficiency did better in unfolding causal chains in that it (or the kind of “added-value” sufficiency used in the crediting causality model) may explain the preference for voluntary and deliberate actions over natural events as causes. However, this was not the case for opportunity chains. A final problem for statistical models of causal judgment is that in a final experiment by Hilton et al. (2003) participants still preferred voluntary and deliberate actions over natural causes as explanations in unfolding chains even when told that both types of event were associated with the same objective probabilities (specified in terms of necessity and sufficiency) of producing the outcome. So why might probabilistic measures of perceived necessity, sufficiency, and conditionalized sufficiency fail to predict causal preferences in the causal chain scenarios studied? One possible reason is methodological: some other means of measuring “mutability” (e.g. the number of mental undoings in response to “if only” probes, or a task which measures the mental ease with which people can generate alternatives that undo an outcome) might provide a stronger association between this factor and causal preferences. For example, generating “if only” counterfactuals may influence a subsequent causal judgment. However effects of mental undoing on causal attribution may occur when the judge is uninvolved with the topic and no preformed mental representations of that topic, but not when she has had high exposure to and involvement with that topic (Douchet et al. 2004). While counterfactual reasoning and probabilistic expectancies may often influence causal attribution, selection of explanations in causal chains may in fact often be driven by other factors, such as pragmatic concerns of an informational or collective kind. These, rather than mutability, may often explain the preference for intentional actions as explanations. For example, intentional actions may carry greater import as causes (e.g. when the second plane flew into the World Trade Center on 9/11, by proving that the first collision was no accident it carried some big news – namely, that the United States had serious enemies), and hence have greater utility as causal explanations. Non-probabilistic counterfactual contrasts may be used if intentional actions are compared to moral (“what people should do”) rather than statistical norms (“what people usually do”; see Catellani and Milesi, Chapter 11 of this volume; Hilton 2001). Finally, the perceived controllability of actions makes the actor subject to social control through institutionalized reward or punishment, and so they are hence highlighted as causes which justify societal responses to wrongdoing and negligence (cf. Smith 1789; Tetlock 2000, 2002). We suspect that future research will do well to attend to the subtle

60

Hilton, McClure, and Slugoski

interplay of cognitive factors (such as the availability of counterfactual alternatives) and social and collective concerns in the generation of attributions of causality and responsibility.

Notes We thank Robbie Sutton for his part in elaborating the ideas developed in this chapter, and Sula Blake for her help in editing the manuscript. 1 We use the term “identified” rather than “real”, as causal generalizations can be applied to understand and explain the behavior of fictional characters, as in novels. For example, the narrative opening formula “Once upon a time” invites a reader to imagine a particular situation with particular actors. 2 Kelley (1973) refers to “multiple necessary causes”. We prefer to call this case “multiple necessary conditions”, as only their conjunction is sufficient to “cause” the outcome (Hilton 1988). 3 Thanks to David Lagnado for this soccer example. 4 This recency effect in the Jones and Cooper scenario is obtained under two conditions: first, the first event must be reported first (Spellman 1997); and second, the judge must be forced to choose which of the two events is the cause rather than evaluate them separately (Mandel 2003c).

4

The mental representation of what might have been Clare R. Walsh and Ruth M.J. Byrne

In your daily life, you think about many things. You may make a choice to visit a favorite place, you may have a chance meeting with an old friend, or you may discover a new book. What do you keep in mind when you think about events? In this chapter we will consider the view that people think by constructing a “mental model.” These mental representations allow people to keep in mind various sorts of possibilities (Johnson-Laird and Byrne 2002). Suppose your choice to visit a favorite place turns out badly: your wallet is stolen while you are there. You may imagine alternatives to the events that led to the bad outcome, such as “If only I had stayed at home . . .” What do you keep in mind when you imagine alternatives? We will consider the view that people rely on the same sorts of mental representations and cognitive processes to imagine alternative events as they do to think about the events that actually happened (Byrne 2005). We begin with a sketch of the idea that people represent events by keeping in mind possibilities. We consider the sorts of possibilities that people think about to understand and reason about an assertion, such as the conditional “If Sam is from Boston then he is a fan of the Red Sox.” Next we sketch the idea that people represent alternatives to reality by relying on the same sorts of mental representations and cognitive processes. We illustrate this suggestion for a counterfactual conditional, such as “If Sam had been from Boston then he would have been a fan of the Red Sox.” We review recent evidence on the way people think about a counterfactual conditional compared to a factual one. Then we sketch the idea that people imagine counterfactual alternatives to reality by relying on the same sorts of mental representations and cognitive processes. We review evidence that most people imagine counterfactual alternatives to their actions, to recent events, to controllable actions, to socially unacceptable actions, and to actions for which they do not have a strong reason. The evidence is consistent with the view that people imagine alternatives by keeping in mind certain kinds of possibilities.

62

C.R. Walsh and R.M.J. Byrne

The mental representation of possibilities Suppose you are told “If Sam is from Boston then he is a fan of the Red Sox.” What do you think about to understand it? You may keep in mind the possibility that Sam is from Boston and he is a Red Sox fan. But your interpretation of the conditional is unlikely to be entirely conjunctive. You may be aware that there are other possibilities that are consistent with it. You may not think about these other possibilities from the outset. But suppose you were asked to list all of the possibilities with which the conditional is consistent, what other possibilities would you list? You might list the possibility that Sam is not from Boston and he is not a Red Sox fan. You may decide that these two possibilities are the only ones consistent with the conditional. Or you might judge that a third possibility is also consistent with it: Sam is not from Boston and he is a Red Sox fan. Suppose you were asked to list the possibilities that are not consistent with the conditional. You might list the possibility that Sam is from Boston and he is not a Red Sox fan. The combination of the two components, Sam is from Boston, and he is a Red Sox fan, results in four possibilities: Sam is from Boston and he is a Red Sox fan. Sam is not from Boston and he is not a Red Sox fan. Sam is not from Boston and he is a Red Sox fan. Sam is from Boston and he is not a Red Sox fan. But when people understand the conditional they do not keep in mind these four possibilities, because of the constraints of working memory (JohnsonLaird and Byrne 1991). Instead, the possibilities they keep in mind are guided by a small set of principles. First, people keep in mind true possibilities (Johnson-Laird and Byrne 2002). They do not tend to think about the false possibility that Sam is from Boston and he is not a Red Sox fan. Second, people keep in mind few possibilities. They do not tend to think about all of the true possibilities. Usually they keep in mind just a single possibility from the outset; namely, Sam is from Boston and he is a Red Sox fan. They are aware that there are other possibilities but these possibilities are akin to an unformed thought. They can think about them if they need to, but they do not think about them from the outset (Johnson-Laird et al. 1992). A considerable body of experimental evidence has been amassed that corroborates the idea that people keep in mind possibilities of these sorts, for example, when they make deductive inferences (for a review, see JohnsonLaird 2001). The theory has been pitted experimentally against alternative accounts; for example, the view that people understand and reason about conditionals by keeping in mind abstract rules of inference (e.g., Braine and O’Brien 1998; Rips 1994), or the view that they keep in mind domain-specific rules of inference (e.g., Fiddick et al. 2000; Holyoak and Cheng 1995).

The representation of what might have been 63 The theory also provides a corroborated account of how people understand and reason with counterfactual conditionals, to which we turn shortly.

Counterfactual conditionals Suppose that you were told, “If Sam had been from Boston, he would have been a Red Sox fan.” What do you think about to understand this assertion? The subjunctive mood helps to convey that the conditional is counterfactual. It seems to mean something very different from the factual conditional. The troublesome nature of counterfactual conditionals led philosophers to suggest a new semantics based on the idea of “possible worlds” (e.g., Lewis 1973; Pollock 1976; Stalnaker 1968). The counterfactual seems to imply the opposite to the factual conditional, Sam is not in Boston and he is not a Red Sox fan. How do you understand a counterfactual conditional? You may keep in mind the possibility that Sam is from Boston and he is a Red Sox fan, and you may appreciate that this possibility is a conjecture that you are being asked to entertain, rather than a fact. Counterfactuals require you to keep in mind a possibility that is false but that you temporarily suppose to be true. You may also keep in mind the possibility that Sam is not from Boston and he is not a Red Sox fan, and you may appreciate that this possibility corresponds to the presupposed facts. The possibilities that people keep in mind to understand a counterfactual conditional are different from the ones they keep in mind to understand a factual conditional. First, to understand a counterfactual conditional, more possibilities must be kept in mind than to understand a factual conditional. You think about a single possibility from the outset to understand the factual conditional; namely, Sam is from Boston and he is a Red Sox fan. However, you keep in mind two possibilities from the outset to understand the counterfactual conditional; namely, Sam is from Boston and he is a Red Sox fan, and Sam is not from Boston and he is not a Red Sox fan. Second, for the counterfactual conditional, you must keep track of the status of the possibilities (Johnson-Laird and Byrne 1991). That is, you keep note of whether they correspond to the imagined conjecture or the presupposed facts. Considerable evidence supports this idea about how people understand counterfactuals (for an earlier review, see Byrne 2002). For example, when people read a set of assertions that includes some counterfactual conditionals, such as “If Sam had been from Boston then he would have been a Red Sox fan,” and then they are given an unexpected memory test, they sometimes believe that they had read “Sam is not from Boston” or “Sam is not a Red Sox fan” (Fillenbaum 1974). Participants asked what someone uttering the counterfactual means to convey often judge that they meant “Sam is not from Boston” or “Sam is not a Red Sox fan” (Thompson and Byrne 2002). When participants read the counterfactual conditional, and then they are given a conjunction, such as “Sam is not from Boston and he is not a Red

64

C.R. Walsh and R.M.J. Byrne

Sox fan,” they read the conjunction much more quickly compared to when it was primed by the factual conditional, “If Sam is from Boston then he is a Red Sox fan” (Santamaria et al. in press). When they read the counterfactual, and then the affirmative conjunction, “Sam is from Boston and he is a Red Sox fan,” they take the same length of time to read it as they do when it was primed by a factual conditional. The result supports the idea that people keep in mind two possibilities to understand the counterfactual and only one possibility to understand the factual conditional. How do people reason about counterfactual conditionals? Consider first how people reason about factual conditionals. Suppose you are told that “if Sam is from Boston then he is a Red Sox fan.” And suppose you discover that, in fact, Sam is not a Red Sox fan. What can you conclude? Most people find the inference difficult (for a review, see Evans et al. 1993). About half of the participants in experiments presented with this sort of information say that nothing follows, whereas the other half say that he is not from Boston. The difficulty of the inference is explained by the possibilities that people keep in mind. They understand the conditional by thinking initially about just a single possibility, Sam is in Boston and he is a Red Sox fan. When they know he is not a Red Sox fan, they find it hard to integrate the new information with the possibility they have thought about: there is no match. Thus, they conclude that nothing follows. To make the inference, they must think about the other true possibilities. When they think about the possibility, Sam is not in Boston and he is not a Red Sox fan, they can integrate the new information with this second possibility: it matches part of the information. Thus, they can conclude that Sam is not in Boston. This inference is much harder than one that can be made on the basis of the single possibility that people have kept in mind from the outset. For example, when participants are told that in fact, Sam is from Boston, almost all of them conclude readily that he is a Red Sox fan. The new information can be integrated with the single possibility they have thought about from the outset. For counterfactual conditionals, people must keep in mind two possibilities from the outset. As a result they can readily make both of the inferences (Byrne and Tasso 1999). When they are told “If Sam had been from Boston, he would have been a Red Sox fan,” they keep in mind two possibilities; namely, Sam is from Boston and he is a Red Sox fan (imagined), and Sam is not from Boston and he is not a Red Sox fan (facts). When they are told that, in fact, Sam is not a Red Sox fan, they can readily conclude that he is not from Boston. The new information matches part of one of the possibilities they have kept in mind from the outset (the presupposed facts). In fact, participants make twice as many of these inferences from a counterfactual conditional compared to a factual conditional (Byrne and Tasso 1999). Perhaps they can make the inference readily just because it corresponds to the facts, and not because they have kept two possibilities in mind (Mandel 2003b)? The data indicate that this is not so. When participants are told that, in fact, Sam is from Boston, they can readily conclude that he is a Red

The representation of what might have been 65 Sox fan. Once again, the new information matches part of one of the possibilities they have kept in mind from the outset; this time the imagined conjecture, not the facts. They make the same frequency of the inference from the two sorts of conditionals. The results show that the initial representation of a counterfactual conditional is richer than the representation of a factual conditional. People keep in mind two possibilities for the counterfactual and so they can make the inferences more readily from them. People keep in mind different possibilities to understand different sorts of conjectures. Consider the assertion “Even if Jane had been from Boston, she would still have been a Red Sox fan.” What do you think about to understand it? The “semi-factual” conditional seems to presuppose different facts than the counterfactual conditional (Chisholm 1946). It presupposes that, in fact, Jane is not from Boston, but she is a Red Sox fan. The two possibilities that people keep in mind to understand an “even if . . .” conditional are the imagined conjecture “Jane is from Boston and she is a Red Sox fan,” and the presupposed facts, Jane is not from Boston and she is a Red Sox fan (Moreno-Ríos et al. 2005). The semifactual may deny a relation between the first part of the conditional, “Jane is from Boston,” and the second part, “Jane is a Red Sox fan” (McCloy and Byrne 2002). When participants read the semifactual conditional, “even if Jane had been from Boston, she would still have been a Red Sox fan,” and then they are given a conjunction, such as “Jane is not from Boston and she is a Red Sox fan,” they read the conjunction much more quickly compared to when it was primed by a factual conditional, “If Jane is from Boston then she is a Red Sox fan” (Santamaria et al. in press). When they read the semifactual, and then the affirmative conjunction, “Jane is from Boston and she is a Red Sox fan,” they take the same length of time to read it as they do when it was primed by a factual conditional. The result supports the idea that people keep in mind two possibilities to understand the semifactual; the possibility corresponding to the imagined conjecture, “Jane is from Boston and she is a Red Sox fan,” and the possibility corresponding to the facts, “Jane is not from Boston and she is a Red Sox fan.” Of course, the interpretation of a conditional depends on its content and context. People can come to as many as ten different interpretations for conditionals, keeping in mind a different set of possibilities for each interpretation (Johnson-Laird and Byrne 2002). They readily interpret a counterfactual with causal content such as “If the butter had been heated then it would have melted” by keeping in mind two possibilities, the butter was heated and it melted (imagined) and the butter was not heated and it did not melt (facts). They keep in mind two possibilities for the causal counterfactual more often than for a counterfactual with arbitrary content such as “If Sarah had gone to Moose Jaw then Tom would have gone to Medicine Hat.” For example, they judge that someone who uttered the causal counterfactual meant to imply that the butter was not heated or it did not melt. They also judge that the situation in which the butter was heated and it

66

C.R. Walsh and R.M.J. Byrne

melted is consistent with the counterfactual. For arbitrary content, they sometimes keep in mind just a single possibility, corresponding to the presupposed fact. For example, they judge that someone who uttered the arbitrary counterfactual meant to imply that Sarah did not go to Moose Jaw or Tom did not go to Medicine Hat. But they tend to judge the situation in which Sarah went to Moose Jaw and Tom went to Medicine Hat as inconsistent with the counterfactual, more often than they make this judgment for the causal counterfactual (Thompson and Byrne 2002). For some factual conditionals, people keep in mind two possibilities from the outset. Suppose you know that “if the nurse cleans blood, she must wear gloves.” What do you think about to understand this obligation? Deontic conditionals such as these are represented differently from indicative conditionals. People keep in mind the permitted possibility, the nurse cleans blood and she wears gloves. But they also keep in mind the forbidden possibility, the nurse cleans blood and she does not wear gloves, and they appreciate that this possibility is forbidden. When they are told the nurse did not wear gloves, they readily make the inference that it is forbidden for her to have cleaned blood. They find it difficult to make this sort of inference for factual conditionals in the indicative mood as we have seen, but for factual conditionals with obligation content, the inference is easier (Quelhas and Byrne 2003). The fact that deontic conditionals are represented by two possibilities may also explain why they facilitate performance on the Wason selection task (e.g. Manktelow and Over 1991). How do people understand a subjunctive obligation, such as “If the nurse had cleaned blood, she would have had to have worn gloves?” The subjunctive mood alone is not enough to ensure that people keep in mind a third possibility corresponding to any presupposed facts. Knowledge that the obligation no longer holds is essential for people to interpret it as a counterfactual obligation. Counterfactual conditionals can seem to mean something very different from factual conditionals. The difference is explained by the sorts of possibilities that people think about. In considering factual conditionals, people tend to think about true possibilities, not false ones. They tend to think about few possibilities, not the full set of true possibilities. People tend to keep in mind a single possibility from the outset to understand a factual conditional. They may be able to think about the other true possibilities if need be, but from the outset, they think about just one. But they think about two possibilities from the outset to understand a counterfactual conditional. They think about the imagined conjecture, and they think about the presupposed facts, and they keep track of the status of these possibilities as imagined or facts. There is considerable evidence to support this idea about how people understand counterfactual conditionals, based on measures of their memory for counterfactuals, their judgments of what counterfactuals imply, their judgments of the situations that are consistent with them, the inferences that they can make readily from them, and the situations that counterfactuals prime people to read quickly. People keep different possi-

The representation of what might have been 67 bilities in mind to understand a counterfactual compared to an “even if” semifactual. Moreover, the content of the conditional – for example, causal, arbitrary, or obligation content – affects the possibilities that people keep in mind. The principles that underlie the way people think about possibilities when they understand counterfactual conditionals are the same principles that underlie the way they think about possibilities when they imagine counterfactual alternatives (Byrne 2005). We turn now to consider how people imagine alternatives.

Imagination of counterfactual alternatives People often imagine counterfactual alternatives to facts. Suppose you know that Bob is not a Red Sox fan. You might imagine, “If only Bob was a Red Sox fan, we could go to the game together this week.” People tend to imagine alternatives to bad outcomes, or outcomes of great personal significance (Davis et al. 1995; Roese and Olson 1995d). The sorts of alternatives that people imagine depend partly on their motivations. For example, they may imagine how events could have turned out better when their purpose is to plan for the future, whereas they may imagine how events could have turned out worst when their purpose is to console someone (Roese et al. 2004; Seelau et al. 1995). Most people tend to imagine alternatives to the same aspects of events. There appear to be “fault lines” in our representation of reality, junctures that most people zoom in on when they imagine how an event could have turned out differently (Kahneman and Tversky 1982a). The alternatives people imagine may be constrained by their cognitive capacity (Byrne 1997). What people have mentally represented about the facts may influence their ability to change the facts (Byrne 1996; Legrenzi et al. 1993). Because of the limitations of working memory, people may tend to make minimal changes to the facts (Byrne 1997). Some facts seem to make available an alternative. For example, when people understand an exceptional event they seem to recruit from memory the normal event (Kahneman and Miller 1986). People keep in mind two possibilities to understand some concepts. For example, to understand a choice between options, such as a choice to have chocolate ice cream, people keep in mind the chosen option, chocolate ice cream, and they also keep in mind an alternative to the choice (Byrne 2005). The alternative option they think about may be the general one, to have a different flavor ice cream, or a specific one, to have strawberry ice cream, or it may be the alternative, not to have chocolate ice cream. It may be easier to imagine a counterfactual alternative when you have kept in mind two possibilities from the outset, compared to when you have kept in mind a single possibility (Byrne 2005), as we shall now discuss.

68

C.R. Walsh and R.M.J. Byrne

Imagined alternatives to actions Suppose you live in London and you think about taking a trip at the weekend. You decide to go to Paris but while you are there your wallet is stolen. It contains a significant amount of money and all of your bank cards. You cannot help thinking that if only you had stayed at home your wallet would not have been stolen. Suppose your friend also lives in London. She too thought about taking the trip to Paris, but she decided to stay in London. But while she is there her wallet is stolen. Do you think she will be plagued by thoughts that if she had gone to Paris, her wallet would not have been stolen? The counterfactual alternative seems somehow less compelling. People tend to focus on actions more than inactions when they imagine alternatives to them (Kahneman and Tversky 1982a; Ritov and Baron 1990). The tendency can be explained by the possibilities that they keep in mind from the outset to understand actions and inactions (Byrne and McEleney 2000). When people think about an action, such as going away for the weekend, they tend to think about two possibilities. They think about the pre-action possibility of being in London and the post-action possibility of being in Paris. They imagine a counterfactual alternative by changing the current post-action possibility to be the same as the now counterfactual pre-action possibility. When they think about an inaction, such as not going away for the weekend, they tend to think about a single possibility. The past and present possibilities are the same (i.e., being in London). Of course, people could imagine a counterfactual alternative to the inaction, such as being in Paris, but this possibility is not readily available in their initial mental representation. In some circumstances, a second possibility may be thought about from the outset to understand an inaction. For example, a long-term perspective may be adopted or there may have been good reasons to act (Gilovich and Medvec 1995; Zeelenberg et al. 2002). However, in general, people think about two possibilities to represent the action and so they can readily imagine an alternative possibility to the facts; they think about a single possibility to represent the inaction and, accordingly, they cannot readily imagine an alternative possibility to the facts (Byrne 2005). Imagined alternatives to recent events When people think about a sequence of events that are not causally related, they tend to imagine an alternative to the most recent one (Miller and Gunasegaram 1990; but see Mandel 2003c). Consider the following scenario (from Byrne et al. 2000). Two individuals (John and Michael) pick a card from a deck of playing cards. If the two cards they pick are of the same color (i.e., both red or both black), each individual wins £1,000. Otherwise, they win nothing. John goes first and picks a red card from his deck; Michael goes next and picks a black card from his deck and so they lose. How do you

The representation of what might have been 69 think the outcome could have turned out differently? Who do you think feels most guilt about the outcome? And who do you think will blame the other more? Most people imagine, “If only Michael had picked a red card too,” they expect Michael will experience more guilt than John, and they expect that Michael will be blamed more by John, than John will be blamed by Michael. The effect has been examined in a range of contexts, including descriptions of games (Spellman 1997), simulated examination situations (Miller and Gunasegaram 1990), judgments of the performance of a basketball team in a league (Sherman and McConnell 1996), and for sequences of different lengths (Segura et al. 2002). The effect may depend on the possibilities that people keep in mind to understand the situation from the outset. They may think about the actual card selections of each player: Facts: John picked red and Michael picked black and they lost. There are three counterfactual possibilities: Imagined: John picked red and Michael picked red and they won. Imagined: John picked black and Michael picked black and they won. Imagined: John picked black and Michael picked red and they lost. But people think about few possibilities. Which of the three imagined possibilities do they think about? The final possibility listed above is a possibility in which the players lose; it does not change the outcome and so they do not entertain it. The tendency to imagine an alternative to the recent event indicates that they keep in mind just the facts and the imagined possibility in which John picked red and Michael picked red and they won. The limitations of working memory may lead people to minimize the number of possibilities that they keep in mind (Johnson-Laird and Byrne 1991). Once they understand the fact that John picked red, they may eliminate the possibility that both pick black. The first player’s selection acts as an “anchor” (Byrne et al. 2000). In fact, when people keep in mind an alternative possibility to the first player’s selection from the outset, they imagine “If only John had picked black.” Consider the situation in which two individuals (John and Michael) take part in a television game show, on which they are offered the attractive proposition described earlier. Each individual is given a shuffled deck of cards, and each one picks a card from his own deck. If the two cards they pick are of the same color (i.e., both from black suits or both from red suits), each individual wins £1,000. However, if the two cards are not the same color, neither individual wins anything. John goes first and picks a black card from his deck. At this point, the game-show host has to stop the game because of a technical hitch. After a few minutes, the technical problem is solved and the game can be restarted. John goes first again, and this time the

70

C.R. Walsh and R.M.J. Byrne

card that he draws is a red card. Michael goes next and the card that he draws is a black card. Thus, the outcome is that neither individual wins anything. To understand the scenario, people may keep in mind several possibilities from the outset. They think about the pre-hitch possibility as well as the post-hitch possibility: Pre-hitch: John picked black. Post-hitch: John picked red and Michael picked black and they lost. They can readily imagine a counterfactual alternative in which John picked black, and so they think “If only . . .” about the first event as often as the most recent event. The possibilities that people keep in mind are influenced by the way in which stories are told (Walsh and Byrne 2004). Suppose you are on a jury and you are told about an accident in which a dog ran onto the road as two individuals, Tom and Bob, were driving in opposite directions towards it. Tom swerved to avoid the dog, and Bob also swerved to avoid it, and as a result their cars collided. The lawyer says, “Each of them could have escaped injury if one or the other but not both had swerved.” You may think about the facts: Facts: Tom swerved and Bob swerved and there was an accident. You may tend to imagine an alternative to the more recent event, if only Bob had not swerved. You could imagine several alternatives to the facts. In one imagined possibility, Tom swerved and Bob did not and there was no accident; in the other imagined possibility, Tom did not swerve and Bob did, and there was no accident. But you may think about the possibilities in an incomplete manner. You may think about the possibility in which Tom swerved and the other possibility in which Bob swerved: Imagined: Tom swerved and there was no accident. Imagined: Bob swerved and there was no accident. In the possibility in which Tom swerved, what did Bob do? The possibility contains complete information about Tom but incomplete information about Bob. You can complete the information in the imagined possibilities: Imagined: Tom swerved and Bob did not swerve and there was no accident. Imagined: Tom did not swerve and Bob swerved and there was no accident. But because of working memory constraints, you are likely to keep in mind the incomplete possibilities.

The representation of what might have been 71 Because people keep in mind incomplete possibilities, the alternatives they imagine depend on the way the story is told. They keep in mind incomplete imagined alternatives: Imagined: Tom swerved and there was no accident. Imagined: Bob swerved and there was no accident. The first imagined possibility in which Tom swerved matches the anchor of the facts in which Tom swerved, and so they may think, “If only Bob had not swerved.” But suppose instead that you are told the same story but this time the lawyer says, “Each of them could have escaped injury if one or the other had continued driving straight on.” You may imagine the incomplete possibilities: Imagined: Tom continued driving straight and there was no accident. Imagined: Bob continued driving straight and there was no accident. The first imagined possibility in which Tom continued driving does not match the anchor of the facts in which Tom swerved. Instead, it provides an alternative. You may imagine “If only Tom had continued driving straight on.” We tested this idea using the card scenario. In the scenario, both players picked black and they lost. We told one group that the players would win if one or the other, but not both, picked a card from a red suit. We told the second group that the players would win if one or the other, but not both, picked a card from a black suit. These winning conditions are identical. But their objective equivalence does not result in their subjective equivalence. People mentally represent them differently. Most of the participants in the group who were told the players would win if one or the other picked a black card (and in fact both picked black), tended to think “If only the second player had picked red.” Most of the participants in the group who were told the players would win if one or the other picked a red card (and in fact both picked black), tended to think “if only the first player had picked red” (Walsh and Byrne 2004). The result shows that people keep in mind incomplete possibilities. Imagined alternatives to controllable actions People tend to imagine alternatives to events that fall within their control, particularly events that result from their own decisions. For example, most participants judge that a man who is delayed on his way home and arrives too late to help his wife who has had a heart attack will think “If only . . .” about his decision to stop for a beer rather than about being stuck in traffic (Girotto et al. 1991; see also Mandel and Lehman 1996; Markman et al. 1995). In fact, people think “If only . . .” more often about controllable

72

C.R. Walsh and R.M.J. Byrne

actions that are socially unacceptable, such as drinking alone, or drinking and driving, than they do about controllable actions that are socially acceptable, such as visiting parents (McCloy and Byrne 2000; N’gbala and Branscombe 1995). Just as abnormal events call to mind their corresponding norms (Kahneman and Miller 1986), certain actions may call to mind the social norm (McCloy and Byrne 2000). People tend to keep in mind two possibilities when they think about controllable actions. For example, they think about stopping for a beer and they also think about not stopping for a beer. When the action is a socially unacceptable one, they keep track of its status as a forbidden possibility. When they imagine a counterfactual alternative, they tend to imagine the forbidden possibility was changed to be the same as the permitted possibility, “If only I had not stopped for a beer” (Byrne 2005). They do not tend to change a permitted possibility to be the same as a forbidden possibility, “If only I had run a red light.” In fact, people seem especially reluctant to imagine a counterfactual alternative to an action that results from an obligation. The reason why an action was carried out places a strong constraint on it. Consider the following scenario (adapted from Klauer et al. 1995). Bernard is a violinist who is performing at a concert in Vienna. The night before the performance he attends a ball. Suppose he is obliged to attend because it is a fund-raiser for the concert. His performance the next day is disappointing and he thinks, “If only . . .” People understand the obligation by keeping in mind two possibilities; namely, that he attends the ball (permitted) and that he does not attend the ball (forbidden). They do not change a permitted possibility to be like a forbidden possibility, and they do not tend to imagine that Bernard will think, “If only I had not attended the ball.” To test this idea, we gave participants one of three versions of the scenario (Walsh and Byrne 2005). In one version, participants were given the obligation as the reason Bernard attended the ball. In a second version, they were given another reason; namely, that Bernard wanted to meet another composer there. In the control condition, they were given no reason, just a description of where the ball was being held. Participants tended to imagine an alternative to Bernard attending the ball more often when they knew of no reason or the reason about meeting another composer than when they knew about his obligation to attend. The result shows that people do not tend to imagine alternatives to obligations. When they imagine an alternative, they do not tend to change a permitted possibility to be the same as a forbidden possibility.

Conclusion People imagine counterfactual alternatives by keeping possibilities in mind. The possibilities they think about are guided by a small set of principles, similar to the principles that guide the way they think and reason about

The representation of what might have been 73 facts (Byrne 2005). People can more readily imagine a counterfactual alternative if they have thought about two possibilities from the outset than if they have thought about a single possibility. These principles help to explain why people imagine alternatives to actions, to the most recent event in a sequence, to controllable events, to socially unacceptable events, and why people do not tend to imagine alternatives to actions constrained by an obligation as much as actions constrained by weaker reasons. The mental models theory described here provides a cognitive framework for understanding counterfactual thinking. The representation of multiple possibilities may provide a common explanation for the differential mutability of events, as we have shown. The theory has led to some novel predictions; for example, the finding that the description of events can change their mutability. These findings also provide some clues regarding how people’s counterfactual thoughts may be modified.

Part II

Functional bases of counterfactual thinking

5

Reflective and evaluative modes of mental simulation Keith D. Markman and Matthew N. McMullen

A news item recently caught our attention. Flight attendant Kim Stroka claimed that she was too distraught to return to work after her co-worker died on United Airlines Flight 93, which was hijacked after taking off from Newark Liberty International Airport en route to San Francisco on 11 September 2001. Of compelling interest to counterfactual researchers, Stroka had apparently traded shifts with her co-worker and, thus, would have died instead of her colleague if she had worked her normal shift. Claiming that she was having difficulty eating and sleeping and that she was being treated by a psychologist for post-traumatic stress disorder, Stroka applied for medical and disability payments but was turned down by the state appellate court. According to the court, Stroka was not entitled to the award because “nothing happened while she was working which led to her current condition” (“No 9-11 Compensation for Flight Attendant,” Associated Press 2003). A number of researchers have focused on the distinction between upward counterfactuals that simulate a better reality and downward counterfactuals that simulate a worse reality (e.g., Mandel 2003a; Markman et al. 1993; McMullen et al. 1995; Roese 1994; Roese and Olson 1995d; Sanna 1996, 2000). These researchers have adopted an approach that describes the possible functions that upward and downward counterfactual thoughts might serve. One function that has been identified is the contrast-based affective function (Roese 1997) – a given outcome will be judged more favorably to the extent that a less desirable alternative is salient. Thus, the strategic generation of downward counterfactuals may serve the function of enhancing coping and feelings of relative well-being by highlighting how the situation or outcome could easily have been worse. Clearly, Kim Stroka has made a downward counterfactual. She did not die, but she can easily imagine how she could have died – indeed, she would have died. Just as clearly, however, generating this downward counterfactual has not made her feel any better. Instead, her consideration of the downward counterfactual world has engendered feelings of sadness, guilt, and fear. The Stroka case helps us make a more general point. Although contrast-based affective reactions to counterfactuals – whereby judgments are displaced

78

K. Markman and M. McMullen

away from the counterfactual standard – are common, they are hardly the rule. Rather, Stroka’s downward counterfactual is assimilative in nature – her affective experience has been pulled toward the counterfactual standard (McMullen 1997). In this chapter we will discuss the important aspects of a model (Markman and McMullen 2003) that attempts to explain how the very same counterfactual can engender dramatically different affective reactions. According to the model, the consequences of simulation direction are moderated by what we have termed simulation mode – relatively stronger tendencies to engage in reflective versus evaluative processing. In turn, we will describe how the interaction between simulation direction and mode produces important consequences for affect, motivation, and behavior.

The reflection and evaluation model Reflection and evaluation Markman and McMullen (2003) developed the Reflection and Evaluation Model (REM) of comparative thinking in order to provide an organizing framework for understanding how assimilation and contrast effects arise following counterfactual, social, and temporal comparisons. At the heart of the model is the assertion that two psychologically distinct modes of mental simulation operate during comparative thinking. The first of these modes is reflection, which is an experiential (“as if”) mode of thinking characterized by vividly simulating that information about the comparison standard is true of, or part of, the self. The second of these modes is evaluation, which is characterized by the use of information about the standard as a reference point against which to evaluate reality (cf. Epstein et al. 1992; Oettingen 1996; Strack 1992). Figure 5.1 depicts the interaction between simulation direction and mode. To illustrate, consider the student who receives a B on an exam but realizes that an A was easily attainable with some additional studying. In Mode Direction

Reflection

Evaluation

Upward

“I almost got an A”

“I got a B … I failed to get an A”

“I nearly got hit by that truck”

“I was fortunate to not have been hit by that truck”

Downward

Figure 5.1 The interaction between simulation direction and mode.

Reflective and evaluative mental simulation 79 the case of upward evaluation (UE), the student switches attention between the outcome (a grade of B) and the counterfactual standard (a grade of A). According to the REM, such attentional switching (“I got a B . . . I could have gotten an A but instead I got a B”) involves using the standard as a reference point and thereby encourages evaluative processing. In the case of upward reflection (UR), however, the student’s attention is focused mainly on the counterfactual itself. According to the REM, focusing on the counterfactual encourages reflective processing whereby the student considers the implications of the counterfactual and temporarily experiences the counterfactual as if it were real (“What if I had actually gotten an A?”). In a sense, the student is “transported” into the counterfactual world (Green and Brock 2000; Kahneman 1995). Likewise, consider the case of a car driver who pulls away from the curb without carefully checking rear and side-view mirrors, and subsequently slams on the brakes as a large truck whizzes by. In the case of downward evaluation (DE), the driver switches attention between the counterfactual standard (being hit by the truck) and the outcome (not being hit by the truck), thereby encouraging evaluative processing (“I was fortunate to not have been hit by that truck”). In the case of downward reflection (DR), however, the driver’s attention is mainly focused on the counterfactual itself, thereby encouraging reflective processing (“I nearly got hit by that truck”). Affect and the accessibility mechanism Reflective processing and evaluative processing of counterfactuals yield predictable affective reactions and, according to the model, this is accomplished through an accessibility mechanism. Work by Mussweiler and his associates (e.g., Mussweiler 2003; Mussweiler and Strack 2000a, b) within the domain of social comparisons suggests that comparative self-evaluation produces two informational consequences. First, comparing oneself to a given standard increases the accessibility of standard-consistent knowledge about the self. Thus, upward comparisons render knowledge indicating a high standard of the self more accessible, whereas downward comparisons render knowledge indicating a low standard of the self more accessible. Second, and in turn, comparative self-evaluation provides a reference point against which the implications of this knowledge can be evaluated. In a similar vein, we propose that counterfactual comparisons can also yield two informational consequences. First, making counterfactual comparisons should enhance the accessibility of cognitions about the self that are evaluatively consistent with the counterfactual standard. In turn, affect should be derived from thoughts about the standard that implicate the self, thereby yielding affective assimilation (Schwarz 1990; Schwarz and Clore 1983; Strack et al. 1985). To illustrate, consider an individual who learns that the aircraft she had originally planned to take crashed with everyone on board killed. Simulating the counterfactual possibility “I could have been on

80

K. Markman and M. McMullen

that plane” (DR) renders standard-consistent cognitions about the self more accessible (e.g., “I could be dead,” “I would never have been able to see my family again,” “I would never have been able to accomplish what I wanted to in life”), and reflecting on these accessible cognitions produces counterfactual-congruent (in this case, negative) affect. On the other hand, employing the counterfactual as a standard against which to evaluate reality (DE) should produce positive affect via a contrast effect: “I’m lucky to be alive”). More generally, the notion that the very same counterfactual can produce both assimilative and contrastive reactions has intriguing implications for affective experience, as it may be that the mixed emotions (Larsen et al. 2001; Larsen et al. in press) that are often felt after events such as switching from the doomed plane flight are the result of reflective and evaluative modes of mental simulation operating in parallel (cf. Biernat and Manis 1994; Biernat et al. 1997; Mussweiler 2003; see also Markman and McMullen 2003 for a more detailed discussion of this issue). In this way, one can feel fortunate to be alive, yet deeply troubled by thoughts of what might have been. Motivational consequences In addition to the contrast-based affective function served by downward counterfactuals, counterfactual researchers have also focused on the preparative function that might be served by upward counterfactuals. Although upward counterfactuals may devalue the actual outcome and make us feel worse (e.g., Markman et al. 1993; Mellers et al. 1997; Roese 1994; Sanna 1996), simulating routes to imagined better realities may help us improve on our outcomes in the future (Johnson and Sherman 1990; Karniol and Ross 1996; Taylor and Schneider 1989). It has been suggested by some that counterfactual thoughts produce causal inferences (e.g., Hilton and Slugoski 1986; Lipe 1991;Wells and Gavanski 1989; but see also Mandel 2003c) and, according to Roese and his colleagues (e.g., Roese 1994, 1997; Roese and Olson 1997), it is this causal inference mechanism that underlies the preparative function. To illustrate, if Jim fails an exam, and then realizes that he would have passed if he had studied the textbook more carefully, he has identified an antecedent action that may trigger an expectancy regarding the consequences of taking that action in the future. In turn, this realization should heighten intentions to perform that action and thereby influence the production of that action. The REM advances previous functional approaches, however, by suggesting that upward and downward counterfactuals can both have affective and preparative (or more generally, motivational) functions via reflective and evaluative processing. One of the key assumptions of the model is that the motivation to act, or not to act, is mediated by one’s affective state, and also depends on the goal that has been adopted for performing a given task. Drawing on both Schwarz’s (1990) feelings-as-information hypothesis (see

Reflective and evaluative mental simulation 81 also Taylor 1991) and the mood-as-input perspective of Martin and his colleagues (e.g., Martin et al. 1993; see also Forgas 1995), the REM posits that negative affect should engender more persistence for tasks pursued to satisfy achievement goals (i.e., by employing the stop rule, “Have I done as well as I can do?”) but lead to less persistence on tasks pursued merely for enjoyment (i.e., by employing the stop rule, “Am I still enjoying this task?”), whereas positive affect should engender more persistence for enjoyment tasks but lead to less persistence for achievement tasks. (See also Apter’s (e.g., Apter 2001; Apter and Larsen 1993) distinction between telic and paratelic states.) Moreover, although the causal inference derived from the counterfactual comparison may suggest specific behaviors that one might perform in the future, we believe that the initial impetus to either change one’s behavior or stay the present course is determined by affect. Overall, then, the REM specifies that affect and cognition make independent contributions to motivation and goal pursuit: affect motivates the individual to either change or maintain the status quo, whereas cognition shapes the specific strategies whereby one will either change or keep things the way they are. Armed with this perspective on the influence of affect and cognition on motivation, specific predictions can be made regarding the motivational implications of upward and downward reflection and evaluation. To begin, UE yields negative affect, and should thus engender more persistence on achievement tasks but less persistence on enjoyment tasks. In addition, the causal inferences derived from UE (e.g., “I should have read the textbook chapters more carefully”) should allow the individual to develop specific behavioral intentions and strategies regarding what actions should or should not be taken (see also Grieve et al. 1999; Morris and Moore 2000; Nasco and Marsh 1999). Expanding on previous functional approaches, however, the REM predicts that DR should also yield negative affect and thus exert effects on persistence similar to those produced by UE. Moreover, DR should produce causal inferences that seek to explain the event that almost happened (e.g., “I almost got hit by that truck because I didn’t check my rear view mirror”). Thus, although DR does not help one to envision a route to a positive outcome, per se, we suggest that it can certainly motivate individuals to discontinue potentially destructive behaviors, in a manner not unlike fear communications that have been used in persuasion studies (e.g., Baron et al. 1994; Janis and Feshbach 1953). On the other hand, DE engenders positive affect and may also yield causal inferences (e.g., “It’s a good thing I studied as much as I did. If I had completely blown off my studying, I would have done much worse”). The causal inference derived here – that “some studying” is the cause of receiving a decent grade – indicates that a moderate amount of studying in the future will help maintain the status quo. Within the achievement domain, then, it is expected that the positive affect and specific causal inferences derived from DE will both contribute to an individual’s complacency.

82

K. Markman and M. McMullen

Expanding on previous functional approaches, however, the REM predicts that UR should also engender positive affect, as well as causal inferences that seek to explain the event that almost happened (e.g., “I almost sunk that thirty-foot putt because I accounted for the upward slope of the green”). Attempts to specify the nature of the influence of UR on motivation bring up two intriguing possibilities. On the one hand, it may be that the positive affect derived from UR will engender less persistence for achievement tasks (and more persistence on enjoyment tasks). On the other hand, the realization that one nearly accomplished the goal (e.g., making the putt), coupled with an understanding of how one nearly accomplished that goal (e.g., by accounting for the upward slope of the green), may instead engender feelings of self-efficacy (e.g., Bandura 1977; Sanna 1997) that empower one to persist and perform better on the task at hand. We will return to this issue a bit later. In addition to producing emotions and suggesting causal inferences, counterfactual thinking may also influence regulatory strategies (Hur 2000; Pennington and Roese 2002; Roese et al. 1999, 2004). Higgins (1998) has argued that both promotion and prevention strategies are important means by which one can achieve desired end states. Promotion-oriented individuals, focused as they are on growth, advancement, and accomplishment, tend to pursue strategies aimed at approaching desirable outcomes, whereas prevention-oriented individuals, focused as they are on protection, safety, and responsibility, tend to pursue strategies aimed at avoiding undesirable outcomes. Recent research has demonstrated how regulatory focus can be temporarily induced by cues in the environment (e.g., Forster et al. 1998; Higgins et al. 1997; Shah et al. 1998), and the salience of counterfactual standards may be one such situational cue. In this regard, an upward counterfactual represents a desirable outcome and thus may activate promotion goals in the service of obtaining that outcome, whereas a downward counterfactual represents an undesirable outcome and thus may activate prevention goals in the service of ensuring that the outcome does not occur (Hur 2000; Lockwood 2002; Lockwood et al. 2002). According to the REM, the promotion focus activated by engaging in upward counterfactual thinking should play an important role in determining the nature of one’s behavioral intentions. In the case of the student who failed to achieve an A, for example, a promotion focus should encourage the student to devise strategies designed to achieve favorable outcomes (e.g., putting more time into school work, attending class on a more regular basis). Conversely, the prevention focus activated by downward counterfactual thinking (e.g., “I almost got hit by that truck”) should encourage the individual to adopt strategies designed to avoid bad outcomes (e.g., checking all rear view and side mirrors). Furthermore, and drawing once again on the feelings-as-information perspective, the REM predicts that prevention goals will be more highly activated after DR than after DE, and promotion

Reflective and evaluative mental simulation 83 goals will be more highly activated after UE than after UR, because both DR and UE focus individuals on their failure to attain desired end-states.

Empirical tests of the reflection–evaluation model Downward counterfactuals and motivation Several laboratory studies have been conducted that test the REM’s predictions in the domain of counterfactual thinking and motivation. To examine the motivational implications of DR and DE, McMullen and Markman (2000) measured students’ responses after receiving their first exam grade in a course. All participants were instructed, in writing, to make a downward counterfactual (i.e., compare their present grade to an imagined worse grade). In the evaluation condition they were instructed to “evaluate your grade in comparison to the worse grade you imagined,” whereas in the reflection condition they were instructed to “vividly imagine receiving that worse grade.” Participants then indicated the extent to which they were experiencing various emotions and then answered several questions regarding their motivation to modify their study habits in the future (e.g., “How much do you feel you should change the way you study for the next exam?”). Consistent with predictions, more negative affect was experienced in the reflection condition, and more positive affect was experienced in the evaluation condition. Furthermore, motivation to modify future study habits was greatest in the reflection condition. Finally, mediational analyses indicated that the mode manipulation initially predicted motivation, but when affect was also entered into the regression equation, the mode coefficient dropped to nonsignificance, whereas the affect coefficient remained significant. Thus, and consistent with one of the key assumptions of the REM, affect mediated the counterfactual’s impact on motivation. Counterfactual thinking and task persistence A second study (Markman et al. 2004b) examined the motivational implications of both downward and upward counterfactuals. Earlier in this chapter, we were equivocal with regard to predicting the effects of UR (e.g., feeling good by imagining having won the lottery) on motivation. If the feelings-asinformation perspective is correct, UE should enhance motivation, whereas UR should lead to complacency. The prediction of a complacency effect following UR is supported by the work of Oettingen and her colleagues (e.g., Oettingen 1996; Oettingen and Mayer 2002; Oettingen et al. 2001). In these studies, engaging in positive fantasy by itself decreased motivation and inhibited success, whereas explicitly contrasting positive fantasies with reality enhanced motivation and facilitated success. According to Oettingen (1996), positive fantasies can be detrimental because they engender anticipatory consumption of motivation that would otherwise be directed toward achieving a given goal.

84

K. Markman and M. McMullen

Markman et al. (2004b) gave participants as much time as they wished to solve an initial set (SET 1) of ten anagrams, with each anagram having multiple potential solutions. Following completion of SET 1, participants received “2” feedback (e.g., if they found ten solutions, the computer informed them that they had found “ten out of the twenty possible solutions”), allowing them equal room to make either upward or downward counterfactuals. Participants in the UE and UR conditions were then asked to think about how their performance could have been better, with those in the UE condition being instructed to “compare their performance to the better performance they imagined,” and those in the UR condition being instructed to “vividly imagine having performed better.” On the other hand, participants in the DE and DR conditions were asked to think about how their performance could have been worse, and received reflection and evaluation instructions equivalent to the UE and UR participants. After generating their counterfactuals, participants responded to a set of mood adjectives, and were then given as much time as they wished to solve ten additional anagrams (SET 2). The dependent variables of interest were mood reports following SET 1 feedback, the amount of time they spent on SET 2 (persistence) relative to SET 1, and the number of anagram solutions correctly found in SET 2 relative to SET 1. Analyses yielded the predicted direction by mode interaction for affect: URs reported more positive affect than did UEs, whereas DEs reported more positive affect than did DRs (see also McMullen 1997). Importantly, the predicted direction by mode interaction for task persistence was also obtained: whereas UEs persisted longer on SET 2 than did URs, DRs persisted longer on SET 2 than did DEs. In addition, UEs actually solved more SET 2 anagrams than did URs, although no differences were found between DRs and DEs. Finally, path analyses (see Figure 5.2) conducted on participants in the upward counterfactual conditions indicated that the relationship between mode (dummy coded: 1  reflection, 2  evaluation) and persistence was mediated by feelings of relaxation – the less relaxed participants felt, the more they persisted. In turn, SET 2 persistence predicted SET 2 Set 1 Performance

0.15 (0.25*)

0.57* Mode

0.47*

Relaxed

0.21*

Set 2 Persistence

0.25*

Set 2 Performance

0.49* Set 1 Persistence 0.21* 0.24*

Figure 5.2 Mediator model of simulation mode, affect, and task persistence. Weights are standardized path coefficients adjusted for all other factors in the model. *p  0.05.

Reflective and evaluative mental simulation 85 performance after controlling for SET 1 persistence and SET 1 performance. Interestingly, however, an independent positive relationship also emerged between relaxation and SET 2 performance, suggesting that the affect derived from upward simulations can affect performance through two distinct mechanisms. On the one hand, UE may lead individuals to feel more aroused, thereby enhancing task persistence and task performance. Alternatively, however, the feelings of relaxation produced by UR may also enhance performance (despite the decrease in persistence), perhaps by facilitating the development of more creative solutions. In support of this possibility, a number of empirical studies have reported a relationship between positive mood and creativity (e.g., Hirt et al. 1997; Martin and Stoner 1996; Murray et al. 1990). Although the analyses described above discovered a positive relationship between relaxation and performance, the independent negative relationship between UR and task persistence demonstrated by the Markman et al. (2004b) study supports Oettingen’s (1996) notion that positive fantasies engender anticipatory consumption of motivation. According to Oettingen (1996: 238–9), in a positive fantasy, a person may “experience” the future event ahead of time and may color the future experience more brightly and joyfully than reality would ever permit. Therefore the need to act is not felt, and the thorny path leading to implementing the fantasy may be easily overlooked. Indeed, the counterfactuals generated by participants in the UR condition were characterized by this sort of flavor. For instance, one participant wrote, “I imagined the letters moving for me, instead of me going through them all individually and crossing them off in my mind. Meaning, I imagined the word appearing for me.” This particular mental simulation may be optimistic, but it is also bereft of the implementation strategies (cf. Gollwitzer et al. 1990) that may be required to achieve the counterfactual outcome. The complex relationships among UR, affect, motivation, and behavior can be further addressed by identifying and contrasting between two different types of counterfactuals. On the one hand, the UR participants in Markman et al. (2004b) engaged in what are essentially positive fantasies – they transported themselves into a better counterfactual world of their own creation. Kahneman and Varey (1990; Kahneman 1995; see also Teigen 1998b), however, have also described the special status of “close counterfactuals” (e.g., “John almost won the lottery,” “Susan almost died”), which are characterized by a strong propensity for the counterfactual outcome to have existed soon before the actual outcome occurred. Propensities “. . . indicate advance toward the focal outcome, or regression away from it” (Kahneman and Varey 1990: 1105). Thus, to the extent that individuals perceive a trajectory toward either a desired or undesired state, assimilative effects

86

K. Markman and M. McMullen

following counterfactual thinking are more likely to occur (see also Carver and Scheier 1990; Hsee et al. 1994; Landman and Petty 2000; Roese and Olson 1995b; Sanna et al. 2003; Tetlock 1998). Indeed, the casino game of Keno takes advantage of this phenomenon in a clever way: The numbers in the near vicinity of the winning number are lit up in addition to the number that won, giving rise to the feeling of “almost winning” (Sherman and McConnell 1995). Thus, close upward counterfactuals may sometimes encourage behavioral persistence. Two of our studies have found evidence of affective assimilation following close counterfactuals. Markman and Tetlock (2000a) had participants engage in a simulated stock investment competition in which they chose between investing in one of two different companies. In the near-win condition, the chosen stock was just barely outperformed by the unchosen stock, whereas in the near-loss condition the chosen stock just barely outperformed the unchosen stock. After viewing the performance of the two stocks across a one-year span, participants indicated being happier when they nearly won (but lost) than when they nearly lost (but won). Similarly, McMullen and Markman (2002) found that fans of a basketball team that was losing by one point at half time but had come back from a substantial deficit felt better about the game than did fans of the team that was winning at half time (cf. Markman et al. 1995; but also Wohl and Enzle 2003 for an alternative perspective). Reminiscent of Kahneman’s (1995) distinction between elaborative and automatic mental simulations, we would suggest that while positive fantasies (i.e., elaborative simulations) of the quality described by Oettingen and her colleagues and elicited by Markman et al. (2004b; see also McMullen 1997) may reduce motivation, close (i.e., automatic) upward counterfactuals that suggest that a better outcome was and, importantly, is plausibly attainable in the future, may actually serve to increase motivation. In fact, Markman and Tetlock (2000a) found initial support for this notion: Participants were less willing to reinvest in their chosen stock when they nearly lost (but won) than when they nearly won (but lost). Counterfactuals, persistence, and goal type Another experiment (McMullen et al. 2004) examined whether the influence of counterfactual thinking on motivation might also interact with the type of goal involved. Participants spent five minutes solving a set of crosswordlike puzzles, and were then instructed to generate either a downward or upward counterfactual about their performance. Next, they were instructed to either vividly imagine the counterfactual (reflection) or to compare the counterfactual to their actual performance (evaluation). They then worked on another set of puzzles, but this time they could spend as much or as little time working as they wished. In the enjoyment condition, they were told that the point of the word puzzles was simply to have fun with the puzzles,

Reflective and evaluative mental simulation 87 and if they were no longer having fun they could stop. In the achievement condition, on the other hand, they were told to perform as best they could, and when they were satisfied with their performance they could stop (cf. Sanna, Meier et al. 2001). The primary dependent variable was the amount of time they spent on the second set of word problems. Consistent with the mood-as-input perspective (e.g., Martin et al. 1993), when participants engaged in the achievement task, UR reduced task persistence relative to UE, whereas DR increased persistence relative to DE. In the enjoyment task, however, this pattern was reversed. That is, UR actually increased task persistence relative to UE, whereas DR decreased persistence relative to DE. More generally then, motivation appears to result from rather complex interactions between simulation direction (upward versus downward), simulation mode (reflection versus evaluation), the affect produced by the simulation (positive versus negative), and the type of goal in question (achievement versus enjoyment). Counterfactual thinking and regulatory focus Finally, we have made an initial attempt to test the REM’s predictions regarding the relationships between simulation direction, simulation mode, and regulatory focus. Within the domain of social comparisons, Lockwood et al. (2002) had participants compare themselves to positive versus negative role models and found that (upward) comparisons to the former were more apt to enhance academic motivation when promotion goals were primed, whereas (downward) comparisons to the latter were more apt to enhance academic motivation when prevention goals were primed. In the counterfactual thinking domain, Markman et al. (2004a) hypothesized that promotion goals would be most highly activated after UE because UE involves a concern with advancement (gains or the presence of positives), whereas prevention goals would be most highly activated after DR because DR involves a concern with safety (non-losses or the absence of negatives). Participants recalled a somewhat negative event, made either an upward or downward counterfactual about it, and then engaged in either reflection (“imagine what might have happened instead”) or evaluation (“think about the event and compare it to what might have happened instead”). Subsequently, participants completed Lockwood et al.’s (2002) strength of regulatory focus questionnaire, which includes items tapping the independent strength of both promotion and prevention foci (e.g., “I frequently think about how I can prevent failures in my life”), as well a measure of academic motivation (e.g., “I plan to keep up with the reading assignments”). Intriguingly, the results indicated that UE and DR enhanced the strength of both promotion and prevention foci relative to the other conditions. In turn, UE and DR also enhanced intentions to study harder and improve upon one’s academic performance. The difference between the present results and those of Lockwood et al. (2002) suggest that

88

K. Markman and M. McMullen

counterfactuals may have more general motivational properties than do more specific and targeted social comparisons. We discuss potential differences between counterfactual and social comparisons at the end of this chapter.

Determinants of reflection or evaluation Automatic versus controlled processing Theorists have recently begun to examine the provocative question of what aspects of counterfactual thinking are more or less automatic versus controlled (cf. Bargh 1994; Kahneman 1995; Shiffrin and Schneider 1977). Roese (1997; Roese and Olson 1997; Roese et al. 2004) has argued that upward counterfactuals represent an automatic default in response to negative affect, whereas downward counterfactuals are constructed effortfully in an attempt to override negative affect. On the other hand, Sanna (2000; Sanna et al. 1999; Sanna, Chang et al. 2001) has suggested that upward and downward counterfactuals can be the result of either automatic or controlled processes, depending on the fit between outcome valence and/or mood and the most salient self-motive. Thus, negative outcomes or moods will automatically activate upward counterfactuals to the extent that selfimprovement motives are salient, whereas positive outcomes or moods will automatically activate downward counterfactuals to the extent that moodmaintenance or mood-repair motives are salient. Conversely, a mismatch between outcome valence and/or mood (e.g., negative) and salient selfmotive (e.g., mood repair) will instead stimulate the effortful construction of (in this case, downward) counterfactuals. We are intrigued by these recent attempts to address this issue and would like to offer a new conceptual piece to the puzzle. The models by Roese and Sanna both assume that counterfactuals are initially and automatically contrasted with reality (see also Gilbert et al. 1995; Wegener and Petty 1997). We would argue, however, that assimilation can also sometimes be the automatic default in counterfactual thinking. To understand how assimilation can be the default, it is first important to consider what activates counterfactual generation in the first place. Roese (1997; Roese and Olson 1997) has argued that negative affect, construed as a response to goal blockage (Roese et al. 2004), is the “engine” that activates counterfactual thinking. We, however, prefer a broader conceptualization of how and why counterfactuals are generated. Drawing on Mandler’s (1964) notion that emotions occur in response to behavioral interruptions, we believe that counterfactual thinking may be automatically activated in response to one’s perception that an interruption has occurred in the natural order and flow of the behavioral “event stream.” Our notion of interruption is conceptually similar to Kahneman and Miller’s (1986) suggestion that counterfactuals are activated in response to violations of normality, but it broadens the normality notion by positing that counterfactuals will be automatically acti-

Reflective and evaluative mental simulation 89 vated in response to interruptions that are perceived in either the world that is (i.e., the actual event), or the world that could have been (i.e., the counterfactual event). To illustrate, consider the student who typically receives A grades on exams but has this time received a B. For this student, a B represents an interruption in the typical event sequence (receiving A grades) and, thus, draws attentional focus. In this case, the grade that was will be automatically contrasted with the grade that could have been. This particular counterfactual would be categorized as an instance of UE, and we believe that UE follows from relatively automatic processing. Perhaps even more interestingly, however, consider the case of the individual who switches from the doomed plane flight at the last minute, only to later learn that the plane has crashed, with all lives lost. Here, we would argue, attention is automatically drawn to the counterfactual because thoughts about what could have been (i.e., being killed on a plane flight) represent interruptions in the typical event sequence (i.e., surviving a plane flight). Thus, affective assimilation will be the default because one initially and automatically reflects on the counterfactual in the absence of any explicit comparison to reality. This counterfactual would be categorized as an instance of DR, and we believe that DR also follows from relatively automatic processing. Conversely, we believe that DE and UR are driven by more effortful processes. As has been suggested by previous researchers (e.g., Markman et al. 1993; Roese 1994; Sanna 2000), DE is probably quite often an effortful attempt to maintain or ameliorate one’s present affect. Likewise, it probably requires some degree of effort to maintain an upward simulation while suppressing (cf. Wegner and Bargh 1998) potentially disturbing comparisons between the simulation and the real world. Indeed, although UR may be “cognitively easier” for the more fantasy-prone individual (Rhue and Lynn 1987), it is likely that some initial intent is required before even these types of individuals can become engrossed in their mental simulations of better possible worlds. Motivational trade-offs An important aspect of our work on counterfactual thinking has been our depiction of a critical tension between seemingly opposing motivations: future improvement versus affective enhancement (e.g., Markman et al. 1993; Markman and McMullen 2003; McMullen and Markman 2000). The road to future improvement via upward counterfactuals may engender inordinate amounts of negative affect (Roese and Olson 1997; Sherman and McConnell 1995), whereas the road to affective enhancement via downward counterfactuals may run the concomitant risk of engendering complacency and poor performance. In our view, the resolution of this preparative–affective trade-off plays an important role in determining whether reflection or evaluation will carry the day.

90

K. Markman and M. McMullen

Indeed, this type of trade-off is not uncommon in psychology. For example, although overconfidence and unrealistic optimism may instantiate positive feelings, they may also be self-defeating (Weinstein and Klein 1995; Buehler et al. 1994; for a different view, see Taylor and Brown 1988). Likewise, positive fantasizing engenders positive affect at the expense of complacency and poor performance (Oettingen 1996). Research on selfhandicapping suggests that people sometimes sabotage their prospects of success in order to enhance the availability of more comforting attributions about themselves and their abilities in the case of failure (Berglas and Jones 1978). In the realm of mental health, narcissists feel that they are superior individuals, yet often display self-defeating behavior patterns (Colvin et al. 1995), and perfectionists hold themselves to very high performance standards but put themselves at risk of depression (Blatt 1995). In a cross-cultural context, Americans exhibit higher academic self-esteem but they are outperformed by the more self-deprecating Japanese (Heine and Lehman 1999). Finally, cognitive dissonance theory asserts that one may either attempt to alter a negative behavior or rationalize it in order to maintain self-integrity (Aronson 1992). This trade-off gives rise to a fundamental dilemma: Should I try to improve my performance, or should I try to feel better about myself? Although a variety of models, such as those just described, have addressed different aspects of this trade-off, the issue comes into particularly clear focus within the context of our reflection–evaluation model: UE and DR generally make people feel worse, but are motivating, while DE and UR generally make people feel good, but induce complacency. What factors might determine which motive – future improvement or affective enhancement – an individual adopts? One factor that Markman et al. (1993; see also Sanna 1996, 1997) identified was whether or not an event was to be repeated in the future. When participants in this study believed that they were going to play future games, they were more likely to engage in upward evaluation – displaying a motivation to improve – but when they did not believe that they were going to play any more, they were more likely to engage in downward evaluation – displaying a motivation to feel good about what they have. More generally, we suggest that perceptions of attainability are critical (Lockwood and Kunda 1997). According to the notion of the unidirectional drive upward (Festinger 1954), it is always preferable to successfully obtain a goal. However, when goal attainment is difficult or seemingly impossible, the only way to drive upward may be to fantasize – it is most often easier to imagine being a millionaire than to actually become one. We propose that when a goal is perceived as attainable, one is more likely to use comparative strategies that improve performance (i.e., UE and DR), whereas when a goal is perceived as unattainable, one is more likely to use comparative strategies that improve affect (i.e., UR and DE). A variety of other situational factors and individual differences undoubtedly play a role in determining the per-

Reflective and evaluative mental simulation 91 ceived attainability of a goal, such as whether the event is to be repeated in the future (Boninger et al. 1994; Markman et al. 1993), the perceived probability of success (Teigen 1998b), feelings of self-efficacy (Sanna 1997), tendencies to engage in positive-constructive versus fear-of-failure daydreaming (Huba et al. 1981), incremental versus entity theories of intelligence (Dweck 2000), and differences in optimism versus pessimism (Scheier and Carver 1992). In turn, we would expect these factors to influence proclivities toward reflecting or evaluating. One of the questions that this discussion raises is whether there must always be a trade-off between performance and affect. Are happy people destined to mediocrity, and successful people destined to depression? Is it not possible to feel both good and perform well? We suspect that both are possible. The strength of our reflection–evaluation approach is that the relationships among comparisons, affect, and motivation are not unidimensional; there are multiple avenues (i.e., simulation direction and mode) leading to both positive and negative affect, as well as to increased or decreased motivation. One fruitful avenue of investigation may be to search for asymmetries in comparisons. For example, it is possible that UR can be both motivating and affectively enhancing, whereas DR may only be motivating to the extent that it engenders negative affect.

Counterfactual versus social comparisons We end our chapter with a brief discussion of differences between counterfactual and social comparisons. Researchers like Markman and McMullen (2003) and Olson et al. (2000) have focused on specifying common mechanisms underlying and common consequences accruing from counterfactual and social comparisons. However, we also believe that it is important for researchers to clarify what makes counterfactual comparisons different from other types of comparisons, and therefore worthy of the attention they have received. First, we remind the reader of the findings obtained by Markman et al. (2004a), and how they were somewhat discrepant from those obtained by Lockwood et al. (2002). Whereas Lockwood et al. found that upward social comparisons enhanced academic motivation when promotion but not prevention goals were primed, while downward social comparisons enhanced academic motivation when prevention but not promotion goals were primed, Markman et al. found that UE and DR activated both promotion and prevention goals. We suggest that this occurred because counterfactuals may energize a broader class of motivations. Social comparisons typically focus on a specific target of comparison. In Lockwood et al., this may have motivated participants to think about specific strategies whereby they could attain the outcomes experienced by the positive role model (i.e., promotion), or avoid the outcomes experienced by the negative role model (i.e., prevention). On the other hand, counterfactual comparisons are quite a bit more

92

K. Markman and M. McMullen

diffuse – simulations are somewhat bounded by plausibility constraints but still remain fairly free to vary. Thus, participants in the Markman et al. study were free to imagine many or multiple ways whereby their academic outcomes could have been different. We speculate that the fewer constraints placed upon the choice or selection of counterfactual comparison targets encourages and allows the activation of more generalized and varied motivational strategies. Thus, after engaging in UE or DR following a negative academic event, an individual might pursue a desired end-state via both promotion (e.g., studying more) and prevention (e.g., reducing procrastination) means. Second, we make note of potential differences between the mechanisms that underlie the processing of counterfactual and social comparisons. In a theoretical paper, Mussweiler (2003) attempts to specify how assimilation and contrast effects arise in comparisons. According to Mussweiler, the perceiver initially engages in a holistic assessment of the shared and unique features of the comparison referent and standard. With regard to social comparisons, if the perceiver decides that the self and the standard share common features, then the perceiver will test the hypothesis “How similar am I to the standard?” Testing for similarity will then heighten the accessibility of standard-consistent knowledge such that self-evaluations will be assimilated toward the standard. If, however, the perceiver decides that self and standard do not share common features, then the perceiver will instead engage in dissimilarity testing by asking, “How different am I from the standard?” In turn, testing for dissimilarity will heighten the accessibility of standard-inconsistent knowledge such that self-evaluations will be contrasted away from the standard. Although the example described above focuses on social comparisons, Mussweiler (2003) suggests that similarity testing is the mechanism that accounts for assimilation and contrast effects in all types of comparisons. We agree that similarity testing may account for a wide range of social comparison phenomena. However, we also believe that it is substantially less useful for accounting for assimilation and contrast effects in counterfactual thinking. This can be illustrated easily enough by considering the student who just misses receiving an A in a class by a tenth of a percentage point. If the similarity-testing mechanism were applied, then the student would presumably arrive at the conclusion that her 89.4 was very similar to the 89.5 that she could have received and thus given her an A for the semester. According to Mussweiler, testing for similarity in this case should engender assimilation, thereby leading the student to feel good about her 89.4. Our intuitions, however, would suggest that the very opposite would occur – our student would probably be quite frustrated, bemoaning the fact that she “just missed” getting an A. Indeed, we would argue that it is the similarity of the real grade to the imagined grade that actually gives rise to such feelings of frustration (i.e., affective contrast)! After all, Kahneman and Tversky’s (1982b) participants judged that Mr Crane would be quite upset,

Reflective and evaluative mental simulation 93 and not at all happy, when he discovered that his plane left just five minutes ago. To illustrate a related point, consider Nike’s “Be Like Mike” (i.e., NBA former player Michael Jordan) advertising campaign a few years ago. Mussweiler (2003) suggests that individuals initially engage in a holistic assessment of one’s similarity to the standard and then engage in similarity or dissimilarity testing depending on the outcome of this initial assessment. Following this logic, most people should conclude that they are dissimilar to Jordan, thereby engendering contrast. However, our observation that children and adults alike pretend that they are “being like Mike” when they get on the basketball court indicates that this is clearly not what is happening. Instead, we argue that it is the act of reflecting on what it would be like to experience the success of Michael Jordan that enhances the accessibility of standard-consistent thoughts about the self – it is hardly necessary to test for similarity between the self and Michael Jordan in order to produce assimilation. In fact, testing the hypothesis that one is similar to Michael Jordan would simply highlight how dissimilar one is to him. Thus we believe that our reflection mechanism can account for these “Be Like Mike” effects in a way that Mussweiler’s hypothesis-testing mechanism cannot. In sum, we hope that our reflection–evaluation approach stimulates novel hypotheses and opens a new window on counterfactual thinking that reveals more of its richness and complexity. In our view, no comprehensive approach to counterfactual thinking can succeed without incorporating assimilation and contrast effects. Many models of comparative thinking include assimilation and contrast as central components, whereas research on counterfactual thinking has lagged behind for substantially too long. By incorporating assimilation and contrast, we would argue, the affective and motivational issues that are so central to the functional approach to counterfactual thinking are fundamentally transformed: Counterfactual thoughts can motivate and discourage, assure and alarm, inspire and depress. In this chapter, we hope to have provided a glimpse into some of the new ideas that arise from this approach to counterfactual thinking.

6

Scenario simulations in learning Forms and functions at the individual and organizational levels Susana Segura and Michael W. Morris

People simulate scenarios of what-might-have-been and what-might-be as they make sense of everyday experiences (Kahneman and Miller 1986; Kahneman and Tversky 1982b; Taylor and Schneider 1989). Researchers have studied these tendencies to simulate counterfactual and prefactual scenarios, asking why some forms of simulation are more likely than others (Kahneman 1995). One answer to this question is that forms of simulation serve distinctive functions, such as learning from experience (Roese and Olson 1995a). A central argument has been that upward-directed simulations (how an outcome could have gone better) foster learning more than downwarddirected simulations (how it could have gone worse). In the first decade of research on scenario simulation, evidence from several studies converged in support of the proposition that upward and not downward simulations function in learning (see Roese and Olson 1995a). This chapter re-examines the relation between simulation form and the learning function in light of an expanded literature. One dimension of expansion is temporal. The rate of research on counterfactual thinking has rapidly grown, and findings have emerged to question the generalizations of early theories. To analyze these complexities, we distinguish three steps of information processing in experiential learning – evaluating the observed consequences of one’s action, inducing changes to one’s operating rules, and then implementing novel strategies (putting the new rules into practice in one’s next encounter with the task or situation). By identifying distinct roles of simulations at these steps, we integrate new and old findings into a more comprehensive account. The literature reviewed here is also expanded in the range of phenomena considered. Whereas early studies of counterfactual thinking and its consequences focused primarily on the context of independent tasks (such as puzzle solving or exam performance), recent research has focused increasingly on interdependent contexts (such as those in social situations or work organizations). Interdependent contexts are those where one individual’s outcome depends not only on this individual’s actions but on those of other people as well. Interdependence makes diagnosing causes of inadequate outcomes, and remedying them, more challenging. Our framework for steps of

Scenario simulations in learning

95

experiential learning is used to analyze which relationships between the form and function are changed by interdependence and which are not. A final dimension of expansion is a greater radical departure from past literature reviews. In addition to reviewing psychological evidence about mental simulation in individual’s learning, we look at evidence about scenario simulation in organization-level learning. Dating from the seminal work of Herbert Simon (1967), researchers of organizations (such as government agencies, corporations, or divisions thereof) have examined how their processing of information mediates changes in their behavior (Herriott et al. 1985). As our framework highlights, organization-level learning from experience proceeds through the same steps of evaluating consequences, inducing rules, and implementing plans.1 Simulation of scenarios figures prominently in organization-level learning. These simulations differ from mental simulations in that they are elaborated in public, sharable formats – in narratives, reports, computer programs, and so forth – yet they can be analyzed on the structural dimensions (i.e. upward versus downward). Hence, as a stimulus to new insights about form and function at the psychological level, we examine the parallels and differences in the forms of simulation that seem to foster experiential learning at the organizational levels.

Experiential learning framework The literatures on experiential learning by individuals and by organizations, respectively, have developed models to describe the information processing involved. A key premise in both literatures is that action occurs through rule following and that rules capture past experience. In psychology and cognitive science, people act by following implicit rules associated with particular situations or tasks, and they update these knowledge structures on the basis of experience (Anderson 1983; Holland et al. 1986). In sociology and administrative science, firms and agencies are understood as acting based on formal policies and informal standard operating procedures, and rules are refined on experienced outcomes (Cyert and March 1992; March 1988; March and Olsen 1989; March and Simon 1993). In highlighting the assumptions of behavioral models, it is instructive to contrast these assumptions with rational models. In rational models, such as those developed in economics, agents make choices based on calculations of expected future consequences, according to a utility maximization algorithm. That is, agents know the utility associated with every possible value of every outcome variable, and they know the probability of its occurrence with a given action alternative, and hence they can compute the overall expected utility for any given action. Before acting, they consider all possible action alternatives in this way, and they choose the option that maximizes utility. After each round of action, agents observe the attained outcome and use this information to update their knowledge of the probabilities (their expectancies) in a Bayesian manner.

96

S. Segura and M.W. Morris

This rational model is easily instantiated in a domain like chess, where a player’s action alternatives at any point are clearly defined. Indeed, chessplaying computer programs act and learn in this manner – making anticipatory calculations about the consequences of various action alternatives and updating probabilities based on experience. Interestingly, however, human chess experts do not act and learn this way; experts and novices do not differ dramatically on the dimension of number of moves ahead considered. Where they differ is in their recognition of tactically relevant patterns when they see a given chessboard (De Groot 1965). These are rules in the sense that they are pieces of procedural knowledge triggered by particular board situations, not necessarily that experts would articulate them as rules. Comparative expert/novice studies find experts have greatly superior memory for possible chessboards (configurations of pieces that could arise in a game and, hence, experts have likely experienced) but not for impossible chessboards (configurations that could not arise). In sum, the chess example illustrates that even in relatively simple domains, people act on the basis of rules capturing past experiences rather than calculations about future consequences. An example of a more complex real-world domain highlights other contrasts between rational and behavioral models. Suppose a restaurant wants to figure out the best way to assign its employees to jobs (i.e., at each location there are bookkeepers, waiters, chefs, janitors, etc.). Its preferences might include: economic interests such as profits, sales, and stock value; reputation concerns, such as customer satisfaction and community respect; and change agendas, such as promoting diversity in management. The rational model holds that the firm has a set of stable, exogenous preferences that are sufficiently precise as to specify the utility of each level of each variable and how they interact. Moreover, the firm knows the probabilities of each outcome given an action. The firm chooses an action by exhaustively searching the space of possible action alternatives (if there are 100 employees and 100 jobs, then there are thousands of such possibilities). This model further assumes that, after acting, the restaurant receives precise feedback from experience on all the outcome variables to update beliefs. That is, the firm can see which value attained on the outcomes variables occurred. Of course, this portrayal strains credulity because feedback on these abstract dimensions is not simply provided by the world, it has to be constructed on the basis of interpretation. Considerable interpretation is required to reach evaluations not only on dimensions such as diversity, but also on seemingly straightforward dimensions such as profitability (as accounting scandals remind us). Compared with the rational model, behavioral models posit a procedure that is less computationally intensive and more robust to incomplete or ambiguous information. As shown in Figure 6.1, three steps of heuristic information processing – evaluating consequences, inducing new rules, and implementing new actions – can be distinguished in the cycle of learning from experience. Evaluating outcomes is a judgment from the perceivable

Scenario simulations in learning Inducing a new rule for action in a situation

Evaluating the experienced consequences

Implementing new action in one’s next encounter with situation

Consequences follow from one’s action

97

Figure 6.1 Experiential learning framework. Dotted lines show steps of information processing.

consequences of one’s action to an assessment of the value of the outcome in light of one’s preferences. Inducing changed rules is a matter of figuring out whether the new evidence from this round of experience warrants a change in rules and, if so, what kind of change. Implementing or putting a novel rule into practice involves planning the changed tactic or strategy concretely so that it can be enacted with the right timing and touch so as to be effective. We now review how these occur in psychological and organizational models and propose the role that scenario simulation plays in each step. Three steps: evaluating outcomes, inducing rules, and implementing actions The first step is evaluating the consequences of one’s action. This involves translating the incomplete and ambiguous information in the perceivable record of events to an assessment of its value on outcomes of interest. One insight about how agents overcome this missing data is they evaluate each preference dimension or attribute separately rather than evaluating all of them in combination. In research on organizational learning, studies find that changes to strategy are sensitive to performances that are poor on any specific attribute, such as stock price, rather than in the aggregate of several attributes, such as stock price, profits, and market share (Cyert and March 1992). The same tendency to evaluate each attribute separately is seen in the “elimination by aspects” model of individual decision making (Tversky 1972). Another insight about how agents heuristically simplify the process is, rather than judging absolute-level performances, they judge performance relative to aspiration levels (March 1988) or reference points (Kahneman and Tversky 1982b). In terms of the example, this means that decision makers at

98

S. Segura and M.W. Morris

the restaurant do not seek to judge the year’s performance on an overall combined metric. Rather, they consider each preference dimension in turn, and look for evidence of discrepancies from an aspiration level or reference point (“Do profits look worse than last year’s?” “Is our diversity progress better than the competitor who was sued last year?”). Behavioral models assume that agents heuristically scan the dimensions of their preferences, looking for signs of a problem. If a problem is detected, the agent is then motivated to progress to the next steps of learning. The role of scenario simulations in the evaluation step is providing dimension-specific reference points. Often reference points come from targets to which an agent has committed in advance – an athlete’s predicted time in the race, a firm’s quarterly revenue projection. Yet in retrospective evaluation, such prospective targets can be displaced by more salient retrospective reference points. For example, research participants asked to imagine that delays had caused them to arrive at an airport late are sensitive not only to the discrepancy from their prospective target but by how closely it turned out that they missed the flight (Kahneman and Tversky 1982b). A large body of evidence from organizational research shows that the reference points firms use in evaluating their performance are not chosen prospectively; they adapt to the experienced level of performance, successes evoking comparisons to higher targets and failures to lower targets (March and Simon 1958). The second step of heuristic processing is inducing changes to rules for a situation. Limits of information processing capacity mean that agents conserve existing rules unless change is imperative. When they do search for alternatives, they follow a “satisficing” rather than maximizing procedure for simulating alternatives (Simon 1967). This is often described as a “thermostatic” process of search for alternative actions: firms invest energy in considering alternative actions only when evaluation indicates a failure, search proceeds until an alternative action that corrects this particular discrepancy is identified, then search ends and this alternative is enacted without searching exhaustively to see if there is an even better alternative (March and Shapira 1999). Similarly, in many psychological tasks, people are conservative in updating their rules (Edwards 1961) and their processing is triggered more by negative than positive experiences (Taylor 1991). The role of counterfactual simulations at the rule induction stage is in drawing attention to the changes that will improve the outcome. This step is the one where most past theorizing about simulations and learning has focused, probably because of the long tradition from Mill (1872) implicating counterfactual scenarios in the induction of causality. Integrating this and other ideas, Hilton and Slugoski (1986) identified that perceivers assign causality to factors that stand out as abnormal conditions, because it is easy to generate the counterfactual scenario (with this factor absent and the outcome absent) from knowledge of normal conditions. A counterfactual scenario that imagines away a past failure turns into a prefactual scenario of

Scenario simulations in learning

99

how a future disaster can be averted (Roese 1997). This counterfactual-cumprefactual thinking is the heuristic means of generating a novel action rule that corrects an attribute-specific failing. The final step is implementing rules in a new action. Rational models are silent about implementation – decision makers calculate the odds, lay the best bet that they can by choosing the utility maximizing action and then await the outcome. Behavioral research draws attention to the performance that is required to enact a new strategy, and that implementation is an ongoing process that involves some “way finding.” Studies of business executives dealing with risk, for instance, show less concern for making the right decision than making the decision right; less effort is made to learn the probabilities to predict risks than to control risks after the decision through adjustments (March and Shapira 1987). Executives reject the very premise that their decisions, like those in games of chance, feature risks that are exogenous and inevitable. Psychological research shows similar patterns – people seek to control situations and even in the case of diseases like cancer believe that risks are not exogenous (Taylor and Schneider 1989). While trying to control risks in a game of chance (e.g. rolling the dice hard to get a high number) is a fallacy, in most problems that people face an active and responsive approach to implementation is helpful in attaining the desired outcome. In the final step of implementing a novel action, scenario thinking helps identify highly feasible plans and sustain one’s confidence. Counterfactual scenarios identify plans that are easy to implement because they look for minimal rewrites of history and the more detail people have in their reasons for an action, the higher their confidence (Dawes 1976).

Functions of upward versus downward simulation in the experiential learning framework How is learning differentially fostered by upward and downward simulation at each step of the present framework? One source of evidence is the comparative prevalence of upward counterfactuals in the conditions where the goal of learning is activated, such as tasks that people expect to repeat as opposed to a task expected to be non-repeatable (Gleicher et al. 1990; Markman et al. 1993). Likewise, upward counterfactuals are relatively prevalent after negative experiences (Nasco and Marsh 1999; Sanna et al. 1999). Other findings come from experiments manipulating instructions to participants to generate upward versus downward simulations and measuring the effect on learning. In this design, Roese (1994) found that participants’ upward counterfactuals about their exam performance led them to generate more lessons (behavioral intentions to improve their study routines), and that participants upward counterfactuals about performance on the first round of an anagram task led them to generate lessons and improve their score on subsequent rounds. On the basis of this evidence, Roese (1994) concludes

100

S. Segura and M.W. Morris

that upward as opposite to downward counterfactuals can prepare for future improvement. These findings have been replicated in a longitudinal study of naturally occurring counterfactual thoughts about academic performances and subsequent academic performance (Nasco and Marsh 1999). However, more recent evidence does not support this claim. McMullen and Markman (2000) studied people’s reflections on risky behaviors, such as driving in bad weather. Motivation to make changes was less engendered by upward simulations (e.g., “If I had not driven in a snowstorm, my traction would have been better”) than downward simulations (“If I had encountered traffic on the road, I probably would have had a collision”). Downward counterfactuals provide a “wake up call,” a signal that one’s behavior is in need of change. To reconcile this and other recent results, it is important to analyze the effect of direction separately in each step of experiential learning. Evaluating outcomes How does direction of simulation affect the initial step of evaluating outcomes? Recall that negative evaluations motivate individuals to pursue the subsequent steps of learning. Upward counterfactuals produce negative evaluations when they are reference points against which experienced outcomes are contrasted. Medvec et al. (1995) found that Olympic silver medalists evaluated their performance negatively in contrast to the salient upward counterfactual of placing first, whereas bronze medalists evaluated theirs positively in contrast to the salient downward counterfactual of placing fourth. However, comparative thinking does not always depend on a contrast process that emphasizes the difference between the stimulus and the referent. It can also involve assimilation, imbuing the stimulus with features of the referent. This seems to be the explanation for McMullen and Markman’s (2000) findings concerning downward counterfactuals in the domain of risk-taking behaviors. Evaluations of the actual experience are assimilated into evaluations of worse possible outcomes. When this is the case, downward counterfactuals should spur learning and, furthermore, upward counterfactuals should impede learning. This latter dynamic may be at work in findings that political forecasters retain their theories after failed predictions because they were “almost right” (Tetlock 1998). Recent theorizing has sought to clarify when each of these modes of comparative judgment will dominate, and one promising idea is that assimilation follows from an experiential (“as if”) mode of simulating oneself in the alternative scenario whereas contrast follows from a more distant, onlooker perspective in imagining the scenario (Markman and McMullen 2003). Research on simulation and learning in the context of interdependent tasks suggest another moderator of the link between upward counterfactuals and negative evaluations. Regret is evoked by contrasts with upward selffocused scenarios (imagined actions one could have taken to improve the

Scenario simulations in learning

101

outcome), whereas disappointment involves thoughts of how better outcomes could have resulted without a change to one’s actions, such as if other persons had acted differently (Zeelenberg, van Dijk et al. 1998c). Mandel (2003a) studied the counterfactual thoughts and emotions that students experienced in reaction to negative life events in an independent context (academic experiences) and an interdependent context (interpersonal experiences). In the independent context, upward self-focused counterfactuals were far more prevalent than in the interdependent context. Mediational analyses revealed that these upward self-focused counterfactuals, rather than upward counterfactuals in general, were the critical ingredient in producing selfblame and self-focused negative emotions (regret, guilt, and shame). These findings converge with those from studies of aviation pilots’ responses to dangerous incidents (Morris and Moore 2000). In dangerous incidents, pilots are interdependent with other planes, controllers, mechanics, and so forth. Two methods were used to study counterfactual thoughts and lessons in this context: laboratory experiments in which students experienced outcomes on a flight simulator, and archival data of narratives by actual pilots filed after experiencing dangerous incidents. Counterfactual thoughts were coded as to direction and focus (self versus other). For example, a self-focused upward thought is “If only I had started my descent earlier, I would have been able to land without skidding” whereas an externally focused upward thought is “If only the other plane had not crossed the runway, I would have been able to land without skidding.” Personal learning was operationalized in terms of behavioral intentions to change one’s approach to a situation. Results showed that learning followed only from self-focused upward counterfactuals, not from all upward counterfactuals. Another aspect of findings – related to self-blame – were the effects of whether or not the pilot was accountable to organizational superiors (manipulated in the experiment and measured in the archival study). Accountability of this sort was predicted to evoke “defensive bolstering” (Lerner and Tetlock 1999) and hence inhibit self-focused upward counterfactuals, which are potentially self-implicating. Results from field and laboratory studies were consistent with this hypothesis: accountability reduced selffocused upward counterfactuals and, consequently, reduced learning from experience. Inducing rules This is the step where Roese’s arguments about the uselessness of downward counterfactuals can be understood. The key is the premise that the agent seeks lessons for improvement from the current level: upward counterfactuals spotlight springboards for improvement whereas downward counterfactuals spotlight pitfalls to be avoided. If an agent simply wanted to induce as much as possible about the key causes of an outcome variable, upward and downward counterfactuals would be equally useful – both enable Mill’s

102

S. Segura and M.W. Morris

method of difference. However, agents do not have a scientist-like interest in identifying the general causes of an outcome; they have a practical interest in identifying what makes the difference between the failure (which spurred their learning) and the reference point on this outcome dimension. Yet, the above logic only works given the assumption that the actor can in fact improve in the sense of approaching the reference point outcome. In some interdependent task contexts this is not the case. Consider the class of competitive tasks in which improvement involves moving up to the next level of competition. In this setting, simply replicating the past performance (under more challenging circumstances) would be improvement, and attempting to better the past performance is futile. For example, a pitcher moving from the minor leagues to his first major league game would be counted a success so long as the first game is not a disaster. Hence, his preparation would not be well served by upward simulations (“Last week’s one-hitter could have been a shut out, if only I had thrown an off-speed pitch”). Instead, attention would be more prudently directed to downward simulations (“Last week’s game could have been lost to a grand slam, had the wind not blown that ball foul”). Pitfalls narrowly avoided in the past are best not forgotten as one moves into circumstances that will make resisting them more difficult. Studies of another interdependent context – negotiation – illustrate a related point. In many negotiations, a tactic that leads to favorable outcomes is proposing the initial offer at an ambitiously self-favoring level. Though counterparts typically reject the offer and counter with their own, their expectations tend to be none the less anchored in favor of the first party’s offer. When a first offer is immediately accepted, this usually indicates that the counterpart was expecting a far worse deal, and hence suggests that one could have done even better. Galinsky et al. (2002) asked subjects to play the role of buyers negotiating over a house, and manipulated whether the seller (a confederate) immediately accepted the first offer or negotiated for a higher price before accepting. Participants in the immediate acceptance had higher economic outcomes but were less satisfied, consistent with notion that upward counterfactuals produce negative evaluations. Furthermore, the availability of upward counterfactuals predicted the time participants planned to spend preparing for their next negotiation, consistent with the negativity to motivation linkage. Yet when asked their planned tactics for next time, upward counterfactuals were associated with plans to switch away from this (generally adaptive) tactic. Other studies of interdependent contexts highlight similar examples, in which upward counterfactuals do not engender good lessons because of the antecedent factors on which they focus. Among the many biases identified in antecedent selection, people tend to focus on human actions rather than states of the environment (Girotto et al. 1991), especially actions that violate social norms (McCloy and Byrne 2000; Segura and McCloy 2003). Such bias inhibits clear thinking in the case of a manager who has to remedy problems

Scenario simulations in learning

103

either by changing the work environment or by changing individual employees. A “human error” fallacy is described in managers’ responses to industrial accidents – managers respond by replacing the individual employee operating the machine and are slow to recognize problems caused by the machine environment, even when the problem recurs across different individual operators (Norman 1990, 1992). Several studies by Morris et al. (1999) linked this human error fallacy to counterfactual thinking, by showing the presence of a request to consider “if only” scenarios increased the rate of human-focused remedies. In summary, habits of antecedent selection lead maladaptive upward counterfactuals in domains with more complex causal structures. Implementing actions Does direction matter in the final step of implementing the new action? Here the answer seems straightforward: upward prefactual scenarios should increase confidence, and this confidence sustains people’s motivation and effort in the process of implementation. In a famous example, James (1948) noted that hikers have a better chance of leaping over a stream when their confidence is high than low. Envisioning a future success is common among athletes and it may provide a mental rehearsal that increases fluency (Taylor and Pham 1996). Taylor (1989) argues that even unrealistic upward scenarios can be adaptive in providing self-efficacy. Granted, research on defensive pessimism has identified seemingly opposite strategies (Norem and Cantor 1986); some individuals succeed by first imaging ways an upcoming performance (say, public speaking) will go badly (losing one’s notes, being stumped by questions, etc.) and then mapping out detailed plans to avert each possibility. Yet in a process in which an initial downward simulation is followed by an upward simulation, the downward-simulation step seems primarily to generate negativity whereas the subsequent upward-simulation step functions to guide implementation. In interdependent tasks, the link between upward simulations and implementation may be even stronger. Studies of executive decision making find that executives gather information not so much to pick one alternative over the others as to elaborate the success story featuring the favored option (see March 1994). This may help implementation by supporting not only the executive’s self-confidence but also the confidence of his or her co-workers, who expect the leader to have “vision” (MacCrimmon and Wehrung 1986). In studies of decision making under uncertainty, March and Shapira (1987) found that executives did not collect information about the full distribution of possible outcome values and their associated probabilities; rather, they formed a detailed picture of the best-case “maximum gain” scenario and usually also the worst case “maximum loss” scenario. If success depended on accuracy in judging risk, assessing options in terms of just these two points would be tremendously costly (as the shape of distribution matters as much

104

S. Segura and M.W. Morris

as its range). Yet if executives deal with risk primarily by controlling it rather than predicting it, by implementing strategies responsively, then holding several vivid scenarios in mind may be useful in monitoring of how the world is unfolding. Compared to managers in large organizations, entrepreneurs may have an even greater need to maintain confidence and optimism in order to implement their plans. The vast majority of ventures fail; entrepreneurs without unshakable confidence in the correctness of their judgment may have difficulty convincing employees and investors. Palich and Bagby (1995) found that entrepreneurs tended to perceive greater potential for gain in highly uncertain situations. Yet, interestingly, entrepreneurs’ upward prefactual scenarios may not derive from upward counterfactuals about past experiences. Compared to individuals in other professions, entrepreneurs report fewer upward self-focused counterfactuals about missed opportunities in the past (Baron 2000). Much like pilots in hierarchical organizations, their accountability may constrain their habits of thought; they may avoid entertaining scenarios that dampen their self-confidence.

Organization-level learning Functions of upward and downward scenario simulations can also be identified in organization-level learning. Organizations respond to experience by adjusting their policies or strategies, through analogous steps of evaluation, induction, and implementation. Organizations have increasingly adopted a variety of simulation practices and techniques to facilitate learning from experience (Senge et al. 1994). Counterfactual and prefactual scenarios are elaborated by members of the organization, as well as by external consultants, and they are stored in external and sharable formats such as narrative reports or computer models. The research literature on these techniques primarily consists of case studies, which trace examples of how forms of simulation produced insights that spurred changed policies and procedures. In reviewing this literature, it is easy to identify examples relevant to the three steps of experiential learning and to classify the direction of the simulation in each case. Hence, although we lack statistical data of the prevalence of upward versus downward simulations and their link to learning, we can review examples to check whether the dynamics of simulation as a learning heuristic seem to function in similar ways at the organizational level. Evaluating outcomes A prominent example of organization-level practices related to the evaluation step of experiential learning is “benchmarking” – setting an ambitious yet attainable standard in the form of a dynamic model (March 1994). Firms want their divisions to evaluate their outcomes in ways that give reliable feedback about their efforts. Retrospectively generated reference points fail

Scenario simulations in learning

105

for several reasons: they anchor on experienced performance, and if they key on absolute numbers or numbers relative to last year they hinge too much on exogenous forces such as economic recession or recovery. These same forces mean that standards in the form of prospective point estimates often end up far below or far above what seems a reasonable standard of excellence. Benchmarking involves a model that controls for exogenous forces, such as indexing the average division or the best division in the firm, or indexing each division against industry leaders in the same product line or geographic region. In any case, a scenario for excellent yet attainable performance derives from the model and this is supposed to take the place of otherwise salient scenarios from intuitive counterfactual thinking. Benchmarking techniques suggest the idea that upward scenarios are motivating when they seem plausibly attainable. Perhaps plausibility is mostly relevant in the case of externally (managerially) imposed scenarios. Yet research on individual counterfactual thinking suggests that people distinguish “might-have-been” scenarios from “would-have-been” scenarios (Segura and Morris 2001), and it may be that it is primarily the latter type that produce negative evaluations and ultimately motivation to learn. Other organization-level simulation techniques have emerged in the area of project management (Pitagorsky 2000). Because complex projects with many unknowns generally overrun deadlines and budgets, firms seek to evaluate whether performance by the contractor is a failure through “project post-mortems,” retrospective investigations by consultants. Typically, the managers involved in the project have an accurate understanding of causes of delay within a limited purview; they sometimes are unaware of interdependences between their work and other components of the project. Hence, consultants use the method of converting individuals’ stories into maps of influence relations among historical events (Eden 1988). These can then be combined across informants to create a holistic model of the chain of events that reveals feedback loops and other causal complexities (Williams et al. 1996). This model can be instantiated with quantitative data from object records of project events, and then simulations can be run exploring various parameter changes to investigate counterfactuals. Simulations reach “if only . . .” conclusions if the parameter appreciably changes the endogenous outcome variables. Equally important, they reach “even if . . .” conclusions if it does not, which may guard against superstitious learning. A final idea about the individual psychology of evaluation comes from the technique of learning from near-histories. Tamuz (1987) found that aviation regulators draw lessons from these downward simulations of how a near accident could have been worse. Yet recall that Morris and Moore (2000) found that individual aviation pilots do not draw lessons after downward simulations (and one of their studies analyzed the same pilot narratives that the regulatory agency compiles). Why do pilots take comfort in the fact that danger was avoided instead of taking the danger as a wake-up call, as the regulatory agencies do? One answer may be that such dangerous incidents

106

S. Segura and M.W. Morris

are rare (expected not to repeat) in the experience of an individual pilot but frequent (expected to repeat) in the experience of a regulatory agency monitoring all the planes in the country. Another answer may be that pilots do not take a stochastic view of the events in the way that regulators do, perhaps cued by their larger sample to think this way. Events which people expect to repeat and assume to be stochastic, like driving in bad weather, may be ones where individuals engage in assimilative processing of downward counterfactuals. Inducing rules Organizational researchers have analyzed the roles of scenario simulation at this step in some detail. These have been noted in studies of organizations that seek better solutions to problems for which history offers meager data, such as problems that are infrequent, novel, or not well recorded (March et al. 1991). One class of techniques is defining and elaborating “near-histories” – events that could have happened had chance circumstances been slightly different. On a view of history as stochastic rather than deterministic, these once possible yet unrealized outcomes are as worthy of consideration as actual outcomes. A practice of military strategists and educators is elaborating alternative histories of battles such as Midway, which hinged as much on luck as on strategy, to caution against overgeneralized lessons (Prange 1982). In industries such as aviation and nuclear power, where it is hard to learn from actual accidents owing to their infrequency and mortality, regulatory agencies collect extensive data about “near accidents” or “safety incidents” (National Research Council 1980). The aviation program, for instance, requires that immediately after a “near collision” (defined objectively in terms of physical distance) pilots and air traffic controllers file narrative accounts of the incident and its perceived causes; analysts enter the incident into a dataset by triangulating on a picture of what happened, and this dataset has spurred many regulatory changes concerning air traffic control systems, cockpit routines, airport layout, pilot training procedures, and so forth (Tamuz 1987). In relation to our thesis, a striking feature of the nearhistory simulations described in the literature is that they are downward rather than upward. Armies do not elaborate “near victories,” nor do research laboratories take comfort in “near discoveries.” Thinking about downward near-histories may attenuate the tendency of organizations to take a positive view of their past performance (Starbuck and Milliken 1988) – a tendency likely due to hindsight (Fischhoff 1975) and self-serving (Miller and Ross 1975) biases. That is, organizations need near-histories to appreciate the possible dangers inherent in the status-quo policies. Perhaps the most ubiquitous organization-level simulations used in inducing changes to policies or strategies are those based on computer models, such as spreadsheets. Such models begin with theories about rela-

Scenario simulations in learning

107

tionships among observables in the organization’s functioning, and are finetuned by fitting equations to the samples of recent history for which the data are available (often quite small samples). Levy (1989) noted how the adoption of spreadsheets increased the facility with which many “what if?” questions were explored by managers before a decision. Counterfactual scenarios are simulated by setting input variables to a recent event and then exploring whether changes to a given parameter appreciably affect the endogenous outcome variable. A difference from intuitive counterfactuals is that computer-based simulations are somewhat released from the processing constraints of the human mind, so more scenarios can be simulated and more complex counterfactuals (such as those changing two parameters at once) can be explored. Prefactual scenarios are simulated by setting input variables to levels corresponding to states expected and actions planned at a point in the future. Note that prefactual scenarios are on shakier grounds in terms of data analysis, in that the model is valid only for making estimates about the range of past events for which it was computed. This may elucidate a parallel between organizational and intuitive induction through simulation, that they involve counterfactual as well as prefactual scenarios. Though the latter are more relevant to future action, the former can be run with more confidence. Implementing actions Several kinds of organization-level simulation are relevant to this final step. Studies of the introduction of fiscal-impact computer models into the process of policy found that, in local government, computer models often foster consensus building, by providing a common representation of the problem that different stakeholders can use to identify their common assumptions established and jointly investigate the implications of different proposals (Northrop et al. 1982). A contrasting picture emerged, however, from studies of national-level politics involving larger organizations and more complex models, where more complex modeling more typically was used in a partisan fashion to justify policies rather than to negotiate them (Kraemer et al. 1987). Such models can have an intimidating complexity and a veneer of objectivity that discourage debate. When abstract models are imposed from outside, insiders who are experts in the domain but not experts in modeling may be less able to correct poor policies. As an example consider Robert McNamara, who succeeded brilliantly in running Ford by attention to production and economic numbers, and then failed dismally as US Secretary of Defense trying to conduct the Vietnam War according to similar models. Few in the army below him had faith that the metrics, such as body-count ratios, were relevant drivers of victory (Pfeffer 1992). Here the prefactual scenarios generated from an abstract model failed to serve as a common vision that would inspire confidence, coordination, and fluency of performance, in part because there was low commitment to the scenario and

108

S. Segura and M.W. Morris

in part because it reflected a model of the domain that failed to correspond to the reality on the ground that insiders experienced. More recent programs in the US Army involve more decentralized simulations aimed at prefactuals that can be more easily implemented. In the “After Action Review” program, facilitators meet with soldiers after a mission or training exercise asking them to consider what could have been done, individually or organizationally, to achieve better outcomes. These sessions function to spur personal learning, yet that is not their only function. The valuable ideas that arise in these discussions are forwarded to a knowledge management program called the “Center for Army Lessons Learned,” where they are analyzed in light of models of particular mission situations and then used to update policies and training procedures for the particular situation (Davenport and Prusak 1998). For example, lessons learned in peacekeeping missions in Somalia and Rwanda in the early 1990s in tasks such as searching houses for weapons led to improved performance of these tasks in the 1994 Haitian mission. The goal is eliciting “grounded lessons” – action rules expressed by soldiers in the situation that will be comprehensible and implementable by their peers in analogous situations. Another aspect of the comparison between mathematical models and narratives as modes of simulation is illustrated by the strategic planning technique known as “scenario analysis” (Schwartz 1991). The premise is that an artificially clear portrait of the future becomes socially constructed in organizations, leaving them exposed and surprised when this future scenario does not come true. The technique aims to ameliorate this, without trying to change the fact that people, especially in organizations, tend to construe the future in narrative scenarios with far more detail than can be rationally predicted. Consultants and managers elaborate several alternative scenarios that challenge the “official future” in the organization, some optimistic and some pessimistic. Owing more to narrative theory than decision analysis, the process identifies driving forces, predetermined elements, and critical uncertainties in the plot of the official future. It then constructs qualitatively different alternative scenarios based on different plot lines. Whereas prefactuals based on counterfactuals tend to vary one parameter at time, the technique produces more radically disparate scenarios. Furthermore, a key feature in each alternative scenario defined is a set of observable implications that the organization could use to monitor whether the scenario is coming true.

Conclusion We have reviewed evidence about forms and functions of counterfactual thinking at the level of individual learning and organizational learning. At each of these levels, simulations function in three steps of learning – evaluating outcomes, inducing rules, and implementing actions. Though our review of these levels and steps has generally supported the rule of learning fostered by upward simulations, it has also revealed some important exceptions.

Scenario simulations in learning

109

At the individual level, evidence suggests that the upward advantage occurs at all three steps in learning process, yet only under certain conditions. At the evaluation step, the upward advantage comes from the generation of negative emotion; hence, it hinges on a contrast rather than assimilation mode of comparative thinking. In addition, it depends on whether upward counterfactuals are self-focused or externally focused. At the induction step, the upward advantage depends on the possibility of performance improvements through tactical shifts; if the individual is currently optimizing then upward scenarios will lead to tactical changes that reduce performance. At the implementation step, the upward advantage holds for prefactual scenarios – they foster confidence, fluency, and coordination. Yet it may not hold for upward counterfactual scenarios, which can undermine an individual’s self-confidence and thus hinder implementation. These qualifications of the upward advantage rule, suggested by our review of the extant evidence, await direct tests in future research on mental simulations and learning. The evidence we have reviewed concerning organizational-level simulation practices and learning suggests a different set of qualifications to the upward advantage rule. As an example of simulation functioning at the evaluation step, benchmarking techniques highlight the issue of the plausibility of the upward counterfactual outcome. At induction step, the practice in hazardous industries of learning from once-possible failures clearly indicates a practice that focuses on downward comparisons. This may be part of a larger difference between strategies used in prevention- as opposed to promotion-focused thinking (Liberman et al. 1999). At the implementation step, the literature suggests there are important differences between mathematical as opposed to narrative simulations of scenarios. Narrative formats seem to have advantages in fostering commitment, coordination, and feasibility. These aspects of the relationship between upward simulations and learning deserve research at the organizational level, and they may point to analogous issues at the individual level, which can be assessed in future psychological research on mental simulation and learning.

Note 1 It should be noted that the organizational literature does not anthropomorphize organizations; it assumes that organizational information processing occurs through complex political processes (March 1988). For the sake of brevity, we will refer to the firm as deciding and learning, but this doesn’t mean that the firm is monolithic or anthropomorphic.

7

Finding meaning from mutability Making sense and deriving significance through counterfactual thinking Adam D. Galinsky, Katie A. Liljenquist, Laura J. Kray, and Neal J. Roese What matters . . . is not the meaning of life in general, but rather the specific meaning of a person’s life at a given moment. (Victor Frankl 1984: 171)

As the opening quotation suggests, at each turn in life, the search for meaning represents an integral part of the human experience. Meaning is acquired in many ways. For some, meaning is sought through spiritual enlightenment. For others, meaning is obtained through status and the adoration of others. Still others cultivate intimate personal relationships as the defining purpose of their lives. Regardless of the particular substance of an individual’s quest for meaning, we contend that the psychological need for meaning is universal. One way in which meaning is derived and enhanced is through the consideration of “what might have been.” In this chapter, we explore this relationship between meaning and counterfactual thinking.

Functional purposes of counterfactual thinking The observation that counterfactual thinking serves a functional purpose is not a new one. The importance of counterfactual thinking for healthy mental and social functioning is underscored by recent observations that in those suffering from brain disorders, such as Parkinson’s disease and schizophrenia, counterfactual thinking is impaired to a more extreme extent than other cognitive skills (Hooker et al. 2000; McNamara et al. 2003). Several theorists have recognized that the variety of counterfactuals thoughts can be classified according to their particular functions (Johnson and Sherman 1990; Markman et al. 1993; Roese 1994). One such classification deals with the direction that counterfactual comparisons take. For example, upward counterfactuals (contemplation of better possible worlds) serve a preparatory function that boosts subsequent performance, whereas downward counterfac-

Finding meaning from mutability 111 tuals (contemplation of worse possible worlds) serve an emotional function that helps people feel better. Thus, upward counterfactuals generally serve a performance-related function, but at a cost to affect, while downward counterfactuals serve an affect-related function, but at a cost to performance (Markman and McMullen 2003; see also Chapter 5, Markman and McMullen, for a review of the functional basis of upward and downward counterfactuals). Counterfactual thoughts can also afford a sense of control, thereby satisfying the fundamental motivation to feel that the world is a predictable place (Markman et al. 1995). Thinking counterfactually not only facilitates learning from specific experience, but it can also activate a mind set that boosts problem solving, produces insights, and even creates group synergies (Galinsky and Kray 2004; Galinsky and Moskowitz 2000; Kray and Galinsky 2003; Liljenquist et al. 2004). When we catch ourselves daydreaming about what might have been, another functional aspect of counterfactual thinking becomes apparent – it provides entertainment. The film industry has capitalized on people’s pleasure in undoing the puzzles of life, scattering the pieces, and seeing what else might have been constructed with them. Audiences are fascinated by movies like Sliding Doors (1998), where one version of Gwyneth Paltrow’s life is juxtaposed against the life she might have lived if she had only arrived at a subway moments earlier and slipped through the sliding doors of a departing train. As the audience witnesses the drastically different chain of events that follows from what seemed to be an inconsequential occurrence, they can’t help but ponder the alternative lives they themselves might have led as a consequence of such minor alterations in the course of events – who else might they have met, what accidents might they have avoided, etc.? Equally intriguing, the three different endings to the movie Clue (1985) attracted repeat audiences who wanted to see how each alternative ending unfolded. Again, the draw was the inherent entertainment of being able to violate the laws that govern our reality by traveling multiple paths at a fork in a road. Just as captivating are literature and movies that present characters, from the Scrooge in A Christmas Carol (1843) to John Anderton in Minority Report (2002), who can peer into the future, and after doing so, realize that knowing the future means one can alter that future. No longer inevitable, the future becomes only probable and open to modification. Although the world in which counterfactuals traffic can provide entertainment, counterfactual thinking can also be dysfunctional (Sherman and McConnell 1995). For example, mutability has been implicated in encouraging dysfunctional behaviors such as gambling (Gilovich 1983). In addition, counterfactuals can spawn unwarranted blame and misplaced sympathies. Because later events in a temporal sequence are more mutable (Miller and Gunasegaram 1990; but see Mandel 2003c for an alternative perspective), the person who engages in the last action will often be subject to venomous and unwarranted blame (e.g., Bill Buckner’s ignominious firstbase blunder in the 1986 World Series in which he uncharacteristically

112

Galinsky et al.

allowed a baseball to pass through his legs, enabling the opposing team to score the winning run). The notion of unwarranted blame and guilt reflects what Miller and Turnbull (1990) termed the “counterfactual fallacy,” a confusion of what might have been with what ought to have been. Engaging in mutable actions that lead to negative but unforeseen events can invite spiteful accusations, but counterfactual thinking and mutability can also lead to misplaced sympathies in which “innocent” bystanders who are not the direct target of an attack are afforded greater sympathy than the intended (and also innocent) targets (Miller and Turnbull 1990). Clearly, counterfactual thinking yields diverse consequences; however, this chapter focuses on a function served by all types of counterfactuals, upward and downward, abnormal and close, reflective and automatic, nostalgic and visceral (Kahneman 1995) – that of providing meaning. By extrapolating from mutability, individuals perceive a structure to their lives and the world in which they live. To be sure, finding meaning from mutability is in a sense ironic; the recognition of multiple possibilities might make the world seem capricious, even random. Yet because our brains work overtime to impose meaning in the face of surprising circumstances, consideration of “what might have been” can produce beliefs that render one’s life and experiences all the more remarkable, and hence, all the more meaningful. Counterfactual conjectures (e.g., “What if Hitler had never risen to power?”) not only underscore the grave significance of historical events (Byrne 2002; Tetlock and Belkin 1996b), but even for the more mundane aspects of life, counterfactuals can embed deeper levels of meaning into our lives as we ponder our good luck, regretted mistakes, and even the tragedies of others. From historical landmarks to the banalities of everyday life, counterfactual thinking plays a critical role in how we experience the world. Although counterfactuals impart greater meaning and understanding to our lives, such thoughts can also contribute to the construction of too much meaning. For example, dwelling on what might have been can result in inaccurate superstitions, exaggerated suspicions, and beliefs in outrageous conspiracy theories. But whether counterfactual thinking inspires some meaning or imposes too much meaning, it is nevertheless clear that such thought processes represent a fundamental ingredient in the psychological brew we concoct to understand our lives and the world around us.

What is meaning? Existential philosophers and psychotherapists have long suggested that humans possess a thirst for meaning, a need to understand and appreciate how the different pieces of the world coalesce into a coherent and recognizable puzzle. We yearn for meaning partially because we are loath to believe that our lives are merely a junction where coincidence and happenstance collide; we instead prefer to see our lives as a stage on which a meaningful agenda can be played out, connecting us to a larger cast of actors. Perhaps

Finding meaning from mutability 113 this quest is so appealing because meaning can highlight the importance of our lives even when we are shadowed by discontent. For example, having children appears to reduce many aspects of marital and life satisfaction (Hurley and Palonen 1967; Kurdek 1999), but at the same time imbues the lives of parents with a profound sense of meaning (White 1987; Shek 1994) and a sense of connectedness (Baumeister and Leary 1995). If finding meaning requires connecting to the larger world, putting oneself into the fabric of that universe, it also requires the somewhat contradictory notion of discovering what makes us a unique part of that world. Searching for and finding meaning is one way that humans deal with the unique burden imposed by self-consciousness and awareness of our own mortality (Simon et al. 1998); by connecting to something larger and more permanent than individual life, a sense of symbolic immortality can be achieved. These competing elements of being both connected to the world and being a unique force within it are apparent in the desire of older generations to engage in work that will symbolize their long-term legacy (McAdams et al. 1997). This desire for generativity (connecting oneself to future generations) provides individuals with a link to humanity while simultaneously highlighting their unique contribution to the world. The drive and desire for meaning has been primarily investigated in the coping literature, which examines how people deal with loss and tragedy (Davis and Nolen-Hoeksema 2001; Davis et al. 1998; Frankl 1984; JanoffBulman 1992). The literature on coping through finding meaning has demonstrated dual purposes served by meaning (Davis et al. 1998; JanoffBulman and Frantz 1997). One constructive element of detecting meaning is that it allows the grief-stricken individual to make sense of events, providing comprehensibility (Janoff-Bulman 1992; McIntosh et al. 1993; Parkes and Weiss 1983; Tait and Silver 1989). This sense-making aspect of meaning allows people to understand how the world works and what its governing principles are, thereby making the world a predictable, ordered, and potentially controllable place. In contrast, meaning also involves the element of supplying and imbuing one’s life and the world with magnified significance (Davis et al. 1998; JanoffBulman and Frantz 1997). In the coping literature, this increased significance is conceptualized in terms of discovering positive implications, finding hidden benefits, emerging from tragedy with a wiser self, a self more in tune with what is important in the world (Lehman et al. 1993; Miles and Crandall 1983; Taylor 1983, 1989; Taylor et al. 2000; Tedeschi and Calhoun 1996). The benefits derived from finding meaning in tragedy are not just of a psychological or emotional nature, but also extend to other realms. For example, a study by Taylor et al. (2000) demonstrated that the progression of HIV among infected male patients depended on their ability to find meaning in the experience; in this way, discovering the meaning or significance of one’s suffering proved to be protective of even physical health. We claim that meaning as significance also carries with it another

114

Galinsky et al.

component, one that is independent of the specific benefit derived for the present and future self – namely, that one’s life itself is significant, a truly original composition of people, choices, and events, and therefore important. Discerning personal growth and significance in a tragedy often depends on mutability and counterfactual conjectures. A victim of misfortune might think, “Had it not been for the tragedy, I wouldn’t have discovered this aspect of myself, changed my lifestyle, focused on this interest, nurtured this talent – I wouldn’t be the person I am today.” Here, mutability provides the cognitive raw materials with which the sense-making process can construct alternative worlds juxtaposed against reality. The fact that alternative paths existed, but that one path ultimately prevailed, makes the particular features of one’s life all the more extraordinary, all the more significant, and thus all the more meaningful.

Mutability and mental simulation We claim that meaning can come from mutability. We use the term mutability to refer to the mental ease with which aspects of reality are imaginatively altered. Although it may seem that the imagination would be able to freely roam within any constructed fantasy, research indicates that the imagination is actually quite disciplined – it follows a set of predictable rules and imposes few changes to reality (Byrne 2002; Kahneman and Miller 1986; Tetlock and Belkin 1996b). The paths that counterfactual thoughts typically tread are those that are created in response to deviations to routine. In other words, when people imagine alternatives to their own actions, they typically change their unusual actions into their more normal form, rather than vice versa. Kahneman and Tversky (1982b) used a skiing metaphor to describe the laws governing mutability: just as it is easier to ski downhill than uphill, it is easier to mutate abnormal (routine violating) events back to normality than to go from normality to exceptionality. For example, imagine getting into a car accident on your way to work. Most people will seize upon a pre-accident action that was unusual for them – the route taken to work, say, or the time of departure – and convert it into its more normal form (e.g., “I should have taken my usual route to work,” or “If only I had left the house on time.”). In Gestalt terms, abnormal actions are salient because they stand out against the backdrop of a base-rate pattern. Any action that is salient (i.e., attracts attention), and thus becomes more accessible from memory, should be more subject to counterfactual conjecture, and hence will be more consequential in subsequent judgments (Miller and Taylor 1995). Counterfactual thoughts are also more likely to be elicited by situations in which an outcome “almost” didn’t happen, such as someone missing a flight by a mere five minutes (Kahneman and Miller 1986; Kahneman and Tversky 1982b; Kahneman and Varey 1990). In this case, the alternative (making one’s flight) is easier to imagine than if the person had missed the flight by

Finding meaning from mutability 115 two hours. These close counterfactuals, Kahneman and Varey pointed out, are often expressed as facts about the world and not viewed as having been merely manufactured in the mind (i.e., people believe that if they had left five minutes earlier they truly would not have missed the flight). These patterns of mutability are important in understanding when and how a sense of meaning is achieved. Mutability helps one find order in what otherwise might seem arbitrary, and therefore assists in the sense-making function of meaning. But mutability in and of itself can make an event seem more remarkable and thereby also magnifies the significance aspect of meaning. The fact that the search for meaning and mutability are both more common after negative events suggests their intimate relationship. Indeed, one of the first responses to any tragedy is to focus on how it might have been avoided and an important purpose of these counterfactual thoughts, of these if-only musings, is to provide scripts for how to avoid that fate in the future (Roese 1994, 1997; Roese and Olson 1997). Although counterfactual thoughts and the search for meaning are both more common after negative events, they are still utilized during and after positive events. In addition, important events, ones that have been ascribed meaning and consequence, should also be more likely to be mutated. Thus, mutability enhances meaning, and meaningful events are more likely to be mutated, suggesting a positive, reciprocal relationship between meaning and mutability. In the following sections, we detail a number of ways that mutability can facilitate the search for meaning.

Fate and destiny “There are ever so many ways that a world might be; and one of these many ways is the way that this world is,” so said philosopher David Lewis in his book on the logical basis of counterfactual inferences (1973: 2). The simple recognition of ever so many ways that the past might have been can dramatize to the individual just how extraordinary it is that things actually happened the way they did – that given the numerous possibilities, the chances of choosing this particular course were so very low. The fact that such daunting odds were overcome, then, can suggest that it must have been meant to be! In this way, the exceptionality of such events can engender a belief in a fate-like force, whether it is deity or destiny. Exceptionality: finding meaning in defying the odds Indeed, exceptionality is intimately connected with mutability. Exceptions to the routines of daily life are not only more mutable, but they can also create seemingly exceptional lives – ones that now seem more meaningful. Deviations from the routines of daily life may or may not be deliberately chosen, yet they often engender a sense that something meaningful and

116

Galinsky et al.

significant has resulted. This is especially likely to occur for positive events in which a tragedy was narrowly avoided (see Chapter 8, Teigen). For example, the pastor of the First United Methodist Church next to the Federal Building that was bombed in Oklahoma City survived the blast because he was not where he was normally supposed to be (Bernstein 1995). Pastor Harris usually recorded a weekly radio address at the time of bombing in the main sanctuary, but on this particular Wednesday, for the first time in four years, Harris stayed in his office because the sound engineer missed the appointment. Believing that the simultaneous occurrence of these two rare events (canceling the radio address and the bombing) was more than coincidental, the pastor and his congregation concluded that God had intervened, that “the hand of God” had steered the sound engineer away from the Church. Embracing this explanation behind such an exceptional course of events, the pastor’s life became more meaningful. Counterfactuals are also a useful sense-making device for understanding events that don’t fit easily into existing schemas. Individuals often times have trouble making sense of events in which the magnitude of the cause does not correspond to the magnitude of the effect or consequence (Nisbett and Ross 1980). Therefore, exceptional and meaningful outcomes ought to have exceptional and meaningful causes. People are sometimes intrigued by individuals who have had near-death experiences not simply because they are so exceptional, but because they now ascribe more significance to these people’s lives – the fact that they came so close to death makes the fact that they survived all the more impressive and endows their lives with a higher level of meaning. The tragedy of 11 September 2001 brought into sharp focus how exceptional events are more mutable and therefore more meaningful. Cantor Fitzgerald, a financial firm located at the top of the World Trade Center, suffered horrific losses, with nearly 80 percent of its work force killed that morning. The company’s CEO, Howard Lutnick, survived because he happened to be at his son’s first day of kindergarten at the time of the disaster rather than in his office where he would usually be at that time of morning. A seemingly ordinary act in the life of a parent was imbued with life-altering meaning: had he not taken his child to kindergarten, he surely would have died along with his co-workers. Lutnick later went to elaborate lengths to provide financial compensation to the families of the lost workers of Cantor Fitzgerald, and it is entirely possible that the vivid counterfactual born of his own exception-to-routine actions energized his efforts (Barbash 2003). Conversely, a number of individuals at the WTC happened to be visiting for once-a-year meetings and the like. Because they weren’t supposed to be there, their deaths seemed more exceptional, more tragic, and therefore more meaningful (see Miller and Turnbull 1990 for a discussion of counterfactual thinking and the notion of “innocent” victims). Whether lives were lost or saved due to the pairing of routine-violating behaviors and this unexpected attack, in both cases one may sense a mysterious force conjoining this pairing, rendering fate as a compelling explanation.

Finding meaning from mutability 117 The notion that counterfactuals can lead people to view their place in life as fated may help explain a puzzling conundrum, which is that counterfactuals ironically exacerbate rather than diminish the hindsight bias (Roese and Olson 1996). The hindsight bias, or the “I knew it all along” effect, refers to the pervasive tendency to see outcomes as more probable in hindsight than in foresight. As Fischhoff (1982: 341) noted, individuals “tend not only to view what has happened as having been inevitable but also to view it as having appeared ‘relatively inevitable’ before it happened.” Fischhoff (1975) referred to this tendency as the “creeping determinism” of the historical perspective in which contingencies and alternative outcomes vanish. Initially, the hindsight bias and counterfactual thinking would seem to be inconsistent with each other. If counterfactuals represent awareness of alternative possibilities, shouldn’t they actually reduce the hindsight bias? If we consider that counterfactuals also provide causal explanations, and causal explanations increase a subjective state of certainty (Anderson et al. 1980; Anderson and Sechler 1986; Sherman et al. 1981), then counterfactual thinking and the hindsight bias become compatible with each other (Roese and Olson 1996). Interestingly, in addition to the notion that counterfactuals can make the past seem predictable and understandable, the role of surprise can push people toward ever more elaborate explanations. Indeed, the more surprising an outcome, the more it inspires sense-making cognitions, such as increased attention, heightened attributional processing, and more effortful counterfactual thinking (Hastie 1984; Sanna and Turley 1996). To the extent that these sense-making activities succeed, the feeling that the past had to be the way it was should also increase (Schkade and Kilbourne 1991; Wasserman et al. 1991). Thus, given a truly surprising event, people may be more prone to conclude that it must have been fated. This sense of inevitability or fate facilitates the psychological conversion of what might have been to what was meant to be; what were merely imaginative musings give way to a new sense of determinism. Counterfactual thinking may especially encourage fate-based explanations of events when the number of potential alternatives renders a single outcome improbable on its own statistical merit. Consider, for example, picking the winning lottery ticket. (Of note, Landman and Petty 2000 argued that lottery advertisers are particularly skilled at utilizing counterfactual thinking.) As lottery winners consider the endless numbers that they could have picked but didn’t, the fact that they did pick the winning numbers – beating the million-to-one odds – makes their win seem all the more fated. They might reason, “How could I have overcome odds like that if it wasn’t meant to be?” It is this perception of defying the odds that can impart meaning to life. In this way, the countless alternatives unleashed by a mutable reality can ultimately give rise to a belief in fate. Fate-based attributions apply not only to the choices that we or others make, but may be utilized even when observing non-agentic entities. Consider the mystifying process of genetic combination – the fact that our genetic make-ups emerged in a particular way

118

Galinsky et al.

among the countless genetic combinations that were possible renders the actual combination all the more meaningful. Again, it is the recognition of such improbabilities that makes reality so compelling. In the same vein, close counterfactuals also play a role in perceptions of luck and fate because they involve the same notion of violating probabilities. Almost encountering a huge and devastating loss often increases perceptions of luck because people believe that only fate could have prevented the enactment of this alternative reality (Wohl and Enzle 2003). Understanding the unexpected Counterfactuals highlight the alternatives that exist at a single point in time (e.g., the various lottery numbers or infinite genetic combinations), but they can also help us explore temporally sequenced series of decisions, each contingent on the other. As one drifts from reality into the subjunctive realm of the imagination, thoughts of “what might have been” can rescript an entire sequence of decisions that have directed the course of our lives. We often engage in counterfactual thoughts as a means of understanding how we came to this point – this juncture in life – but when we manipulate the precursors of our current situation, the resulting possibilities can be overwhelming as we contrast our reality to the infinite episodes that might have otherwise defined our lives. In thinking back about the more significant relationships in our lives, such as how we met our spouse, we might muse, “If I hadn’t graduated from college a semester early, I never would’ve moved back home for the spring and taken the job at the coffee shop. And if I hadn’t agreed to cover the weekend shifts, I never would’ve become friends with my supervisor, Nancy. And if she hadn’t introduced me to her nephew, I never would’ve met John . . .” As the sequence of choices and events spools through our minds, each decision point dominos to the next, one choice predicated upon another, and we recognize the path-dependent nature of the choices that culminated in our current reality. The realization that just one break in this sequence of events would have fundamentally altered life as we know it emphasizes just how easily a particular event might not have happened and makes it seem all the more remarkable that it did. Again, such realizations can bolster the notion that it must have been “meant to be,” for surely, amidst the vast sea of alternative pasts, presents, and futures, one never could have anticipated what would be. In this way, while true love is special in and of itself, recognizing the improbability of the many contingencies underlying our most significant relationships can further accentuate their significance. The temporal component of counterfactual thinking may also influence decisions that people make in the present and their expectations of the future. If an individual infers that a particular outcome from the past was destined, it can help shape beliefs and actions in the present. This consideration of what might have been can provide direction about one’s life calling.

Finding meaning from mutability 119 For example, one of the authors knew of an individual who nearly died from meningitis, an infectious disease. The fact that she defied the odds led her to impart meaning to the illness to such a degree that she subsequently perceived a calling to attend medical school and became an infectious disease doctor herself. This perception of fate and the concomitant perception that life forces beyond her control were at play likely contributed to her perseverance and ultimate attainment of a medical degree. We may be particularly prone to cling to a fate-based explanation of events when we could not have foreseen the weighty consequences of a particular decision, especially decisions that seemed unimportant or inconsequential at the time they were made. The trivial becomes consequential as we realize that we might not have met our significant other if we hadn’t carried an extra shift at work, taken the bus when our car was in the shop, or held the elevator for our new neighbor. The notion of trivial behaviors resulting in profound outcomes resembles “chaos theory,” which contends that complex systems interact in unpredictable ways, thereby making precise calculations of future states virtually impossible. This “sensitive dependence” (Barton 1994; Mandel 1995) means that minor differences in the initial conditions of two entities will often lead to dramatically divergent outcomes. A compelling symbol for chaos and sensitive dependence is the butterfly effect: the notion that the beating of a butterfly’s wings in a faraway land can set off a tornado here at home (Lorenz 1993). Our own small decisions and choices, that impact not only our own lives but also the lives of others, are not unlike the flapping wings of a butterfly. The very fact that the world is unpredictable and the consequences of these decisions could not have been foreseen makes them all the more remarkable and easier to attribute to fate. The connection between mutability and fate may naturally occur for positive events: finding one’s soul mate, landing the perfect job, choosing an inspiring career. However, counterfactuals, which more commonly follow negative events, can also lead to perceptions of fate even for despairing outcomes, perceptions that can be soothing and comforting. In this way, mutability can increase both how tragic an event was and the meaningfulness of that event. A poignant example helps to illustrate the appeal of fate-based explanations when counterfactual thoughts yield a host of alternatives that might have circumvented a seemingly random tragedy. The friend of one of the authors lost her brother when he fell out of a loose window of a skyscraper, dying on impact, while touring through Thailand. Although “if only” thoughts were the inevitable consequence of such a painful and shocking tragedy, such counterfactuals did little to help in terms of understanding why his death occurred, for nothing could have predicted such a random event. However, constructing the millions of ways that this tragedy might not have happened – how truly unlikely it was that it did, in fact, occur – left his sister only to conclude that his death was meant to be, that her brother’s time had come. It was the sheer number of counterfactuals

120

Galinsky et al.

generated (that he could’ve delayed the trip, could’ve reordered the destinations, could’ve peered through a different window, etc.) that convinced her that his untimely death must have been woven into the fabric of destiny. Certainly, finding meaning in both fortune and tragedy carries both physical and emotional benefits, but these benefits are not without liabilities. Although a belief in fate can be reassuring, it can sometimes be demotivating or disenfranchising. Likewise, counterfactual thinking can result in denials of responsibility based on the belief that the course of life is beyond an individual’s control (Markman and Tetlock 2000b). Taking these beliefs forward, it can lead to rationalizations and justifications for unseemly behaviors in support of the notion that one is merely an agent carrying out destiny’s work. And in some cases, it may set people up for disillusionment. Consider our earlier discussion of how people come to believe they were meant to meet their spouse after considering the many ways they almost didn’t meet their spouse. The idea of romantic love is a modern phenomenon because choice or the notion of having numerous options has entered into the business of marriage only in recent history. Only when we perceive endless possibilities do we then focus on and single out one person as our fated soul mate. As we discussed earlier, beating the odds can give reassurance that we made the right decision, but when things go awry, we may feel betrayed by our own intuition. The belief in an idealized soul mate can anchor expectations so high that divorce is often the consequence of violating these unrealistic expectations. It can be devastating when what was once meaningful and meant to be comes to a disappointing demise.

Misled by mutability: finding too much meaning As perceptions of personal autonomy surrender to fate, life outcomes, both pleasant and tragic, can be infused with greater meaning at the realization that even menial decisions are overseen by fate – an enigmatic force that eludes comprehension but imbues our reality with a certain hue of determinism. The path from mutability to meaning is often a comforting one, endowing ordinary lives with extraordinary significance. However, mutability can lead to perceptions of too much meaning, implanting suspicions of hidden agendas and covert schemes lurking in the shadows. Seeing meaning everywhere can be exhausting. Sometimes it is easier and even comforting when a cigar is simply a cigar, and not a metaphor or portent of something hidden and dark. In fact, counterfactuals can be punishing, leading to obsessive undoing and disabling ruminations (Davis and Lehman 1995). In some cases, we may construct elaborate conspiracy theories; other times, as we undo events and simulate alternative realities, we may impute faulty causal links, thereby sowing beliefs in fallacious superstitions.

Finding meaning from mutability 121 Suspicion and conspiracy Although counterfactuals are useful in the processes of sense making and in understanding what caused an event to occur, mutable events can lead people to read too much meaning into events and see spurious connections and hidden driving forces. In some of the earlier examples regarding the attack on the World Trade Center, we highlighted how the coincidence of two exceptional events (e.g., the attack and taking a child to his first day of kindergarten) can enhance a sense of meaning and destiny. But the coinciding of two rare events can also lead to suspicion and even conspiracy theories. Generally, when there are few ways that a low-probability event could have occurred by chance, people are apt to suspect that the event did not occur randomly, but was the direct result of intentional, and sometimes nefarious, behavior. Miller et al. (1989) posited that when it is difficult to imagine multiple, plausible pathways to a rare outcome, individuals react with suspicion. In other words, when the natural occurrence of a highly improbable event seems inconceivable, we resort to alternative explanations that often implicate actors with dubious intentions. To explain Miller et al.’s approach to testing this idea, imagine a dartboard dappled with one red dot and nineteen blue dots. You then witness a blindfolded player (who will be rewarded for hitting the red dot) take a shot at random, unbelievably hitting the one red dot on the board! Is it mere luck? Or do you suspect that he peeked? If you’re suspicious, you’re not alone. In a similar scenario, Miller et al.’s participants were far more likely to suspect cheating behavior when the player nailed the one and only desired object (in their experiment, the lone object was a coveted chocolate chip cookie amidst a bin of nineteen oatmeal counterparts) than if they successfully targeted the desired object on a dartboard that contained ten red dots out of a total of 200. Although the chances of hitting one of those ten red dots reduces to the same probability, because there are ten ways to win, or at least ten ways to simulate winning, a positive outcome in the latter condition is deemed more probable and therefore less suspicious simply because it is easier to imagine. At the other end of the spectrum, if it is too easy to imagine alternative explanations, we may also become suspicious simply because of the abundance of available plausible alternatives. This carries implications for the inferences we draw about people’s characters and the alluring logic behind the necessity of a police state. For example, Miller et al. (2004) examined how our impressions of others are a reflection of not only what those others do, but also what we can imagine them doing. In their research, they constructed a video scenario in which participants saw ten-year-old boys taking a math test. When both the temptation to cheat and the possibility of being caught were high, the non-cheating actor was judged less trustworthy than an actor who resisted high temptation under low likelihood of detection. Miller et al. suggested that observers witnessing the actor in the

122

Galinsky et al.

high-temptation condition generated counterfactual thoughts that the actor would have cheated had the threat of detection not been high. Therefore, observers’ inferences were based on what they had imagined rather than the behavior they had actually witnessed. Through counterfactual conjectures about what might have happened, it is easy to understand the allure of surveillance systems and the erosion of personal liberties when we suspect the worst in others. Operating on similar principles, mutability can also give rise to conspiracy theories, making it easy to imagine alternatives to the way things allegedly “happened”. For example, the concurrence of two rare events can lead to mental simulations that might cause one to conclude that something – or someone – masterminded the uncanny coincidence. The notion of conspiracy taps into the representative bias in that we expect big effects to have been produced by big causes. Sim and Morris (1998) noted that the assassination of John F. Kennedy was such fodder for conspiracy theories because it seemed improbable that a man as insignificant as Lee Harvey Oswald with such a simple plan could have produced an event with such farreaching consequences and significance. Thus, counterfactual thoughts allow us to construct more elaborate meanings to satisfy our expectation of congruence between the magnitude of causes and effects (Einhorn and Hogarth 1986; Nisbett and Ross 1980; Sim and Morris 1998). Seemingly incongruous events are rescripted in a more sensible way by invoking a conspiracy, which implies a far grander causal force of a particular event than meets the eye. Superstition Superstition is another side effect of mutability. Superstition is a belief resulting from a false conception of causation. Sometimes, when we deconstruct the past, we detect causal links in the sequence of events, but many times these causal attributions may contribute to an erroneous belief structure. If people believe that an event, negative or positive, did not occur by chance but that something or some behavior caused that event to occur, they may construct superstitions in an attempt to replicate the positive events or avoid the negative events. Superstitions, both prohibitions and compulsions, run a wide gamut – from what many consider to be foolish old wives’ tales to revered rules of thumb. We highlight just a few to demonstrate how, in our quest to understand why things happen, counterfactual thoughts can provide unfounded and spurious meaning. Many people subscribe to the superstition that you shouldn’t switch lines at supermarket checkouts, because your new line will slow down and your old line will turn out to have proceeded more quickly. It seems implausible that the mere act of switching lines affects the speed with which cashiers and customers interact; thus, we turn to counterfactual thinking to understand the superstition that warns against line switching. We can assume

Finding meaning from mutability 123 that enduring a painfully slow queue occurs just as frequently following line switching as line staying. But line switching is more likely to activate a selfrecriminating counterfactual (“If only I had stayed put . . .”) than line staying because actions are more mutable, at least over the short term, than inaction, and therefore any direct actions tend to be the focus of sharp, vivid regret (Miller and Taylor 1995). In other words, counterfactuals that undo action rather than add new action are more vivid, more frequent, and more mentally influential. Consequently, this counterfactual, being more active in mind, becomes more accessible in memory and ultimately produces more regret (Gilovich et al. 1995), leading people to subsequently conclude that line switching is a dubious strategy for dealing with long lines, when in fact it has no impact on duration of waiting in the long run. The sting of the increased regret following this act of commission is what produces the foundation for superstitious admonishments. There is some empirical evidence for the counterfactual-induced development of superstitious beliefs. Galinsky et al. (2002) found that negotiators whose first offers were immediately accepted expressed counterfactual recriminations despite the fact that they received objectively superior outcomes (they didn’t have to make any concessions). These counterfactual recriminations turned into the basis for superstition as the amount of counterfactual expression predicted the avoidance of making subsequent first offers. Like many superstitions, this subsequent avoidance is maladaptive because not making a first offer precludes negotiators from gaining an anchoring advantage and maximizing their negotiated outcomes (Galinsky and Mussweiler 2001). By granting too much meaning to our trivial behaviors, we collectively create prohibitions against tempting fate and condemn ourselves to longer lines and sub-par negotiated outcomes. Superstitions also abound in the scholastic sphere. For example, a rule of thumb that test takers often follow is to stick to their first answer when completing multiple-choice exams. Even Kaplan (1999: 3–7), an American company that specializes in preparing students for standardized tests, advises its clients: “Exercise great caution if you decide to change an answer. Experience indicates that many students who change answers change to the wrong answer.” In surveys of college students, about three out of four students say they agree with this advice. But this advice turns out be flat wrong, as confirmed by dozens of studies spanning over half a century (Benjamin et al. 1984; Lehman 1928; Reile and Briggs 1952; Vispoel 1998). Kruger et al. (2005) suggest that counterfactual thinking contributes to this first-instinct fallacy. Like switching lines in a supermarket, the mere act of switching, of doing something tangible and obvious, makes any subsequent problems more likely to be regretted. In one of their studies, Kruger et al. found that several weeks after taking an SAT exam, students misrecalled the consequences of switching answers versus sticking to their original answers. Students mistakenly remembered that switching answers was worse for their exam performance than it actually was, and that sticking to

124

Galinsky et al.

the original answer was better for their exam performance than it actually was. Superstitions not only prescribe prohibitions but also sometimes require one to engage in certain behaviors to avoid a negative state. By knocking on wood, one supposedly guards against the possibility that the mere mention of words can summon tragedy. Even obsessive-compulsive behaviors can be targeted against the worrisome and negative possibilities that emerge from mutability. From obsessively washing one’s hands to disinfecting door knobs, seeing one’s behavior (or lack of a behavior) as having disproportionate influence on the outcomes of life can be a disturbing and debilitating downside of counterfactual thinking. Prejudice and shortsighted legislation Taken further, counterfactual-induced superstitions can even fuel our prejudices and inspire acts of discrimination. By uncovering alternative “if–then” sequences, counterfactuals help specify the necessary conditions to avoid a given negative outcome in the future. If we incorrectly attribute a particular outcome to the wrong conditions, these specified conditions lay ground rules from which the seeds of discrimination can take root. Focusing on easily imagined causes of an event can lead to legislation or preventative actions that may or may not be appropriate. For example, even after the Oklahoma City bombing was linked to domestic terrorism and the TWA flight was linked to mechanical failure, legislators still proceeded with antiimmigration legislation. The fact that it was so easy to simulate foreign terrorism engendered beliefs that it could have been foreign terrorists and maybe would be the next time around. Similarly, Miller et al. (1990) linked the very processes that lead to suspicion as laying the groundwork for the development and maintenance of stereotypes; when a member of a numerical minority group engages in reprehensible behavior, people may be suspicious that such co-occurrence is rare and instead conclude that most members of group engage in such unseemliness. Mutability may not only lead to legislation with prejudicial intent and discriminatory consequences but may also result, through the counterfactual fallacy, in initiatives designed to undo the particular counterfactuals. These specific mutability-induced remedies can be mere band-aids and shortsighted fixes that don’t solve the larger, less mutable problem. For example, the efforts devoted to improving airport security (e.g., taking off shoes through security checkpoints) partially occurred because the most mutable component of the 9/11 attacks was the preventative actions that the airlines could have taken to deny the terrorists boarding. However, such efforts didn’t deal with the deeper problem and roots of anti-Americanism that led to the attacks in the first place. Similarly rooted in counterfactual alterations is the Brady Bill that resulted from John Hinkley’s assassination attempt on Ronald Reagan and the subsequent wounding of Jim Brady. Hinkley’s

Finding meaning from mutability 125 history of mental illness provoked the feeling that the attempt could have been avoided had he been denied a gun. Though the efficacy of such measures may be debatable, they demonstrate how counterfactual thoughts can powerfully instigate legislative imperatives.

Conclusion Counterfactual thinking and searching for meaning are not only ubiquitous parts of our mental experience, but are intimately connected due to the relationship between meaning and mutability. Counterfactual constructions can highlight the singular significance of our lives as we recognize the exceptionality of particular events. Whether it is defying the odds or peering into a parallel world, counterfactual thinking can lead individuals to acknowledge the unfathomable forces of fate. As people strive to make sense of the world around them, the relationship between meaning and mutability can also become perverted; counterfactual thoughts can create a montage of fallacies, spawning paranoid suspicions, crippling superstitions, and prejudicial ascriptions. Whether counterfactual thinking imbues our lives with significance or engenders erroneous delusions, the thirst for meaning entices many to explore the rabbit hole of possibilities.

Note Preparation of this chapter was supported by a National Science Foundation grant to the first and third authors (SES-0136931 and SES-0233294) and by a National Science Foundation Graduate Fellowship to the second author.

Part III

Counterfactual thinking and emotion

8

When a small difference makes a big difference Counterfactual thinking and luck Karl Halvor Teigen

Counterfactuals affect the way we feel about events. Failures become worse when we consider how they could have been avoided, and successes become more special and memorable when contrasted with the possibility of failure. Ben Ze’ev (1996, 2000) claims that all emotions are basically of a comparative nature. They arise when we compare our current situation to our prior state, to our goals and expectations, to other people’s conditions, and to purely imagined, counterfactual outcomes. Evaluations based on downward comparisons make us usually feel better, and upward comparisons make us usually feel worse, by a mechanism of affective contrast (Markman and McMullen 2003). We may, however, distinguish between counterfactual comparisons that primarily affect the intensity of an experience, and counterfactual comparisons that create a qualitative difference. In Kahneman and Tversky’s (1982b) classic case of Mr Crane and Mr Tees, who both miss their flights after being stuck in a traffic jam, Mr Tees is supposed to be more upset than Mr Crane because he arrived only five minutes too late, and can more easily imagine a counterfactual success. Mr Tees’s counterfactual thoughts are in this case supposed to amplify his feelings of disappointment and frustration. In other cases, as with the emotion of regret, counterfactuals constitute an integral part of the emotion itself. The experience of regret hinges critically upon the thought: “If I only had . . .” or “I shouldn’t have . . . .” (Landman 1993). The thesis of the present chapter is that the experience of being lucky or unlucky belongs to the second rather than the first of these categories. Without a counterfactual alternative, there would be no experience of good or bad luck. One does not simply become more lucky by considering the alternative, one becomes lucky by considering the alternative. Indeed, the same outcome can be described as either lucky or unlucky depending on the activated counterfactual. For instance, it is not uncommon that accident victims describe themselves in one moment as extremely unlucky (to have been involved in the accident) and in the next as extremely lucky (to have survived). The counterfactual alternatives are in the first case life without an accident, and in the second case, death. Anat Ben-Tov, survivor of two Tel Aviv bus bomb attacks, expressed this succinctly in an interview, given from

130

Karl Halvor Teigen

her hospital bed: “I have no luck, or I have all the luck – I am not sure which” (“Perspectives ‘95” 1995: 42). There is, however, little in the standard definitions of luck that explicitly suggests the importance of counterfactuals. Luck is often described in causal terms. To attain an outcome by luck means that it is produced by accident, or chance; that is, by unpredictable factors beyond the control of the individual. In attribution theory, luck is defined as an external, uncontrollable, and unstable cause (Weiner 1986). Even so, luck is sometimes believed to accompany special individuals, namely those “born under a lucky star,” or in the words of Hector Berlioz: “The luck of talent is not enough, one must also have a talent for luck” (Strauss 1979). Folk wisdom also regards luck as not completely unpredictable; there are “streaks of luck.” Wagenaar and Keren (1988) asked people to assess the relative importance of chance and skill in games like soccer, only to find that luck emerged as a third factor, distinguishable from skill by being more capricious, but also distinct from chance by obeying certain rules. Gamblers are known to use a variety of methods to evoke and make the most out of their luck before it “runs out” (Langer 1975; Proctor 1887). To many people, luck is more than meets the eye; they think of luck as a secret power creating meaningful coincidences and producing winners and losers (Darke and Freedman 1997a, b). In the words of Paulo Coelho’s alchemist: “If I could, I’d write a huge encyclopedia just about the words luck and coincidence. It’s with those words that the universal language is written” (Coelho 1999: 73). But even those who believe in a deeper significance of luck, or hold non-normative views of the controllability and predictability of lucky outcomes, will probably agree that a lucky event is one that at least superficially has the character of happening “by accident.”

Luck as an unexpected benefit As a starting point for analyzing luck, let us consider the definition offered by the philosopher Nicolas Rescher (1995), in one of the few scholarly books on the luck concept. According to Rescher, in characterizing a certain development as lucky for someone, we make two pivotal claims: • •

that as far as the affected person is concerned, the outcome came about “by accident.” that the outcome at issue has a significantly evaluative status in representing a good or bad result, a benefit or loss. If X wins the lottery, it is good luck; if Z is struck by a falling meteorite, that is bad luck. (Rescher 1995: 32)

Rescher further points out that good and bad luck allow for gradations. We can speak of someone as extremely lucky or only a little lucky. This can

Counterfactual thinking and luck

131

refer to the value of the outcome – the winner of a large sum of money will be considered luckier than the winner of a small amount. It can also refer to its predictability. A totally unexpected win is luckier than one that is expected, so outcome probabilities are also involved. These two factors are brought together in Rescher’s suggested formula for luck, where the value of outcome, E, is multiplied by its improbability. Paraphrasing expected utility models, the degree of luck, L, associated with E equals its “unexpected utility”: L(E)  U(E)  (1  p(E)) How well does this analysis capture people’s intuitions about good and bad luck? From our own research, it appears that neither expectancies nor values are unproblematic determinants of luck. Degree of luck does not always correlate with outcome probability, and it does not always reflect the magnitude of gains and losses. 1. Not all events with the same probability are equally lucky. Consider the gambling wheels displayed in Figure 8.1 (from Teigen 1996). • •

Wheel A is divided in three equal sectors, one red, one yellow, and one blue. The winning color is red (dark). Wheel B is divided in eighteen equal sectors, six red, six yellow, and six blue. The winning color is red here, too.

Liv plays wheel A, gets red, and wins. Anne plays wheel B, gets red, and wins. Who will feel more lucky? When this question was posed to a group of eighty-nine students, 85.4 percent said Anne would feel more lucky, 4.5 percent chose Liv (10.1 percent said they would feel equally lucky). In this case, Anne’s chances are as good as Liv’s (p  1/3). According to the ratio-bias phenomenon (DenesRaj and Epstein 1994; Denes-Raj et al. 1995), people may even feel that A

B

Figure 8.1 Fortune wheels presented in a study of luck and closeness to failure. Note that both wheels offer the same chances of winning, but wheel B winners are luckier (source: Teigen (1966: figure 1)).

132

Karl Halvor Teigen

Anne has a better chance to win simply because she is offered a greater number of favorable outcomes (six out of eighteen is somehow more impressive than one out of three). This was confirmed in a follow-up study where a group of students were asked which gamble they would prefer, granted that the objective probabilities were the same. Of fifty participants, thirty-two (64 percent) chose wheel B, whereas fourteen (28 percent) chose wheel A (four had no preference). So Anne cannot be luckier than Liv because of her a priori odds of winning. But if she wins, she will win with a smaller margin, as the red sections on wheel B are narrow and surrounded by sections of other colors. A winner on this wheel will be a near loser, whereas a winner on wheel A may land “safely” inside the wide red sector. This suggests that luck is a matter of margins, rather than probabilities. The lucky winner is, to paraphrase Kahneman and Varey (1990), the winner that almost lost. 2. Not all good luck incidents imply a benefit or a gain. Winning in the lottery obviously implies a gain. But surviving an accident is also considered lucky, even when nothing gained. Consider Rescher’s own dramatic example, introducing the notion of luck. The second atomic bomb to be dropped over Japan in World War II was intended for the arsenal city of Kokura. But the aiming point was obscured by clouds and haze. So the bomber continued to the secondary target, the port city of Nagasaki. “And what was an incredible piece of good luck for the inhabitants of Kokura turned equally bad for those of Nagasaki” (Rescher 1995: 3). In this example, the inhabitants of Kokura are described as beneficiaries of good luck, even if (or precisely because) nothing happened to them. They did not gain anything from the accidental clouds and haze, but they avoided annihilation. Under the circumstances, this made for “an incredible piece of good luck,” comparable (according to Rescher) to the awful bad luck striking the inhabitants of Nagasaki. If we accept this example, we must also accept a form of good luck that does not entail a positive benefit or a gift that is thrown into your lap. Good luck can also mean that a catastrophe is avoided. Perhaps the prototypical example of good luck as “X wins in the lottery” is not so representative after all. Good luck may be equally well illustrated by “X was not struck” (by a falling meteorite, or a bomb, as the case might be).

Luck as closeness to disaster Some years ago, a major Norwegian newspaper carried a story about “The luckiest man in the world” (Kristiansen 1993). This champion of luck was a German motor cyclist touring the Dominican Republic, who collided with a truck, was thrown off his bike, hit a tree, and was impaled on a branch. A picture showed him being carried to the hospital, sitting on a stretcher (he could not lie down), with the branch sticking into his chest and out of his back, but otherwise he appeared calm and collected. No vital organs were hurt.

Counterfactual thinking and luck

133

A more systematic perusal of two Norwegian newspapers over a onemonth period revealed that the term “lucky” was frequently used in connection with accidents or near-accidents (Teigen 1988). With the exception of an occasional sports champion and a double lottery winner, the typical lucky person had survived a plane or car crash, had been stabbed or shot, fallen off a cliff or a bridge, been shipwrecked, or surrounded by flames. Some lucky people manage to escape uninjured, whereas others (like Amat Ben-Tov) ended up in hospital. Their luck was often characterized as “incredible,” presumably because death would have been a more “normal” or expected outcome of the situation. Only one instance of “bad luck” was reported in the search period, namely that of a UN soldier who stepped on a land mine. But even this unlucky soldier added that he had been lucky to lose only one leg. The newspaper examples are admittedly extreme, partly because of their sensational character, and perhaps also because of journalists’ reputed inclination toward disasters. To ensure a more “normal” collection of good- and bad-luck incidents, we asked Norwegian and Polish students to describe episodes from their own lives where they had been lucky or unlucky (Teigen 1995). They also rated the events for attractiveness and probability. Badluck events were, unsurprisingly, always unattractive (car crashes, lost objects, missed travel connections), whereas good-luck events were more mixed. Although less spectacular than those in the newspapers, many “everyday” luck episodes were clearly unpleasant, involving traffic accidents, close calls, trains and flights nearly missed, and objects lost and recovered. Here are two illustrative cases: I was driving a car before I got my driver’s license. There was a police checkpoint, and the car in front of me was stopped for control – whereas I was allowed to drive on. I was cooking potatoes. Suddenly I brushed against the pan with my left arm, and the pan fell crashing to the floor, without spilling any of the boiling water on me. Outcome probabilities were consistently low, but the correlation between estimated outcome probability and degree of luck was essentially zero. Thus the study did not support Rescher’s claim that degree of luck is simply a function of outcome attractiveness, on one hand, and low outcome probability, on the other. Rather, it supported the idea that luck is “post-computed” (Kahneman and Miller 1986) as a function of what happened compared to what could have happened, as suggested by the situation. A counterfactual interpretation of luck posits that good luck is dependent upon the difference in attractiveness between the factual (F) and the counterfactual outcome (Cf), as well as by their closeness. In the potato-cooking example, the obtained outcome is not very attractive (dropping the pan), but it is clearly better than the counterfactual alternative (being splashed with

134

Karl Halvor Teigen

boiling water). Moreover, the counterfactual is close at hand; with boiling water all over the place the protagonist could “easily” have been scalded. This interpretation was tested in a subsequent experiment (Teigen 1995: Study 2), where thirty good-luck and thirty bad-luck stories from Study 1 were rewritten as brief vignettes and given to a peer group, who were to write down what “could have happened” in each case, and to rate the stories for degree of luck, attractiveness of F, attractiveness of Cf, and Cf closeness. It is evident from Table 8.1 that good-luck events are not always rated as pleasant, but the alternatives are clearly worse. For bad-luck events the alternatives are clearly better. The Cf outcomes received in both cases very high closeness ratings, meaning that the outcome could have easily been worse (good luck) or better (bad luck). Degree of good luck was in this study highly correlated with Cf closeness, and with the F  Cf difference in attractiveness. The F  Cf difference in attractiveness was also the best predictor of bad luck. As a control, we asked another group to give examples of positive (pleasant, good) or negative (unpleasant, painful) events from their own lives, to be rated by a peer group for attractiveness and counterfactual closeness. The positive events differed from the lucky ones in being clearly more attractive, whereas the negative events and the unlucky events were equally unattractive. Upon request, the participants could produce counterfactual outcomes also for these stories, which were generally negative for the positive events and positive for the negative events. These counterfactuals were, however, not regarded as close. This study indicates that lucky outcomes differ from other positive outcomes by virtue of their counterfactuals. Lucky events do not have to be pleasant, but they must be more attractive than their Cf alternatives. How lucky one is perceived to be depends upon how easily the situation could have taken a worse turn, as well as how much worse it could have been. Table 8.1 Peer ratings (1–9) of outcome attractiveness and counterfactual closeness of thirty “good luck” and thirty “bad luck” stories, correlated with degree of good and bad luck Correlate

Attractiveness of actual event Attractiveness of counterfactual event Attractiveness difference Closeness of counterfactual event

Good luck

Bad luck

M

r

M

r

5.09 2.81 2.28 7.07

0.37* 0.44* 0.71** 0.63**

2.81 5.91 3.01 6.55

0.53** 0.29 0.61** 0.24

Source: Adapted from Teigen (1995: tables 2 and 3). Notes *p  0.05; **p  0.01.

Counterfactual thinking and luck

135

Manipulating closeness If luck signifies closeness to disaster, people should be luckier the closer they are. Closeness can sometimes mean proximity in time and space. Consider this situation, observed in downtown Tromsø (from Teigen et al. 1999): An icicle falls from the roof, hitting the sidewalk next to two young boys waiting for the bus. One of them turns to the other, exclaiming: “There you were lucky!” Now who made this remark: the boy closer to the icicle or the one farther away from it? When this vignette was presented to a group of forty-five students, all agreed that the boy closer to the icicle was the lucky one, so the remark had to be made by the one farther away. The example has the advantage of being authentic, showing how luck is spontaneously used to characterize a close call. It may feel better to be standing at a safe distance, but it is luckier to be close. Closeness can also be conceived in more symbolic terms. Participants in a different study (Teigen 1996) were told about two lottery players who won the same amount of money. John bought only one ticket, whereas Henry bought two tickets and gave one of them away, winning with the one he kept for himself. Who was more lucky, John or Henry? As predicted, a majority (80 percent) perceived Henry as the luckier one. It is easy to imagine Henry giving the winning ticket away, which makes him closer to failure than John, the other winner. Bad luck is also affected by counterfactual closeness. A lottery player who happens to give the winning ticket away to her friend will be regarded as extremely unlucky, even if there is no objective loss involved. The friend, who is in possession of the winning ticket, is very lucky, but more than 90 percent of the participants classified the event as a “bad luck” story. Apparently, the near win (by the player who for a while had held the winning ticket in her hand), is the dominant feature of this story. It follows that one can alleviate bad luck by making the missed opportunity less accessible, as in the following example (Teigen 1996): Petter is going to Africa, and has been advised to take a vaccine against yellow fever, but the vaccination makes him so ill that he is not able to go. While lying in bed, feeling bad, since his vacation plans went down the drain, he receives the news that the whole trip has been cancelled because of an airline strike. This makes Petter feel: • •

Still more unlucky? Yes/No. A little lucky after all? Yes/No.

About 95 percent of our respondents answered “No” to the first question, and nearly 85 percent answered “Yes” to the second. There are apparently

136

Karl Halvor Teigen

situations in which two misfortunes feel better than one. When two obstacles prevent Petter from traveling, he cannot say: “if only . . .” (Moreover, from a social-comparison point of view, his situation has improved since he is no longer the only person grounded.)

How lucky are you? Thinking back on his life and career, Fritz Heider concluded: “I cannot help believing that it has been marked by many strokes of good luck” (1983: 188). This is an interesting statement coming from the founder of attribution theory, who more than anyone else has formed our way of thinking about personal versus situational explanations, including luck. If the present analysis of luck is a correct one, Heider must have had personal incidents in mind that could “easily” have taken a worse turn. True enough, Heider bolsters his description with worse counterfactuals. “I shudder to think of the factors related to this meeting [with his future wife] . . . and how easily any one of them might have happened differently . . . For instance, if I had decided in 1927 to accept Bühler’s offer of an assistantship in Vienna instead of Stern’s for Hamburg, I would almost certainly never have been offered the position in Northampton, and I would probably never have come to America” (Heider 1983: 190). A life with only pleasant memories would be remembered as a good and happy life, but not necessarily a lucky one. Lucky people must be hindsight worriers, with an eye for events that could have gone awry in their lives. Selective recall of life episodes has been found to be effective in inducing positive or negative moods (Westermann et al. 1996) and for inferring personality traits (Schwarz et al. 1991). What kind of episodes will make people think of themselves as lucky? In one experiment Teigen et al. (1999) asked two “factual” groups of respondents to produce examples from their daily life of situations they would judge to be clearly positive (A) or negative (B). Two other “counterfactual” groups were asked to recall situations in which something positive (C) or negative (D) could easily have happened, which did not happen after all. Participants in all groups were then asked to rate on seven-point scales to which degree they perceived themselves as lucky and unlucky persons, the prediction being that the priming intervention would affect the ratings primarily in the counterfactual groups. The results showed that all participants, except those in condition C, considered themselves as more lucky than unlucky. Recall of positive or negative events (groups A and B) did not make a significant difference. But recall of narrow escapes from negative outcomes, or missed positive outcomes (groups C and D), did, leading to a highly significant interaction between story valence (good versus bad end results) and story type (factual versus counterfactual conditions).

Counterfactual thinking and luck

137

Risk and luck If luck implies closeness to disaster, one might turn the question around and ask whether close disasters also imply luck. Situations characterized by closeness to disaster are often described as risky, or dangerous. People who do not take proper precautions against risk are often characterized as negligent, or careless. In both cases, the final outcome will be heavily dependent upon chance; it could end well, but it could also go terribly wrong. There are no statistics telling us, in general, how often dangers end in disasters, or how many risk takers are punished by their carelessness, compared to those who get away with it. We may, however guess that there are more near misses than actual misses (although in the case of carelessness, people may not realize that they are taking chances until something goes wrong). If a majority of hazards ends better than imagined, risk takers should have many opportunities to feel lucky, and have more luck stories to tell than prudent people. This prediction was tested in two parallel studies (Teigen 1998a), in which students were asked to write sketches of situations from their own lives which they had experienced as dangerous or risky (Experiment 1), or in which they had behaved in a careless or incautious manner (Experiment 3). Both instructions yielded thirty-eight stories, which were subsequently judged by a peer group following a similar procedure as in the study about luck in everyday life. Luck and danger The collection of danger episodes ranged from traffic accidents, or near accidents, leaking boats, and parachuting to skiing, climbing, and being assaulted. They were distinctly unpleasant, but were rated both by the actors themselves and by the peer group as more lucky than unlucky. Mean peer ratings are displayed in Table 8.2, showing that the episodes were Table 8.2 Peer ratings of luck, outcome attractiveness, and counterfactual closeness of thirty-eight “dangerous” stories, correlated with dangerousness and good luck Correlate

Good luck Attractiveness of actual event Attractiveness of counterfactual event Closeness of counterfactual event

Mean ratings (1–9) 6.29 2.92 2.07 6.90

Source: Adapted from Teigen (1998a: tables 1 and 2). Notes *p  0.05; **p  0.01.

Correlations with Dangerousness

Good luck

0.69** 0.73** 0.78** 0.76**

– 0.40* 0.81** 0.77**

138

Karl Halvor Teigen

characterized by closeness to a worse counterfactual, explaining their high scores on good luck. The episodes were also rated for degree of dangerousness. It appears that the most dangerous episodes were rated as the luckiest ones (after all, the actors had survived to tell the story). The correlations reported in Table 8.2 show a close parallel between danger and luck, which are both highly correlated with the closeness of counterfactuals as well as with their aversiveness (indicated by low attractiveness scores). Luck and carelessness The situations in which students admitted that they had behaved in a careless manner ranged from poor preparation for exams to illegal activities, reckless driving, and unprotected sex. Despite the negative content, most situations could have ended even worse than they did, so most actors described themselves as more lucky than unlucky. From the peer ratings, displayed in Table 8.3, it is apparent that carelessness is also characterized by close negative outcomes, although less negative than for dangerous situations, making the protagonists correspondingly less lucky. But again we have a set of events that are generally unattractive and lucky at the same time. There is further a tendency for episodes with high scores on carelessness to receive high scores on good luck. Two examples: a paraglide jumper without any training who lost control over his paraglider, and a motorist finding himself driving in the wrong direction on a freeway lane; both survived. Conversely, stories with low scores on carelessness were more unlucky than lucky, as for instance a girl spilling coffee on her white sweater, and another cutting her finger on the new vegetable knife. Not very serious situations, but as there are no imaginable worse outcomes, there is no good luck involved. The correlations in Table 8.3 are generally more moderate than in Table 8.3 Peer ratings of luck, outcome attractiveness, and counterfactual closeness of thirty-eight “careless” behaviors, correlated with carelessness and good luck Correlate

Good luck Attractiveness of actual event Attractiveness of counterfactual event Closeness of counterfactual event

Mean ratings (1–9) 5.51 3.76 3.07 6.78

Source: Adapted from Teigen (1998a: tables 4 and 5). Notes *p  0.05; **p  0.01.

Correlations with Carelessness

Good luck

0.36* 0.36* 0.34* 0.59**

– 0.34* 0.87** 0.47**

Counterfactual thinking and luck

139

the case of dangers, the highest correlation being between good luck and aversiveness (unattractiveness) of counterfactual outcome, confirming the hypothesis that degree of luck is more dependent upon what does not happen than upon what actually takes place.

The narrative structure of luck When are we led to think about worse outcomes? Neutral or positive events do not by themselves activate thoughts about failures. There must be something in the situation that specifically suggests losses and misses. Such suggestions can arise in two distinct ways: 1

Post-computed closeness. In many cases, people discover only in retrospect that they “nearly” failed. While the factual outcome is instantly realized, the counterfactual alternative, indicating the luck involved, might require additional, post-hoc information, as in this observation: At 4:29 p.m. a customer runs out of breath into the local liquor store, his first question being: “When is closing time?” “Fourthirty.” (With emphasis): “Then I was lucky.” For this customer it was not enough to observe that he had reached his goal in time. To feel lucky, he needed another piece of information, namely how close he was to have had the door shut in his face. He was not lucky just because he made it, he was lucky because he just made it. (Teigen 1996: 158)

2

Situations taking an unexpected turn. In other cases, the situation unfolds in a negative direction and seems to be heading for disaster. If the envisaged failure is averted, the protagonist will be considered lucky. Here, a negative outcome is expected before the final turn of events.

Lucky events of this second kind should follow a special narrative trajectory, with a downward slope, pointing towards doom and disaster, concluded with a surprising last-minute upward turn. This prediction was tested by asking students to recall a lucky situation from their own life, and describe it in words and pictures as a comic strip with three frames (Teigen et al. 1999: Experiment 2). This format was chosen to encourage participants to structure the event into distinct scenes. After sketching the scenes, they were asked to rate all three frames on separate scales from very unpleasant to very pleasant. A good-luck story in three scenes could run like this: 1 2 3

I was at a party. When I came home I could not find my keys. The next day I found them in an alley.

140

Karl Halvor Teigen

Like most other good-luck stories, this story follows a downward trajectory from scene 1 to scene 2. At this point it looks like a story about bad luck. But in scene 3 it takes a sharp upward turn, whereas the trajectory for bad-luck stories continues downward. If good and bad-luck stories have distinct profiles, or narrative structures, we may turn a good-luck situation into a bad-luck situation, or vice versa, by rearranging the order of the component events. Consider the following vignette (Teigen et al. 1999): Peter goes to the casino for the first time in his life. He plays twice. (Condition A) First he wins $150, then he loses $150. (Condition B) First he loses $150, then he wins $150. In both conditions, Peter left the casino with the same amount of money as he entered it. No loss, no gain. So from a financial point of view, nothing happened. Yet story A was perceived as a bad-luck story. Peter did not stop in time but wasted his gain. In contrast, story B was perceived (by a different group) as a good-luck story. He recovered his loss instead of going on losing. Other vignette stories revealed that this is not just a matter of temporal order of how things happen. It is also a matter of narrative order; that is, the order in which things are told. If you want to give the impression that you have been lucky, you should start with the bad news and end with the good news, regardless of what came actually first and last.

Luck and good fortune The term “lucky” is sometimes employed in situations where worse counterfactuals do not seem imminent. If I say: “I am lucky to be in good health,” or “You are lucky to have a job,” I do not necessarily imply that my health is a matter of accident, or that you were at risk of losing work. One could argue that “lucky” is used here in a figurative sense, referring to more permanent aspects of life, which perhaps should have been described as fortunate rather than lucky (Rescher 1995; Pritchard and Smith 2004). But the fact that people appear to use “lucky” and “fortunate” interchangeably indicates that these concepts have a common core: namely, a comparison with worse counterfactuals. This was tested in a study (Teigen 1997) in which respondents were asked to compare pairs of sentences, where one member of the pair was of the form “It is lucky that . . .” and the other “It is good that . . .” Otherwise, the sentences were identical. One experiment used an open-ended format, where respondents simply were asked to guess the communicative intentions of the speakers. For the sentence “It is lucky that you have a job” 90 percent of respondents spontaneously commented that it implied a downward comparison with others (who are not so lucky), whereas “It is good that you have

Counterfactual thinking and luck

141

a job” very rarely (5 percent) was believed to imply a comparison. This sentence stressed, according to our respondents, instead the advantages or usefulness of having a job. Comparisons with others point in turn to the counterfactual possibility of not having a job. Two other interesting themes emerged. About half of the respondents spontaneously remarked that a person using the formulation: “You are lucky that ...” might be a little envious. Several also remarked that luck statements imply that the person should feel grateful for his or her lot in life. These are evidently feelings that presuppose counterfactuals. Envious people compare themselves to somebody better off, thinking: “It should (or could) have been me.” A grateful person, on the other hand, realizes that the privileges he or she enjoys are “gifts” (from other people or from a kind fate) that cannot be taken for granted. To explore these two themes, two further experiments were conducted asking participants to give autobiographical descriptions of situations in which they had been envious or felt grateful (Teigen 1997: Studies 3 and 4). They were also asked about counterfactual thoughts and the degree of luck involved, using a similar procedure as in the studies of autobiographical luck and risk, reported above. Gratitude appeared, as expected, to have a strong counterfactual component. One has received a favor or an advantage that could have been withheld, being unearned or even undeserved. People who feel grateful consider themselves very lucky. Envy implies that someone else has received a favor or an advantage that could equally well have been mine. So envy also implies an upward counterfactual comparison. The study showed that when people envy somebody they describe themselves as being unlucky, whereas the envied person is in most cases perceived as extremely lucky.

Are good and bad luck opposite concepts? The results and analyses reviewed in this chapter have had good luck as their main focus. Perceptions of bad luck are assumed to follow a similar mechanism, with the difference that bad-luck situations are characterized by upward rather than downward counterfactuals. At the same time, the circumstances that produce good and bad luck may not be completely symmetrical. We have seen that autobiographical “good luck” episodes are distinct from simply “positive” events, whereas “bad luck” and “negative” events are much more similar (Teigen 1995). In the narrative structure study (Teigen et al. 1999), good-luck stories had a characteristic curve, with a contrasting final outcome, whereas bad-luck stories continued downward. This study also contained an experiment where participants were presented with several brief phrases allegedly taken from good luck or bad luck stories. None of the phrases indicated outcome valence, yet most respondents agreed that, for instance, the phrase “It ended with . . .” belonged to a bad-luck story, whereas “It could have ended

142

Karl Halvor Teigen

with . . .” was taken from a story about good luck. See Table 8.4 for further examples. From these observations, it is tempting to conclude that bad-luck stories refer less to counterfactuals, and more to what really happened, whereas good luck is more strongly determined by what did not happen than what actually took place. Post-computed counterfactuals are a case in point, as illustrated by the statement “Only later, I realized how —— I had been,” which nearly everyone felt should be completed with “lucky” rather than “unlucky.” The typical bad-luck story appears to be one that starts out innocently, before minor things go wrong, and the snowball starts rolling downhill, with a disastrous ending. Statements referring to trifles and normal, everyday behavior (“As I was about to enter the bus . . .”) are accordingly believed to occur in bad-luck stories. Thus we are left with an apparent paradox. Several of the vignette studies (like Peter at the casino, described above) show good and bad luck to be symmetrical, involving opposite counterfactuals. A near loser is lucky, whereas a near winner is unlucky, even if nothing is in fact won or lost. At the same time, other lines of evidence, particularly those based on autobiographical material, suggest that they are, in real life, not mirror images. Bad-luck incidents are typically bad (negative, painful) in an absolute sense, whereas good-luck incidents are typically good (positive, pleasant) only in a relative sense, compared to what might have been. The solution to this apparent contradiction comes from an analysis of what typically activates counterfactual thinking. According to Roese (1997), counterfactual thoughts are primarily activated by (1) negative outcomes, and (2) outcome closeness. Negative events will, as a rule, give rise to upward counterfactuals, following a suggestion from norm theory (Kahneman and Miller 1986) that we tend to mentally compare deviant outcomes to “normal” outcomes. In most areas, normal outcomes are positive or neutral (Bohner et al. 1988), whereas negative outcomes represent a deviaTable 8.4 Percentage of respondents assigning statements to “lucky” and “unlucky” stories Statement

Lucky

Unlucky

It all started with a trifle As I was about to answer the phone As I was about to enter the bus It ended with . . .

6.7 6.2 25.4 34.4

93.3 93.8 76.5 65.6

I could not believe my own eyes But at the last moment . . . It could have ended with . . . Only later I understood how . . . I had been

73.3 88.9 91.2 95.6

26.7 11.1 8.8 4.4

Source: Adapted from Teigen et al. (1999: table 1).

Counterfactual thinking and luck

143

tion from the norm. This would explain why so many bad-luck incidents simply consist in reports of accidents and disappointments. The upward counterfactuals need in these cases not to be explicitly specified, as it is tacitly understood that accidents should be avoided, and actions should achieve their aims. Upward counterfactuals also have a preparative function (Roese and Olson 1995c), suggesting how similar outcomes can be prevented or improved in the future. Positive events do not routinely activate counterfactuals to the same extent (Sanna and Turley 1996). Downward counterfactuals arise only when negative outcomes are close, either in retrospect (post-computed counterfactuals), or by being suggested by the development of the situation (trajectories with a downward slope). So, good-luck situations may in practice differ from bad-luck situations because they require a more narrow set of conditions: namely, close counterfactuals. Downward counterfactuals can also be elicited in negative situations as a form of affective repair. Thus even misfortunes may elicit a search for downward counterfactuals in order to make the victim feel better (Roese and Olson 1995c). To conclude: good and bad luck are symmetrical in the sense that they both are based on the same underlying counterfactual comparison process, but asymmetrical in the sense that this process serve different functions and can be triggered by different determinants.

Conclusion In the beginning of this chapter we cited Rescher’s (1995) analysis of luck. Rescher claimed that luck is partly dependent on the valence of the outcome, and partly on its probability. We are now in a position to modify both these claims. 1. Luck does not simply depend on the valence (or utility) of the factual outcome for the individual, but on the signed difference between the utility of the obtained outcome, U(F), and utility of the counterfactual outcome, U(Cf). Thus, L is directly proportional to U(F)  U(Cf). A positive difference indicates good luck, a negative difference bad luck. The larger this difference, the greater the luck. A positive difference can be obtained in more than one way: by a highly positive F (winning a million in the lottery) compared to a relatively neutral Cf (no prize), or by a less spectacular F compared to a negative Cf (avoiding a collision). Good-luck events in daily life are often of the second kind. A negative difference, signifying bad luck, can similarly be obtained either by experiencing a highly negative F (being the victim of an accident), or by missing a highly positive Cf, like the dissatisfied Olympic silver medalists of Medvec et al. (1995). Bad-luck events in daily life are often of the first kind. 2. Rather than being determined by the probability of the obtained outcome, p(F), luck depends on the distance between F and Cf. People are

144

Karl Halvor Teigen

perceived as especially lucky if Cf appears close. Thus L is inversely proportional with the |F  Cf| distance. Subjective distances and subjective probabilities are, however, not completely unrelated concepts. Subjective probabilities are sometimes assessed by a kind of “closeness heuristic” (Teigen 1998b). For instance, people who have “almost” won or “almost” lost will typically describe their probabilities of winning or losing as higher than people who were never close. Interestingly, this applies more to the probability of counterfactual than of factual outcomes, even in cases where F and Cf form an exhaustive pair, and p(F) and p(Cf) should be complementary. Probability estimates for F and for Cf appear in these cases to be based on different judgmental strategies, leading sometimes to paradoxical estimates. For instance, if team A loses a game against team B by a very small margin, people will, retrospectively, estimate team A’s probability of winning as high (A “could have” won), in some cases even higher than team B’s probability, despite the wisdom of hindsight. In line with this analysis, the thirty-eight dangerous episodes collected in the “risk and luck” study were presented to a peer group who were asked to estimate the probabilities of either the factual outcomes, p(F), or the counterfactual outcomes, p(Cf) (Teigen 1998a: Experiment 2). The post-computed counterfactual probabilities were highly correlated with ratings of luck, r  0.81, indicating that people avoiding a disaster that appears probable in retrospect, are perceived to be more lucky than those who avoid a less probable disaster. The correlation between luck and estimated probability for the actual outcome was significantly lower. In a review of philosophical and psychological treatments of luck, Pritchard and Smith (2004) agree with the present analysis about the central role of counterfactuals for the perception of luck. They identify two conditions, which in their view capture the “core” notion of luck: (L1) If an outcome is lucky, then it is an outcome which occurs in the actual world but which does not occur in most of the nearest possible worlds to the actual world (worlds which most resemble the actual world). (L2) If an outcome is lucky, then it is an outcome that is significant to the agent concerned. The first of these conditions is clearly related to the present concept of counterfactual closeness, or the F  Cf distance. The second condition, personal significance, is, in our view, better captured by the U(F)  U(Cf) difference. A highly significant outcome is, according to the present conceptualization, a highly positive or a highly negative outcome, relative to its alternative; in other words, an outcome that makes a great difference for the person involved. This gives us a model for perceived luck based on two links between observed and counterfactual outcomes: A small distance in terms of situation

Counterfactual thinking and luck

145

structure, and a large difference in terms of personal evaluations. A lucky event can accordingly be described as a small difference that makes a great difference. Or, combining the symbolic expressions developed above: U(F)  U(Cf) L   |F  Cf | Traditional definitions of luck see luck and chance as closely related concepts. Pritchard and Smith (2004) point out that chance is not an unproblematic concept, being sometimes, but not always synonymous with concepts like by accident, coincidental, unpredictable, and uncontrollable, which have also been evoked to explain luck. Perhaps concepts like these are not so essential for the definition of luck. If the core meaning of luck is to stress the contingent nature of F, and the possibility of an alternative, the causes of F become a secondary question. Factual–counterfactual comparisons will, however, be especially salient in situations where F did not have to occur, but can be viewed as one among several equally plausible, or more plausible, outcomes. Thus, chance may imply luck, not by definition, but in its capacity of facilitating counterfactual thoughts. In his popular book, The Luck Factor, Wiseman (2003) tried to identify and outline the main differences between individuals who describe themselves as extremely lucky and those who feel haunted by bad luck. The characteristics of lucky individuals can in his view be summarized in four major points: (1) lucky people are outgoing and sociable, exposing themselves to all kinds of new experiences and chance encounters; (2) they are open-minded and able to listen to their gut feelings; (3) they are persistent optimists; and (4) they do not dwell on mistakes and failures, but succeed in seeing the positive side even of bad luck events. The research reported in the present chapter is in good agreement with Wiseman’s last point, but extends it further by claiming that all good-luck experiences, even in individuals who do not view themselves as children of fortune, rest on a perceived contrast between a real and an imagined outcome. To be lucky, according to this analysis, presupposes an ability to obtain positive outcomes (in line with Wiseman’s second and third points), while at the same time imagining far worse counterfactual outcomes. This will normally require exposure to new, not entirely controllable, and sometimes even risky situations (as suggested by Wiseman’s first point). It follows that traditional, one-sided recommendations of prudent behaviors, like careful planning, urgings to “stay on the safe side,” and to “look before you leap” may increase one’s chances of success without increasing one’s chances of having been lucky, while “Nothing ventured, nothing gained” may be a better motto for those seeking luck (but not necessarily gains). Throughout this chapter, good and bad luck have been discussed as a contrast phenomenon, where the factual outcome is observed to deviate sharply from what could or should have occurred. However, it is not clear whether

146

Karl Halvor Teigen

this contrast is primarily of an affective or a cognitive nature. On the one hand, people are asked to report their luck experiences, and to what extent they feel lucky or not lucky, suggesting that luck has an emotional core. On the other hand, it also makes sense to say about a person that “he does not realize how lucky he is” (cf. the statement: “Only later did I realize how lucky I had been”), whereas we would hesitate to say that he did not understand how happy (or sad) he was. This suggests that a lucky event is perceived to be a situational rather than an emotional characteristic, even when it is held to have emotional consequences. Research suggests that close counterfactuals do not always create affective contrast, but can occasionally lead to affective assimilation, particularly with repeatable events or events that have implications for the future. Thus downward comparisons after a narrow escape can give reasons for alarm and function as a wake-up call (McMullen and Markman 2000). Upward comparisons, elicited by a close but lost game, will be described in positive or negative terms dependent upon whether it is the first or the last game of a series (McMullen and Markman 2002). McMullen and Markman (2002: Study 2) discovered that fans of a basketball team actually felt better at halftime when their team was down by one point than fans of the team that was up by one point. Fans of a team that is trailing closely behind may feel they have good reasons for being hopeful, whereas a narrow lead indicates uncertainty about the final outcome. Reports about past lucky events have usually referred to final outcomes, suggesting affective contrast. Future research on luck should also include reports on luck in ongoing activities and in fields where the final outcomes are not settled. Goals that are not yet achieved, but are attainable “with a little bit of luck” (to quote Mr Doolittle in My Fair Lady) may be more appealing than temporary successes that are barely attained, but not definitely secured. This would give an occasion to further explore the relation between perceived luck and the affective reactions it prompts.

Note The research reported in this chapter has been supported by grants from the Research Council of Norway. Students from the universities of Bergen, Tromsø, and Trondheim (NTNU) contributed in important ways to the original experiments.

9

On the comparative nature of regret Marcel Zeelenberg and Eric van Dijk

On 8 April 1995 a fifty-one-year-old man, who lived in Liverpool, England, decided to end his life. He did so after learning that he had missed out on a £2 million prize in the national lottery. While watching television he saw the numbers of his winning combination, 14, 17, 22, 24, 42 and 47, appearing one by one on the screen. He always played these numbers, but on this occasion he had forgotten to renew his five-week ticket on time. It had expired the previous Saturday. The reason why this man responded so emotionally has to do with the human ability to think counterfactually; that is, the ability to reflect not only on the things that happen, but also on the things that might have happened differently. For the lottery player, it was easy to imagine a different world in which he would have been a millionaire if only he had made a near effortless change in his own behavior – namely, simply doing what he normally had done in the past. Hence he suffered the consequences of comparing his current self with his counterfactual millionaire self. As this example illustrates, counterfactual thinking has a profound influence on emotional experiences. It can elicit and amplify emotions. Any volume on the psychology of counterfactual thinking would be incomplete without a discussion regarding one of the emotions most clearly associated with it: namely, regret. Regret is a negative, cognition-based emotion that we experience when realizing or imagining that our present situation would have been better had we decided differently. In early psychological research on counterfactual thinking by Kahneman and Tversky (1982b) and Kahneman and Miller (1986), the consequences of counterfactual thinking for regret already received attention. As Kahneman and Miller (1986: 136) stated, “Constructed elements also play a crucial role in counterfactual emotions such as frustration or regret, in which reality is compared to an imagined view of what might have been.” Later on, researchers also studied the effects of counterfactual thinking on other emotions, such as anger, disappointment, distrust, guilt, sadness and shame (e.g., Mandel 2003a; Niedenthal et al. 1994; Zeelenberg et al. 1998d). A study by Mandel (2003a), which focused on a range of emotions and their relation with counterfactual thinking, showed that the correlation between counterfactual thinking and emotion was

148

M. Zeelenberg and E van Dijk

strongest for regret, thus underscoring our claim that regret is the prototypical counterfactual emotion. In this chapter, we discuss the link between counterfactual thinking and regret. We will first explain what regret is and why the study of regret is important. Then, we describe a set of studies that demonstrate the specific comparisons necessary for regret and also show how this emotion differs from the related counterfactual emotion of disappointment. Next, we discuss a set of studies that tests the comparison dependence of regret directly. In these studies, we manipulated in several ways the extent to which a decision maker can compare the obtained outcome to forgone outcomes. These manipulations enabled us to investigate and demonstrate the direct link between regret and counterfactual comparisons. Finally, we discuss a second set of studies in which we manipulated the expectation of future feedback on forgone alternatives and thereby the possibility to compare outcomes in the future. We show that these manipulations of counterfactual comparisons clearly impact actual choice behavior.

What is regret? One may argue that regret is a hybrid of cognition and emotion that cannot exist without both components. We conceive of regret as a cognitive emotion instead of an emotional cognition. The reason for this is that, as we discuss later, regret contains all the elements that are typical of emotional experiences (Roseman et al. 1994; Zeelenberg et al. 1998c). Regret is an emotion that requires cognition to be experienced and that may influence and produce cognitions as well. In order to feel regret one has to think (see also Roese 2001). One has to think about one’s choices and the outcomes generated by these choices, but also about what other outcomes might have been obtained by making a different choice. As a result, regret is typically felt in response to decisions that produce unfavorable outcomes compared to the outcomes that the rejected option would have produced. That is, we decide to do X, but in retrospect we discover that we would have preferred doing Y because we think or know that Y would have resulted in a better outcome. An important issue is the question of how regret feels. What is the phenomenology of this emotion? What is its experiential content? Regret can be differentiated from several other negative emotions on the basis of the phenomenological characteristics (Roseman et al. 1994; Zeelenberg et al. 1998c). These characteristics are the different components that make up the emotional experiences; namely, feelings, thoughts, action tendencies, actions and motivational goals. It was found that regret is accompanied by feelings that one should have known better and by having a sinking feeling, by thoughts about the mistake one has made and the opportunities lost, by feeling a tendency to kick oneself and to correct one’s mistake and wanting to undo the event and to get a second chance.

On the comparative nature of regret 149 An important element of regret is also its relation to responsibility for the negative decision outcome (see Zeelenberg et al. 2000b). This makes regret uniquely linked to decisions. If there were no alternatives to choose from, there would be no regret. The closely related emotion of disappointment, for example, is also connected to negative outcomes and also stems from thinking counterfactually, but this emotion may be felt in absence of a decision (Zeelenberg et al. 1998c). One may be disappointed by the rainy weather (which is beyond our control), but one cannot regret it (although one may regret the decision of not carrying an umbrella).

Why study regret? The importance of the study of regret is related to its prevalence. In a study of verbal expressions of emotions, Shimanoff (1984) found that regret was the second most frequently named emotion (only love was mentioned more frequently). That “regret is a common, if not universal, experience,” as Landman (1993: 110) stated, by itself justifies the current inquiry into the comparative nature of regret, but there are other reasons. One reason is that the experience of regret has some clear psychological consequences, such as effects on goal setting (Kinner and Metha 1989; Lecci et al. 1994) and rumination (Handgraaf et al. 1997; Savitsky et al. 1997; Wrosch and Heckhausen 2002). Landman et al. (1995) report that counterfactual thinking about missed opportunities (which could be considered as a proxy for regret) is associated with emotional distress in the short run but with motivational benefits in the long run. Although they found that counterfactual thoughts were associated with higher levels of depression and anxiety, they also found that “compared to those who report no such counterfactuals, those who acknowledge thoughts of past missed opportunities are more likely to envision future changes in their lives” (Landman et al. 1995: 94). These psychological consequences of regret may be dependent on age, given that the opportunities to overcome regrets decline with age. Wrosch and Heckhausen (2002) found that for young adults internal control attributions were associated with active change of regrettable behavior, attenuating the regret and lowering rumination. For older adults, internal control attributions were related to high levels of regret and more intrusive thoughts related to these regrets. Another reason why regret research is important is because regret is a powerful predictor of behavior. For example, consumers who regret purchasing a particular service are likely to switch to another service provider, and the more regret decision makers feel after having chosen the “losing” gamble the more likely they are to choose the other gamble in a next round. (For a review of the behavioral consequences of regret, see Zeelenberg et al. 2001.) But also the mere expectation of future regret may impact our decision making. This is the notion that is central in the regret theories that were developed by the economists Bell (1982) and Loomes and Sugden (1982). In order to explain

150

M. Zeelenberg and E van Dijk

why people sometimes choose the opposite of what classic economic theory would prescribe, these theorists proposed that decision makers might do so in order to avoid future regret. Traditionally, it was assumed that when decision makers have two options, A and B, they calculate the expected utility of both options and choose the one with the highest expected utility. The intuition behind regret theory is that the expected utility of A is additionally dependent on what one misses by not choosing B. Regret theories are based on the following assumptions. First, it is assumed that we experience emotions, such as regret and rejoicing, as a consequence of our choice. We feel regret when the rejected option would have resulted in a better outcome and rejoicing when it would have resulted in a worse outcome. Next, it is assumed that these emotions have a clear impact on the utility that we derive from the decision outcomes. The most important assumption in the regret theories is that these effects of regret on utility are anticipated beforehand and taken into account when making a choice. Hence, our decision making can be characterized as regret averse. We choose in such a way that we believe will minimize regret in the future. Regret aversion should be distinguished from the better-known phenomena of risk aversion and loss aversion (Kahneman and Tversky 1979). For example it has been documented that people make risk seeking choices in order to minimize regret (Zeelenberg et al. 1996).

Why study the comparative nature of regret? As we have seen, the comparative nature of regret is discussed in both the early psychological theorizing on counterfactual thinking and the early economic theorizing on decision making. Interestingly, although research has reported findings consistent with the assumptions of regret theory (for a review, see Zeelenberg 1999), the notion that comparability is vital for regret has not yet been tested empirically. Up to now, regret researchers have simply assumed that decision makers are able to compare outcomes despite the fact that in real life these counterfactual comparison processes may be difficult or impossible. The effects of comparability on experiences of regret and anticipations of future regret are the topic of the remainder of this chapter. The relevance of studying the comparative nature of regret is further stressed by Frijda’s (1988: 353) Law of Comparative Feeling, according to which “The intensity of emotion depends on the relationship between an event and some frame of reference against which the event is evaluated.” Frijda compiled a set of such laws in order to describe the regularities of emotional experience. As he pointed out, the laws of emotion were not all equally well established, but they could be used as a program of research. As will be apparent in this chapter, these laws have inspired us and we hope to contribute to their further development.

On the comparative nature of regret 151

Regret and thinking counterfactually In the present section of this chapter, we combine ideas about counterfactual thinking with ideas from regret theory and disappointment theory. Disappointment theory is theoretically related to regret theory in that it is assumed that decision makers can experience disappointment as a result of a decision and that they take this into account before they decide (Bell 1985; Loomes and Sugden 1986). In these decision theories, regret and disappointment are conceptualized in different ways. The theories assume that there is a difference in the source of comparison from which the two emotions arise. Although regret and disappointment both stem from a comparison between “what is” and “what might have been,” regret is assumed to originate from comparisons between the factual decision outcome and a counterfactual outcome that might have been had one chosen differently. By contrast, disappointment is assumed to originate from a comparison between the factual decision outcome and a counterfactual outcome that might have been had another state of the world occurred. This other state of the world is typically the expected state of the world. We tested whether regret indeed follows from counterfactuals in which an obtained outcome is compared to an outcome that would have been produced by another choice (Zeelenberg et al. 1998d). We contrasted this prediction with one concerning disappointment. Disappointment should follow from counterfactuals in which an obtained outcome is compared to an outcome that would have been produced by a different state of the world. In a first study, we asked people to recall either an experience of regret or one of disappointment. We next asked them to indicate in what way the situation could have been better. That is, we asked them to generate counterfactuals. The counterfactuals produced were classified as addressing aspects pertaining to the behavior of the participant, to the participant’s character, or to the situation. This classification was based upon the research by Niedenthal et al. (1994). As we expected, the participants who were asked to recall a regretful experience tended to generate counterfactuals in which they mutated aspects that were under their own control (i.e., behaviors). In contrast, participants who were asked to recall a disappointing experience typically generated counterfactuals about aspects that were not under their own control (i.e., aspects of the situation and to a lesser extent their character). This finding that regret and disappointment are associated with different types of counterfactuals provided support for the idea that regret is related to specific counterfactuals that mutate the choices or decisions made by the focal actor. In a follow-up study (Zeelenberg et al. 1998d: Study 3), we sought to replicate and extend the finding that qualitatively different counterfactuals shape the experience of regret and disappointment. Although the study reported above was consistent with this hypothesis, it did not provide a direct test of the causal relationship. The follow-up study was designed to

152

M. Zeelenberg and E van Dijk

provide such a test. Participants in this study were presented with a detailed vignette describing a situation that resulted in a negative outcome. They had to imagine themselves being in this situation. There were various elements in the vignette that contributed to the negative outcome, including the protagonist’s own choices and also uncontrollable aspects of the situation. Hence, the situation could elicit either regret or disappointment (or a combination of the two). After having read the scenario, participants were asked to generate counterfactuals for the outcome. Half the participants were instructed to undo the event by mutating aspects of their behavior, whereas the other half were instructed to undo the event by mutating aspects of the situation. Next, all participants were asked to indicate the level of regret and disappointment they would feel in that situation. In support of our hypothesis, participants indicated feeling significantly more regret than disappointment when the counterfactuals they generated mutated their behavior. Moreover, they indicated that they would feel more disappointment than regret when the counterfactuals they generated mutated situational aspects. Together these findings extend the ideas formulated in norm theory about emotional amplification and counterfactual thinking (Kahneman and Miller 1986). The influence of counterfactual thinking on emotion that has received most attention is that counterfactual thoughts affect the intensity of affective reactions to outcomes (Kahneman and Tversky 1982b; Roese 1994, 1997). The easier it is to generate a counterfactual outcome, the stronger is the affective reaction to the factual outcome. Counterfactual thoughts also affect the valence of these reactions. We feel bad if the counterfactuals are better than reality, we feel good if they are worse than reality (Roese 1994). The research described here, together with that of Niedenthal et al. (1994) and Mandel (2003a), showed that counterfactual thinking influences not only the valence and intensity of the affective reactions, but also their specific tone. That is, the type of counterfactual thoughts (i.e., their content and focus) determines which specific emotion is felt. The dependence of regret on these specific counterfactuals is consistent with the assumptions in regret theory regarding the causes of this emotion. In the following section, we will report on three studies that directly tested the comparison dependence of regret.

Regret and the comparability of what is and what might have been Given the fact that our lives are fraught with possibilities to regret our decisions, it should come as no surprise that regret is one of the most prevalent emotions. Indeed, with such an abundance of possibilities to experience regret, one might wonder why we do not constantly beat ourselves over the head over missed opportunities. This raises the question of when we are most likely to engage in counterfactual comparisons. Surprisingly, research

On the comparative nature of regret 153 on regret has remained relatively silent on this matter. Kahneman and Miller (1986: 142) noted that “The experienced facts of reality evoke counterfactual alternatives and are compared to these alternatives” and that “Outcomes that are easily undone by constructing an alternative scenario tend to elicit strong affective reactions” (1986: 145). Norm theory thus predicts that the more difficult it is to generate a counterfactual that would have yielded better outcomes, the less regret one will feel over one’s current decisions. But what determines whether or not a regret-producing counterfactual is difficult to make? Research on the near-miss effect (Johnson 1986; Kahneman and Varey 1990), the temporal order effect (Segura et al. 2002) and action–inaction differences in regret (Kahneman and Tversky 1982a; Zeelenberg et al. 2002) suggests that one of the main factors to consider is the ease with which one can imagine a possible alternative action that would have yielded a better outcome (for a review, see Byrne 2002). For an example of what determines which counterfactual comparisons are easy to make, we consider one of our recent studies (Zeelenberg et al. 2002). We asked participants how much regret a soccer coach would feel if his team lost after he either changed or did not change the team. The study’s design thus mimicked the traditional action–inaction studies. In addition, however, we manipulated prior experiences: the coach’s team had either a winning or losing record. The findings revealed that more regret was reported for the active coach than for the inactive coach (replicating the traditional action–inaction difference), but only if the team had a winning record. If the team had been losing, then the inactive coach was seen as experiencing more regret. These results support our theorizing that prior outcomes and events may call for action and hence make inaction the more normal decision. In other words, the ease with which a decision is compared to another course of action appears to be influenced by events related to the decision process. We are likely to compare unexpected, nonroutine, nonrequired behaviors to their expected, routine and required counterparts and to feel regret as a result. We maintain that there are new insights to be gained by focusing more closely on the comparison that is so central to regret: the comparison between “what is” and “what could have been.” In most of regret research, it is relatively easy to compare the two because the outcomes are simply provided. Participants may learn that one course of action yielded them with a particular outcome, whereas they would have obtained more had they decided differently. For example, in Kahneman and Tversky’s (1982a: 173) classic “two investors” problem, the participants in the action condition read that: “George owned shares in company B. During the past year he switched to stock in company A. He now finds out that he would have been better off by $1,200 if he had kept his stock in company B.” It doesn’t take much cognitive effort to assess that inaction would have been the more attractive alternative here and that, apparently, the situation calls for regret. But is the life we live that simple? Two points need to be considered

154

M. Zeelenberg and E van Dijk

here. First, we may not always be certain of the outcomes we could miss or have missed out on. What outcomes would we have obtained in life had we studied law instead of psychology? Second, even if we could be certain of the alternative outcomes, they may not always be easily compared with what we have got. Even if we are certain that we would have become lawyers if we had not become psychologists, it is not clear how “being a lawyer” should be compared to “being a psychologist.” In the following sections, we will further elaborate on these two aspects of the comparison process. Comparing the known with the unknown Do we always know for certain what outcomes we would have generated had we decided differently? As the fictional movie character Forrest Gump reminded us, “Life’s like a box of chocolates, you never know what you’re gonna get.” Eventually, we will know what we have, but this does not mean that we will know with certainty what “chocolates” we would have gotten had we chosen differently. It is this type of uncertainty that may shield us from experiencing regret even more frequently than we do. Why might uncertainty keep us from experiencing regret? One of the main reasons, we suggest, is that it may keep us from thinking through the consequences of what might have been. This reasoning is related to prior research on the disjunction effect (Shafir and Tversky 1992; see also van Dijk and Zeelenberg 2003) that suggests that uncertainty and ambiguity may induce people to engage in non-consequential reasoning. This can best be illustrated by discussing an example taken from Tversky and Shafir (1992). In a scenario study, participants had to imagine that they had taken an exam. Next, they had to imagine that they failed the exam, passed the exam, or did not know whether they had passed or failed the exam. The main dependent variable was the willingness to purchase a vacation to Hawaii. The results showed that both participants who had learned that they passed the test and participants who had learned that they failed the test were likely to purchase the vacation. Interestingly, those who were still ignorant about their test result were unlikely to purchase the vacation. These findings clearly suggest that people may not think through the consequences of uncertain outcomes. After all, had participants taken the consequences into account, they should have purchased the vacation because they were willing to buy the vacation if they failed and also if they passed the exam. Research on the disjunction effect, therefore, suggests that people are reluctant to base their decisions on ambiguous information. In a similar vein, we suggest that people may not think through the consequences of what could have been if what could have been remains uncertain to them. If so, this would imply that people may become less likely to compare what is with what could have been. As the lid remains on the chocolate box of missed opportunities, people may not be troubled by feelings of regret. In a first experimental study (van Dijk and Zeelenberg 2005) we put this

On the comparative nature of regret 155 reasoning to the test by having participants imagine that they participated in a game in which several prizes could be won. In this game, the prizes were behind closed doors and participants had to choose one out of two doors. Whatever was behind the door they chose would be theirs. All participants learned that behind the door of their choosing they found a stress ball (a little rubber ball that one can squeeze in order to relieve stress). They also received some information about the prize behind the other door. Some participants read that the missed prize was a CD of their choosing. Others read that they missed out on a dinner for two and some that they missed out on a Walkman. We also included a condition in which participants were uncertain about their missed prize and where all they knew was that the missed prize was either a CD of their choosing, a dinner for two, or a Walkman. In agreement with the notion that uncertainty about missed opportunities may sometimes be a blessing, participants reported less regret when they were uncertain about the prize they had missed than when they knew the nature of the missed prize. The disjunction effect has been explained by suggesting that people find it cognitively too complex to think through the consequences of ambiguous and uncertain outcomes. This would imply that in our study, participants were less likely to experience regret when being uncertain of what they missed out on because it was too complex for them to think through how they would feel if they either won a CD of their choosing, a dinner for two, or a Walkman. This interpretation may be related to Frijda’s Law of Apparent Reality, which states that “Emotions are elicited by events appraised as real and their intensity corresponds to the degree to which this is the case” (Frijda 1988: 352). In the case of uncertainty about the missed opportunity, the counterfactual outcome may lack a sense of realness. Consequently, it would evoke less regret than when the forgone alternative is known. In addition to these cognitive explanations, motivational explanations may also account for the observed findings. For example, people may dislike thinking about how they would feel if they miss out on a CD, if they know that the reality could be that they would not miss out on this CD but on a Walkman (cf. Tversky and Shafir 1992). This interpretation is related to studies by Tykocinski (2001) and Gilovich et al. (1995). Tykocinski has shown that when faced with an unfavorable outcome, people tend to exhibit “retroactive pessimism”; namely, they attempt to attenuate the outcome’s emotional impact by telling themselves that it was more or less inevitable. In Gilovich et al.’s research, outcomes were rigged such that participants, apparently as the result of their decision, ended up with a small prize and missed out on a big prize. The more regret these participants felt, the more likely they were to engage in dissonance-reduction strategies (i.e., they comforted themselves by increasing the valuation of the small prize). These two lines of research show that decision makers may be motivated to “distort” their thinking about what happened in order to mitigate their negative emotional experiences. In a similar vein, one could argue that people distort

156

M. Zeelenberg and E van Dijk

their thinking about what could have happened in order to regulate what they feel. Comparing the incomparable From what we have stated thus far, one might get the impression that we maintain that certainty about the missed opportunity will inevitably lead to regret because, with certainty about what is and about what could have been, people can easily make the comparison. Of course, sometimes this can be the case. If you know that your decision led you to earn $100 whereas you would have earned $200 if you decided differently, you are facing a relatively easy comparison and, consequently, you are likely to experience regret. But again, this view may be too simplistic. What if you know that you have got an orange and you also know with certainty that you would have gotten two apples should you have decided differently? Comparing apples to oranges is a difficult task. More generally, this implies that comparisons may differ in complexity. As Medin et al. (1995: 8) point out: Not all comparisons are equally easy to make; comparisons that involve substantially different properties are difficult. It is easier, for example, to compare the merits of Mendelssohn and Schumann than to compare the merits of Schumann and the Beatles. It may be possible to convert both Schumann and the Beatles into generic utilities, but this process seems to require more work than comparing items that have similar aspects. If comparability lies at the heart of regret, this implies that incomparability may be another feature of the comparison process that could explain why we do not constantly experience the intense feeling of regret. In a second study (van Dijk and Zeelenberg 2005), we investigated the moderating effect of comparability on regret by presenting participants with a scenario in which they encountered a small fair. At the fair, they took part in a lottery with instant scratch-card lottery tickets. They were just in time because only two scratch cards were left and, after some deliberation, they bought one of them. One group of the participants learned that they had won a C15 liquor-store token, whereas another group learned that they had won a C15 bookstore token. After this, participants were informed that someone else bought the last remaining scratch card. We manipulated the prize of this other person. It was either a C50 bookstore token or a C50 liquor-store token. Thus, we were able to examine the effect of the complexity of the comparison on regret (Johnson 1984, 1989). The comparison is relatively easy if the prize won and the missed prize were in the same product category (either both bookstore tokens or both liquor-store tokens) and relatively complex if they were in a different product category (i.e., you win a bookstore token but miss out on a liquor-store token, or vice versa). The results

On the comparative nature of regret 157 of our experiment showed that, in agreement with the notion that comparability is a determinant of regret, participants winning the bookstore token reported more regret when missing out on the C50 bookstore token than when missing out on the C50 liquor-store token. Similarly, participants winning the liquor-store token reported more regret when missing out on the C50 liquor-store token than when missing out on the C50 bookstore token. Thus, reduced comparability attenuates regret. Individual differences in tendencies to compare The previous two experiments provided support for the idea that regret is related to outcome comparability. Another issue to consider here is that individuals differ in the extent to which they compare outcomes, especially in the situations that we have described here. Note that in the scratch-card study reported above, the information about the counterfactual outcome was provided by the outcome of another person. In this case, counterfactual thinking and social comparison go hand in hand (Olson et al. 2000). In real life, we often learn about the outcomes forgone by comparing our own outcomes to those of others and their outcomes may serve as counterfactual reference points. Such social comparison information is important in many decisions, given that people are very sensitive to the outcomes of others. Decision makers can be especially dissatisfied when others receive a better outcome. Previous research has shown that these social comparison effects can also contribute to the regret that people may feel in response to a decision that goes awry (Boles and Messick 1995; Larrick 1993; Zeelenberg and Pieters 2004). According to Larrick, regret and dissatisfaction are strongest when social comparison information reflects badly on one’s one decision ability. The observation that social comparisons play a role suggests an additional opportunity for testing our reasoning by studying individual differences. Gibbons and Buunk (1999) recently constructed a reliable scale for measuring individual differences in social comparison orientation. If our reasoning about regret is correct, people who score high on this scale should be more vulnerable to regret in situations in which feedback provided by the decisions of others because they are more likely to compare themselves to others. We recently tested this line of reasoning (van Dijk and Zeelenberg 2005: Study 3). In the beginning of the academic year we administered the social comparison orientation scale to a group of first-year students. Eight months later, we contacted the students whose score fell in the highest thirtieth percentile range of the social comparison scores and the students whose score fell in the lowest thirtieth percentile range. We asked them to participate in a study and we presented them with the scratch-card scenario that was also used in study described above. All participants read that they had won a C15 bookstore token. Half of them read that they missed out on a C50 bookstore token, whereas the other half read that they missed out on a

158

M. Zeelenberg and E van Dijk

C50 liquor-store token. The study had the following design: 2 (Missed Prize: C50 bookstore token versus C50 liquor-store token)  2 (Comparison

Orientation: low versus high). After the students had read the scenario, they were asked to indicate the regret they would experience over their choice and their happiness with the C15 bookstore token. Overall, this study replicated the findings reported earlier. Participants reported feeling more regret when both the obtained and the missed outcome were from the same category (and thus easy to compare) than when they were from different categories (and thus more difficult to compare). However, when the data were split on the basis of participants’ social comparison orientation, it was apparent that this result was observed only among participants who scored low in social comparison orientation. Those who scored high on social comparison orientation tended to report high regret irrespective of whether the outcomes were from the same category or not. It seems that people who are high on social comparison orientation compare their outcomes with those obtained by others even when these outcomes are in a different category. This finding underscores the earlier findings that regret is highly dependent on comparison processes. Taken together, the results of the three studies reported in this section paint a clear picture of the comparative nature of regret. We have provided first evidence that shows a direct link between the regret that may stem from decisions that go awry and the comparability of obtained and forgone decision outcomes. Let us now turn to the behavioral implications of these findings. If comparison of decision outcomes is indeed important, one may expect that future opportunities for comparing outcomes may influence current decisions. This has been a central issue in our research on anticipated regret.

Avoiding comparisons as a regret-minimizing strategy In real-life decisions people may occasionally receive information about counterfactual outcomes. For example, people choosing to invest in particular stocks will learn about future stock prices for the chosen stocks, but also for the non-chosen stocks. Likewise, gamblers who decide not to bet on the long shot in a horse race will learn the position at which this horse finished after the race and, thus, whether this option would have been better. For many other decisions, however, such counterfactual feedback is frequently absent. If you decide to enter into a business venture with someone or to marry someone (else), you will never find out how successful each enterprise would have been had you chosen another partner or spouse, or none at all. In these cases, only the chosen option produces an outcome. On the basis of the studies reported in the previous section, we expect that in the cases in which the counterfactual outcome is unknown, one cannot compare the outcome of the chosen alternative to that of rejected alternatives and the regret experienced over a bad decision outcome would

On the comparative nature of regret 159 be less intense. According to regret theory, the impact of regret on decision making will then also be less profound. Bell (1983: 1165) already proposed that the effect of expected counterfactual feedback “is the predicted phenomenon on which experimentation should be concentrated.” Indeed, research has shown that manipulations of counterfactual information about the nonchosen alternatives influences the extent to which people experience regret or its positive counterpart (Ritov and Baron 1995; Zeelenberg and Pieters 2004) or more general outcome satisfaction (Boles and Messick 1995; Inman et al. 1997; Mellers et al. 1999). The question boils down to whether decision makers make choices that shield them from possible regret-causing comparisons on forgone alternatives. Below we describe research by Zeelenberg et al. (1996) that has investigated this question. The assumption in Zeelenberg et al.’s (1996) research was that people are regret-averse and that they will choose in order to minimize future regret. In their studies, participants could do so by opting for an alternative that made it impossible to compare the outcomes. The precise design of the three experiments was as follows. Participants were given a choice between two alternatives (i.e., two monetary gambles), one being relatively risky and the other being relatively safe. A matching procedure (cf. Slovic 1975) ensured that these alternatives were of roughly equal attractiveness. This was done in the following manner. Participants always knew the outcome of the risky alternative and the probability of winning it. For example, in Experiment 1, the relatively risky alternative would result in a gain of 130 Dutch guilders with a probability of 35 percent, or it would result in no gain with a probability of 65 percent. The relatively safe alternative would provide a gain of an unknown amount X with a probability of 65 percent, or no gain with a probability of 35 percent. The participants’ task was to write down the value of X for which they found the alternatives equally attractive. Next, feedback on one of the alternatives was manipulated independently of the riskiness of the alternatives. All three experiments included a Feedback Safe Alternative condition and a Feedback Risky Alternative condition. In the Feedback Safe Alternative condition participants would always learn the outcomes of the safer alternative, regardless of their choice. In the Feedback Risky Alternative condition participants would always learn the outcomes of the riskier alternative, regardless of their choice. In addition to this feedback, all participants always expected to learn the outcome of the chosen alternative. Based on the idea that comparisons with rejected alternatives may result in regret, it was predicted that participants would prefer not being able to compare their outcome with the counterfactual outcome. Thus, participants in the Feedback Safe Alternative condition were predicted to choose the safer alternative. This would provide them with feedback on the chosen alternative only and protect them from threatening counterfactual comparisons with the outcome of the riskier alternative. Likewise, participants in the Feedback Risky Alternative condition, who would always learn the outcome of the

160

M. Zeelenberg and E van Dijk

riskier alternative, were predicted to opt for the riskier alternative, thereby avoiding comparisons with the outcome of the safer alternative. The results supported the predictions. The majority of participants who expected feedback on the risky alternative opted for this alternative, thereby avoiding counterfactual comparisons. Similarly, participants expecting feedback on the safe alternative showed a clear preference for the safe alternative. This pattern was found in all three studies, in which included alternatives involving both gains and losses. One of the studies reported in Zeelenberg et al. (1996) provided additional information that confirmed the role of regret in these studies. Participants in this study were asked for justifications of their choices. Participants in the two counterfactual feedback conditions reported significantly more regret-related justifications than participants in a no-feedback control condition. This research on expected counterfactual feedback clearly demonstrates that the anticipation of regret can induce decision makers to protect themselves from counterfactual information. By now this tendency to avoid counterfactual comparisons has been found in a number of studies (e.g., Guthrie 1999; Larrick and Boles 1995; Ritov 1996; Zeelenberg and Beattie 1997; Zeelenberg and Pieters 2004). This tendency to avoid regret could have serious implications for real-life decision making. Our research shows that regret aversion results in choices that shield people from threatening feedback on forgone alternatives and we suggest that this may generalize to post-decisional stages. Active avoidance of feedback can deter people from learning from experience.

Final thoughts Decisions, by their very nature, are the necessary element for regret. As the research reviewed in this chapter has shown, regret stems from comparing the outcome of a decision with what the outcome would have been had one chosen differently. This close connection between decisions, regret, and counterfactual comparisons has attracted the attention of psychologists and economists. Because of its relevance for decision making and counterfactual thinking, the emotion regret has also spurred much empirical research to investigate its correlates and consequences. None of this previous research, however, addressed explicitly the comparative nature of the emotion regret. In this chapter, we have aimed to fill this gap and have provided first tests of this notion. The results of these tests support the assumptions of regret theory and contribute to testing the laws of emotion proposed by Frijda (1988). Taken together, in our study of the role of counterfactual thinking in regret we have tried to integrate theories, paradigms, and findings from decision research, emotion research, and social cognition research. We believe there are many interesting research questions that could be addressed more effectively by this combined approach. One question concerns the

On the comparative nature of regret 161 dynamics of regret over time and its relation to counterfactual thinking. It would be interesting to study how behavioral decisions and their outcomes influence counterfactual comparison processes, which in turn result in the labeling and experience of regret. It may further be of interest to study how behavioral and cognitive attempts to cope with this regret (i.e., the avoidance of counterfactual information) may affect the experience. Investigating these dynamics should further our understanding of decision processes, emotional experiences, and the psychology of counterfactual thinking.

Note This chapter is based on equal contribution. We thank Jacqueline Tanghe for valuable comments on an earlier version of this chapter.

Part IV

Counterfactual thinking in the context of crime, justice, and political history

10 Escape from reality Prisoners’ counterfactual thinking about crime, justice, and punishment Mandeep K. Dhami, David R. Mandel, and Karen A. Souza I’m a non-violent, first-time offender with a 24-year sentence. I could have been a productive member of society, given the chance, but I will spend over two decades in prison. (Anonymous prisoner)

Our understanding of an event is influenced not only by what actually happened, but also by what “coulda, woulda, or shoulda” happened. These thoughts about how the past might have happened differently are known as counterfactuals. Counterfactual thinking is prevalent in domains of ordinary personal life such as career and romance (Landman and Manis 1992), after traumatic life experiences such as bereavement (Davis et al. 1995), and in public life as observed during public inquiries (Reiss 2001) and court cases (Kassin et al. 1990). Studies have found that counterfactual thinking is involved in a variety of psychological processes, including attributions of blame and responsibility (e.g., Branscombe et al. 1996; Miller and Gunasegaram 1990), perceptions of fairness (Buck and Miller 1994; Folger and Kass 2000), and feelings of guilt and shame (e.g., Nario-Redmond and Branscombe 1996; Niedenthal et al. 1994). The counterfactual thoughts of offenders, defendants, or prisoners are likely to center on issues of blame and fairness, and feelings of guilt and shame, much like victims, criminal justice agents, the media, and public focus on these issues when considering crime, justice, and punishment. In this chapter, we review research on counterfactual thinking in prisoners. In the first section, we review and critique past research on counterfactual thinking in the legal domain. In the second section, we present an overview of our research on counterfactual thinking in prisoners, focusing on three issues. First, what effect does upward counterfactual thinking have on prisoners’ attributions of blame and their feelings of guilt and shame? Second, what is the relation between prisoners’ upward counterfactuals and their perceptions of fairness and feelings of anger? Third, what is the semantic content of prisoners’ upward counterfactual thoughts? In the third section, we highlight theoretical implications of this research for the study

166

Dhami, Mandel, and Souza

of counterfactual thinking and we propose future research in the legal domain. In the final section, we outline implications of our research for improvements to the justice system.

Past research on counterfactual thinking in the legal domain A growing body of literature has examined the role of counterfactual thinking in judgments, attributions, and emotional responses to cases involving acts of intentional or unintentional wrongdoing (e.g., Bothwell and Duhon 1994; Branscombe et al. 1996; Catellani and Milesi 2001; Macrae and Milne 1992; Macrae et al. 1993; Miller and McFarland 1986; Nario-Redmond and Branscombe 1996; Turley et al. 1995). Such research has shown that factors including agency, normality, and the direction of counterfactual thinking can influence attributions of blame and responsibility to offenders and victims. For example, Branscombe et al. (1996) found that blame assigned to the victim in a hypothetical rape case was greater when participants undid the outcome by changing her actions rather than the offender’s actions. Similarly, participants assigned more blame to the offender when the outcome was undone by focusing on his actions rather than the victim’s actions. In addition, Turley et al. (1995) reported that a hypothetical rape victim was considered more responsible for the offence when her preceding actions were unusual rather than usual. Nario-Redmond and Branscombe (1996) found that downward counterfactuals about how things could have been worse for the victim in a hypothetical rape case led to the offender being judged less culpable for his actions, whereas downward counterfactuals focused on the offender led to him being attributed greater culpability. Research has also demonstrated that factors such as normality, direction of counterfactual thought, and perspective can influence the harshness of the penalty recommended for an offender. Turley et al. (1995) reported that participants proposed longer prison sentences for an offender when they focused on unusual behaviors of the victim, and shorter sentences when they focused on unusual behaviors of the offender. Similar results have been obtained by studies involving civil cases. For instance, Wiener et al. (1994) found that determinations of negligence were related to mock jurors’ ability to mutate the negligent act, which in turn was related to perceptions of the abnormality of the defendant’s behavior. Antecedent abnormality may influence mock jurors’ award compensations such that compensations are higher when a negative event is experienced after uncommon rather than common circumstances (Macrae 1992; Macrae and Milne 1992; Miller and McFarland 1986). Macrae et al. (1993) reported that the availability of upward counterfactuals suggesting how things could have been better were associated with recommendations for more severe punishment of a hypothetical offender. Finally, Bothwell and Duhon (1994) demonstrated that compensation awards to the plaintiff were lower if mock jurors took the perspective of the

Prisoners’ counterfactual thinking

167

plaintiff rather than the defendant when imagining how the event could have been avoided. Critique of past research Although advances in this research area have been made, there are also important limitations worth noting. First, most studies have used hypothetical cases, and virtually all involved events whose descriptions were highly simplified. Real crime and victimization tend to be much more complex than those depicted in brief vignettes in terms of, for example, the intentions, motivations, and capabilities of offenders and the emotional and physical harm experienced by victims. Because researchers have rarely manipulated such contextual factors it is unclear how they influence the counterfactual thoughts of offenders and victims. Second, researchers have mostly studied counterfactual thinking in the context of serious crimes against the person, especially rape. However, most crimes are less serious and are committed against property (e.g., theft or burglary). Thus, it is unclear whether the counterfactual research findings involving crimes against the person would generalize to crimes against property. For instance, counterfactual thoughts in the context of crimes against the person may be more likely to focus on other people as a self-serving strategy of deflecting blame than in the context of crimes against property. Third, participants in most studies are university students who are required to merely imagine being the offender, victim, judge, or juror. Whereas real offenders and victims are likely to be deeply influenced by their experiences of criminality and victimization, respectively, and to be highly involved in their cases, student participants have no personal involvement in the cases presented to them. Moreover, whereas real judges and jurors are likely to feel accountable for their decisions, and, in addition, judges’ views may be influenced by their training and experience, mock judges and jurors will tend to be lacking in these features. Finally, in those studies where participants were instructed to imagine being a judge or juror, they were not given the type of guidelines that real judges and jurors would be required to consider in the course of making judgments about a case. These guidelines are important because they set constraints on the nature of admissible evidence (e.g., a prior criminal record may be ruled inadmissible), on judgment procedures (e.g., in a criminal trial there is a presumption of innocence), and on the range of possible judgments that can be rendered (e.g., lack of intention to kill rules out a conviction for murder but not manslaughter). Moreover, mock judges or jurors are not provided with response formats that are representative of how real judges and jurors would be required to respond, and it is unclear how such measures would be translated into legal decisions. For example, even a severe judgment of blame provided on a rating scale would not necessarily be translated into a guilty verdict. Given the threats to external validity that we have highlighted, the

168

Dhami, Mandel, and Souza

question remains: how, if at all, does counterfactual thinking impact the judgments and feelings of real offenders, victims, and legal decision makers when they consider real criminal or civil cases? The aim of our research program has been to examine the role of counterfactual thinking in prisoners with a view to testing and expanding theories of counterfactual thinking in a real-world, high-stakes context, and with a view to revealing how prisons can enhance their rehabilitative functions. Employing both quantitative and qualitative methodologies, we have examined the role of counterfactual thinking in prisoners’ attributions of blame, their feelings of shame, guilt, and anger, and their perceptions of fairness.

Upward counterfactual thinking in prisoners Like social comparisons, counterfactual comparisons are directional. Upward counterfactuals bring to mind possible worlds that are better than reality, whereas downward counterfactuals bring to mind worse possible worlds. Although people also construct downward counterfactuals, studies indicate that upward counterfactuals are much more prevalent (e.g., Mandel 2003a; Sanna et al. 1999). Moreover, people are more likely to generate upward counterfactuals after negative events than after positive ones (e.g., Gleicher et al. 1990; Grieve et al. 1999; Markman et al. 1993). Clearly, being sentenced to imprisonment is a negative event that can have harmful consequences for an individual (Bukstel and Kilmann 1980). Hence, we thought it likely that prisoners would engage in upward counterfactual thinking about the chain of events that led to their imprisonment. Functional accounts posit that upward counterfactuals tend to prepare individuals for avoiding negative outcomes in the future (Roese 1994; Roese and Olson 1995b, 1997; see Chapter 5, Markman and McMullen, for a more recent account) by identifying conditions that would have been sufficient to prevent a negative outcome from having occurred. Accordingly, upward counterfactuals have been shown to influence judgments of causality (Wells and Gavanski 1989), preventability (Mandel and Lehman 1996), and blame (Branscombe et al. 1996; Miller and Gunasegaram 1990). Upward counterfactual thinking can also amplify negative affect (Gleicher et al. 1990; Roese and Olson 1997). According to Kahneman and Miller’s (1986) emotional amplification hypothesis, people feel worse after contemplating better possible worlds due to a contrast effect. That is, by contrasting reality to a more positive (or less negative) alternative our affective reactions to that reality are likely to become more negative. Specifically, past research has shown that upward counterfactual thinking can heighten negative emotions such as regret (Landman 1987; Zeelenberg et al. 1998d), distress (Davis et al. 1995), shame and guilt (Niedenthal et al. 1994), and disappointment and sadness (Mandel 2003a). Our research examines the relation between upward counterfactual thinking, attributions of blame and fairness, and feelings of guilt, shame, and anger.

Prisoners’ counterfactual thinking

169

Effect of thinking focus on blame, guilt, and shame Two important claims in the counterfactual literature are that counterfactual thinking influences attributions of blame and emotional reactions to outcomes. Nevertheless, there is a lack of direct evidence to support these claims. The aim of our first study (Mandel and Dhami in press) was to directly compare the effect of factual versus counterfactual thinking on prisoners’ attributions of blame and their feelings of guilt and shame. Past research has found that blame assigned to an actor was more severe when an actor’s behavior was exceptional rather than routine (e.g., Kahneman and Tversky 1982b; Macrae 1992; Macrae and Milne 1992; Miller and McFarland 1986). These studies were motivated by norm theory (Kahneman and Miller 1986), which proposes that counterfactual thinking is activated more strongly in cases where a negative outcome is preceded by an abnormal act rather than a normal one. Other studies have reported significant positive correlations between self-blame and frequency of self-implicating upward counterfactuals (e.g., Davis et al. 1996; Mandel 2003a). Moreover, as noted earlier, Branscombe et al. (1996) found that the focus of counterfactual thinking influences the severity of blame assigned to victims and offenders. Although these studies suggest an effect of counterfactual thinking on blame assignment, none manipulated whether participants were directed to think counterfactually as opposed to factually about a case and then measured the effect of that manipulation on blame assignment. Indeed, the only previous study that directly manipulated thinking focus did not find a differential effect of counterfactual versus factual thinking on a composite measure of attributional judgments (Mandel 2003b). An important goal of our research, then, was to seek direct evidence that counterfactual thinking per se influences blame. We predicted that prisoners engaged in counterfactual thinking about how they might have prevented the events leading up to their imprisonment would assign more blame to themselves than prisoners who engaged in thoughts about how they actually brought about those events. Our prediction is based on the idea that counterfactual thinking can identify a broader range of blame-relevant factors than a factual analysis of causes (Davis et al. 1996; Mandel and Lehman 1996). According to judgment dissociation theory (Mandel 2003c), upward counterfactuals tend to focus on the functional goal of identifying ways in which a negative outcome could have been prevented. These thoughts can undo outcomes not only by negating direct causes, but also by negating enabling conditions or adding in disabling conditions. This suggests that there are more ways in which an actor could have prevented an outcome than ways in which the actor could have caused it. Thus, self-implicating upward counterfactuals are likely to draw attention to blame-implicating actions. There has also been no direct test of the effect of counterfactual thinking on emotion. The emotional amplification hypothesis (Kahneman and Miller

170

Dhami, Mandel, and Souza

1986) states that negative emotion tends to be heightened by upward counterfactuals. Studies have varied factors such as normality, action, and outcome closeness, which are believed to influence counterfactual thinking (see Roese and Olson 1997; Sanna 2000). Although past findings are generally consistent with the emotional amplification hypothesis (e.g., Johnson 1986; Kahneman and Tversky 1982a; Landman 1987), none of the studies directly manipulated thinking focus. Furthermore, as others have noted (Lerner and Keltner 2000; van der Pligt et al. 1998), much of the research linking cognition and emotion has distinguished only between positive and negative emotions. We conducted a direct test of thinking focus on prisoners’ emotions, and sought to refine the emotional amplification hypothesis by using an emotion-specific approach to examine how two emotions that share the same valence – guilt and shame – may be differentially affected by counterfactual thinking. Whereas regret and disappointment, which are likely to stem from intrapersonal harm (Berndsen et al. 2004), have often been compared in the counterfactual literature (for a review, see Chapter 8, Zeelenberg and van Dijk), guilt and shame, which are likely to be triggered by interpersonal harm, have received less attention. Our study of prisoners was an ideal context within which to examine these emotions because all of these individuals were found guilty of committing some form of harm to other persons, their property, or the state. Our predictions build on studies showing that the effect of thinking on emotion is often mediated by attributional judgments such as blame assignment (Branscombe et al. 2003; Mandel 2003a; Roese and Olson 1997). For example, Zeelenberg et al. (2000a) found that the magnitude of the actor effect (i.e., the tendency usually attributed to the mediating role of counterfactual thinking for action to elicit more intense emotion than inaction) was predicted by the degree to which active versus passive actors were assigned responsibility for the outcomes. Similarly, we predicted that the effect of thinking focus on emotion would be mediated by blame. We further hypothesized that blame would be more strongly related to guilt than to shame. Blame and guilt are both believed to be elicited by moral transgressions (e.g., Alicke 2000; Smith et al. 2002), whereas shame typically implies a painful feeling that stems from having lost the respect of others due to improper or incompetent behavior. Smith et al. (2002) found that attributions of self-blame were more likely to be inferred from literary passages referring to guilt than to shame; and guilt, but not shame, was significantly correlated with a composite measure of blame and remorse. Our key positive prediction, then, was that thinking focus would have a significant effect on guilt that, in turn, would be mediated by blame. Our key negative prediction, by contrast, was that the main effect of thinking focus on shame would be unreliable due to the latter variable’s weak relationship with blame. We sampled ninety adult male prisoners from a medium-security UK prison, who were serving an average sentence of 5.15 years for crimes includ-

Prisoners’ counterfactual thinking

171

ing burglary, drugs, and violence. Thinking focus and stage were manipulated in a fully-crossed design. The first factor was manipulated betweensubjects via instructions in the survey. Participants in the counterfactual condition were asked to think about how things might have turned out better if only they “had done something differently” or “were a different kind of person.” Participants in the factual condition were asked to think about how things turned out the way they did because of “something they had done” or “the kind of person they are.”1 Stage was manipulated withinsubjects by asking participants to complete the relevant blame, guilt, and shame rating scales first with respect to the time they were caught, then with respect to the time they were convicted, and finally with respect to the time they were sentenced. Blame Providing direct empirical support for the hypothesis that counterfactual thinking has a causal effect on blame assignment, we found a significant main effect of thinking focus on blame. On average, prisoners who were directed to think counterfactually about being caught, convicted, and sentenced assigned significantly more blame to themselves than prisoners who were directed to think factually about the same events (see Figure 10.1). We 10

Assignment of self-blame

9

8

7 Thought focus Factual Counterfactual 6

Caught

Convicted

Sentenced

Stage

Figure 10.1 Mean self-blame assigned as a function of thinking focus and stage (source: Adapted from Mandel and Dhami (in press)).

172

Dhami, Mandel, and Souza

explain the effect of thinking focus on blame in terms of the differential emphasis that counterfactual and factual thinking place on identifying selffocused means of preventing negative outcomes (Branscombe et al. 2003; Mandel and Lehman 1996; Morris et al. 1999). People may blame themselves for failing to prevent a negative outcome even if they do not view themselves as the primary cause of the outcome. In hindsight, it is easier to imagine ways in which one could have prevented a negative outcome than to imagine ways in which one caused it. Interestingly, prisoners’ self-blame was also related to the type of offence for which they were serving a sentence and to their past record with the criminal justice system. Consistent with a pattern of self-serving attributions (Miller and Ross 1975), prisoners convicted of a crime against the person (e.g., assault) assigned significantly less blame to themselves than those convicted of a crime against property (e.g., burglary). It may be easier to deflect blame on to a visible victim than on to a piece of property or its unseen owner. We also found that prisoners who had been previously tried for an offence reported feeling significantly more blameworthy than prisoners who had not been tried previously. This finding is compatible with Kelley’s (1967) ANOVA model of attribution, which posits that attributions to an actor will be more likely when the actor exhibited consistent behavior in the past. Prisoners who have been tried previously have higher consistency in terms of their criminal behavior. The correlation between blame and length of sentence was nonsignificant. Guilt and shame There was a strong positive correlation (r  0.80) between prisoners’ feelings of guilt and shame. This relationship was stronger than that reported in other studies (e.g., Mandel 2003a), and may be explained by the fact that prisoners are publicly labeled and stigmatized for their actions, thus inducing feelings of shame in them (Braithwaite 1989). Both emotions were also significantly correlated with self-blame, controlling for the other emotion. However, blame was positively related to guilt (r  0.47), whereas it was negatively related to shame (r  0.26). Finally, feelings of guilt and shame were not significantly related to sentence length, time served, or prisoners’ past record. Turning to our key predictions, we found a significant main effect of thinking focus on emotion. On average, the reported intensity of emotion was greater in the counterfactual condition than in the factual condition. However, as Figure 10.2 shows, the effect of thinking focus, as we predicted, was significantly greater on prisoner’s feelings of guilt than on their feelings of shame (the interactions with stage were nonsignificant). That is, prisoners who were directed to generate self-implicating upward counterfactuals about their arrest, conviction, and sentence reported feeling significantly guiltier than prisoners who were directed to generate self-implicating factual thoughts about the stages leading up to their imprisonment.

Prisoners’ counterfactual thinking

173

7.0

Emotion intensity

6.5

6.0

5.5 Thought focus Factual Counterfactual 5.0

Guilt

Shame Emotion type

Figure 10.2 Mean emotional intensity as a function of emotion type and thinking focus (source: Adapted from Mandel and Dhami (in press)).

In support of our mediational hypothesis, we found that blame significantly predicted guilt and, as already noted, both blame and guilt were influenced by thinking focus. If our mediational hypothesis is correct, then, we should find that the predictive effect of thinking focus on guilt is significantly attenuated when blame is controlled. Indeed, this was the case. After controlling for blame, thinking focus was no longer a significant predictor of guilt. (Conversely, the predictive effect of thinking focus on blame was not mediated by guilt.) Thus, the present findings suggest that the impact of counterfactual thinking on guilt is due to the mediating role of blame. These findings cohere with recent research indicating that blame and guilt are closely related (Mandel 2003a; Smith et al. 2002), and are consistent with the findings of Zeelenberg et al. (2000a), which revealed that responsibility attributions mediated the effect of counterfactual thinking on regret. Taken together, these findings suggest that the effect of counterfactual thinking on emotion is not only due to affective contrast – emotional amplification also appears to occur via an attributional route of influence.

174

Dhami, Mandel, and Souza

Upward counterfactual thinking, fairness perceptions, and anger In our second study (Mandel and Dhami 2005), we examined the relations between upward counterfactual thinking, prisoners’ perceptions of fairness, and their feelings of anger. As noted earlier, literature suggests that upward counterfactual thinking can heighten negative emotions such as regret, guilt, distress, and sadness. However, little research has examined the relation between upward counterfactual thinking and anger. Prisoners often experience anger, and prisons typically provide counseling services to help prisoners manage their anger (see McDougall et al. 1987). Unlike selffocused emotions such as guilt, shame, and regret, anger is often directed at other people or factors, and it tends to be elicited by external attributions of responsibility for negative outcomes (Keltner et al. 1993). Consistent with this view, Mandel (2003a) found that anger was negatively correlated with perceived control and self-blame, and positively correlated with distrust of others. One objective of our second study was to explore the role of situational context as a potential moderator of the relationship between cognitive and affective variables – in this case, the upward counterfactuals and feelings of anger reported by our prisoner sample. Demonstrating a moderating effect of context on cognition and emotion, Mandel (2003a) found that whether a negative experience occurred in either an interpersonal or academic context affected the content of counterfactuals and the likelihood of counterfactual activation. Context also influenced emotion, such that negative interpersonal experiences heightened other-focused emotions (namely, anger and distrust), whereas negative academic experiences heightened self-focused emotions (namely, regret, shame, and guilt). We examined an overall context (i.e., the criminal justice system) that can be decomposed into sub-contexts or stages (e.g., committing the crime, being caught, convicted, and sentenced) that tend to unfold as a causal chain of events. Given that the control offenders have over their environment is increasingly reduced as they “progress” through the justice system, we predicted that they would be more likely to focus on themselves at early stages (i.e., when committing the crime) and that they would tend to focus on other people (e.g., police, witnesses, judge, jury, attorney) or external factors (e.g., laws) that exert greater control over outcomes at later stages of the justice process (i.e., being caught, convicted, and sentenced). Accordingly, we predicted that the relation between counterfactual thinking and anger would vary as a function of context, such that upward counterfactuals would be associated with less anger at the crime stage and with greater anger at the arrest, conviction, and sentencing stages. A second objective of our study was to examine the relation between upward counterfactual thinking and perceived fairness. As Buck and Miller (1994: 29) pointed out, “people react not only to the nature and severity of a victim’s fate but also by its perceived deservingness.” Thus, when consider-

Prisoners’ counterfactual thinking

175

ing outcomes, people not only think about whether they could have been better or worse, they also reflect on the fairness of the outcomes (distributive justice) and the fairness of the procedures used to allocate the outcomes (procedural justice). Nario-Redmond and Branscombe (1996) found that downward counterfactuals that focused on the offender in a hypothetical rape case led to feelings of justice. By contrast, upward counterfactuals about conviction and sentence may bring to mind ways in which legal procedures could have been different such that they would have led to a more favorable outcome for the defendant (viz. an acquittal and/or non-custodial or shorter sentence). Thus, we predicted that prisoners who reported upward counterfactuals about their conviction and sentence would perceive their trial and sentence as less fair than prisoners who did not report such thoughts. People tend to believe in a just world (Lerner 1980). They expect to receive what they deserve and believe that they deserved what they received. Buck and Miller (1994) found that incongruous outcomes were perceived as more unfair and undeserved than equally severe and improbable congruous outcomes. Congruity was defined as the extent to which the event was consistent with existing knowledge. Adams’s (1965) equity theory stipulates that people evaluate their outcomes via comparison (either with oneself or others) to form a sense of deservingness. Outcomes that are perceived as being below an equitable level may lead to attributions of unfairness. Therefore, we predicted that prisoners with no previous convictions would perceive their trial and sentence to be less fair than those with previous convictions because the former group has more incongruent information. To test our predictions, we surveyed approximately 500 adult male prisoners serving sentences in three (low-, medium-, and high-security) US federal prisons. The prisoners were serving an average sentence of 8.61 years for crimes including fraud and forgery, drugs, and violence. In response to closed-ended questions, prisoners indicated whether they had specific thoughts about how things could have turned out better at different stages of the justice process. The upward counterfactual thoughts listed were about committing the crime (i.e., “I should not have committed the crime or should have committed a less serious crime”), being caught (i.e., “I should not have been caught”), being convicted (i.e., “I should have entered a different plea or have been acquitted”), and being sentenced (i.e., “I should have received a noncustodial sentence or a shorter prison sentence”). Prisoners also rated how angry they had been feeling lately compared to before on a scale ranging from “much less” through “same as before” to “much more.” They also rated the perceived fairness of their trial and sentence, separately. In addition, we used open-ended questions to probe prisoners’ counterfactual thoughts. In the remainder of this subsection, we report the findings that address the aforementioned predictions. In the subsequent subsection, we summarize our study of prisoners’ open-ended responses.

176

Dhami, Mandel, and Souza

Anger Consistent with our prediction, prisoners who had upward counterfactuals about their crime felt significantly less angry than those who did not report having such thoughts. By contrast, prisoners who had upward counterfactuals about being caught, convicted, or sentenced felt significantly angrier than those who did not report having such thoughts (see Figure 10.3). These findings are consistent with the notion that anger is an emotion triggered by external attributions of responsibility (e.g., Keltner et al. 1993; Mandel 2003a) because upward counterfactual thinking was associated with more intense anger in contexts where we would expect there to be a higher degree of focus on other people or factors. Furthermore, these findings support the idea that the situational context to which counterfactual thinking refers can shape the content of these thoughts and their subsequent effects on emotion. Fairness As predicted, prisoners who reported having upward counterfactuals about being convicted perceived their trial to be significantly less fair than those who did not report such thoughts (see Figure 10.4). Similarly, prisoners who reported upward counterfactuals regarding their sentence perceived their 0.6

Mean intensity of anger

0.4

0.2

0.0

0.2

0.4 Had upward counterfactual Not had upward counterfactual

0.6

0.8

Crime

Caught

Convicted

Sentenced

Stage

Figure 10.3 Mean intensity of anger as a function of upward counterfactuals and stage (source: Adapted from Mandel and Dhami (2005)).

Prisoners’ counterfactual thinking

177

5

Perception of fairness

4

3

2 Had upward counterfactual Not had upward counterfactual 1

Trial

Sentence Stage

Figure 10.4 Mean perception of fairness as a function of upward counterfactuals about trial and sentence (source: Adapted from Mandel and Dhami (2005)).

sentence to be less fair than those who did not report such thoughts. Although our findings are correlational, they are consistent with the hypothesis that thinking about how one might have entered a different plea or have been acquitted and how one might have received a noncustodial or shorter sentence can lead to perceptions of unfairness. Consistent with Buck and Miller’s (1994) “incongruence hypothesis,” we also found that prisoners who had previous convictions rated both their trial and sentence as significantly fairer than prisoners who had no previous convictions. Content of prisoners’ upward counterfactual thoughts As noted earlier, the prisoners in our US sample (Mandel and Dhami 2005) responded to open-ended questions asking “How could things have turned out better?” and “What do you think could have made your trial (and sentence) fairer?” We coded prisoners’ initially listed counterfactuals for mutability features and we coded the semantic content of prisoners’ fairness-related responses. Some features of reality are more mutable than others (Kahneman and Miller 1986). For instance, research suggests that people are more likely to

178 Dhami, Mandel, and Souza mutate their own behaviors than those of others (e.g., Davis et al. 1996). Studies on the effect of controllability on counterfactual thinking suggest that people most often mutate features of their own behavior rather than their character (e.g., Mandel and Lehman 1996). There is also evidence that after a negative event people are more likely to generate counterfactuals that specify the addition of an antecedent that was not present in reality rather than the subtraction of an antecedent that was present (e.g., Roese and Olson 1993). We found that the content of prisoners’ upward counterfactual thoughts generally coincided with that reported in previous research. Forty-three percent of the prisoners’ upward counterfactuals focused on the self. Of these self-focused counterfactuals, 80 percent had a behavioral focus. Finally, in line with Roese and Olson’s (1993) earlier findings, the majority (63 percent) of prisoners’ counterfactuals were additive. As Table 10.1 shows, when asked about how things could have turned out better, prisoners thought that the criminal justice system could have been better in terms of them having a more effective defense lawyer, a fairer judge, and an unbiased jury. Prisoners also thought they should have received a noncustodial sentence or a shorter prison sentence. Some prisoners believed that they could have been better people, for instance, by having an education, employment, not abusing drugs and alcohol, and having more strength of character. Other prisoners focused on their crime by suggesting that they should not have committed the crime as it occurred, they could have committed a less serious crime, or they should not have been caught. Finally, some prisoners reported that they could have had better relationships with their family and friends. Consideration of how their trial could have been fairer led prisoners to report upward counterfactuals focused on how the criminal justice system could have been less corrupt and unbiased in terms of prejudice and discrimination, the witnesses not telling lies, the police not setting up the prisoner, and the police and prosecutors not threatening or intimidating the prisoner (see Table 10.1). By contrast, prisoners’ counterfactual thoughts concerning how their sentence could have been fairer centered primarily on them receiving a less punitive sentence (i.e., noncustodial sentence or shorter sentence) or a sentence that was proportionate to the seriousness of the offense, to the responsibility of the offender, or congruent with sentences for other similar offenders and crimes (see Table 10.1). In addition, prisoners tended to report that their trial and sentence could have been fairer if there were better people working in the justice system, such as more competent and effective defense lawyers, more reasonable and unbiased judges, and more representative juries. They also tended to report that their trial and sentence could have been fairer if there were better laws, guidelines, and procedures. For instance, prisoners thought that only factual (rather than circumstantial) evidence should have been presented in court, that mitigating circumstances should have been taken into account, that

20 19 15 14 11 21 26 26 18 30 42 24 11 23

Better justice system Shorter/non-custodial sentence Better self Crime-related factor Better relationships with family/friends Other

Less corruption/bias Better justice system Better laws/guidelines/procedures Other

Sentence-related factor Better laws/guidelines/procedures Better justice system

Other

Source: Adapted from Mandel and Dhami (2005).

Prisoners (%)

Content category

Sentence could have been fairer if . . . “If ran the state and federal charges concurrent” “I got less time for a first offence” “A judge who is trying to get elected to a higher position and is tough on crime for his election benefit is unfair. If only I had a more reasonable judge”

Trial could have been fairer if . . . “There was no perjury by the prosecution’s expert witness” “I had a better lawyer. I had a court appointed one” “If due process of law was followed”

Things could have turned out better if . . . “The prosecution would act justly instead of manipulating people for convictions” “I could have been released to do work in the community instead of prison time” “I didn’t start doing drugs” “I had just broken his legs” “I had a better family life as a child”

Examples

Table 10.1 Prisoners’ counterfactuals about how things could have turned out better and fairer

180

Dhami, Mandel, and Souza

there should not have been any plea bargains, and that certain laws should have been different, such as the mandatory minimum sentencing guidelines. The content of prisoners’ upward counterfactuals, therefore, coincided with theorists’ characterizations of distributive and procedural justice (e.g., Adams 1965; Leventhal 1980). Adams (1965) stated that distributive justice includes features such as equity. A just outcome is expected in light of the perceived input, and inequity refers to any inequality between one input–output ratio and another input–output ratio. According to the prisoners in our sample, a fairer sentence would have been proportionate to the seriousness of the offence and the responsibility of the offender, and should have been similar to that of other offenders who were convicted of similar crimes. Leventhal (1980) lists features of procedural justice that were mentioned by prisoners in our sample. For instance, prisoners reported that judges and jurors should have been unbiased, jurors should have been more representative, criminal justice agents should have behaved ethically (e.g., by not lying), legal professionals should have relied on valid information and well informed opinion, and that sentencing decisions should have been consistent across offenders. Finally, although few prisoners said that their conviction or sentence should be reviewed, revised, or reversed in response to the questions that probed their counterfactual thoughts, many did mention this “correctability” feature of procedural justice to us when we conducted the research.

Implications The study of counterfactual thinking Our research on counterfactual thinking in prisoners overcomes some of the methodological limitations of past research on counterfactual thinking in the legal domain. There is some evidence that counterfactual availability is heightened by involvement (e.g., Macrae and Milne 1992), and that the generation of counterfactuals may be strongly influenced by internal states of the individual such as intention and motivation (e.g., Catellani and Milesi 2001). The prisoners in our studies were directly involved in the criminal justice system, and were asked to consider real, and often complex, events that were personally relevant to them such as committing the crime, being caught, convicted, and sentenced to imprisonment. We studied crimes that varied in severity, and compared the effect of crimes against the person and other crimes, as well as the effect of having previous convictions or not, on measures of cognition and emotion. Finally, we examined counterfactual thinking at different stages of the justice process. The findings of our research contribute to theoretical developments in the area of counterfactual thinking in several ways. First, we demonstrated that counterfactual thinking has a stronger effect than factual thinking on the assignment of self-blame and on feelings of guilt. Second, we found that the

Prisoners’ counterfactual thinking

181

relation between counterfactual thinking and emotion is mediated by attribution. Third, we showed that the context in which counterfactual thinking occurs can moderate the relation between counterfactual thinking and emotion (specifically, anger). Finally, we demonstrated that the availability of counterfactual thoughts is related to perceptions of fairness. Our research extended the range of dependent variables examined in counterfactual thinking research to include anger and fairness. Future research could examine the relation between counterfactual thinking and perceptions of fairness in such a manner that clearly distinguishes between distributive and procedural fairness (e.g., Folger and Kass 2000). From our research (see also Mandel 2003a), it is clear that context can have important moderating effects on the interplay of cognition and emotion, and future research is needed to more fully explore the effect of context on cognition–emotion interactions. Researchers could also manipulate factors that are relevant to the study of criminal justice processes such as the intentions, motivations, and capabilities of the offender, and the harm experienced by the victim. Future research could also examine how legal guidelines and procedures constrain the content of counterfactual thoughts and the effect of thinking focus manipulations on judgments made by others involved in the criminal justice system such as victims, judges, and juries. Crime control and rehabilitation Our research on prisoners’ counterfactual thinking has also addressed several issues pertaining to the effectiveness of prison sentences. Imprisonment has several functions (Mathiesen 1990). As a method of crime control, it is theorized to work in at least two ways. First, severe sanctions such as long prison sentences are meant to deter people from committing premeditated crimes. Second, prisons are meant to provide offenders with an opportunity for reeducation and rehabilitation leading to effective reintegration into society and reduced rates of recidivism. Beyond these functions, imprisonment may also serve a retributive function by punishing offenders for their crimes. Our findings suggest that some of the objectives of imprisonment are not being effectively met. First, we did not find a significant correlation between length of sentence and either the intensity of guilt or shame, or the degree of self-blame, experienced by prisoners. This is contrary to the belief underlying the retributive function of imprisonment that punishment involves inducing negative self-attributions and emotions in prisoners. Second, we did not find a significant correlation between the duration of time served by prisoners and either the intensity of guilt or shame, or the degree of selfblame, experienced by prisoners. This is contrary to the notion that prisons can rehabilitate offenders by providing them with the time and means for reflecting on their moral wrongdoings and acknowledging their responsibility for the offense. Finally, contrary to the notion that prisons act as

182

Dhami, Mandel, and Souza

specific deterrents to crime, 54 percent of the UK prisoners in our research had been previously convicted and 48 percent had served a prison sentence in the past for over eight years on average. Similarly, 60 percent of the US prisoners in our research had at least one previous conviction, and they had served, on average, one prison sentence in the past. There is a need to find methods of crime control that are more effective than incarcerating offenders for long periods of time at high economic and social costs to society (Haney and Zimbardo 1998). In the meantime, it may be worth while exploring how counterfactual thinking can contribute to offender rehabilitation. For instance, parole boards require that an offender acknowledges guilt and accepts blame for the offense. Our research suggests that prison programs designed to stimulate and explore prisoners’ upward counterfactual thoughts about their crime, arrest, conviction, and sentence may increase prisoners’ attributions of self-blame, and enhance their feelings of guilt. Upward counterfactual thinking can function to help people gain mastery over their environment by avoiding the recurrence of negative outcomes and by increasing the chances of achieving positive outcomes in the future (e.g., Roese and Olson 1997). It is important, however, to recognize that such thoughts may lead prisoners to either abstain from reoffending or to be more effective at avoiding capture and penalties in the future. For instance, a prisoner told us that “I should have taken a second [mortgage] on my house instead of taking up bank robbery.” By contrast, another prisoner said that things could have turned out better if “I’d worn a mask.” While upward counterfactual thinking may be functional for the individual prisoner, it may, in some cases, be dysfunctional for society. The consequences of counterfactual escapes from reality ultimately depend on the chosen escape route: some counterfactual escapes might facilitate lawful reintegration, whereas others might facilitate escaping detection for future crimes.

Notes Preparation of this chapter was facilitated by a City University (London) Research Fellowship and research grant SG-34166 from the British Academy to M.K.D. and by research grant 249537-02 from the Natural Sciences and Engineering Council of Canada to D.R.M. We thank Lucia Bianco and Katy Sothmann for their research assistance. 1 As the reader may have noted, we also manipulated the content focus of prisoners’ sentence completion stems so that they either focused on a behavioral aspect of self or a characterological aspect. Niedenthal et al. (1994) proposed and found partial support for the idea that guilt is more likely to be influenced by behaviorfocused upward counterfactuals than by character-focused upward counterfactuals, and that the opposite was true for shame. In the present study, however, we did not find support for this hypothesis. The content focus by emotion type interaction effect was nonsignificant and, moreover, content focus did not interact with thinking focus or stage. Therefore, we do not discuss this factor further in this chapter.

11 When the social context frames the case Counterfactuals in the courtroom Patrizia Catellani and Patrizia Milesi

“If only the victim had not actively cooperated, the perpetrator could not have taken her jeans off and the rape would not have happened.” A counterfactual thought of this type is implicitly referred to in the written motivation of an acquittal verdict given by the Italian Supreme Court (Corte di Cassazione) some years ago. It was a case of “acquaintance rape,” in which a young woman was raped by her instructor during a driving lesson. In the verdict, the judges pointed out that: “it is almost impossible to take off another person’s jeans without the person’s active cooperation, because it is a difficult action even for the person who wears them” (Italian Supreme Court, section III, 6 November 1998–10 February 1999, no. 1636). The verdict, which reversed a previous verdict of guilt, caused great sensation in Italy along with several protests from the feminist movement. The above verdict is just one of several possible examples of the use of counterfactual thinking in judicial decision making. It shows how interpretation and evaluation of judicial cases are heavily influenced not only by considerations of what actually happened, but also by considerations of what might have happened “if only. . . .” Much research has shown that focusing counterfactuals on one of the actors of a judicial case is likely to increase the amount of responsibility attributed to that actor. Investigating what factors may constrain counterfactual focus in the judicial context is therefore of relevance to a better comprehension of how legal cases are interpreted and evaluated. This issue is dealt with in the present chapter. Unlike previous research, which was mainly focused on intrapersonal, context-independent constraints (for a review, see Seelau et al. 1995), we focus our attention on psychosocial, context-dependent constraints that have been less investigated so far (see also Mandel 2003a). Most empirical studies on counterfactual constraints have referred to norm theory (Kahneman and Miller 1986), according to which abnormal events are more likely to be counterfactually mutated. In these studies, attention has been mainly focused on the actor’s behavior, and abnormality has been intended as deviation from routine behavior. According to this approach, the actor’s behavior is compared with the actor’s own behavioral standard, and behaviors showing low consistency with this intrapersonal norm

184

P. Catellani and P. Milesi

are more likely to be counterfactually mutated. However, in a socially embedded context like the judicial one, the actor of an event is likely to be perceived not only as an individual but also as a member of a social category (e.g., a woman, an old person, a gipsy). Consequently, the actor’s behavior is likely to be compared not so much with the actor’s own behavioral standard as with a social category’s behavioral standard; namely, with a social norm. Behaviors that do not conform to that social norm are more likely to be counterfactually mutated. Thus, social or stereotype-based norms that are evoked by the actors’ social memberships are likely to form a relevant category of psychosocial constraints influencing counterfactual mutability. Other psychosocial constraints that are likely to influence counterfactual mutability are related not so much to the context of the event as to the context in which the event is interpreted. In particular, roles played by people when interpreting the event (e.g., perpetrator, victim, attorney, juror), together with role-related expectations and aims (e.g., defense, accusation, neutrality) may have an influence on counterfactual mutability. On the one hand, these factors may induce people who generate counterfactuals to focus attention on some antecedents of the event instead of others. On the other hand, they may induce people who are exposed to counterfactuals generated by others to take them into account or not in their own interpretation of the event. In consideration of the above points, in this chapter we propose a Social Context Model of Counterfactual Constraints, according to which two categories of psychosocial constraints influencing counterfactual mutability may be envisaged, one related to the social context of the event and the other to the social context in which the event is interpreted. In doing this, we aim at extending previous application of norm theory to the study of counterfactual thinking, showing how norms triggered by the social context can influence counterfactuals and, as a consequence, social judgment. After starting the chapter with a brief survey of studies showing the links between counterfactual thinking and responsibility attribution in the judicial context, we focus our attention on studies investigating the so-called exceptionality effect; namely, how counterfactual mutability may be constrained by abnormality in the sense of violation of intrapersonal norms. We then devote the major part of the chapter to the two categories of psychosocial constraints envisaged by our Social Context Model of Counterfactual Constraints. First, we present recent research results supporting the existence and the strength of a nonconformity effect, according to which, under given circumstances, people would be especially inclined to focus counterfactuals on actors’ behaviors that do not conform to social norms. Then, our attention shifts from the context of the event to the context in which the event is reconstructed. We offer empirical evidence of how role-related expectations and aims of people reconstructing the event may affect counterfactual mutability. In the final discussion, we suggest that the Social Context Model of Counterfactual Constraints might be usefully extended

Counterfactuals in the courtroom

185

from the judicial context to other real-life contexts in which counterfactual thinking is widely employed.

Counterfactual focus and evaluation of judicial cases Several studies have shown the existence of a link between counterfactual thinking and evaluation of judicial cases (inter alia Bothwell and Duhon 1994; Branscombe et al. 1996; Nario-Redmond and Branscombe 1996; Wiener et al. 1994). In these studies, participants are usually told to think how a crime episode might have turned out differently and to complete open-ended counterfactual stems. Some studies have demonstrated that, when counterfactual alternatives to the negative outcome are readily available, participants feel greater sympathy towards the victim, envisage a more severe punishment for the perpetrator, and judge the case as more serious than when such alternatives are not so readily available (e.g., Macrae and Milne 1992; Macrae et al. 1993; see also Miller and McFarland 1986). Other studies have established that thinking about how a crime episode might have had a better outcome (upward counterfactual) if only an actor had behaved differently is likely to increase the degree of responsibility attributed to that actor. This means that the amount of blame assigned to the protagonists of a judicial case may depend on whose behaviors counterfactuals are focused on: more blame is usually attributed to the protagonist (either the victim or the perpetrator) who might have behaved differently and improved the outcome. For example, it has been shown that in a rape case the amount of blame attributed to the victim increases as the number of counterfactuals focused on the victim increases (Nario-Redmond and Branscombe 1996; see also Branscombe et al. 1996). Similar results have been observed in a variety of judicial cases: car accident cases (Branscombe et al. 1996), negligence and burglary cases (inter alia Branscombe et al. 1993; Wiener et al. 1994). Furthermore, the link between counterfactual focus and responsibility attribution has been observed both when participants generate counterfactuals on their own and when they listen to counterfactuals presented by an attorney (NarioRedmond and Branscombe 1996). All the above studies (and others that will be discussed in the following sections) suggest that judicial events are not judged in isolation. Rather, event interpretation and responsibility attribution are influenced not only by considerations of what actually happened but also by considerations of what might have happened “if only. . .” That is, these interpretations and attributions depend on what counterfactual alternatives are used as a comparison.

186

P. Catellani and P. Milesi

The influence of intrapersonal norms: the exceptionality effect In consideration of the observed link between counterfactual focus and responsibility attribution in the judicial context, increasing our knowledge of what behaviors are more likely to be focused on may be of some relevance. According to norm theory, event antecedents perceived as abnormal or exceptional are more likely to be counterfactually mutated. In their classic experiment, Kahneman and Tversky (1982b) presented participants with the story of Mr Jones, who died while driving home because a truck driver failed to stop at a red light and crashed into his car. Participants were asked to complete open-ended counterfactual stems (“if only . . .”) from the perspective of the Jones family and their friends. Over 80 percent of the participants were more likely to mutate an exceptional event, such as the victim leaving work earlier than usual, than a routine event, such as the victim taking the usual route home (the exceptionality effect). Thus, in Kahneman and Tversky’s work, and in subsequent studies inspired by that work (inter alia Bouts et al. 1992; Gavanski and Wells 1989; Klauer et al. 1995; Wells et al. 1987), abnormality has been interpreted as low consistency of the actor’s behavior compared with the actor’s own behavioral standard (an intrapersonal norm). In a similar vein, some studies carried out in the judicial context have also focused on abnormality intended as a violation of an intrapersonal norm, and have verified its consequences in terms of emotional response and responsibility attribution (Macrae and Milne 1992; Miller and McFarland 1986; Turley et al. 1995). These studies are based on the hypothesis of emotional amplification, according to which “affective response to an event is enhanced if its causes are abnormal” (Kahneman and Miller 1986: 145). It is assumed that people, faced with a judicial case including abnormal behaviors, are more likely to generate counterfactuals and, therefore, to exhibit extreme emotional and judgmental responses. In these studies, two groups of mock jurors were presented with two scenarios that differed only for one feature in the behavior of one of the actors, which was either consistent or inconsistent with the actor’s routine behavior. For example, in a study by Turley et al. (1995: Study 3) one group of participants were presented with a scenario of a rape case in which a woman had been raped in a health club parking lot at night after going, as usual, to her aerobics class. Another group of participants were presented with the same scenario, except for the fact that the victim that night had unusually gone to aerobics class. Participants believed that the rape victim would feel more responsible and would experience greater regret when the rape was preceded by the victim’s exceptional behavior. Several studies dealing with a variety of judicial cases have shown that high availability of counterfactual alternatives for exceptional antecedents in a crime episode affects a whole range of incident-related judgments. Among

Counterfactuals in the courtroom

187

others, in the case of a man who was mugged while walking home, Macrae et al. (1993: Study 2) found that participants recommended a harsher punishment for the perpetrator, rated the crime as more serious, and felt greater sympathy for the victim when the incident was preceded by exceptional circumstances (i.e., the man had taken a new route home) than by routine circumstances (i.e., the man had taken his regular route home). Thus, when an event has been preceded by exceptional antecedents, judgments tend to become extreme. This may give rise to paradoxical consequences: the victim of a crime may receive higher compensation than another victim of a similar crime only because the former was victimized in exceptional rather than routine circumstances. For example, a man who is shot on a robbery in a convenience store and loses the use of his right arm receives higher compensation by participants who are told that the man rarely frequents that convenience store than by participants who are told that he is a regular customer (Miller and McFarland 1986: Study 1; see also Macrae and Milne 1992).

The influence of social norms: the nonconformity effect The studies reported so far have investigated the effect of perceived abnormality intended as deviation from the actor’s routine behavior. In real life, however, the actor of an event is often perceived not just as an individual, but also as a member of a social category (e.g., a woman, a student, or a White person). Hence, actors’ behaviors may be compared not only with the actors’ own behavioral standards (i.e., their routine behaviors), but also with the perceived behavioral standards of the social categories evoked by the actors. In this case, the reference norm is not rooted in the past frequency of an individual’s given behavior, but in the perceived frequency of that behavior in a specific social category. Psychosocial research has clearly shown that behaviors perceived as frequent or normal within a given community are likely to evolve into social norms, behaviors perceived as right and proper for that community (see Thibaut and Kelley 1959). In other words, normal behaviors are likely to become normative behaviors. This entails that when the actor of an event evokes a social norm people may be more likely to generate counterfactual scenarios in which observed behaviors are substituted by expected normative behaviors. Let us consider the following case. A woman who usually goes to work by train decides to go by car for a change. Her car has a breakdown and she accepts a lift from a male stranger who afterwards rapes her. The victim’s behavior is likely to be compared with her routine behavior (i.e., “if only she had taken the train”). However, the victim’s behavior is also likely to be compared with the standard behavior of a (nonraped) woman, which implies not accepting a lift from a stranger (i.e., “if only she had not accepted a lift from a stranger”). Something similar happens when people explain intergroup differences

188

P. Catellani and P. Milesi

(Hegarty and Pratto 2001). For example, when people are asked to explain gender differences in voting behavior, they first think of a typical voter and call to mind a male exemplar. Consequently, male voters constitute the normative group, while female voters constitute the nonnormative group, whose behaviors need to be explained (Miller et al. 1991). Thus, behaviors of the nonnormative group attract people’s attention and are compared with the ones of the normative group assumed as a standard reference. In a similar vein, McGill (1993) has observed that people asked to explain why a woman is not successful in a typically male task (e.g., shooting a pool) tend to compare the woman with successful men in the same task, while people asked to explain why a man is not successful in a typically female task (e.g., sewing) tend to compare the man not with successful women but, still, with successful men. According to McGill, this is because people are more likely to generate counterfactuals having men as protagonists instead of women, because men are treated as a default reference group (see also McGill 2000; McGill and Klein 1995). Contrasting the observed behavior with the expected normative behavior may be even more likely to occur in an evaluative context such as the judicial one. However, a fundamental legal principle states: “all men are equal before the law.” In principle, in the judicial context the observed behavior should be compared not with one that is normative of a given social category, but with one that is normative for everybody. Actually, in a number of lawsuits (e.g., tort, criminal, or discrimination) courts have relied upon the “reasonable person” standard (Keeton et al. 1984). For example, in a negligence suit involving a person who is blind, jurors were told to expect that person to “take the precautions . . . which the ordinary reasonable person would take if he were blind” (Keeton et al. 1984: 174). The behavior of the blind person is evaluated for its conformity to the reasonable person standard, and subsequent sanctions or compensation is established accordingly. Referring to what the average person may do in a given context brings up the point of how some contexts might evoke behavioral norms per se; that is, norms that might be perceived as valid across different social categories. Some evidence in this regard may be found in a recent study by Mandel (2003a; see also Chapter 10, Dhami et al.). Participants asked to generate counterfactuals about a negative event they experienced either in an academic context (e.g., failing an exam) or in an interpersonal context (e.g., having problems with a friend) have generated more self-focused counterfactuals in the first case than in the second one. According to Mandel, the difference is because the academic context evokes a norm of personal responsibility and control, while the interpersonal context evokes norms of shared responsibility and reciprocity; in generating counterfactuals, participants would therefore have referred to context-related norms. In many cases, however, a given context is likely to evoke not only generic context-related norms, but also more specific social norms related to the different categories of people who may act in that context. For example,

Counterfactuals in the courtroom

189

the academic context may evoke a generic norm of personal responsibility but also social norms regarding the different degree of personal responsibility of female versus male students, of younger versus older students, of psychology versus law students, and so on. If these social categories are made salient by the context, the observed behavior is likely to be compared with the expected behavior in the salient contrast category. Similarly, the judicial context may evoke a general (although very strong) norm of personal responsibility, but is also likely to evoke more specific social norms related to the social categories active in the context. For example, previous research has shown that people have different expectancies regarding crimes that may be committed by Black people as compared with White people (Gordon 1990). As we will see in more detail in the following section, previous research has also shown that people have different expectancies regarding behaviors of male crime victims as compared to female crime victims. Thus, when interpreting events in a social context perceived abnormality of the actors’ behaviors may regard not only behaviors that are inconsistent with what the actor is used to do (an intrapersonal norm), but also behaviors that do not conform to behavioral standards of a contrast social category (a social norm). Accordingly, besides the much studied exceptionality effect mentioned above, the existence of a nonconformity effect (Catellani et al. 2004) may be envisaged, according to which behaviors that do not conform to a relevant social norm are perceived as more likely to be mentally mutated in counterfactuals than behaviors that conform to that norm. Endorsement of social or stereotype-based norms may vary from one person to another. Hence, the strength of the nonconformity effect will depend on the degree of the perceiver’s endorsement of social norms that are relevant in a given context, being highest in case of high social norm endorsement. For this reason, research aimed at assessing the nonconformity effect should also include measures of social norm or stereotype endorsement. The presence of a nonconformity effect in counterfactual thinking is likely to have a strong influence on the attribution of responsibility and blame. As research on attribution processes by Jones and McGillis (1976) clearly highlighted, the actor’s behaviors that do not conform to our expectancies are diagnostic of the actor’s dispositions and are therefore likely to orient our evaluation of the actor. Similar to our argument in this chapter, in Jones and McGillis’s approach expectancies are based not only on information regarding the specific actor (target-based expectancies) but also on information regarding the target social category (category-based expectancies). Social norms in rape cases Although previous research has suggested that “stereotype-inconsistent” behaviors might especially evoke counterfactual thoughts (Branscombe and Weir 1992; Branscombe et al. 1993; Hegarty and Pratto 2001), only recent

190

P. Catellani and P. Milesi

studies (Catellani and Milesi 2001, 2004; Catellani et al. 2004) have offered empirical evidence of the influence of social or stereotype-based norms on counterfactual mutability. These studies have focused attention on rape cases and in particular on how rape victims’ behaviors are likely to be counterfactually mutated. This is because there are strong stereotypes about rape victims (the so-called rape myth; Burt 1980) and these, by contrast, prescribe what a woman should or should not do to avoid rape. For example, according to stereotype-based norms, a woman should not engage in a host of risky behaviors, such as accepting lifts from strangers (Acock and Ireland 1983), walking late at night (Pallak and Davies 1982), or drinking on her own in a pub (Krahè 1988). At the same time, a woman is expected to assume a number of preventive behaviors, such as trying to escape and opposing appropriate dissent and resistance (e.g., saying “no” to unwanted advances and fighting back, Howard 1984), in order to avoid being raped. In fact, such a rich set of prescriptions has been shown to direct people’s attention in rape cases to the victim rather than to the perpetrator, contrary to what happens in most other judicial cases (inter alia, Atkinson and Drew 1979; Borgida and Brekke 1985). Consistently, in a study on counterfactual thinking in two different judicial cases, Catellani and Milesi (2001) have shown that in a rape case counterfactual focus is more on the victim than on the perpetrator, while it is more on the perpetrator than on the victim in an assault case involving two men. This result has been explained by the fact that the rape victim’s behavior is compared with a consolidated set of social norms, while the same does not hold for victims of different crimes. Generation and evaluation of explicit counterfactuals Further studies (Catellani et al. 2004) have investigated the nonconformity effect in a more direct way, assessing whether people with higher endorsement of the rape victim stereotype are especially inclined to focus counterfactuals on the victim’s behaviors that do not conform to the stereotype-based norm. The distinction between conforming and nonconforming behaviors was based on the prototypical profile of the rape situation described by Krahè (1991) and was corroborated by the results of a pilot study where participants presented with a rape report were asked to rate how the victim’s behaviors conformed to the way a woman should behave with a stranger. Participants in the main studies were asked to play the role of mock jurors and were presented with the same rape report employed in the pilot study. The report described a case involving a woman who had a car breakdown on her way home and accepted a lift from a police officer, who afterwards raped her. The victim’s behaviors were balanced as regards their conformity versus nonconformity to stereotypic norms concerning women’s behavior. For example, one of the normative behaviors was “she got frightened when the man took off the gun and laid it aside,” while one of the non-

Counterfactuals in the courtroom

191

normative behaviors was “she was pleasant with the man.” The victim’s behaviors were also balanced as regards their being actions (e.g., “she talked freely to the man”) versus inactions (e.g., “she did not say ‘no’ clearly”). In Study 1, after reading the report participants were asked to complete open-ended counterfactual stems, starting with “The outcome might have been better, if only. . .” In Study 2, participants were asked to rate their agreement with the most frequent counterfactual statements generated in Study 1, again balanced as regards their focus on conforming versus nonconforming behaviors and on actions versus inactions. In both studies, participants’ endorsement of the rape victim stereotype was also assessed, using a scale based on Burt’s (1980) Rape Myth Acceptance Scale, reviewed by Lonsway and Fitzgerald (1995). Results of both studies confirmed the presence of a nonconformity effect in high stereotypers. These participants generated more counterfactuals focused on the victim’s nonconforming behaviors (Study 1) and showed higher agreement with the same type of counterfactuals (Study 2). Such counterfactuals were also shown to have the highest correlation ratings with the victim’s responsibility. A further result of our research is that high stereotypers focused a greater number of counterfactuals on the victim’s nonconforming inactions. That is, they focused on what the victim did not do but might have done according to stereotype-based norms regarding a woman’s behavior with a stranger (see Table 11.1). As mentioned above, stereotype-based norms regarding rape victims include a series of actions aimed at preventing an assault, for example “crying out for help.” It is, therefore, not surprising if the absence of one or more of these normative actions in a specific episode may especially catch the attention of high stereotypers, leading them to generate counterfactuals like “if only she had cried out for help.” Such enhanced counterfactual focus on inactions observed by Catellani et al. (see also Zeelenberg et al. Table 11.1 Counterfactuals focused on the rape victim’s nonconforming behaviors as a function of stereotype endorsement Stereotype endorsement Lower

Higher

Mean proportion of self-generated counterfactuals (Study 1) Actions Inactions

0.13a 0.09a

0.18a 0.29b

Mean agreement with other-generated counterfactuals (Study 2) Actions Inactions

2.87a 2.68a

4.10a 5.59b

Source: Adapted from Catellani et al. (2004). Note Means within rows not having a common subscript differ at p  0.01.

192

P. Catellani and P. Milesi

2002) becomes especially interesting, as it contrasts a general tendency often observed in previous research on counterfactual thinking; namely, the tendency to focus attention on actions more than on inactions (the so called action–inaction effect; inter alia Kahneman and Miller 1986; Landman 1987; Lundberg and Frost 1992; Miller and Taylor 1995; Zeelenberg et al. 1998b). Kahneman and Miller (1986) suggested that actions are usually perceived as more abnormal than inactions and therefore more likely to be focused on in counterfactuals. However, Catellani et al.’s research has shown that the nonconformity effect may be strong enough to moderate the action–inaction effect. When stereotype-based norms that prescribe actions are evoked, inactions may be perceived as more abnormal than actions and may therefore be more likely to be mutated in counterfactuals. In Catellani et al.’s research, counterfactuals focused on the victim’s nonconforming inactions turned out to be highly related to responsibility assigned to the rape victim. This result extends what was found by previous research on the link between counterfactuals and responsibility attribution, as it shows that this link may hold not only for action-focused counterfactuals (Turley et al. 1995), but also for inaction-focused counterfactuals. This is likely to be the case when, as for rape victims, stereotype-based expectations include the adoption of preventive behaviors aimed at avoiding the crime. Consistently, in Catellani et al.’s research perceived crime avoidability turned out to be a significant mediator of the relationship between inaction-focused counterfactuals and victim’s responsibility (see also Mandel and Lehman 1996): people thought of what the victim might have done, but did not do, to avoid the rape and this led to increased responsibility attributed to the victim. Overall, the above findings demonstrate that abnormal behaviors, in the sense of nonconforming to stereotype-based norms, may stimulate the generation of counterfactual alternatives focused on those behaviors and, consequently, may increase the attribution of responsibility to the actors. Perception and reproduction of implicit counterfactuals In Catellani et al.’s Study 2 people were presented with and asked to evaluate other-generated counterfactuals expressed in the explicit form “if . . . then.” However, in real life, counterfactuals are often conveyed implicitly, through linguistic indicators like even, at least, without, next time (Catellani and Milesi 2001; Sanna and Turley 1996; Sanna and Turley-Ames 2000). For example, the adverb “even” may convey a counterfactual focused on a person’s action. This means that a sentence like “She even talked intimately to him” implicitly hints at the counterfactual hypothesis that “if she had not talked intimately to him, things might have ended differently.” Similarly, the adverb “without” may convey a counterfactual focused on a person’s inaction. Therefore, a sentence like “She accepted the kiss without any resistance” implicitly hints at the counterfactual hypothesis that “if she had put up some resistance, things might have ended differently.”

Counterfactuals in the courtroom

193

What happens when people are presented with implicit counterfactuals? More specifically, what kind of relation exists between implicit counterfactuals “suggested” by others and counterfactuals that people generate on their own? Such an issue, which has not been previously investigated by counterfactual research, may be of some interest to research aimed at reproducing the conditions in which counterfactual thinking is employed in courtrooms. During a trial, jurors are likely to be exposed to counterfactuals more or less deliberately suggested by other protagonists of the trial. For example, attorneys have been shown to be especially inclined to employ counterfactuals as a means of influencing jurors (Conley and O’Barr 1990; Kassin et al. 1990). Catellani and Milesi (2004: Study 1) have investigated whether the nonconformity effect may still be observed when people generate counterfactuals following exposure to other-generated counterfactuals. Mock jurors with higher versus lower endorsement of the rape victim stereotype were presented with the report of a rape case that included four implicit counterfactuals focused on the victim’s nonconforming actions (e.g., “She even told him that he was nice”), and four implicit counterfactuals focused on the victim’s nonconforming inactions (e.g., “. . . without showing embarrassment”). After reading the report, participants were asked to complete openended counterfactual stems, starting with “The outcome might have been better, if only. . .” Results showed that high stereotypers were significantly more inclined than low stereotypers to reproduce counterfactuals “suggested” in the report (e.g., “. . . if only she hadn’t told him that he was nice”). Moreover, high stereotypers generated a higher number of “original” counterfactuals of the same type, that were not referred to in the report but might be inferred from it (e.g., “. . . if only she hadn’t been pleasant to him”). These findings offer a further confirmation of the presence of the nonconformity effect in counterfactual thinking. First, they show that high stereotypers pay special attention to counterfactuals suggested by others when they are focused on nonconforming behaviors, even if these counterfactuals are only implicitly conveyed. Second, they show that after exposure to these counterfactuals high stereotypers are also specially inclined to generate further counterfactuals of the same type. This suggests that, once evoked, stereotype-based norms are likely to become the main reference norm in generating counterfactuals. The studies presented in this section thus offer converging evidence that social norms may influence counterfactual thinking similarly to intrapersonal norms. In the interpretation of complex social events, the actors’ behaviors are likely to evoke and be compared to social norms. Even behaviors that per se would be unlikely to attract the perceiver’s attention and be mutated in counterfactuals (for example, the actor’s inactions), if they deviate from an evoked social norm may become the focus of perceiver’s attention and, consequently, mutable in counterfactual thinking.

194

P. Catellani and P. Milesi

The influence of the communicative context So far, our analysis of counterfactual constraints in the judicial context has focused on factors related to the context of the event being interpreted (e.g., a rape case). However, counterfactual thinking is often a social practice carried out through interpersonal communication (see Hilton 1991). Therefore, further constraining factors may be taken into account, related to the context in which counterfactual thinking is expressed (e.g., reconstruction of a rape episode during a trial). In this case what is under examination are not intrapersonal or social norms concerning people acting in the event, but other psychosocial and communication-related factors concerning people interpreting the event. What people say is usually constrained by the roles people play in a given context. Role-related expectations and goals are therefore among the factors that are likely to influence counterfactuals expressed in a given communicative context. The judicial context is particularly fit for investigating the presence of such an influence as it is characterized by higher formalization than other everyday contexts. In a trial, the participants’ roles are fixed (e.g., perpetrator, victim, attorney, juror) and so are the goals they pursue (e.g., defense, accusation, neutrality). Accordingly, strong expectations prescribe what participants should or should not say as a function of their judicial role (inter alia Atkinson and Drew 1979; Mannetti et al. 1991). Empirical evidence of the influence that the role played by people in the judicial context may have on counterfactual mutability may come from examining counterfactuals by people who give an account of a judicial case from opposite perspectives. In research by Catellani and Milesi (2001: Studies 1 and 2), male and female participants were first presented with the report of a rape case and then invited to give their own accounts of the same event, playing the role of either the female victim or the male perpetrator. Analysis of counterfactuals implicit in the participants’ accounts revealed that the number of counterfactuals focused on actions versus inactions (e.g., “if Julia hadn’t joked” versus “if Julia had put up some resistance”) and on controllable versus uncontrollable behaviors (e.g., “if Julia hadn’t been so kind” versus “if Julia hadn’t been scared”) was influenced by role-related aims. For example, the often observed tendency to focus counterfactuals on controllable instead of uncontrollable behaviors (controllable–uncontrollable effect, Davis et al. 1995; Girotto et al. 1991; Markman et al. 1995) was still visible in these studies, but it was stronger in the case of other-focused counterfactuals than in the case of self-focused counterfactuals (cf. Mandel 2003a). Role-related expectations and goals may also influence the perception and subsequent reproduction of other-generated counterfactuals. What happens, for example, when people are exposed to implicit counterfactuals generated either by the police – supposedly a trustworthy source – or by the perpetrator – supposedly an untrustworthy source? In Catellani and Milesi (2004:

Counterfactuals in the courtroom

195

Study 2), participants with higher endorsement of the rape victim stereotype were presented with the same rape report used in Catellani and Milesi’s Study 1 (see above), including implicit counterfactuals focused on the rape victim’s nonconforming behaviors. This time, however, half the participants were told that the report came from the police, while the other half were told that the report described the facts as the perpetrator had reconstructed them. After reading the report, participants were asked to play the role of mock jurors and to complete “if only . . .” sentence completion stems. In the case of the perpetrator reporting the event, participants were expected to experience a conflict regarding the norm or rule to be referred to in expressing their counterfactuals. On the one hand, reference to stereotype-based norms would induce them to reproduce “suggested” counterfactuals focused on the victim’s nonconforming behaviors. On the other hand, reference to the communicative rule according to which the source of the report is unreliable would induce them to discard the same “suggested” counterfactuals. Such a conflict is likely to evoke cognitive dissonance: the individual endorses stereotypic beliefs that are also endorsed by someone else the individual wants to keep his/her distance from. Results showed that participants faced with the untrustworthy source found a way to overcome this conflict. As compared to participants faced with the trustworthy source, they generated a significantly higher number of “original” stereotype-consistent counterfactuals, not suggested by the suspected source, but still consistent with stereotype-based norms regarding the victim (Figure 11.1). In this way, they could confirm their stereotypes without seeming to follow an untrustworthy source. The outcome is somehow disconcerting, as the fact of being exposed to stereotype-consistent counterfactuals suggested by an untrustworthy source may lead to an increase, instead of a reduction, of stereotype-based counterfactuals. This result may remind the reader of the “rebound effect” described by Macrae et al. (1994), according to which people who have been initially asked to suppress their stereotypes engage in a systematic cognitive activity that induces them to reaffirm these stereotypes even more strongly afterwards. What we have observed in our research, however, differs from the rebound effect in two respects. First, our participants were not induced to suppress stereotypes by an explicit request, but by a communicative situation evoking suspicion (see Fein and Hilton 1994). Second, our participants did not go through the two separate steps of stereotype suppression and stereotype reaffirmation. Rather, we suspect that, right from the beginning, participants engaged in systematic cognitive activity to overcome their cognitive dissonance, which led them to find a way of expressing their stereotypes without violating the rules of the communicative context. To sum up, the studies described in this section offer converging evidence that role-related goals and expectations regarding people who report on a given event (e.g., their being trustworthy versus untrustworthy) may have an influence on counterfactual thinking regarding the same event. More

196

P. Catellani and P. Milesi

60

Proportion of counterfactuals

50

40

30

20

Suggested counterfactuals Original counterfactuals

10

0

Trustworthy source Police

Untrustworthy source Perpetrator Source

Figure 11.1 Mean proportion of counterfactuals focused on the rape victim’s nonconforming behaviors as a function of source trustworthiness (source: Adapted from Catellani and Milesi (2004)).

generally, they suggest that counterfactual thinking may be influenced by psychosocial factors related to the context in which the interpretation of the event takes place.

Conclusion In the present chapter, we have examined the factors that influence counterfactual mutability in a socially embedded context such as the judicial one. While previous research has mostly focused attention on context-independent, intrapersonal constraints, we have provided evidence in support of a Social Context Model of Counterfactual Constraints. According to our model, two categories of psychosocial constraints may play a relevant role in counterfactual mutability: one related to the context of the event that is interpreted, the other related to the context in which the interpretation of the event takes place. Although the research presented in this chapter has been carried out in the judicial context, we argue that the two categories of psychosocial constraints are also likely to play a role in other socially embedded contexts.

Counterfactuals in the courtroom

197

As for the psychosocial constraints regarding the context of the event, our model suggests that the social categories evoked by the actors of an event are likely to have an influence on the kind of norm referred to in counterfactuals. The assumption is that the study of counterfactual thinking in applied real-life contexts requires a wider conceptualization of norms – and, therefore, of abnormality – than the one often implied by previous research (see Hilton 2001). Behaviors perceived as abnormal may be not only those deviating from intrapersonal norms, but also those deviating from social norms. Consistently, we have reported several research results supporting the existence and strength of a nonconformity effect, according to which behaviors perceived as nonconforming to social norms may be especially likely to be mentally undone. Observed results may be summed up in the following six points. First, the nonconformity effect is more likely to manifest itself when people generate counterfactuals about scripted events for which a well developed set of social behavioral prescriptions is available. Second, the nonconformity effect is stronger in people who endorse the social norms that are relevant in the context of the event. Third, the nonconformity effect may be found in the generation of counterfactuals, as well as in the perception and evaluation of counterfactuals generated by other individuals. Fourth, the nonconformity effect has a significant impact on the evaluation of the event, given that counterfactuals focused on an actor’s nonconforming behaviors are especially related to the attribution of responsibility to that actor. Fifth, the nonconformity effect may moderate the often observed action–inaction effect. When inactions deviate from stereotypic expectations regarding the actor they become especially likely to be counterfactually mutated. Sixth, the nonconformity effect does not disappear even when the source of counterfactuals is presented as unreliable. (Paradoxically, as we have seen, it may even be enhanced.) This persistence of the nonconformity effect suggests that comparing reality with its alternatives often results into a confirmation of one’s preconceptions, instead of favoring the creation of new options and a change in one’s preconceptions (see also Tetlock 1998; Chapter 12, Tetlock and Henik). The second category of psychosocial constraints implied by our Social Context Model of Counterfactual Constraints pertains to the communicative context in which the event is interpreted. In particular, the studies presented in this chapter have shown that role-related expectations and aims of people reconstructing the event influence the generation, selection, and evaluation of counterfactuals. These results may be summed up in the following three points. First, the expectations and aims of people generating counterfactuals may be so strong as to moderate two “basic” effects in counterfactual mutability, such as the action–inaction effect and the controllable–uncontrollable effect. Second, a role-related expectation, such as trustworthiness, may influence the perception of counterfactuals generated by other individuals and the subsequent generation of further

198

P. Catellani and P. Milesi

counterfactuals. Third, role-related expectations regarding the people who interpret the event may interact with expectations regarding the people who act in the event, as exemplified by the study on how the nonconformity effect may be influenced by a source’s trustworthiness. In conclusion, our Social Context Model of Counterfactual Constraints implies that counterfactual thinking is constrained not only by contextindependent, intrapersonal factors, but also by psychosocial factors that are activated by the context of the event and by the context in which the event is interpreted. Further examination of these factors and of their interaction may contribute to a better comprehension of counterfactual thinking and, therefore, of the interpretation of events in various real-life contexts.

12 Theory- versus imaginationdriven thinking about historical counterfactuals Are we prisoners of our preconceptions? Philip E. Tetlock and Erika Henik Many thoughtful observers have wondered why it is so often so difficult to reach consensus on the correct policy lessons to be drawn from history (Jervis 1976; Neustadt and May 1986; Vertzberger 1991). This chapter argues that the answer is part ontological (wired into the quirky path-dependent structure of the world we inhabit) and part psychological (wired into the operating routines and structures of the human mind). This chapter also reports a series of empirical studies that illustrate how easily historical reasoning can slip into ideologically self-serving tautology and how difficult it can be to avoid becoming prisoners of our preconceptions. The ontological argument builds on an obvious observation: there is enormous room for reasonable people to disagree over why key historical events unfolded when they did and whether they had to occur in the forms they did (Fearon 1991; Ferguson 1997; Tetlock and Belkin 1996a). Casual inspection of the voluminous historical literature on the twentieth century reveals that historians have yet to agree on whether the outbreak of World War I was the inevitable byproduct of powerful laws operating on welldefined geopolitical conditions or the fluky byproduct of an almost-botched assassination; whether World War II had to end in Nazi defeat or that ending was the fortuitous byproduct of strategic blunders by the German high command; and whether the non-occurrence of World War III was the inevitable consequence of the powerful incentives that nuclear weapons created for restraint or the byproduct of a series of lucky breaks in crises that could easily have escalated out of control. Consensus has been elusive for the simple reason that historians lack the methodological tools that laboratory scientists possess for answering questions about cause and effect (Fogel 1964; Tetlock and Belkin 1996a). Historians cannot rerun history to determine how often we would avoid a major war in the early twentieth century if we were, through some time-travel intervention, to undo a particular assassination or decision. They have to rely on counterfactual thought experiments in which they try to make as strong a case as they can, with the fragmentary evidence at their disposal, for how events would have unfolded along historical paths humanity never traveled and now never can.

200

P.E. Tetlock and E. Henik

The psychological argument also builds on some uncontroversial observations. It is well known that people, especially people who fit the cognitivestylistic profile of those high in need for closure, dislike ambiguity and try to impose structure on ambiguous stimuli by fitting them into existing schemata (Kruglanski and Webster 1996). Indeed, there is now a small warehouse of experimental evidence that attests to our collective willingness to jump too quickly to conclusions on the basis of incomplete evidence, as well as our collective reluctance to revise our opinions in response to disconfirming evidence (Gilovich et al. 2002). When we combine the ontological argument (history is a terrible teacher) with the psychological argument (we are imperfect pupils), we should expect that human observers, especially those with the strongest needs for closure, will try to escape the oppressive ambiguity that is the natural consequence of acknowledging both the multiplicity of paths that history could have taken and the fragility of claims to knowledge about what would have happened along each path. We will try to escape by filling in the missing counterfactual “control” conditions of history with elaborate stories grounded in our deepest assumptions about how the world works (Tetlock and Belkin 1996a). “Epistemic pessimists” sometimes draw an even stronger conclusion from this combination of arguments: the impossibility of our learning any lessons from history that we were not ideologically predisposed to learn in the first place (Tetlock 1999). The bulk of this chapter is devoted to presenting the results of a long program of empirical work that supports a rather pessimistic assessment of the capacity of human beings to draw ideologically dissonant or surprising conclusions from history. However, we close on a somewhat more hopeful note by exploring the possibility of using imaginative scenario exercises to pry open otherwise-closed minds and of pitting vividness/salience biases grounded in the availability heuristic against cognitive-conservatism biases grounded in theory-driven modes of thinking. Still, there is no cognitive equivalent of a “free lunch.” The evidence suggests that the price of imaginative liberation is often intellectual chaos, the undisciplined proliferation of scenarios and the inflation of subjective probabilities beyond the bounds of reason.

Evidence for the theory-driven portrait of counterfactual reasoning in history Research on how people in real-world settings construct mental representations of possible worlds has thus far focused on testing three categories of hypotheses: 1

People who believe in laws of history – be they hawks or doves, Marxists or capitalists, etc. – should reject close-call counterfactuals that undermine the application of those laws by implying that other things could

Are we prisoners of our preconceptions? 201

2

3

easily have happened. In cognitive-consistency theory terms, people should try to neutralize dissonant close-call scenarios. Building on work on cognitive style (e.g., Kruglanski and Webster 1996; Suedfeld and Tetlock 2001), we should expect people with high needs for closure to be drawn to top-down approaches to history that deduce “what had to be” from theoretical first principles, and especially intolerant of dissonant close-call counterfactuals that imply that there are no deductive certainties in history. Building on work on the logic of counterfactual thought experiments in history and other disciplines (Fearon 1991; Tetlock and Belkin 1996a), we should expect theory-driven, closure-seeking thinkers to resort to three logically distinct lines of defense against dissonant close-call counterfactuals. Consider, as an example, the possible responses to one of the most sharply debated counterfactual claims of twentieth-century history: “If the assassination of Archduke Ferdinand in June 1914 had been thwarted, then World War I would not have broken out in August of that year.”

The first defense is to challenge the mutability of the antecedent. In this example, one could argue that the assassination cannot be undone because the archduke was so detested in Sarajevo and the assassins were so determined. Most challenges to the mutability of an antecedent claim that the proposed counterfactual violates the “minimal rewrite” rule, which stipulates that as little violence as possible should be done to the historical record (Tetlock and Belkin 1996a). Few experts endorse this defense in this particular case, however, because the assassination depended on a series of such improbable coincidences of wrong turns and misjudgments that it is easily “undoable.” A second defense is to challenge the adequacy of the causal connecting principles linking the antecedent to the hypothesized consequent. In this case, experts could claim that the assassination had no significant impact on the unfolding of history because Austro-Hungary was determined to attack Serbia in the summer of 1914 in any event. This defense is also not popular in this particular case because there is no evidence of a pre-assassination intention to invade Serbia. A third defense is to concede the mutability of the antecedent and the soundness of the connecting principles, but to invoke second-order counterfactuals that put history back on track. Second-order counterfactuals allow for deviations from reality but minimize their significance by invoking additional forces that soon bring events in the (simulated) counterfactual world back toward the (observed) historical path. Thus, they “undo the undoing” of the original counterfactual. In this example, one could allow that the archduke could have escaped assassination and that war could have not broken out that summer but insist that another incident would have sparked a war in the not-distant future. Incidentally, this defense is the most popular of the three in this case.

202

P.E. Tetlock and E. Henik

To test these hypotheses, previous work needed to satisfy an array of methodological preconditions. They included reliable and valid measures of cognitive style (adaptations of the need-for-closure scale of Kruglanski and Webster 1996, described in detail in Tetlock 1998), measures of broad ideological or theoretical convictions (described in Tetlock 1998) and measures of reactions to counterfactual scenarios that are in tension with these convictions (described in Tetlock 1998, 1999; Tetlock and Lebow 2001). For example, the adapted need-for-closure scales included such items as “Even after I have made up my mind about something, I am always eager to consider a different opinion” and “I dislike questions that can be answered in many different ways.” Reactions to counterfactual scenarios were measured by responses to three questions that corresponded to the three belief-system defenses: (1) How plausible was the antecedent condition of the argument? (2) Assuming the plausibility of the antecedent, how likely was the hypothesized consequence? (3) Assuming the plausibility of the hypothesized consequence, would the effects on subsequent developments have been profound (leading to a radically different world) or insignificant (because other historical forces would have brought events back in the long run to an end state similar to the real world)? Counterfactual cognition and policy controversies One series of studies examined counterfactual reasoning in a wide range of historical domains: the Soviet Union (1920s–1980s), South Africa (1990s), World Wars I and II and the Cold War. Below we summarize the historical laws that clash in each one, the counterfactual probes we selected for their potential to provoke irritated rejection from experts who subscribed to those laws, and our principal findings. Each study’s participants, also described below, were sophisticated experts who had devoted years of study (at least fifteen years, on average) to the respective entities and topics examined. History of the Soviet Union PARTICIPANTS

Experts had doctoral degrees or were “all but dissertation” in political science, specialized in Russian history or Soviet domestic or foreign policy, and were either affiliated with major research universities and think tanks or worked for branches of the US government devoted to monitoring the Soviet Union. COMPETING HISTORICAL SCHEMAS

“Essentialist” observers of the USSR depicted the state, from its Bolshevik beginnings, as intrinsically totalitarian and wedded to the use of terror to

Are we prisoners of our preconceptions? 203 eliminate enemies. This view portrayed Stalinism as the natural outgrowth of Leninism, not an aberration, and was more popular among conservatives than liberals. Advocates of more plural conceptions of the Soviet polity saw cleavages between the more doctrinaire and reformist factions of the Party dating back to the 1920s. In this view, more popular among liberals, there was nothing foreordained about the evolution of the Soviet polity (Cohen et al. 1985). These observers also suspected that the system had some legitimacy among the population and that dissolution was not the inevitable consequence of Gorbachev’s policies of glasnost and perestroika. COUNTERFACTUAL PROBES

These broad stances have direct implications for the acceptability of specific close-call scenarios. The essentialists should see less wiggle room than the pluralists for rerouting Soviet political history after 1917 via internal-tothe-regime causal pathways. Sharp disagreements should arise over such counterfactuals as, “If the Communist Party of the Soviet Union had deposed Stalin in the late 1920s or early 1930s, the Soviet Union would have moved toward a kinder, gentler version of socialism fifty years earlier than it actually did,” “If Malenkov had prevailed in the post-Stalin succession struggle, the Cold War would have ended in the 1950s rather than the 1980s” and “If Gorbachev had been a shrewder tactician, pacing reforms differently, the Soviet Union would exist today.” Only external causes, like World War I at the beginning of the Soviet regime or the hard-line policies of the Reagan administration at the end of the regime, should be judged to have made a difference. FINDINGS

The connections between abstract beliefs and reactions to close-call scenarios were most pronounced for the more theory-grounded (and less contextspecific) belief-system defenses listed above: challenging connecting principles and invoking second-order counterfactuals. Predictability fell off precipitously for the first strategy, challenging the mutability of the antecedents. Moreover, it can be shown that this differential predictability is not due to differential reliability of measures or restriction-of-range artifacts. DISCUSSION

This pattern of results is consistent with the notion that people use more context-specific standards for judging the feasibility of tweaking history (the “if” starting point of all counterfactuals) and more abstract, portable, theorygrounded standards for judging the consequences likely to follow from the tweaking (Tetlock and Visser 2000). What does this mean operationally? In the Soviet domain, it means that

204

P.E. Tetlock and E. Henik

it is hard to get a good political brawl going over whether Lenin, Stalin or Brezhnev could have lived longer or whether Gorbachev could have been a shrewder tactician, and easy to get one going over the long-term significance of these minimal-rewrite alterations of history. All but one of the big correlations between liberalism–conservatism and counterfactual beliefs were with judgments of what would have happened once we enter the hypothetical world activated by altering antecedent conditions. Thus, conservatives were more accepting than liberals that the Bolshevik revolution could have been avoided but for the chaos of Russia’s defeat in World War I and that Soviet foreign policy in the late 1980s would not have been nearly so accommodating if Reagan had not taken so tough a stand in the early 1980s. Conversely, liberals were more likely than conservatives to believe that a kinder, gentler communism could have emerged earlier by purging Stalin or adding ten years to Lenin’s life. Liberals were also more likely to suspect that Gorbachev was critical in steering the Soviet Union down a reformist path and that Gorbachev could have held together some form of Soviet federation if he had been a shrewder tactician. Although ideological disagreement over the mutability of antecedent conditions was subdued, there was a conspicuous exception: liberals and conservatives strongly disagreed over the plausibility of the antecedent and the conditional linkage for the Stalinism counterfactual. Conservatives had a harder time than liberals imagining that the Soviet Communist Party could have purged, or would have wanted to purge, Stalin in the late 1920s or early 1930s. From an essentialist perspective, which views Stalinism as the natural next step of Leninism, the deletion-of-Stalin counterfactual violates the minimal-rewrite rule. But this counterfactual may well pass the minimal-rewrite test for those with a more pluralistic perspective on the Soviet polity (Cohen et al. 1985). Liberals and conservatives also disagreed on what would have happened if Stalin had been deposed. Like most counterfactuals, this one does not spell out the complex connecting principles necessary for bridging the logical gap between antecedent and consequent. To hold the counterfactual together, it is necessary to posit that advocates of Gorbachev-style socialism in the CPSU would have seized the reins of power and guided the Soviet state toward social democracy. Conservatives, who view the Bolshevik Party of the time as monolithically oppressive, regard such arguments as fanciful. These data are open to rival interpretations. One hypothesis is that those on the left view history as more fluid, contingent and indeterminate than those on the right. A harsher variant of the hypothesis depicts more conservative observers as more prone to hindsight bias, a cognitive failure to recall how uncertain one once was about what was going to happen before being contaminated by outcome knowledge. An alternative hypothesis is that liberal and conservative experts reason in fundamentally similar ways, but that there is something special about the Soviet Union that motivates wistful perceptions of lost possibilities on the

Are we prisoners of our preconceptions? 205 left and angry accusations of inevitable repression and expansion on the right. If we could identify a state that excites fear and loathing on the left comparable to those once excited by the Soviet Union on the right, we would observe a sign reversal of the relationship between ideological sympathies and counterfactual beliefs. South Africa was the ideal case for teasing these hypotheses apart. Demise of white-minority rule in South Africa PARTICIPANTS

Experts included academics with similar degree qualifications as in the Soviet exercise and people who worked in think tanks, the media and branches of the US government devoted to monitoring trends in South Africa. COMPETING HISTORICAL SCHEMAS

Observers on the political left now leaned toward essentialism. They portrayed the white-minority regime as incorrigibly racist and as likely to cede power only under enormous pressure. Observers on the political right favored a more plural view of politics inside Pretoria. They sensed that some factions in the leadership were enlightened and eager to explore flexible power-sharing arrangements. COUNTERFACTUAL PROBES

Since this disagreement between essentialists and pluralists was as close to a mirror image of the controversy over the Soviet Union as nature was going to provide, a reversal in patterns of openness to close-call counterfactuals was hypothesized, with the right more willing to entertain counterfactuals that assign a key role to political personalities within the regime (e.g., “if no de Klerk, then continued impasse”), and the left more willing to entertain counterfactuals that assign a key role to external pressure (e.g., “if no Western sanctions, then continued minority rule”). The operative psychological principle here is old-fashioned dissonance reduction: the more we hate a regime, the more repugnant it becomes to attribute anything good to redemptive dispositions of the regime (such as a capacity for self-correction). FINDINGS

Consistent with the two-tiered model of counterfactual inference, political ideology was again an anemic predictor of the mutability of historical antecedents but a robust predictor of antecedent–consequent linkages. Conservatives assigned more credit to de Klerk and to the collapse of

206

P.E. Tetlock and E. Henik

Soviet-style communism, whereas liberals assigned more credit to Western economic sanctions. Indeed, the debate over the impact of sanctions on South Africa serves as an instructive mirror image of the debate over the impact of Reagan’s defense build-up on the Soviet Union. Paraphrasing sentiments some liberal Western observers attributed to the Soviet elite at the dawn of the Gorbachev period in 1985, one conservative argued that the domestic momentum for change in South Africa had become overwhelming because white elites had concluded from the bloody township revolts and demographic trends that they “could not go on living this way.” Another conservative argued that credit for ending white minority rule should go to Reagan, whose policies precipitated the implosion of Soviet communism, allowing de Klerk to convince his followers that releasing Mandela and negotiating with the ANC were not tantamount to surrender to the Kremlin. This argument brings us full circle. The observer refuted the argument that “even if you guys were right about the Soviet Union, you were wrong about South Africa” by arguing that “it was because we were right about the Soviet Union that we were also right about South Africa.” DISCUSSION

The openness of conservatives to the de Klerk counterfactual in the South African case parallels the openness of liberals to the Stalin, Malenkov and Gorbachev counterfactuals in the Soviet case; liberal skepticism toward the Reagan-pressure counterfactual in the Soviet case parallels conservative skepticism toward the economic-sanctions counterfactual in the South African case. On balance, the data undermine the sweeping claim that liberals subscribe to a more contingent view of philosophy of history than conservatives. Much hinges on whose “policy ox” is being gored. We caution, however, that our data do not speak strongly to the reactivated debate over the link between cognitive style and political ideology (e.g., Jost et al. 2003), as the correlations between our cognitive style and abstract ideological schemata measures were generally weak (r  0.2). These two studies do show that experts’ beliefs about specific counterfactual possibilities were tightly coupled to their ideological outlooks and policy preferences. But what happens when scholars contemplate counterfactuals that undo events far removed from present-day controversies? Does temporal and political distance reduce the iron grip of our preconceptions on our judgments of what could have been? The next four studies examine this question and explore whether people with high versus low needs for closure respond differently to dissonant versus consonant what-if scenarios. Specifically, is there a cognitive-style-bytheoretical-worldview interaction: are high-need-for-closure individuals especially likely to reject close-call counterfactuals that undercut their pet theories?

Are we prisoners of our preconceptions? 207 Rerouting history at earlier choice points (1) Unmaking the West PARTICIPANTS

Experts were drawn from a random sample of the membership roster of the World History Association. COMPETING HISTORICAL SCHEMAS

Historians have long puzzled over how a relatively small number of Europeans, and their colonial offshoots, came to exert such disproportionate influence around the globe. The resulting debate has polarized scholars into feuding philosophical and ideological camps. Determinists view the West’s geopolitical ascendancy as having been inevitable for a long time, easily back to 1500 CE, and, for a few, even further back. Many embrace the notion that history efficiently winnows out maladaptive institutional forms and that the triumph of capitalism has been in the cards all along. Some scholars note other advantages of European polities: more deeply rooted legal and historical traditions of private property and individual rights, a religion that encouraged achievement in this world, and a fractious multi-state system that prevented any single power from dominating all others and bringing all innovation to a grinding halt whenever reactionary whims struck the ruling elite. At the other pole are the radical antideterminists who believe, to adapt Gould’s famous thought experiment, that if we were to rerun world history repeatedly from the same conditions that prevailed as recently as 1500 CE, European dominance would be one of the least likely outcomes. These scholars resent “Eurocentrism and neoconservative triumphalism.” They believe that the European achievement was a precarious one that can be unraveled easily. COUNTERFACTUAL PROBES

Antideterminists have generated a long list of close-call counterfactuals designed to puncture triumphalism: South Asia, East Africa and perhaps the Americas might have been conquered and colonized by an invincible Chinese armada in the fifteenth century if there had been more support in the imperial court for technological innovation and territorial expansion; Europe might have been conquered and Islamicized in the eighth century if the Moors had cared to launch a serious invasion of southern France and Italy; and European civilization might have been devastated by Mongol armies in the thirteenth century if not for Genghis Khan’s death in the nick of time.

208

P.E. Tetlock and E. Henik

FINDINGS

The more experts embraced long-range deterministic explanations for Western dominance, the more dismissive they were of counterfactuals that implied that the West was just luckier than the Rest, and the more prone they were to reject counterfactuals that implied that other civilizations could have achieved dominance or at least blocked Western hegemony. As in the Soviet and South Africa exercises, responses here were consistent with the two-tiered model of counterfactual inference. Ideology was a weak predictor of antecedent mutability but a robust predictor of antecedent– consequent linkages and the use of second-order counterfactuals. The hypothesized interaction also emerged: the power of preconceptions to predict reactions to counterfactuals was greater among experts with high needs for closure than low. Those with high needs for closure had an unusually strong propensity to make history fit their ideological frameworks, whereas those with low needs for closure were, relatively speaking, tolerant of counterfactuals that poked holes in their ideological frameworks. This interactive relationship also held in the final three studies, which explored reactions to rewrites of twentieth-century history. (Note: Findings are presented in aggregate after the three study summaries.) Rerouting history at earlier choice points (2) The outbreak of World War I PARTICIPANTS

Experts in the World War and Cold War exercises were drawn from the Society for Military History and those divisions of the American Political Science Association that focused on international relations and security policy. COMPETING HISTORICAL SCHEMAS

Some scholars believe that war among the great powers of Europe in the early twentieth century was inevitable. This thesis is often grounded in causal arguments that stress the inherent instability of multiethnic empires and multipolar balances of power, as well as the “cult of the offensive” (the widespread perception among general staffs that the side that struck first would gain a decisive advantage). COUNTERFACTUAL PROBES

The more experts endorse these “macro” causal arguments, the more ill disposed they should be toward counterfactuals that imply that war could have been avoided by undoing one of the bizarre coincidences preceding the assassination of Archduke Ferdinand or by permitting minor alterations of the content or timing of diplomatic messages exchanged among the great powers in the six weeks preceding the outbreak of war.

Are we prisoners of our preconceptions? 209 Rerouting history at earlier choice points (3) The outcomes of World Wars I and II COMPETING HISTORICAL SCHEMAS

Neorealist balancing, one of the most influential explanatory schemas in world politics, asserts that when one state threatens to become too powerful and capable of dominating the entire international system, other states coalesce against it and preserve the balance of power (Vasquez 1997; Waltz 1979). From this standpoint, it is no accident that would-be world conquerors such as Philip II, Napoleon and Hitler failed, as their failures were predetermined by a fundamental law of world politics. COUNTERFACTUAL PROBES

The more experts endorse neorealist balancing, the more ill disposed they should be to close-call counterfactuals that imply that the Germans could easily have emerged victorious in either of the two world wars and achieved at least continental hegemony if they had made better strategic decisions at key junctures in the conflicts (e.g., “If Germany had proceeded with its invasion of France on 2 August 1914, but had respected Belgian neutrality, Britain would not have entered the war, and France would have fallen quickly”). Rerouting history at earlier choice points (4) Why the Cold War never went “hot” COMPETING HISTORICAL SCHEMAS

Some scholars believe in the robustness of nuclear deterrence and mutual assured destruction: rational actors do not commit suicide (Sagan and Waltz 1995). When these scholars look back on the Cold War, they have a hard time imagining that crises could have escalated out of control (just as they have a hard time getting agitated about future dangers of nuclear proliferation). COUNTERFACTUAL PROBES

These scholars should be dismissive of close-call counterfactuals in which the United States and the Soviet Union slip into nuclear war at various junctures in the Cold War (e.g., if Kennedy had heeded the advice of his hawkish advisors and launched air strikes against Soviet missile sites in Cuba in October 1962, or if Eisenhower had followed through on the threat to use nuclear weapons to break the stalemate in negotiations to end the Korean War).

210

P.E. Tetlock and E. Henik

FINDINGS

In all three twentieth-century scenarios, the more committed scholars were to a generalization undercut by a counterfactual, the more dismissive they were of that counterfactual. And in all three scenarios, scholars with strong needs for closure were especially likely to dismiss counterfactuals that undercut their theoretical commitments. Counterfactuals were viewed as a nuisance at best, and a threat at worst, to theory-driven historical analysts on the prowl for ways of achieving closure by assimilating past events into favored explanatory frameworks. Also, again, we observe tight links between theoretical orientations to world politics and the two more abstract belief-system defenses: challenging the logic of connecting principles and generating second-order counterfactuals. Challenging the mutability of antecedents was more weakly correlated with the other two strategies, as well as with abstract orientations toward world politics. These results support the prediction that judgments of the mutability of antecedents should be governed by highly context-bound factors of the sort identified by Kahneman and Miller (1986) and not tightly coupled to abstract orientations toward world politics. People see events as more easily altered when they violate strong expectations of what is normal in a given locale and summon up vivid images of how nearly an outcome was averted. Indeed, there is no good reason why one’s theoretical position on the macro causes of war should predict whether one believes the assassination of the archduke could have been thwarted or why one’s position on the robustness of nuclear deterrence should predict whether one believes Stalin could have survived his cerebral hemorrhage. The plausibility of these antecedents hinges on specific facts tied to particular times, places, people, and events. Although the strategy of resisting close-call counterfactuals by challenging the mutability of the antecedent was not as closely coupled with abstract beliefs as other strategies, it was not completely decoupled from abstract beliefs either. Indeed, one may find it disturbing that abstract theoretical orientations predicted as much of the variance as they did in judgments of the mutability of antecedents (8–12 percent). These results are a sign of how even the most innocently apolitical fact about the world can become politicized as soon as rival schools of thought discover an advantage in showing a downstream outcome to be either easy or hard to undo.

Reactions to historical discoveries At this juncture, one might ask, what is to stop politically motivated observers from positing counterfactuals that justify whatever causal assertions they find it expedient to make? We probed for answers in a series of scenario experiments that explored how willing experts were to change their minds when new archival evidence came to light that either challenged or

Are we prisoners of our preconceptions? 211 reinforced causal lessons they had already drawn (Tetlock and Belkin 1996a). In one such study, we asked Sovietologists how they would react if a research team working in the Kremlin archives announced the discovery of evidence that shed light on three choice points in Soviet history. The evidence had either a liberal or conservative tilt and either featured or lacked methodological checks on ideological bias. After reading about each discovery, participants rated the credibility of the hypothetical researchers’ conclusions and three distinct grounds for impugning the team’s credibility: dismissing the motives of the research team as political rather than scholarly, disputing the authenticity of documents, and arguing that key documents had been taken out of context. Regardless of announced checks on bias, both liberals and conservatives rated consonant evidence as highly credible and dissonant evidence as relatively incredible. When reacting to dissonant data discovered by a team that did not implement precautions, experts used all three belief-system defenses, challenging the authenticity and representativeness of the archival documents and the motives of the investigators. The open-ended data underscore this point. The same tactics of data neutralization were almost four times as likely to appear in spontaneous comments on dissonant evidence as on consonant evidence. The composite-scale measure of the tendency to endorse all three tactics of data neutralization consistently predicted rejection of the conclusions that the investigators wanted to draw from their “discovery” (cf. Lord et al. 1979). Cognitive-style effects also emerged. Respondents with high needs for closure were more likely than those with low needs for closure to deploy double standards in the evaluation of evidence. Far from changing their minds in response to dissonant discoveries, high-need-for-closure individuals increased their confidence in their prior positions, whereas low-need-forclosure individuals at least made small adjustments in the direction of the new evidence. Moreover, high-need-for-closure individuals were not embarrassed by their judgments. In debriefing, experts were asked whether they thought their evaluations of the credibility of the study would have been different if the investigators had made the opposite archival discovery. Lowneed-for-closure individuals were reluctant to acknowledge that they might be keeping two sets of epistemological books and maintained that their reactions would have been similar. By contrast, high-need-for-closure individuals generally acknowledged that their reactions would have been strikingly different and defended “setting the bar higher” for more “farfetched” claims. The key empirical point of this turnabout experiment is the pervasiveness of double standards in assessments of evidence. Experts, especially highneed-for-closure ones, searched more vigorously for flaws in the face of dissonant results. They posed “Must I believe this?” questions of dissonant data and “Can I believe this?” questions of consonant data. We have another

212

P.E. Tetlock and E. Henik

illustration of how beliefs about what could or might have been are often theory-driven and self-perpetuating, insulated by defensive maneuvers that attribute awkward findings to methodological sloppiness or partisan bias. Selective openness to close-call counterfactuals Should we ever expect theory-driven thinkers to be more open to close-call counterfactuals? From a cognitive-consistency perspective, there are at least two sets of conditions under which we should: (1) when the close-call counterfactual undercuts the historical cause–effect claims promoted by rival schools of thought (“Those guys think they can explain outcomes B, C and D, but the best cases for their position were just aberrations”); and (2) when the close-call counterfactual helps to rescue their own floundering forecasts. Little systematic research has examined the first possibility, but Tetlock (1998) has looked in depth at the second. In one study, experts were asked to make five-year predictions (by choosing from a menu of scenarios) for the future of Soviet communism in 1988, the future of white minority rule in South Africa in 1989 and the future of Canadian confederation in 1992. They also rated their confidence in their forecasts on subjective probability scales. After the specified intervals had elapsed, participants were recontacted and reminded of the options they were given, their original forecasts, and their confidence estimates. They then rated on a nine-point agree–disagree scale the degree to which they believed (among other things) that the antecedent conditions for their forecasts had been fairly satisfied and that alterations in previously unspecified “background conditions” could easily have altered the predicted outcomes. Results showed that experts who made accurate predictions credited their sound reading of the “basic forces” at play in the situation and had little interest in entertaining the counterfactual hypothesis that other things almost happened. However, those who made inaccurate predictions were almost as likely to believe that their reading of the political situation was fundamentally sound. They preserved confidence in their worldviews, in large part, by advancing close-call counterfactuals in which the worlds they had predicted almost occurred. Cognitive closure was significantly related to the endorsement of belief-system-defense tactics among inaccurate forecasters, for whom higher scores on need for closure translated into stronger endorsements of close-call counterfactuals. In brief, inaccurate forecasters with high scores on need for closure cushioned their forecasts and their worldviews against disconfirmation by generating close-call counterfactuals. We see here an abrupt reversal from earlier studies in the signs of the statistical relationships between openness to close-call counterfactuals and cognitive style. Experts with high needs for closure reject close-call counterfactuals that undercut their pet historical theories, but invoke them to rescue their forecasts from refutation. There is nothing inconsistent about rejecting close-call counterfactuals that challenge one’s preferred explana-

Are we prisoners of our preconceptions? 213 tions of the past and embracing those that buffer one’s expectations about the future from refutation.

Unpacking imaginative alternatives to reality The recurring theme thus far is that people, especially those with high needs for closure, are theory-driven information processors who fill in the missing counterfactual conditions of history by relying on ideologically scripted cause–effect sequences. Although we may seem fated, at this point, to be prisoners of our preconceptions, closed minds can be pried open. Drawing on Tversky’s support theory (Tversky and Fox 1995; Tversky and Koehler 1994), Tetlock and Lebow (2001) find that encouraging people to imagine specific counterfactual scenarios can induce them to become more circumspect about the power of their favorite causal generalizations to delimit historical possibilities. Specifically, searching for possible alternative pathways to historical episodes and then “unpacking” these alternatives into progressively more differentiated subsets can inflate the subjective probabilities of those alternatives (cf. Koehler 1991). One study focused on the “rise of the West” scenario described earlier in this chapter. (A second study, which focused on the Cuban missile crisis of 1962, provided a conceptual replication of the key results.) Experts drawn from a random sample of the membership roster of the World History Association were randomly assigned to either a no-unpacking control condition or an “unpacking of alternative outcomes” condition. Participants in the no-unpacking group simply judged the likelihood of Western geopolitical domination (inevitability curve) and its complement, all possible alternatives to Western geopolitical domination (impossibility curve), over an 850-year stretch of time (1000–1850 CE). Experts in the unpacking condition were asked, before making any judgments, to break the set of all possible alternatives to Western geopolitical domination into progressively more detailed (mutually exclusive and exhaustive) subsets in which either no region achieved global hegemony or a non-Western civilization, such as China or Islam, achieved hegemony. They then completed the same inevitability and impossibility curves over the same 850-year period. The shaded area in Figure 12.1 represents the cumulative effect of unpacking: a massive (across-time) increase in the subjective probabilities attached to counterfactual alternatives. Although not displayed here, a significant unpacking-by-cognitive-style interaction also emerged. The differences between low- and high-need-for-closure respondents grew more pronounced when respondents performed the unpacking exercise. Low-needfor-closure experts were more influenced by the unpacking manipulation and more prone to inflate their subjective probability judgments of counterfactual scenarios. Finally, unpacking undermined the power of theoretical beliefs to con-

214

P.E. Tetlock and E. Henik

Retrospective subjective probability

1.0 0.9

Impossibility curve with unpacking of alternatives to Western domination (n36)

0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 1000

Inevitability curve for Western domination (n27) 1200

Impossibility curve with no unpacking of alternatives to Western domination (n27) 1400

1600

1800

Year

Figure 12.1 Inevitability and impossibility curves from the “rise of the West” scenario. The diagram presents inevitability and impossibility curves that historians generated for the “rise of the West.” The inevitability curve displays gradually rising likelihood judgments of Western geopolitical dominance. The lower impossibility curve displays gradually declining likelihood judgments of all possible alternatives to Western geopolitical dominance. The higher impossibility curve was derived by adding experts’ likelihood judgments of specific subsets of possible alternatives to Western domination. Adding values of the lower impossibility curve to the corresponding values of the inevitability curve yields sums around 1.0 (affirming Tversky’s principle of binary complementarity). Inserting values from the higher impossibility curve yields sums well above 1.0 (affirming Tversky’s subadditivity prediction). The area between the two impossibility curves represents the cumulative impact of unpacking on the subjective probability of counterfactual alternatives to reality.

strain perceptions of specific historical possibilities. Correlations plummeted between abstract beliefs about cause-and-effect and specific beliefs about what was possible at specific times and places. Prior to unpacking, history appeared to be a smooth monotonic progression toward a foreordained outcome (a simple linear equation captures 80 percent of the variance in the lower impossibility curve); after unpacking, history appeared to be a more meandering and unpredictable process (a fifth-order polynomial equation is needed to explain 80 percent of the variance in this bumpy journey through time). This study highlights the importance of the specificity of the counterfactual posed: the more detailed the unpacking of counterfactual alternatives, the stronger the tendency to judge the whole class of alternatives to be less likely than the sum of its exclusive and exhaustive components (Koehler

Are we prisoners of our preconceptions? 215 1991). Unpacking created pockets of turbulence in otherwise-gracefully descending impossibility curves by highlighting ways in which history could have been rerouted. The more unpacking, the higher-order the polynomial function we need to capture the more numerous ups and downs in the perceived likelihood of alternatives to reality. Of course, there is a price to be paid for our liberation. The more effort we devote to unpacking counterfactual scenarios, the more contradictory are our resulting judgments. Our most “open-minded” participants, the lowneed-for-closure experts, were most likely to assign incoherent subjective probabilities that, in the terminology of Tversky’s support theory, violated “extensionality” and exhibited “subadditivity” (i.e., the judged likelihood of the whole set of possibilities was less than the sum of the probabilities of their exhaustive and exclusive constituent parts). Historical observers would seem, then, to confront a trade-off between theory-driven and imagination-driven modes of making sense of the past. Theory-driven strategies confer the benefits of explanatory closure and parsimony but desensitize us to nuance, complexity, contingency, and the possibility that our theory is wrong. Imagination-driven strategies sensitize us to possible worlds but exact a price in increased confusion and even incoherence.

Conclusion The theme of this chapter has been the power of preconceptions to shape observers’ views of reality. We fill in the missing control conditions of history with agreeable scenarios, reject dissonant scenarios, and only reluctantly reconsider our judgment in light of fresh evidence. Given the ontological inadequacies of history as teacher and our psychological inadequacies as pupils, some readers may be tempted to embrace some pretty strong forms of “epistemic pessimism:” the depressing doctrine that stresses the enormous obstacles to our learning any theoretical or policy lessons from history that we were not already ideologically predisposed to learn. Some readers may also find it tempting to generalize the current results. Although we draw our examples from political and military history, it is not hard – mutatis mutandis – to imagine parallel demonstrations using economic historians (how easy is it to imagine the industrial revolution being delayed or never occurring in England? How hard is it to imagine an industrial revolution occurring instead in China in the twelfth or thirteenth century?), historians of science (if Newton or Einstein, etc., had not advanced his revolutionary ideas, how long would it have taken someone else to fill the void?), historians of business (if Bill Gates had not created a Microsoft-like entity, would someone else have done so in fairly short order?) and even evolutionary theorists (is intelligent life a statistical fluke, or would it almost surely have risen in some form, even if we counterfactually alter the timing and size of collisions of our planet with extraterrestrial bodies?). Indeed, we are

216

P.E. Tetlock and E. Henik

pursuing some of these empirical questions elsewhere, and there are preliminary indications that similar processes are at work (see, for instance, the chapters in Tetlock et al. 2004). We do not, however, classify ourselves as epistemic pessimists. In our view, awareness of past possibilities need not be irretrievably buried under the layers of deterministic thinking that build up as professional communities convince themselves, by piling one causal argument on another, that whatever happened had to happen. In most communities of historically minded scholars, there is a vigorous dialectic between proponents of more and less deterministic mindsets. And although it is true that determinists have built-in psychological advantages in these debates, antideterminists are far from defenseless. Tetlock and Lebow (2001) demonstrate that unpacking “alternative worlds” can be a powerful check on deterministic thinking. And, ironically, unpacking may well work by pitting offsetting cognitive biases against one another. The pre-eminent bias in historical reasoning is, as many studies now attest, certainty of hindsight: the tendency ex post to portray the past as more predictable than it was ex ante (Hawkins and Hastie 1990). The leading explanation of hindsight bias attributes it to the “automaticity” of theory-driven thought, to the rapidity with which people assimilate known outcomes into their favorite covering laws, in the process demoting once-plausible futures to the status of implausible counterfactuals. Unpacking counterfactual alternatives to reality may be such an effective tool for checking hindsight bias because unpacking effects are powered by arguably equally powerful psychological forces: our capacity to talk ourselves into hoping for, or fearing, improbable outcomes as long as we can construct good stories leading to those outcomes (Koehler 1991; Tversky and Kahneman 1983). In the end, then, those of us who want to draw lessons from history (i.e., all of us) are left with a delicate balancing act, one that some learning theorists have characterized as the tension between exploitation (of what we know) and exploration (of what we do not know) (March 1991) and that we have characterized as the tension between theory-driven and imaginationdriven thinking. These metacognitive trade-offs – however exactly formulated – have been thus far been little explored. So we are not in a strong position to offer prescriptive advice. But we can say this. We strongly suspect that, insofar as historical observers err more often in one direction or the other, it is more often in the direction of excessively deterministic thinking that overapplies neat theories to messy historical situations and that blinds us to other plausible paths that events could have taken. From this standpoint, imaginative unpacking of counterfactual alternatives to reality does indeed have considerable potential as a judgment-enhancing tool.

References

Acock, A.C. and Ireland, N.K. (1983) “Attribution of blame in pare cases: The impact of norm violation, gender, and sex role attitude,” Sex Roles 9: 179–93. Adams, J.S. (1965) “Inequity in social exchange,” in L. Berkowitz (ed.) Advances in Experimental Social Psychology (vol. 2, pp. 267–99), New York: Academic Press. Ahn, W. and Kalish, C.W. (2000) “The role of mechanism beliefs in causal reasoning,” in F.C. Keil and R.A. Wilson (eds) Explanation and Cognition (pp. 199–225), Cambridge, MA: Harvard University Press. Alicke, M.D. (1991) “Culpable causation,” Journal of Personality and Social Psychology 63: 368–78. —— (2000) “Culpable control and the psychology of blame,” Psychological Bulletin 126: 556–74. Anderson, C.A. and Sechler, E.S. (1986) “Effects of explanation and counterexplanation on the development and use of statistical theories,” Journal of Personality and Social Psychology 50: 24–34. Anderson, C.A., Lepper, M.R. and Ross, L. (1980) “Perseverance of social theories: The role of explanation in the persistence of discredited information,” Journal of Personality and Social Psychology 39: 1037–47. Anderson, J.R. (1983) The Architecture of Cognition, Cambridge, MA: Harvard University Press. Anderson, J.R. and Sheu, C.-F. (1995) “Causal inferences as perceptual judgments,” Memory and Cognition 23: 510–24. Apter, M.J. (2001) Motivational Styles in Everyday Life: A Guide to Reversal Theory, Washington, DC: American Psychological Association. Apter, M.J. and Larsen, R. (1993) “Sixty consecutive days: Telic and paratelic states in everyday life,” in J.H. Kerr, M.J. Apter and S. Murgatroyd (eds) Advances in Reversal Theory (pp. 107–22), Amsterdam: Swets & Zeitlinger. Aronson, E. (1992) “The return of the repressed: Dissonance theory makes a comeback,” Psychological Inquiry 3: 303–11. Atkinson, J.M. and Drew, P. (1979) Order in Court: The Organization of Verbal Interaction in Judicial Settings, London: Macmillan. Bandura, A. (1977) “Self-efficacy: Toward a unifying theory of behavioral change,” Psychological Review 84: 191–215. Barbash, T. (2003) On Top of the World: Cantor Fitzgerald, Howard Lutnick, and 9/11: A Story of Loss and Renewal, New York: Harper Collins. Bargh, J.A. (1994) “The four horsemen of automaticity: Awareness, intention,

218

References

efficiency, and control in social cognition,” in R.S. Wyer, Jr and T.K. Srull (eds) Handbook of Social Cognition (vol. 1, pp. 1–40), Mahwah, NJ: Erlbaum. Baron, R. (2000) “Counterfactual thinking and venture formation: The potential effects of thinking about what might have been,” Journal of Business Venturing 15: 79–92. Baron, R.A., Logan, H., Lilly, J., Inman, M. and Brennan, M. (1994) “Negative emotion and message processing,” Journal of Experimental Social Psychology 30: 181–201. Barsalou, L.W. (1983) “Ad hoc categories,” Memory and Cognition 11: 211–27. —— (1991) “Deriving categories to achieve goals,” in G.H. Bower (ed.) The Psychology of Learning and Motivation (vol. 27, pp. 1–64), New York: Academic Press. Barton, S. (1994) “Chaos, self-organization and psychology,” American Psychologist 49: 5–14. Baumeister, R.F. and Leary, M.R. (1995) “The need to belong: Desire for interpersonal attachments as a fundamental human motivation,” Psychological Bulletin 117: 497–529. Bell, D.E. (1982) “Regret in decision-making under uncertainty,” Operations Research 30: 961–81. —— (1983) “Risk premiums for decision regret,” Management Science 29: 1156–66. —— (1985) “Disappointment in decision-making under uncertainty,” Operations Research 33: 1–27. Benjamin, L.T., Cavell, T.A. and Shallenberger, W.R. (1984) “Staying with initial answers on objective tests: Is it a myth?” Teaching of Psychology 11: 133–41. Ben-Ze’ev, A. (1996) “Emotional intensity,” Theory and Psychology 6: 509–32. —— (2000) The Subtlety of Emotions, Cambridge, MA: MIT Press. Berglas, S.E. and Jones, E.E. (1978) “Drug use as a self-handicapping strategy in response to noncontingent success,” Journal of Personality and Social Psychology 36: 405–17. Berndsen, M., van der Pligt, J., Doosje, B. and Manstead, A.S.R. (2004) “Guilt and regret: The determining role of interpersonal and intrapersonal harm,” Cognition and Emotion 18: 55–70. Bernstein, E.M. (1995) “Terror in Oklahoma: The displaced; Its building is shattered, but church survives,” New York Times, 24 April, p. B7. Biernat, M. and Manis, M. (1994) “Shifting standards and stereotype-consistent judgments,” Journal of Personality and Social Psychology 66: 5–20. Biernat, M., Manis, M. and Kobrynowicz, D. (1997) “Simultaneous assimilation and contrast effects in judgments of self and others,” Journal of Personality and Social Psychology 73: 254–69. Blatt, S.J. (1995) “The destructiveness of perfectionism: Implications for the treatment of depression,” American Psychologist 50: 1003–20. Bohner, G., Bless, H., Schwarz, N. and Strack, F. (1988) “What triggers causal attributions? The impact of valence and subjective probability,” European Journal of Social Psychology 18: 335–45. Boles, T.L. and Messick, D.M. (1995) “A reverse outcome bias: The influence of multiple reference points on the evaluation of outcomes and decisions,” Organizational Behavior and Human Decision Processes 61: 262–75. Boninger, D.S., Gleicher, F. and Strathman, A.J. (1994) “Counterfactual thinking: From what might have been to what may be,” Journal of Personality and Social Psychology 67: 297–307.

References

219

Borgida, E. and Brekke, N. (1985) “Psycholegal research on rape trials,” in A. Burgess (ed.) Research Handbook on Rape and Sexual Assault (pp. 313–42), New York: Garland. Bothwell, R.K. and Duhon, K.W. (1994) “Counterfactual thinking and plaintiff compensation,” Journal of Social Psychology 134: 705–6. Bouts, P., Spears, R. and van der Pligt, J. (1992) “Counterfactual processes and the correspondence between events and outcomes: Normality versus value,” European Journal of Social Psychology 22: 387–96. Braine, M.D.S. and O’Brien, D.P. (eds) (1998) Mental Logic, Mahwah, NJ: Erlbaum. Braithwaite, J. (1989) Crime, Shame and Reintegration, Cambridge: Cambridge University Press. Branscombe, N.R. and Weir, J.A. (1992) “Resistance as stereotype-inconsistency: Consequences for judgments of rape victims,” Journal of Social and Clinical Psychology 11: 80–102. Branscombe, N.R., Crosby, P. and Weir, J.A. (1993) “Social inferences concerning male and female homeowners who use a gun to shoot an intruder,” Aggressive Behavior 19: 113–24. Branscombe, N.R., Owen, S., Garstka, T.A. and Coleman, J. (1996) “Rape and accident counterfactuals: Who might have done otherwise and would it have changed the outcome?” Journal of Applied Social Psychology 26: 1042–67. Brewer, M.B. (1977) “An information-processing approach to attribution of responsibility,” Journal of Experimental Social Psychology 13: 58–69. Brickman, P., Ryan, K. and Wortman, C.B. (1975) “Causal chains: attribution of responsibility as a function of immediate and prior causes,” Journal of Personality and Social Psychology 32: 1060–7. Brown, R. and Fish, D. (1983) “The psychological causality implicit in language,” Cognition 14: 237–73. Buck, M.L. and Miller, D.T. (1994) “Reactions to incongruous negative life events,” Social Justice Research 7: 29–46. Buehler, R., Griffin, D. and Ross, M. (1994) “Exploring the ‘planning fallacy’: Why people underestimate their task completion times,” Journal of Personality and Social Psychology 67: 366–81. Bukstel, L.H. and Kilmann, P.R. (1980) “Psychological effects of imprisonment on confined individuals,” Psychological Bulletin 88: 469–93. Burt, M.R. (1980) “Cultural myths and supports for rape,” Journal of Personality and Social Psychology 38: 217–30. Byrne, R.M.J. (1996) “Towards a model theory of imaginary thinking,” in J. Oakhill and A. Garnham (eds) Mental Models in Cognitive Science: Essays in Honour of Phil Johnson-Laird (pp. 155–74), Hove: Erlbaum. —— (1997) “Cognitive processes in counterfactual thinking about what might have been,” in D.L. Medin (ed.) The Psychology of Learning and Motivation: Advances in Research and Theory (vol. 37, pp. 105–54), San Diego, CA: Academic Press. —— (2002) “Mental models and counterfactual thoughts about what might have been,” Trends in Cognitive Sciences 6: 426–31. —— (2005) The Rational Imagination: How People Create Alternatives to Reality, Boston, MA: MIT Press, in preparation. Byrne, R.M.J. and McEleney, A. (2000) “Counterfactual thinking about actions and failures to act,” Journal of Experimental Psychology: Learning, Memory, and Cognition 26: 1318–31.

220

References

Byrne, R.M.J. and Tasso, A. (1999) “Deductions from factual, possible, and counterfactual conditionals,” Memory and Cognition 27: 726–40. Byrne, R.M.J., Segura, S., Culhane, R., Tasso, A. and Berrocal, P. (2000) “The temporality effect in counterfactual thinking about what might have been,” Memory and Cognition 28: 264–81. Carver, C.S. and Scheier, M.F. (1990) “Origins of positive and negative affect: A control-process view,” Psychological Review 97: 19–35. Catellani, P. and Milesi, P. (2001) “Counterfactuals and roles: mock victims’ and perpetrators’ accounts of judicial cases,” European Journal of Social Psychology 31: 247–64. —— (2004) “Does your counterfactual become my counterfactual? Stereotyped expectations and suspicion,” manuscript submitted for publication. Catellani, P., Alberici, I.A. and Milesi, P. (2004) “Counterfactual thinking and stereotypes: The nonconformity effect,” European Journal of Social Psychology 34: 421–36. Cheng, P.W. (1997) “From covariation to causation: A causal power theory,” Psychological Review 104: 367–405. Cheng, P.W. and Novick, L.R. (1990) “A probabilistic contrast model of causal induction,” Journal of Personality and Social Psychology 58: 545–67. —— (1992) “Covariation in natural causal induction,” Psychological Review 99: 365–82. Chisholm, R.M. (1946) “The contrary-to-fact conditional,” Mind 55: 289–307. Coelho, P. (1999) The Alchemist, London: Harper Collins. Cohen, S.F., Rabinowitch, A. and Sharlet, R. (eds) (1985) The Soviet Union since Stalin, Bloomington, IN: Indiana University Press. Collingwood, R.G. (1940) An Essay on Metaphysics, Oxford: Clarendon Press. Colvin, C., Block, J. and Funder, D.C. (1995) “Overly positive self-evaluations and personality: Negative implications for mental health,” Journal of Personality and Social Psychology 68: 1152–62. Conley, J.M. and O’Barr, W.M. (1990) Rules versus Relationships: Ethnography of Legal Discourse, Chicago: University of Chicago Press. Cummins, D.D. (1995) “Naïve theories and causal deduction,” Memory and Cognition 23: 646–58. Cyert, R.M. and March, J.G. (1992) A Behavioral Theory of the Firm, Cambridge, MA: Blackwell. Darke, P.R. and Freedman, J.L. (1997a) “Lucky events and beliefs in luck: Paradoxical effects on risk-taking,” Personality and Social Psychology Bulletin 23: 378–88. —— (1997b) “The belief in good luck scale,” Journal of Research in Personality 31: 486–511. Davenport, T.H. and Prusak, L. (1998) Working Knowledge: How Organizations Manage What they Know, Boston, MA: Harvard Business School Press. Davis, C.G. and Lehman, D.R. (1995) “Counterfactual thinking and coping with traumatic life events,” in N.J. Roese and J.M. Olson (eds) What Might Have Been: The Social Psychology of Counterfactual Thinking (pp. 353–74), Mahwah, NJ: Erlbaum. Davis, C.G. and Nolen-Hoeksema, S. (2001) “Loss and meaning: How do people make sense of loss?” American Behavioral Scientist 44: 726–41. Davis, C.G., Nolen-Hoeksema, S. and Larson, J. (1998) “Making sense of loss and

References

221

benefiting from the experience: Two construals of meaning,” Journal of Personality and Social Psychology 75: 561–74. Davis, C.G., Lehman, D.R., Silver, R.C., Wortman, C.B. and Ellard, J.H. (1996) “Self-blame following a traumatic life event: The role of perceived avoidability,” Personality and Social Psychology Bulletin 22: 557–67. Davis, C.G., Lehman, D.R., Wortman, C.B., Silver, R.C. and Thompson, S.C. (1995) “The undoing of traumatic life events,” Personality and Social Psychology Bulletin 21: 109–24. Dawes, R.M. (1976) “Shallow psychology,” in J.S. Carroll and J.W. Payne (eds) Cognition and Social Behavior (pp. 3–12), Hillsdale, NJ: Erlbaum. —— (1996) “Counterfactual inferences as instances of statistical inferences,” in P.E. Tetlock and A. Belkin (eds) Counterfactual Thought Experiments in World Politics: Logical, Methodological, and Psychological Perspectives (pp. 301–8), Princeton, NJ: Princeton University Press. De Groot, A.D. (1965) Thought and Choice in Chess, New York: Basic Books. Denes-Raj, V. and Epstein, S. (1994) “Conflict between intuitive and rational processing: When people behave against their better judgment,” Journal of Personality and Social Psychology 66: 819–29. Denes-Raj, V., Epstein, S. and Cole, J. (1995) “The generality of the ratio-bias phenomenon,” Personality and Social Psychology Bulletin 21: 1083–92. Douchet, K., Giudicelli, M.-N. and Hilton, D.J. (2004) “Counterfactual thinking and attributions of responsibility for wrongdoing in Vichy France,” manuscript in preparation. Dweck, C.S. (2000) Self-theories: Their Role in Motivation, Personality, and Development. Philadelphia, PA: Taylor & Francis. Eden, C. (1988) “Cognitive mapping: A review,” European Journal of Operational Research 36 (1): 1–13. Edwards, W. (1961) “Behavioral decision theory,” Annual Review of Psychology 12: 473–86. Einhorn, H.J. and Hogarth, R.M. (1986) “Judging probable cause,” Psychological Bulletin 99: 3–19. Epstein, S., Lipson, A., Holstein, C. and Huh, E. (1992) “Irrational reactions to negative outcomes: Evidence for two conceptual systems,” Journal of Personality and Social Psychology 62: 328–39. Evans, J. St B.T., Handley, S.J. and Over, D. (2003) “Conditionals and conditional probability,” Journal of Experimental Psychology: Learning, Memory, and Cognition 29: 321–35. Evans, J. St B.T., Newstead, S. and Byrne, R.M.J. (1993) Human Reasoning: The Psychology of Deduction, Hillsdale, NJ: Erlbaum. Fairley, N., Manktelow, K. and Over, D. (1999) “Necessity, sufficiency, and perspective effects in causal conditional reasoning,” Quarterly Journal of Experimental Psychology 52A: 771–90. Fearon, J. (1991) “Counterfactuals and hypothesis testing in political science,” World Politics 43: 474–84. Fein, S. and Hilton, J.L. (1994) “Judging others in the shadow of suspicion,” Motivation and Emotion 18: 167–98. Ferguson, N. (ed.) (1997) Virtual History: Alternatives and Counterfactuals, London: Picador.

222

References

Festinger, L. (1954) “A theory of social comparison processes,” Human Relations 7: 117–40. Fiddick, L., Cosmides, L. and Tooby, J. (2000) “No interpretation without representation: The role of domain-specific representations and inferences in the Wason selection task,” Cognition 77: 1–79. Fillenbaum, S. (1974) “Information amplified: memory for counterfactual conditionals,” Journal of Experimental Psychology 102: 44–9. Fincham, F.D. and Jaspars, J.M.F. (1983) “Subjective probability approach to responsibility attribution,” British Journal of Social Psychology 22: 145–62. Fincham, F.D. and Shultz, T.R. (1981) “Intervening causation and the mitigation of responsibility for harm,” British Journal of Social Psychology 20: 113–20. Fischhoff, B. (1975) “Hindsight  foresight: The effects of outcome knowledge on judgment under uncertainty,” Journal of Experimental Psychology: Human Perception and Performance 1: 288–99. —— (1982) “For those condemned to study the past: Heuristics and biases in hindsight,” in D. Kahneman, P. Slovic and A. Tversky (eds) Judgment under Uncertainty: Heuristics and Biases (pp. 335–54), New York: Cambridge University Press. Fischhoff, B., Slovic, P., and Lichtenstein, S. (1978) “Fault trees: Sensitivity of estimated failure probabilities to problem representation,” Journal of Experimental Psychology: Human Perception and Performance 4: 330–44. Foersterling, F. (1989) “Models of covariation and attribution: How do they relate to the analysis of variance?” Journal of Personality and Social Psychology 57: 615–25. Fogel, R. (1964) Railroads and American Economic Growth: Essays in Econometric History, Baltimore, MD: Johns Hopkins University Press. Folger, R. and Kass, E.E. (2000) “Social comparison and fairness: A counterfactual simulations of perspective,” in J.M. Suls and L. Wheeler (eds) Handbook of Social Comparison: Theory and Research (pp. 423–41), New York: Kluwer Academic/ Plenum Publishers. Forgas, J.P. (1995) “Mood and judgment: The affect infusion model (AIM),” Psychological Bulletin 117: 39–66. Forster, J., Higgins, E.T. and Idson, L.C. (1998) “Approach and avoidance strength through goal attainment: Regulatory focus and the ‘goal looms larger’ effect,” Journal of Personality and Social Psychology 75: 1115–31. Frankl, V.E. (1984) Man’s Search for Meaning: An Introduction to Logotherapy (3rd edn), New York: Simon & Schuster (original work published in 1959). Frijda, N.H. (1988) “The laws of emotion,” American Psychologist 43: 249–358. Galinsky, A.D. and Kray, L.J. (2004) “From thinking about what might have been to sharing what we know: The role of counterfactual mind-sets in information sharing in groups,” Journal of Experimental Social Psychology 40: 606–18. Galinsky, A.D. and Moskowitz, G.B. (2000) “Counterfactuals as behavioral primes: Priming the simulation heuristic and consideration of alternatives,” Journal of Experimental Social Psychology 36: 257–383. Galinsky, A.D. and Mussweiler, T. (2001) “First offers as anchors: The role of perspective-taking and negotiator focus,” Journal of Personality and Social Psychology 81: 657–69. Galinsky, A.D., Seiden, V.L., Kim, P.H. and Medvec, V.H. (2002) “The dissatisfaction of having your first offer accepted: The role of counterfactual thinking in negotiations,” Personality and Social Psychology Bulletin 28: 271–83.

References

223

Gavanski, I. and Wells, G.L. (1989) “Counterfactual processing of normal and exceptional events,” Journal of Experimental Social Psychology 25: 314–25. Gibbons, F.X. and Buunk, B.P. (1999) “Individual differences in social comparison: Development and validation of a measure of social comparison orientation,” Journal of Personality and Social Psychology 76: 129–42. Gilbert, D.T., Giesler, R.B. and Morris, K.A. (1995) “When comparisons arise,” Journal of Personality and Social Psychology 69: 227–36. Gilovich, T. (1983) “Biased evaluation and persistence in gambling,” Journal of Personality and Social Psychology 44: 1110–26. Gilovich, T., Griffin, D. and Kahneman, D. (eds) (2002) Heuristics and Biases: The Psychology of Intuitive Judgment, New York: Cambridge University Press. Gilovich, T. and Medvec, V.H. (1994) “The temporal pattern to the experience of regret,” Journal of Personality and Social Psychology 67: 357–65. —— (1995) “The experience of regret: what, when, and why,” Psychological Review 102: 379–95. Gilovich T., Medvec V.H. and Chen S. (1995) “Commission, omission, and dissonance reduction: Coping with regret in the Monty Hall problem,” Personality and Social Psychology Bulletin 21: 182–90. Girotto, V., Legrenzi, P. and Rizzo, A. (1991) “Event controllability in counterfactual thinking,” Acta Psychologica 78: 111–33. Gleicher, F., Kost, K.A., Baker, S.M., Strathman, A.J., Richman, S.A. and Sherman, S.J. (1990) “The role of counterfactual thinking in judgments of affect,” Personality and Social Psychology Bulletin 16: 284–95. Goldinger, S.D., Kleider, H.M., Azuma, T. and Beike, D.R. (2003) “‘Blaming the victim’ under memory load,” Psychological Science 14: 81–5. Goldvarg, E. and Johnson-Laird, P.N. (2001) “Naïve causality: A mental model theory of causal meaning and reasoning,” Cognitive Science 25: 565–610. Gollwitzer, P.M., Heckhausen, H. and Steller, B. (1990) “Deliberative vs. implemental mind-sets: Cognitive tuning toward congruous thoughts and information,” Journal of Personality and Social Psychology 59: 1119–27. Gordon, R.A. (1990) “Attributions for blue-collar and white-collar crime: The effect of subjects and defendant race on simulated juror decisions,” Journal of Applied Social Psychology 128: 971–83. Gorovitz, S. (1965) “Causal judgments and causal explanations,” Journal of Philosophy 62: 695–711. Graesser, A.C., Robertson, S.P. and Anderson, P.A. (1981) “Incorporating inferences in narrative representations: A study of how and why,” Cognitive Psychology 13: 1–26. Green, M.C. and Brock, T.C. (2000) “The role of transportation in the persuasiveness of public narratives,” Journal of Personality and Social Psychology 79: 701–21. Greene, E.J. and Darley, J.M. (1998) “Effects of necessary, sufficient, and indirect causation on judgments of criminal liability,” Law and Human Behavior 22: 429–51. Grice, H.P. (1975) “Logic and conversation,” in P. Cole and J.L. Morgan (eds) Syntax and Semantics 3: Speech Acts (pp. 41–58), New York: Academic Press. Grieve, F.G., Houston, D.A., Dupuis, S.E. and Eddy, D. (1999) “Counterfactual production and achievement orientation in competitive athletic settings,” Journal of Applied Social Psychology 29: 2177–202. Guthrie, C. (1999) “Better settle than sorry: The regret aversion theory of litigation behavior,” University of Illinois Law Review 1999: 43–90.

224

References

Halpern, J. and Pearl, J. (in press-a) “Causes and explanations. A structural model approach. Part I: Causes,” British Journal for the Philosophy of Science. —— (in press-b) “Causes and explanations. A structural model approach. Part II: Explanations,” British Journal for the Philosophy of Science. Handgraaf, M., Zeelenberg, M. and Manstead, A.S.R. (1997, November) “Piekeren en het sociaal delen van emoties [Rumination and social sharing of emotions],” paper presented at the twelfth conference of the Association of Social Psychological Researchers, Enschede, Netherlands. Haney, C. and Zimbardo, P. (1998) “The past and future of US prison policy: Twenty-five years after the Stanford Prison Experiment,” American Psychologist 53: 709–27. Hart, H.L.A. and Honoré, A.M. (1985) Causation in the Law (2nd edn), Oxford: Clarendon Press. Hastie, R. (1984) “Causes and effects of causal attributions,” Journal of Personality and Social Psychology 46: 44–56. Hawkins, S. and Hastie, R. (1990) “Hindsight: Biased judgment of past events after the outcomes are known,” Psychological Bulletin 107: 311–27. Hegarty, P. and Pratto, F. (2001) “The effects of social category norms and stereotypes on explanations for intergroup differences,” Journal of Personality and Social Psychology 80: 723–35. Heider, F. (1958) The Psychology of Interpersonal Relations, New York: Wiley. —— (1983) The Life of a Psychologist: An Autobiography, Lawrence, KA: University Press of Kansas. Heine, S.J. and Lehman, D.R. (1999) “Culture, self-discrepancies, and self-satisfaction,” Personality and Social Psychology Bulletin 25: 915–25. Hempel, C.G. (1965) Aspects of Scientific Explanation, New York: Macmillan. Herriott, S.R., Levinthal, D. and March, J.G. (1985) “Learning from experience in organizations,” American Economic Review 75: 298–302. Hesslow, G. (1976) “Discussion: Two notes on the probabilistic approach to causality,” Philosophy of Science 43: 290–2. —— (1983) “Explaining differences and weighting causes,” Theoria 49: 87–111. —— (1988) “The problem of causal selection,” in D. Hilton (ed.) Contemporary Science and Natural Explanation: Commonsense Conceptions of Causality (pp. 33–65), Brighton: Harvester Press. Higgins, E.T. (1998) “Promotion and prevention: Regulatory focus as a motivational principle,” in M.P. Zanna (ed.) Advances in Experimental Social Psychology (vol. 30, pp. 1–46), San Diego, CA: Academic Press. Higgins, E.T., Shah, J. and Friedman, R. (1997) “Emotional responses to goal attainment: Strength of regulatory focus as a moderator,” Journal of Personality and Social Psychology 72: 515–25. Hilton, D.J. (1985) “Causal beliefs: From attribution theory to cognitive science,” in J. Allwood and E. Hjelmquist (eds) Foregrounding Background, Lund, Sweden: Doxa. —— (1988) “Logic and causal attribution,” in D. Hilton (ed.) Contemporary Science and Natural Explanation: Commonsense Conceptions of Causality (pp. 33–65), Brighton: Harvester Press. —— (1990) “Conversational processes and causal explanation,” Psychological Bulletin 107: 65–81. —— (1991) “A conversational model of causal explanation,” European Review of Social Psychology 2: 51–81.

References

225

—— (1995) “Logic and language in causal explanation,” in D. Sperber, D. Premack and A.J. Premack (eds) Causal Understanding in Cognition and Culture (pp. 495–529), Oxford: Oxford University Press. —— (2001) “Norms and commonsense explanation: Types of contrast and abnormal conditions,” in R. Boudon, P. Demeulenaere and R. Viale (eds) L’Explication des normes sociales (pp. 205–14), Paris: Presses Universitaires de France. Hilton, D.J. and Erb, H.-P. (1996) “Mental models and causal explanation: Judgements of probable cause and explanatory relevance,” Thinking and Reasoning 2: 273–308. Hilton, D.J. and Jaspars, J.M.F. (1987) “The explanation of occurrences and nonoccurrences: A test of the inductive logic model of causal attribution,” British Journal of Social Psychology 26: 189–201. Hilton, D.J. and Slugoski, B.R. (1986) “Knowledge-based causal attribution: The abnormal conditions focus model,” Psychological Review 93: 75–88. Hilton, D.J., Mathes, R.H. and Trabasso, T. (1992) “The study of causal explanation in natural language: Analysing reports of the Challenger disaster in the New York Times,” in M.L. McLaughlin, M. J. Cody and S.J. Read (eds) Explaining One’s Self to Others: Reason-giving in a Social Context (pp. 41–60), Hillsdale, NJ: Erlbaum. Hilton, D.J., McClure, J.L., Sutton, R.M., Baroux, A. and Magaorou, I. (2003) “Selecting ‘the’ explanation from a causal chain: On the pragmatic selection of attributional principles,” unpublished manuscript. Hirt, E.R., Levine, G.M., McDonald, H.E., Melton, R.J. and Martin, L.L. (1997) “The role of mood in quantitative and qualitative aspects of performance: Single or multiple mechanisms?” Journal of Experimental Social Psychology 33: 602–29. Holland, J.H., Holyoak, K.J., Nisbett, R.E. and Thagard, P. (1986). Induction: Processes in Inference, Learning and Discovery, Cambridge, MA: MIT Press. Holyoak, K.J. and Cheng, P.W. (1995) “Pragmatic reasoning with a point of view,” Thinking and Reasoning 1: 289–313. Hooker, C., Roese, N.J. and Park, S. (2000) “Impoverished counterfactual thinking is associated with schizophrenia,” Psychiatry: Interpersonal and Biological Processes 63: 326–35. Howard, J.A. (1984) “The ‘normal’ victim: The effects of gender stereotypes on reactions to victims,” Social Psychology Quarterly 47: 270–81. Hsee, C.K., Salovey, P. and Abelson, R.P. (1994) “The quasi-acceleration relation: Satisfaction as a function of the change of velocity of outcome over time,” Journal of Experimental Social Psychology 30: 96–111. Huba, G.J., Aneshensel, C.S. and Singer, J.L. (1981) “Development of scales for three second-order factors of inner experience,” Multivariate Behavioral Research 16: 181–206. Hur, T. (2000) “Counterfactual thinking and regulatory focus: Upward versus downward counterfactuals and promotion versus prevention,” unpublished M.Phil. thesis, Evanston, IL: Northwestern University. Hurley, J.R. and Palonen, D.P. (1967) “Marital satisfaction and child density among university student parents,” Journal of Marriage and the Family 29: 483–4. Inman, J.J., Dyer, J.S. and Jia, J. (1997) “A generalized utility model of disappointment and regret effects on post-choice valuation,” Marketing Science 16: 97–111. James, W. (1948) Psychology, Cleveland, OH: Longman. Janis, I.L. and Feshbach, S. (1953) “Effects of fear arousing communications,” Journal of Abnormal and Social Psychology 48: 78–92.

226

References

Janoff-Bulman, R. (1992) Shattered Assumptions: Towards a New Psychology of Trauma, New York: Free Press. Janoff-Bulman, R. and Frantz, C.M. (1997) “The impact of trauma on meaning: From meaningless world to meaningful life,” in M. Power and C.R. Brewin (eds) The Transformation of Meaning in Psychological Therapies (pp. 91–106), New York: Wiley. Jervis, R. (1976) Perception and Misperception in International Politics, Princeton, NJ: Princeton University Press. —— (1996) “Counterfactuals, causation, and complexity,” in P.E. Tetlock and A. Belkin (eds) Counterfactual Thought Experiments in World Politics: Logical, Methodological, and Psychological Perspectives (pp. 309–16), Princeton, NJ: Princeton University Press. Johnson, J.T. (1986) “The knowledge of what might have been: Affective and attributional consequences of near outcomes,” Personality and Social Psychology Bulletin 12: 51–62. Johnson, J.T., Ogawa, K.H., Delforge, A. and Early, D. (1989) “Causal primacy and comparative fault: the effect of position in a causal chain on judgments of legal responsibility,” Personality and Social Psychology Bulletin 15: 161–74. Johnson, M.D. (1984) “Consumer choice strategies for comparing noncomparable alternatives,” Journal of Consumer Research 11: 741–53. —— (1989) “The differential processing of product category and noncomparable choice alternatives,” Journal of Consumer Research 16: 300–9. Johnson, M.K. and Sherman, S.J. (1990) “Constructing and reconstructing the past and the future in the present,” in E.T. Higgins and R.M. Sorrentino (eds) Handbook of Motivation and Cognition: Foundations of Social Behavior (vol. 2, pp. 482–526), New York: Guilford Press. Johnson-Laird, P.N. (2001) “Mental models and deductive reasoning,” Trends in Cognitive Sciences 5: 434–42. Johnson-Laird, P.N. and Byrne, R.M.J. (1991) Deduction, Hove: Erlbaum. —— (2002) “Conditionals: a theory of meaning, pragmatics, and inference,” Psychological Review 109: 646–78. Johnson-Laird, P.N., Byrne, R.M.J. and Schaeken, W.S. (1992) “Propositional reasoning by model,” Psychological Review 99: 418–39. Johnston, D. and Dwyer, J. (2004) “Pre-9/11 files show warnings were more dire and persistent,” New York Times. Online, available at: www.nytimes.com (18 April 2004). Jones, E.E. and McGillis, D. (1976). “Correspondent inferences and the attribution cube: A comparative appraisal,” in J.H. Harvey, W.J. Ickes and R.F. Kidd (eds) New Directions in Attribution Research (pp. 389–420), Hillsdale, NJ: Erlbaum. Jost, J.T., Glaser, J., Kruglanski, A.W. and Sulloway, F.J. (2003) “Political conservatism as motivated social cognition,” Psychological Bulletin 129: 339–75. Kahneman, D. (1995) “Varieties of counterfactual thinking,” in N.J. Roese and J.M. Olson (eds) What Might Have Been: The Social Psychology of Counterfactual Thinking (pp. 375–96), Mahwah, NJ: Erlbaum. Kahneman, D. and Miller, D.T. (1986) “Norm theory: Comparing reality to its alternatives,” Psychological Review 93: 136–53. Kahneman, D. and Tversky, A. (1972) “Subjective probability: A judgment of representativeness,” Cognitive Psychology 3: 430–54. —— (1979) “Prospect theory: An analysis of decision under risk,” Econometrica, 47: 263–91.

References

227

—— (1982a) “The psychology of preferences,” Scientific American 246: 160–73. —— (1982b) “The simulation heuristic,” in D. Kahneman, P. Slovic and A. Tversky (eds) Judgment under Uncertainty: Heuristics and Biases (pp. 201–8), New York: Cambridge University Press. —— (1982c) “Variants of uncertainty,” Cognition 11: 143–57. Kahneman, D. and Varey, C.A. (1990) “Propensities and counterfactuals: The loser that almost won,” Journal of Personality and Social Psychology 59: 1101–10. Kahneman, D., Slovic, P. and Tversky, A. (eds) (1982) Judgment under Uncertainty: Heuristics and Biases, New York: Cambridge University Press. Kamin, K.A. and Rachlinski, J.J. (1995) “Ex post  ex ante: Determining liability in hindsight,” Law and Human Behavior 19: 89–104. Kanazawa, S. (1992) “Outcome or expectancy? Antecedent of spontaneous causal attribution,” Personality and Social Psychology Bulletin 18: 659–68. Kao, S.-F. and Wasserman, E.A. (1993) “Assessment of an information integration account of contingency judgment with examination of subjective cell importance and method of information presentation,” Journal of Experimental Psychology: Learning, Memory, and Cognition 19: 1363–86. Kaplan (1999) A Systematic Approach to the GRE, New York: Kaplan. Karniol, R. and Ross, M. (1996) “The motivational impact of temporal focus: Thinking about the future and the past,” Annual Review of Psychology 47: 593–620. Kassin, S.M., Williams, L.N. and Saunders, C.L. (1990) “Dirty tricks of cross-examination: The influence of conjectural evidence on the jury,” Law and Human Behavior 14: 373–84. Katz, L. (1987) Bad Acts and Guilty Minds: Conundrums of the Criminal Law, Chicago: University of Chicago Press. Keeton, W.P., Dobbs, D.B., Keeton, R.E. and Owen, D.B. (1984) Prosser and Keeton on the Law of Torts (5th edn), St Paul, MN: West Publishing. Kelley, H.H. (1967) “Attribution theory in social psychology,” in D. Levine (ed.) Nebraska Symposium on Motivation (vol. 15, pp. 192–238), Lincoln, NE: University of Nebraska Press. —— (1973) “The process of causal attribution,” American Psychologist 28: 107–28. —— (1983) “Perceived causal structures,” in J.M.F. Jaspars, M.R.C. Hewstone and F.D. Fincham (eds) Attribution Theory and Research: Conceptual, Developmental and Social Dimensions (pp. 343–59), New York: Academic Press. Keltner, D., Ellsworth, P.C. and Edwards, K. (1993) “Beyond simple pessimism: Effects of sadness and anger on social perception,” Journal of Personality and Social Psychology 64: 740–52. Kinner, R.T. and Metha, A.T. (1989) “Regrets and priorities at three stages of life,” Counseling and Values 33: 182–93. Klauer, K.C., Jacobsen, T. and Migulla, G. (1995) “Counterfactual processing: Test of an hierarchical correspondence model,” European Journal of Social Psychology 25: 577–95. Koehler, D.J. (1991) “Explanation, imagination, and confidence in judgment,” Psychological Bulletin 110: 499–519. Kraemer, K.L., Dickhoven, S., Tierney, S.F. and King, K.L. (1987) Datawars: The Politics of Modeling in Federal Policymaking, New York: Columbia University Press. Krahè, B. (1988) “Victim and observer characteristics as determinants of responsibility attribution to victims of rape,” Journal of Applied Social Psychology 18: 50–8.

228

References

—— (1991) “Social psychological issues in the study of rape,” European Review of Social Psychology 2: 279–309. Kray, L.J. and Galinsky, A.D. (2003) “The debiasing effect of counterfactual mindsets: Increasing the search for disconfirmatory information in group decisions,” Organizational Behavior and Human Decision Processes 91: 69–81. Kristiansen, S.K. (1993) “Verdens heldigste,” VG: 19, 24 February. Kruger, J., Wirtz, D. and Miller, D.T. (2005) “Counterfactual thinking and the first instinct fallacy,” Journal of Personality and Social Psychology 88: 725–35. Kruglanski, A.W. and Webster, D.M. (1996) “Motivated closing of the mind: ‘Seizing’ and ‘freezing’,” Psychological Review 103: 263–8. Kurdek, L.A. (1999) “The nature and predictors of the trajectory of change in marital quality for husbands and wives over the first ten years of marriage,” Developmental Psychology 35: 1283–96. Landman, J. (1987) “Regret and elation following action and inaction: Affective responses to positive versus negative outcomes,” Personality and Social Psychology Bulletin 13: 524–36. —— (1993) Regret: The Persistence of the Possible, New York: Oxford University Press. Landman, J. and Manis, J.D. (1992) “What might have been: Counterfactual thought concerning personal decisions,” British Journal of Psychology 83: 473–7. Landman, J. and Petty, R. (2000) “‘It could have been you’: How states exploit counterfactual thought to market lotteries,” Psychology and Marketing 17: 299–321. Landman, J., Vandewater, A.A., Stewart, A.J. and Malley, J.E. (1995) “Missed opportunities: Psychological ramifications of counterfactual thought in midlife women,” Journal of Adult Development 2: 87–97. Langer, E.J. (1975) “The illusion of control,” Journal of Personality and Social Psychology 32: 311–28. Larrick, R.P. (1993) “Motivational factors in decision theories: The role of self-protection,” Psychological Bulletin 113: 440–50. Larrick, R.P. and Boles, T.L. (1995) “Avoiding regret in decisions with feedback: a negotiation example,” Organizational Behavior and Human Decision Processes 63: 87–97. Larsen, J.T., McGraw, A.P. and Cacioppo, J.T. (2001) “Can people feel happy and sad at the same time?” Journal of Personality and Social Psychology 81: 684–96. Larsen, J.T., McGraw, A.P., Mellers, B.A. and Cacioppo, J.T. (in press) “The agony of victory and the thrill of defeat: Mixed emotional reactions to disappointing wins and relieving losses,” Psychological Science. Lecci, L., Okun, M.A. and Karoly, P. (1994) “Life regrets and current goals as predictors of psychological adjustment,” Journal of Personality and Social Psychology 66: 731–41. Legrenzi, P., Girotto, V. and Johnson-Laird, P.N. (1993) “Focusing in reasoning and decision-making,” Cognition 49: 37–66. Lehman, D.R., Davis, C.G., DeLongis, A., Wortman, C.B., Bluck, S., Mandel, D.R. and Ellard, J.H. (1993) “Positive and negative life changes following bereavement and their relations to adjustment,” Journal of Social and Clinical Psychology 12: 90–112. Lehman, H.C. (1928) “Does it pay to change initial decisions in a true-false test?” School and Society 28: 456–8.

References

229

Lerner, J.S. and Keltner, D. (2000) “Beyond valence: Toward a model of emotionspecific influences on judgement and choice,” Cognition and Emotion 14: 473–93. Lerner, J.S. and Tetlock, P.E. (1999) “Accounting for the effects of accountability,” Psychological Bulletin 125: 255–75. Lerner, M.J. (1980) The Belief in a Just World: A Fundamental Delusion, New York: Plenum. Leventhal, G.S. (1980) “What should be done with equity theory? New approaches to the study of fairness in social relationships,” in K.J. Gergen, M.S. Greenbers and R.H. Willis (eds) Social Exchange: Advances in Theory and Research (pp. 27–53), New York: Plenum. Levy, L. (1989) “Community in recreational settings,” Leisure Sciences 11: 303–22. Lewis, D. (1973) Counterfactuals, Oxford: Blackwell. Lewis, N.A. (2002) “FBI chief admits 9/11 might have been detectable,” New York Times. Online, available at: www.nytimes.com (30 May 2002). Liberman, N., Idson, L.C., Camacho, C.J. and Higgins, E.T. (1999) “Promotion and prevention choices between stability and change,” Journal of Personality and Social Psychology 77: 1135–45. Liljenquist, K.A., Galinsky, A.D. and Kray, L.J. (2004) “Exploring the rabbit hole of possibilities by myself or with my group: Harnessing the benefits of the counterfactual mind-set for information sharing, judgment accuracy, and group coordination,” manuscript submitted for publication, Journal of Behavioral Decision Making 17: 263–79. Lipe, M.G. (1991) “Counterfactual reasoning as a framework for attribution theories,” Psychological Bulletin 109: 456–71. Lockwood, P. (2002) “Could it happen to you? Predicting the impact of downward comparisons on the self,” Journal of Personality and Social Psychology 82: 343–58. Lockwood, P. and Kunda, Z. (1997) “Superstars and me: Predicting the impact of role models on the self,” Journal of Personality and Social Psychology 73: 91–103. Lockwood, P., Jordan, C.H. and Kunda, Z. (2002) “Motivation by positive or negative role models: Regulatory focus determines who best will inspire us,” Journal of Personality and Social Psychology 83: 854–64. Lombard, L.M. (1990) “Causes, enablers, and the counterfactual analysis,” Philosophical Studies 59: 195–211. Lonsway, K.A. and Fitzgerald, L.F. (1995) “Attitudinal antecedents of rape myth acceptance: A theoretical and empirical re-examination,” Journal of Personality and Social Psychology 68: 704–11. Loomes, G. and Sugden, R. (1982) “Regret theory: an alternative theory of rational choice under uncertainty,” Economic Journal 92: 805–24. —— (1986) “Disappointment and dynamic consistency in choice under uncertainty,” Review of Economic Studies 53: 271–82. Lord, C., Ross, L. and Lepper, M. (1979) “Biased assimilation and attitude polarization: The effects of prior theories on subsequently considered evidence,” Journal of Personality and Social Psychology 37: 2098–109. Lorenz, E.N. (1993) The Essence of Chaos, Seattle, WA: University of Washington Press. Lundberg, C.G. and Frost, D.E. (1992) “Counterfactual in financial decision making,” Acta Psychologica 79: 227–44. MacCrimmon, K.R. and Wehrung, D.A. (1986) Taking Risks: The Management of Uncertainty, New York: Free Press.

230

References

Mackie, J.L. (1974) The Cement of the Universe: A Study of Causation, Oxford: Oxford University Press. Macrae, C.N. (1992) “A tale of two curries: Counterfactual thinking and accidentrelated judgments,” Personality and Social Psychology Bulletin 18: 84–7. Macrae, C.N. and Milne, A.B. (1992) “A curry for your thoughts: Empathic effects on counterfactual thinking,” Personality and Social Psychology Bulletin 18: 625–30. Macrae, C.N., Bodenhausen, G.V., Milne, A.B. and Jetten, J. (1994) “Out of mind but back in sight: Stereotypes on the rebound,” Journal of Personality and Social Psychology 67: 808–17. Macrae, C.N., Milne, A.B. and Griffiths, R.J. (1993) “Counterfactual thinking and the perception of criminal behaviour,” British Journal of Social Psychology 84: 221–6. Mandel, D.R. (1995) “Chaos theory, sensitive dependence, and the logistic equation,” American Psychologist 50: 106–7. —— (2003a) “Counterfactuals, emotion, and context,” Cognition and Emotion 17: 139–59. —— (2003b) “Effect of counterfactual and factual thinking on causal judgments,” Thinking and Reasoning 9: 246–65. —— (2003c) “Judgment dissociation theory: An analysis of differences in causal, counterfactual, and covariational reasoning,” Journal of Experimental Psychology: General 132: 419–34. —— (2005) “FINS: A theoretical framework for inferences about necessity and sufficiency,” manuscript submitted for publication. Mandel, D.R. and Dhami, M.K. (2005) “Relation between prisoners’ upward counterfactual thoughts, perceived fairness, and anger,” manuscript submitted for publication. —— (in press) “‘What I did’ versus ‘What I might have done’: Effect of factual and counterfactual thinking on blame, guilt, and shame in prisoners,” Journal of Experimental Social Psychology. Mandel, D.R. and Lehman, D.R. (1996) “Counterfactual thinking and ascriptions of cause and preventability,” Journal of Personality and Social Psychology 71: 450–63. —— (1998) “Integration of contingency information in judgments of cause, covariation, and probability,” Journal of Experimental Psychology: General 127: 269–85. Mandler, G. (1964) “The interruption of behavior,” in E. Levine (ed.) Nebraska Symposium on Motivation (pp. 163–219), Lincoln, NE: University of Nebraska Press. Manktelow, K.I. and Over, D.E. (1991). “Social roles and utilities in reasoning with deontic conditionals,” Cognition 39: 85–105. Mannetti, L., Catellani, P. Fasulo, A. and Pajardi, D. (1991) “Resoconti della condotta nelle deposizioni giudiziarie: analisi dei repertori linguistici,” Giornale Italiano di Psicologia 18: 579–600. March, J.G. (1988) “Variable risk preferences and adaptative aspirations,” Journal of Economic Behavior and Organization 9: 5–24. —— (1991) “Exploration and exploitation in organizational learning,” Organization Science 2: 71–87. —— (1994) A Primer of Decision Making: How Decisions Happen, New York: Macmillan. March, J.G. and Olsen, J.P. (1989) Rediscovering Institutions: The Organizational Basis of Politics, New York: Free Press. March, J.G and Shapira, Z. (1987) “Managerial perspective on risk and risk taking,”

References

231

in J. G. March (ed.) Decisions and Organizations (pp. 76–97), New York: Blackwell. —— (1999) “Variable risk preferences and the focus of attention,” Psychological Review 99: 172–83. March, J.G. and Simon, H.A. (1958) Organizations, New York: Wiley. —— (1993) “Organizations revisited,” Industrial and Corporate Change 23: 299–317. March, J.G., Sproull, L.S. and Tamuz, M. (1991) “Learning from samples of one or fewer,” Organization Science 2: 1–13. Markman, K.D. and McMullen, M.N. (2003) “A reflection and evaluation model of comparative thinking,” Personality and Social Psychology Review 7: 244–67. Markman, K.D. and Tetlock, P.E. (2000a) “Accountability and close counterfactuals: The loser that nearly won and the winner that nearly lost,” Personality and Social Psychology Bulletin 26: 1213–24. —— (2000b) “‘I couldn’t have known’: Accountability, foreseeability and counterfactual denials of responsibility,” British Journal of Social Psychology 39: 313–25. Markman, K.D., Elizaga, R.A. and McMullen, M.N. (2004a) “Counterfactual thinking and regulatory focus,” manuscript in preparation. —— (2004b) “Mental simulation and task persistence,” manuscript in preparation. Markman, K.D., Gavanski, I., Sherman, S.J. and McMullen, M.N. (1993) “The mental simulation of better and worse possible worlds,” Journal of Experimental Social Psychology 29: 87–109. —— (1995) “The impact of perceived control on the imagination of better and worse possible worlds,” Personality and Social Psychology Bulletin 21: 588–95. Martin, L.L. and Stoner, P. (1996) “Mood as input: What we think about how we feel determines how we think,” in L.L. Martin and A. Tesser (eds) Striving and Feeling: Interactions among Goals, Affect, and Self-regulation, Hillsdale, NJ: Erlbaum. Martin, L.L., Ward, D.W., Achee, J.W. and Wyer, R.S. (1993) “Mood as input: People have to interpret the motivational implications of their moods,” Journal of Personality and Social Psychology 64: 317–26. Mathiesen, T. (1990) Prison on Trial, London: Sage Publications. McAdams, D.P., Diamond, A., Aubin, E.D. and Mansfield, E. (1997) “Stories of commitment: The psychosocial construction of generative lives,” Journal of Personality and Social Psychology 72: 678–94. McArthur, L.A. (1972) “The how and what of why: Some determinants and consequences of causal attributions,” Journal of Personality and Social Psychology 22: 171–88. McCloy, R. and Byrne, R.M.J. (2000) “Counterfactual thinking about controllable events,” Memory and Cognition 28: 1071–8. —— (2002) “Semifactual ‘even if’ thinking,” Thinking and Reasoning 8: 41–67. McClure, J.L. and Hilton, D.J. (1997) “For you can’t always get what you want: When script-preconditions make good explanations,” British Journal of Social Psychology 36: 223–40. McClure, J.L., Hilton, D.J. and Sutton, R.M. (2003) “Judgments of intentional and physical causes in causal chains,” unpublished manuscript. McDougall, C., Barnett, R.M., Ashurst, B. and Willis, B. (1987) “Cognitive control of anger,” in B.J. McGurk, D.M. Thornton and M. Williams (eds) Applying Psychology to Imprisonment (pp. 303–13), London: HMSO. McGill, A.L. (1993) “Selection of a causal background: Role of expectation versus feature mutability,” Journal of Personality and Social Psychology 64: 701–7.

232

References

—— (1998) “Relative use of necessity and sufficiency information in causal judgments about natural categories,” Journal of Personality and Social Psychology 75: 70–81. —— (2000) “Counterfactual reasoning in causal judgments: Implications for marketing,” Psychology and Marketing 17: 323–43. McGill, A.L. and Klein, J.G. (1993) “Contrastive and counterfactual thinking in causal judgment,” Journal of Personality and Social Psychology 64: 897–905. —— (1995) “Counterfactual and contrastive reasoning in explanations for performance: Implications for gender bias,” in N.J. Roese and J.M. Olson (eds) What Might Have Been: The Social Psychology of Counterfactual Thinking (pp. 333–51), Mahwah, NJ: Erlbaum. McGill, A.L. and Tenbrunsel, A.E. (2000) “Mutability and propensity in causal selection,” Journal of Personality and Social Psychology 79: 677–89. McIntosh, D.N., Silver, R.C. and Wortman, C.B. (1993) “Religion’s role in adjustment to a negative life event: Coping with the loss of a child,” Journal of Personality and Social Psychology 65: 812–21. McMullen, M.N. (1997) “Affective contrast and assimilation in counterfactual thinking,” Journal of Experimental Social Psychology 33: 77–100. McMullen, M.N. and Markman, K.D. (2000) “Downward counterfactuals and motivation: the wake-up call and the Pangloss effect,” Personality and Social Psychology Bulletin 26: 575–84. —— (2002) “Affective impact of close counterfactuals: Implications of possible futures for possible pasts,” Journal of Experimental Social Psychology 38: 64–70. McMullen, M.N., Eppers, N. and Markman, K.D. (2004) “The impact of counterfactual thinking on motivation and task persistence,” unpublished data. McMullen, M.N., Markman, K.D. and Gavanski, I. (1995) “Living in neither the best nor worst of all possible worlds: Antecedents and consequences of upward and downward counterfactual thinking,” in N.J. Roese and J.M. Olson (eds) What Might Have Been: The Social Psychology of Counterfactual Thinking (pp. 133–67), Mahwah, NJ: Erlbaum. McNamara, P., Durso, R., Brown, A. and Lynch, A. (2003) “Counterfactual cognitive deficit in persons with Parkinson’s disease,” Journal of Neurology, Neurosurgery, and Psychiatry 74: 1065–70. Medin, D.L., Goldstone, R.L. and Markman, A.B. (1995) “Comparison and choice: relations between similarity processes and decision processes,” Psychonomic Bulletin and Review 2: 1–19. Medvec, V.H., Madey, S.F. and Gilovich, T. (1995) “When less is more: Counterfactual thinking and satisfaction among Olympic medalists,” Journal of Personality and Social Psychology, 69: 603–10. Mellers, B.A., Schwartz, A., Ho, K. and Ritov, I. (1997) “Decision affect theory: Emotional reactions to the outcomes of risky options,” Psychological Science 8: 423–9. Mellers, B.A., Schwartz, A. and Ritov, I. (1999) “Emotion-based choice,” Journal of Experimental Psychology: General 128: 332–45. Michotte, A.E. (1963) The Perception of Causality, trans. T.R. Miles and E. Miles, New York: Basic Books (original work published in 1946). Miles, M.S. and Crandall, E.K.B. (1983) “The search for meaning and its potential for affecting growth in bereaved parents,” Health Values 7: 19–23. Mill, J.S. (1872) A System of Logic, Racionative and Inductive (8th edn), London: Longman.

References

233

Miller, D.T. and Gunasegaram, S. (1990) “Temporal order and the perceived mutability of events: Implications for blame assignment,” Journal of Personality and Social Psychology 59: 1111–18. Miller, D.T. and McFarland, C. (1986) “Counterfactual thinking and victim compensation: A test of norm theory,” Personality and Social Psychology Bulletin 12: 513–19. Miller, D.T. and Ross, M. (1975) “Self-serving biases in the attribution of causality: Fact or fiction?” Psychological Bulletin 82: 213–25. Miller, D.T. and Taylor, B.R. (1995) “Counterfactual thought, regret, and superstition: How to avoid kicking yourself,” in N.J. Roese and J.M. Olson (eds) What Might Have Been: The Social Psychology of Counterfactual Thinking (pp. 305–31), Mahwah, NJ: Erlbaum. Miller, D.T. and Turnbull, W. (1990) “The counterfactual fallacy: Confusing what might have been with what ought to have been,” Social Justice Research 4: 1–19. Miller, D.T., Taylor, B. and Buck, M.L. (1991) “Gender gaps: Who needs to be explained?” Journal of Personality and Social Psychology 61: 5–12. Miller, D.T., Turnbull, W. and McFarland, C. (1989) “When a coincidence is suspicious: The role of mental simulation,” Journal of Personality and Social Psychology 57: 581–9. —— (1990) “Counterfactual thinking and social perception: Thinking about what might have been,” in M.P. Zanna (ed.) Advances in Experimental Social Psychology (vol. 23, pp. 305–31), Orlando, FL: Academic Press. Miller, D.T., Visser, P.S. and Staub, B.D. (2004) “When surveillance engenders perceptions of dishonesty: The case of the counterfactual sinner,” unpublished manuscript. Moreno-Ríos, S., García-Madruga, J.A. and Byrne, R.M.J. (2005) “Reasoning with ‘Even if . . .’ and ‘If . . . also’ conditionals,” manuscript submitted for publication. Morris, M.W. and Moore, P.C. (2000) “The lessons we (don’t) learn: Counterfactual thinking and organizational accountability after a close call,” Administrative Science Quarterly 45: 737–65. Morris, M.W., Moore, P.C. and Sim, D.L.H. (1999) “Choosing remedies after accidents: Counterfactual thoughts and the focus on fixing ‘human error’,” Psychonomic Bulletin and Review 6: 579–85. Murray, N., Sujan, H., Hirt, E.R. and Sujan, M. (1990) “The influence of mood on categorization: A cognitive flexibility interpretation,” Journal of Personality and Social Psychology 59: 411–25. Mussweiler, T. (2003) “Comparison processes in social judgment: Mechanisms and consequences,” Psychological Review 110: 472–89. Mussweiler, T. and Strack, F. (2000a) “Consequences of social comparison: Selective accessibility, assimilation, and contrast,” in J. Suls and L. Wheeler (eds) Handbook of Social Comparisons: Theory and Research (pp. 253–70), New York: Kluwer/Plenum. —— (2000b) “The ‘relative self’: Informational and judgmental consequences of comparative self-evaluation,” Journal of Personality and Social Psychology 79: 23–38. Nario, M.R. and Branscombe, N.R. (1995) “Comparison processes in hindsight and causal attribution,” Personality and Social Psychology Bulletin 21: 1244–55. Nario-Redmond, M.R. and Branscombe, N.R. (1996) “It could have been better or it might have been worse: Implications for blame assignment in rape cases,” Basic and Applied Psychology 18: 347–66.

234

References

Nasco, S.A. and Marsh, K.L. (1999) “Gaining control through counterfactual thinking,” Personality and Social Psychology Bulletin 25: 556–68. National Research Council (Assembly of Engineering Committee on FAA Airworthiness Certification Procedures) (1980) Improving Aircraft Safety: FAA Certification of Commercial Passenger Aircraft, Washington, DC: National Academy of Sciences. Neustadt, R.E. and May, E.R. (1986) Thinking in Time: The Uses of History for Decision-Makers, New York: Free Press. N’gbala, A. and Branscombe, N.R. (1995) “Mental simulation and causal attribution: When simulating an event does not affect fault assignment,” Journal of Experimental Social Psychology 31: 139–62. —— (1997) “When does action elicit more regret than inaction and is counterfactual thinking the mediator of this effect?” Journal of Experimental Social Psychology 33: 324–43. Niedenthal, P.M., Tangney, J.P. and Gavanski, I. (1994) “‘If only I weren’t’ versus ‘if only I hadn’t’: Distinguishing shame and guilt in counterfactual thinking,” Journal of Personality and Social Psychology 67: 585–95. Nisbett, R.E. and Ross, L. (1980) Human Inference: Strategies and Shortcomings of Social Judgment, Englewood Cliffs, NJ: Prentice-Hall. “No 9–11 Compensation for Flight Attendant” (2003) Associated Press. Online, available at: www.softcom.net/webnews/wed/do/Aattacks-workers-compensation. RzIU_DNR.html (27 November 2003). Norem, J.K. and Cantor, N. (1986) “Defensive pessimism: ‘Harnessing’ anxiety as motivation,” Journal of Personality and Social Psychology 52: 1208–17. Norman, D.A. (1990) “Cognitive science in the cockpit,” paper presented at the Aerospace Human Factors Symposium, Moffet Field, CA: NASA Ames Research Center. —— (1992) Turn Signals are the Facial Expressions of Automobiles, Reading, MA: Addison-Wesley. Northrop, A., Dutton, W.H. and Kraemer, K.L. (1982) “The management of computer applications in local government,” Public Administration Review 42: 234–44. Oettingen, G. (1996) “Positive fantasy and motivation,” in P.M. Gollwitzer and J.A. Bargh (eds) The Psychology of Action: Linking Cognition and Motivation to Action (pp. 236–59), New York: Guilford Press. Oettingen, G. and Mayer, D. (2002) “The motivating function of thinking about the future: Expectations versus fantasies,” Journal of Personality and Social Psychology 83: 1198–212. Oettingen, G., Pak, H. and Schnetter, K. (2001) “Self-regulation of goal setting: Turning free fantasies about the future into binding goals,” Journal of Personality and Social Psychology 80: 736–53. Olson, J.M., Buhrmann, O. and Roese, N.J. (2000) “Comparing comparisons: An integrative perspective on social comparison and counterfactual thinking,” in J. Suls and L. Wheeler (eds) Handbook of Social Comparison: Theory and Research (pp. 379–98), New York: Kluwer/Plenum. Palich, L.E. and Bagby, D.R. (1995) “Using cognitive theory to explain entrepreneurial risk-taking: Challenging conventional wisdom,” Journal of Business Venturing 10: 425–38. Pallak, S.R. and Davies, J.M. (1982) “Finding fault versus attributing responsibility: Using facts differently,” Personality and Social Psychology Bulletin 8: 454–9.

References

235

Parkes, C.M. and Weiss, R.S. (1983) Recovery from Bereavement, New York: Basic Books. Pennington, G.L. and Roese, N.J. (2002) “Regulatory focus and mental simulation,” in S.J. Spencer, J.M. Olson and M.P. Zanna (eds) Motivated Social Perception: The Ontario Symposium (vol. 9, pp. 277–98), Mahwah, NJ: Erlbaum. “Perspectives ’95” (1995), Newsweek: 42, 25 December. Pfeffer, J. (1992) Managing with Power, Boston, MA: Harvard Business School Press. Pitagorsky, G. (2000) “Lessons learned through process thinking and review,” PM Network 14: 35–40. Pollock, J.L. (1976) Subjunctive Reasoning, Dordrecht, Netherlands: Reidel. Prange, G.W. (1982) Miracle at Midway, New York: McGraw-Hill. Pritchard, D. and Smith, M. (2004) “The psychology and philosophy of luck,” New Ideas in Psychology 22: 1–28. Proctor, R.A. (1887) Chance and Luck, London: Longman. Quelhas, A.C. and Byrne, R.M.J. (2003) “Reasoning with deontic and counterfactual conditionals,” Thinking and Reasoning 9: 43–65. Read, S.J. (1987) “Constructing causal scenarios: A knowledge structure approach to causal reasoning,” Journal of Personality and Social Psychology 52: 288–302. Rehder, B. and Hastie, R. (2001) “Causal knowledge and categories: The effects of causal beliefs on categorization, induction, and similarity,” Journal of Experimental Psychology: General 130: 323–60. Reile, P.J. and Briggs, L.J. (1952) “Should students change their initial answers on objective-type tests? More evidence regarding an old problem,” Journal of Educational Psychology 43: 110–15. Reiss, D. (2001) “Counterfactuals and inquiries after homicide,” Journal of Forensic Psychiatry 12: 169–81. Rescher, N. (1995) Luck: The Brilliant Randomness of Everyday Life, New York: Farrar Straus & Giroux. Rhue, J.W. and Lynn, S.J. (1987) “Fantasy proneness and psychopathology,” Journal of Personality and Social Psychology 53: 327–36. Rips, L.J. (1994) The Psychology of Proof, Cambridge, MA: MIT Press. Ritov, I. (1996) “Probability of regret: Anticipation of uncertainty resolution in choice,” Organizational Behavior and Human Decision Processes 66: 228–36. Ritov, I. and Baron, J. (1990) “Reluctance to vaccinate: omission bias and ambiguity,” Journal of Behavioral Decision Making 3: 263–77. —— (1995) “Outcome knowledge, regret and omission bias,” Organizational Behavior and Human Decision Processes 64: 119–27. Roese, N.J. (1994) “The functional basis of counterfactual thinking,” Journal of Personality and Social Psychology 66: 805–18. —— (1997) “Counterfactual thinking,” Psychological Bulletin 121: 133–48. —— (2001) “The crossroads of affect and cognition: Counterfactuals as compensatory cognitions,” in G. Moskowitz (ed.) Cognitive Social Psychology: The Princeton Symposium on the Legacy and Future of Social Cognition (pp. 307–16), Mahwah, NJ: Erlbaum. Roese, N.J. and Olson, J.M. (1993) “The structure of counterfactual thought,” Personality and Social Psychology Bulletin 19: 312–19. —— (1995a) “Counterfactual thinking: A critical overview,” in N.J. Roese and J.M. Olson (eds) What Might Have Been: The Social Psychology of Counterfactual Thinking (pp. 1–55), Mahwah, NJ: Erlbaum.

236

References

—— (1995b) “Functions of counterfactual thinking,” in N.J. Roese and J.M. Olson (eds) What Might Have Been: The Social Psychology of Counterfactual Thinking (pp. 169–97), Mahwah, NJ: Erlbaum. —— (1995c) “Outcome controllability and counterfactual thinking,” Personality and Social Psychology Bulletin 21: 620–8. —— (eds) (1995d) What Might Have Been: The Social Psychology of Counterfactual Thinking, Mahwah, NJ: Erlbaum. —— (1996) “Counterfactuals, causal attributions, and the hindsight bias: A conceptual integration,” Journal of Experimental Social Psychology 32: 197–227. —— (1997) “Counterfactual thinking: The intersection of affect and function,” in M.P. Zanna (ed.) Advances in Experimental Social Psychology (vol. 29, pp. 1–59), San Diego, CA: Academic Press. Roese, N.J., Hur, T. and Pennington, G.L. (1999) “Counterfactual thinking and regulatory focus: Implications for action versus inaction and sufficiency versus necessity,” Journal of Personality and Social Psychology 77: 1109–20. Roese, N.J., Sanna, L.J. and Galinsky, A.D. (2004) “The mechanics of imagination: Automaticity and control in counterfactual thinking,” in J.A. Bargh, J.S. Uleman and R. Hassin (eds) The New Unconscious (pp. 138–70), New York: Oxford University Press. Roseman, I.J., Wiest, C. and Swartz, T.S. (1994) “Phenomenology, behaviors and goals differentiate discrete emotions,” Journal of Personality and Social Psychology 67: 206–11. Sagan, S. and Waltz, K. (1995) The Spread of Nuclear Weapons: A Debate, New York: Norton. Salmon, W.C. (1984) Scientific Explanation and the Causal Structure of the World, Princeton, NJ: Princeton University Press. —— (1998) Causality and Explanation, Oxford: Oxford University Press. Sanna, L.J. (1996) “Defensive pessimism, optimism, and simulating alternatives: Some ups and downs of prefactual and counterfactual thinking,” Journal of Personality and Social Psychology 71: 1020–36. —— (1997) “Self-efficacy and counterfactual thinking: Up the creek with and without a paddle,” Personality and Social Psychology Bulletin 23: 654–66. —— (2000) “Mental simulation, affect, and personality: A conceptual framework,” Current Directions in Psychological Science 9: 168–73. Sanna, L.J. and Turley, K.J. (1996) “Antecedents to spontaneous counterfactual thinking: Effects of expectancy violation and outcome valence,” Personality and Social Psychology Bulletin 22: 906–19. Sanna, L.J. and Turley-Ames, K.J. (2000) “Counterfactual intensity,” European Journal of Social Psychology 30: 273–96. Sanna, L.J., Chang, E.C. and Meier, S. (2001) “Counterfactual thinking and selfmotives,” Personality and Social Psychology Bulletin 27: 1023–34. Sanna, L.J., Meier, S. and Wegner, E. (2001) “Counterfactuals and motivation: Mood as input to affective enjoyment and preparation,” British Journal of Social Psychology 40: 235–56. Sanna, L.J., Parks, C.D., Meier, S., Chang, E.C., Kassin, B.R., Lechter, J.L., TurleyAmes, K.J. and Miyake, T.M. (2003) “A game of inches: Spontaneous use of counterfactuals by broadcasters during major league baseball playoffs,” Journal of Applied Social Psychology 33: 455–75. Sanna, L.J., Schwarz, N. and Stocker, S.L. (2002) “When debiasing backfires:

References

237

Accessible content and accessibility experiences in debiasing hindsight,” Journal of Experimental Psychology: Learning, Memory, and Cognition 28: 497–502. Sanna, L.J., Turley-Ames, K.J. and Meier, S. (1999) “Mood, self-esteem, and simulated alternatives: Thought-provoking affective influences on counterfactual direction,” Journal of Personality and Social Psychology 76: 543–58. Santamaria, C., Espino, O. and Byrne, R.M.J. (in press) “Counterfactual primes,” Journal of Experimental Psychology: Learning, Memory and Cognition. Savitsky, K., Medvec, V.H. and Gilovich, T. (1997) “Remembering and regretting: The Zeigarnik effect and the cognitive availability of regrettable actions and inactions,” Personality and Social Psychology Bulletin 23: 248–57. Schank, R.C. and Abelson, R.P. (1977) Scripts, Plans, Goals and Understanding: An Enquiry into Human Knowledge Structures, Hillsdale, NJ: Erlbaum. Scheier, M.F. and Carver, C.S. (1992) “Effects of optimism on psychological and physical well-being: Theoretical overview and empirical update,” Cognitive Therapy and Research 16: 201–28. Schkade, D.A. and Kahneman, D. (1998) “Does living in California make people happy? A focusing illusion in judgments of life satisfaction,” Psychological Science 9: 340–6. Schkade, D.A. and Kilbourne, L.M. (1991) “Expectation–outcome consistency and hindsight bias,” Organizational Behavior and Human Decision Processes 49: 105–23. Schustack, M.W. and Sternberg, R.J. (1981) “Evaluation of evidence in causal inference,” Journal of Experimental Psychology: General 110: 101–20. Schwartz, P. (1991) The Art of the Long View, New York: Doubleday/Currency. Schwarz, N. (1990) “Feelings as information: Informational and motivational functions of affective states,” in E.T. Higgins and R.M. Sorrentino (eds) Handbook of Motivation and Cognition: Foundations of Social Behavior (vol. 2, pp. 527–61), New York: Guilford Press. Schwarz, N. and Clore, G.L. (1983) “Mood, misattribution, and judgments of wellbeing: Informative and directive functions of affective states,” Journal of Personality and Social Psychology 45: 513–23. Schwarz, N., Bless, H., Strack, F., Klumpp, G., Rittenauer-Schatka, H. and Simons, A. (1991) “Ease of retrieval as information: Another look at the availability heuristic,” Journal of Personality and Social Psychology, 61: 195–202. Seelau, E.P., Seelau, S.M., Wells, G.L. and Windschitl, P.D. (1995) “Counterfactual constraints”, in N.J. Roese and J.M. Olson (eds) What Might Have Been: The Social Psychology of Counterfactual Thinking (pp. 57–79), Mahwah, NJ: Erlbaum. Segura, S. and McCloy, R. (2003) “Counterfactual thinking in everyday life situations: Temporal order effects and social norms,” Psicologica 24: 1–15. Segura, S. and Morris, M.W. (2001) “Might have been versus would have been: Perceived causal relationship as determinants of asserted likelihood of counterfactual outcomes,” European Bulletin of Social Psychology 13: 80. Segura, S., Fernandez-Berrocal, P. and Byrne, R.M.J. (2002) “Temporal and causal order effects in thinking about what might have been,” Quarterly Journal of Experimental Psychology: Human Experimental Psychology 55A: 1295–305. Semin, G. and Fiedler, K. (1991) “The Linguistic Category Model, its bases, applications and range,” in W. Stroebe and M. Hewstone (eds) European Review of Social Psychology (vol. 2, pp. 1–30), Chichester: Wiley. Senge, P., Kleiner, A., Roberts, C., Ross, R. and Smith, B. (1994) The Fifth

238

References

Discipline Fieldbook: Strategies and Tools for Building a Learning Organization, New York: Doubleday. Shafir, E. and Tversky, A. (1992) “Thinking through uncertainty: Nonconsequential reasoning and choice,” Cognitive Psychology 24: 449–74. Shah, J., Higgins, E.T. and Friedman, R. (1998) “Performance incentives and means: How regulatory focus influences goal attainment,” Journal of Personality and Social Psychology 74: 285–93. Shek, D.T.L. (1994) “Meaning in life and adjustment amongst midlife parents in Hong Kong,” International Forum for Logotherapy 17: 102–7. Sherman, S.J. and McConnell, A.R. (1995) “Dysfunctional implications of counterfactual thinking: When alternatives to reality fail us,” in N.J. Roese and J.M. Olson (eds) What Might Have Been: The Social Psychology of Counterfactual Thinking (pp. 199–231), Mahwah, NJ: Erlbaum. —— (1996) “Counterfactual thinking in reasoning,” Applied Cognitive Psychology 10: 113–24. Sherman, S.J., Skov, R.B., Hervitz, E.F. and Stock, C.B. (1981) “The effects of explaining hypothetical future events: From possibility to probability to actuality and beyond,” Journal of Experimental Social Psychology 17: 142–58. Shiffrin, R.M. and Schneider, W. (1977) “Controlled and automatic human information processing: II. Perceptual learning, automatic attending, and general theory,” Psychological Review 84: 127–90. Shimanoff, S.B. (1984) “Commonly named emotions in everyday conversations,” Perceptual and Motor Skills 58: 514. Sim, D.L.H. and Morris, M.W. (1998) “Representativeness and counterfactual thinking: The principle that antecedent and outcome correspond in magnitude,” Personality and Social Psychology Bulletin 24: 596–609. Simon, H.A. (1967) “Motivational and emotional controls of cognition,” Psychological Review 74: 29–39. Simon, L., Arndt, J., Greenberg, J., Pyszczynski, T. and Solomon, S. (1998) “Terror management and meaning: Evidence that the opportunity to defend the worldview in response to mortality salience increases the meaningfulness of life in the mildly depressed,” Journal of Personality 66: 359–82. Sloman, S. and Lagnado, D. (2004) “Causal invariance in reasoning and learning”, in B. Ross (ed.) Handbook of Learning and Motivation (vol. 44, pp. 387–425), New York: Academic Press. Slovic, P. (1975) “Choice between equally valued alternatives,” Journal of Experimental Psychology: Human Perception and Performance 1: 280–7. Smith, A. (1789/2002) The Theory of Moral Sentiments, ed. K. Haakonssen, Cambridge: Cambridge University Press. Smith, R.H., Webster, J.M, Parrott, W.G. and Eyre, H.L. (2002) “The role of public exposure in moral and nonmoral shame and guilt,” Journal of Personality and Social Psychology 83: 138–59. Spellman, B.A. (1997) “Crediting causality,” Journal of Experimental Psychology: General 126: 1–26. —— (2001) “Wine, women, and Wells: Why thinking about more (consequentchanging) counterfactuals leads to greater attributions of causality,” paper presented at the European Association for Experimental Social Psychology Small Group Meeting on Counterfactual Thinking (Chairs: D.R. Mandel, D.J. Hilton and P. Catellani), Aix-en-Provence, France, May.

References

239

—— (2003) “The relation between counterfactual and causal reasoning: Availability mediates some of the similarities and differences in judgments,” unpublished manuscript. Spellman, B.A. and Kincannon, A. (2001) “The relation between counterfactual (‘but for’) and causal reasoning: Experimental findings and implications for jurors’ decisions,” Law and Contemporary Problems 64: 241–64. Spellman, B.A. and Mandel, D.R. (1999) “When possibility informs reality: Counterfactual thinking as a cue to causality,” Current Directions in Psychological Science 8: 120–3. —— (2003) “Causal reasoning, Psychology of,” in L. Nadel (ed.) Encyclopedia of Cognitive Science (vol. 1, pp. 461–6), London: Nature Publishing. Spellman, B.A. and Meyers, N.M. (2003) “Wine, women, and food poisoning: The more-choices leads to more-causality effect in the relation between counterfactual and causal reasoning,” manuscript in preparation. Sperber, D. and Wilson, D. (1986) Relevance: Communication and Cognition, Cambridge, MA: Harvard University Press. Stalnaker, R.C. (1968) “A theory of conditionals,” in N. Rescher (ed.) Studies in Logical Theory, Oxford: Blackwell. Starbuck, W.H. and Milliken, F.J. (1988) “Challenger: Fine-tuning the odds until something breaks,” Journal of Management Studies 25: 319–40. Stose, S.J. and Spellman, B.A. (2002) “A causal definition of counterfactual regret,” paper presented at the meeting of the Society for Judgment and Decision Making, Kansas City, MO, November. Strack, F. (1992) “The different routes to social judgments: Experiential versus informational strategies,” in L.L. Martin and A. Tesser (eds) The Construction of Social Judgments (pp. 249–75), Mahwah, NJ: Erlbaum. Strack, F., Schwarz, N. and Gschneidinger, E. (1985) “Happiness and reminiscing: The role of time perspective, affect, and mode of thinking,” Journal of Personality and Social Psychology 49: 1460–9. Strassfeld, R.N. (1992) “If . . . : Counterfactuals in the law,” George Washington Law Review 60: 339–416. Strauss, H.M. (1979) A Talent for Luck, New York: Random House. Suedfeld, P. and Tetlock, P.E. (2001) “Cognitive styles,” in A. Tesser and N. Schwarz (eds) Blackwell International Handbook of Social Psychology: Intra-individual Processes (vol. 1, pp. 282–304), London: Blackwell. Sutton, R.M. and McClure, J.L. (2001) “Covariational influences on goal-based attribution: An integrative model,” Journal of Personality and Social Psychology 80: 222–36. Tait, R. and Silver, R.C. (1989) “Coming to terms with major negative life events,” in J.S. Uleman and J.A. Bargh (eds) Unintended Thought (pp. 357–81), New York: Guilford Press. Tamuz, M. (1987) “The impact of computer surveillance on air safety reporting,” Columbia Journal of World Business 22: 69–77. Taylor, S.E. (1983) “Adjustment to threatening events: A theory of cognitive adaptation,” American Psychologist 38: 1161–73. —— (1989) Positive Illusions: Creative Self-deception and the Healthy Mind, New York: Basic Books. —— (1991) “Asymmetrical effects of positive and negative events: The mobilization–minimization hypothesis,” Psychological Bulletin 110: 67–85.

240

References

Taylor, S.E. and Brown, J.D. (1988) “Illusion and well-being: A social psychological perspective on mental health,” Psychological Bulletin 103: 193–210. Taylor, S.E. and Pham, L.B. (1996) “Mental simulation, motivation, and action,” in P.M. Gollwitzer and J.A. Bargh (eds) The Psychology of Action: Linking Cognition and Motivation to Behavior (pp. 219–35), New York: Guilford Press. Taylor, S.E. and Schneider, S.K. (1989) “Coping and the simulation of events,” Social Cognition 7: 174–94. Taylor, S.E., Kemeny, M.E., Reed, G.M., Bower, J.E. and Gruenewald, T.L. (2000) “Psychological resources, positive illusions, and health,” American Psychologist 55: 99–109. Tedeschi, R.G. and Calhoun, L.G. (1996) “The Posttraumatic Growth Inventory: Measuring the positive legacy of trauma,” Journal of Traumatic Stress 9: 455–72. Teigen, K.H. (1988) “Reddet av flaks [Saved by luck],” Tidsskrift for Norsk Psykologforening 25: 439–46. —— (1995) “How good is good luck? The role of counterfactual thinking in the perception of lucky and unlucky events,” European Journal of Social Psychology 25: 281–302. —— (1996) “Luck: The art of a near miss,” Scandinavian Journal of Psychology 37: 156–71. —— (1997) “Luck, envy, and gratitude: It could have been different,” Scandinavian Journal of Psychology 38: 313–23. —— (1998a) “Hazards mean luck: Counterfactual thinking and perceptions of good and bad fortune in reports of dangerous situations and careless behavior,” Scandinavian Journal of Psychology 39: 235–48. —— (1998b) “When the unreal is more likely than the real: Post-hoc probability judgments and counterfactual closeness,” Thinking and Reasoning 4: 147–77. Teigen, K.H., Evensen, P.C., Samoilow, D.K. and Vatne, K.B. (1999) “Good and bad luck: How to tell the difference,” European Journal of Social Psychology 29: 981–1010. Tetlock, P.E. (1998) “Close-call counterfactuals and belief system defense: I was not almost wrong but I was almost right,” Journal of Personality and Social Psychology 75: 639–52. —— (1999) “Theory-driven reasoning about possible pasts and probable futures: Are we prisoners of our preconceptions?” American Journal of Political Science 43: 335–66. —— (2000) “Cognitive biases and organizational correctives: Do both disease and cure depend on the politics of the beholder?” Administrative Science Quarterly 45: 293–326. —— (2002) “Social functionalist frameworks for judgment and choice: Intuitive politicians, theologians and prosecutors,” Psychological Review 109: 451–71. Tetlock, P.E. and Belkin, A. (1996a) Counterfactual Thought Experiments in World Politics: Logical, Methodological, and Psychological Perspectives, Princeton, NJ: Princeton University Press. —— (1996b) “Counterfactual thought experiments in world politics: Logical, methodological, and psychological perspectives,” in P.E. Tetlock and A. Belkin (eds) Counterfactual Thought Experiments in World Politics: Logical, Methodological, and Psychological Perspectives (pp. 3–38), Princeton, NJ: Princeton University Press. Tetlock, P.E. and Lebow, R.N. (2001) “Poking counterfactual holes in covering

References

241

laws: Cognitive styles and historical reasoning,” American Political Science Review 95: 829–43. Tetlock, P.E. and Visser, P. (2000) “Thinking about Russia: Possible pasts and probable futures,” British Journal of Social Psychology 39: 173–96. Tetlock, P.E., Lebow, R. and Parker, G.P. (eds) (2004) Unmaking the West: Exploring Alternative Histories of Counterfactual Worlds, Ann Arbor, MI: University of Michigan Press. Thibaut, J.J. and Kelley, H.H. (1959) The Social Psychology of Groups, New York: Wiley. Thompson, V.A. and Byrne, R.M.J. (2002) “Reasoning counterfactually: Making inferences about things that didn’t happen,” Journal of Experimental Psychology: Learning, Memory, and Cognition 28: 1154–70. Trabasso, T. and Bartolone, J. (2003) “Story understanding and counterfactual reasoning,” Journal of Experimental Psychology: Learning, Memory, and Cognition 29: 904–23. Trabasso, T.R. and Sperry, L.L. (1985) “The causal basis for deciding the importance of story events,” Journal of Memory and Language 24: 595–611. Trabasso, T.R. and van den Broek, P. (1985) “Causal thinking and story comprehension,” Journal of Memory and Language 24: 612–30. Tulving, E. (2004) Memory, consciousness, and time, Sixteenth annual convention of the American psychological society keynote address, Chicago, May. Turley, K.J., Sanna, L.J. and Reiter, R.L. (1995) “Counterfactual thinking and perceptions of rape,” Basic and Applied Social Psychology 17: 285–303. Turnbull, W.M. (1986) “Everyday explanation: The pragmatics of puzzle resolution,” Journal for the Theory of Social Behaviour 16: 141–60. Turnbull, W.M. and Slugoski, B.R. (1988) “Conversational and linguistic processes in causal attribution,” in D. Hilton (ed.) Contemporary Science and Natural Explanation: Commonsense Conceptions of Causality (pp. 66–93), Brighton: Harvester Press. Tversky, A. (1972). “Elimination by aspects: A theory of choice,” Psychological Review 79: 281–99. Tversky, A. and Fox, C. (1995) “Weighting risk and uncertainty,” Psychological Review 102: 269–83. Tversky, A. and Kahneman, D. (1982a) “Causal schemas in judgment under uncertainty,” in D. Kahneman, P. Slovic and A. Tversky (eds) Judgment under Uncertainty: Heuristics and Biases (pp. 117–28), New York: Cambridge University Press. —— (1982b) “Evidential impact of base rates,” in D. Kahneman, P. Slovic and A. Tversky (eds) Judgment under Uncertainty: Heuristics and Biases (pp. 153–60), New York: Cambridge University Press. —— (1983) “Extensional versus intuitive reasoning: The conjunction fallacy in probability judgment,” Psychological Review 91: 293–315. Tversky, A. and Koehler, D.J. (1994) “Support theory: A nonextensional representation of subjective probability,” Psychological Review 101: 547–67. Tversky, A. and Shafir, E. (1992) “The disjunction effect in choice under uncertainty,” Psychological Science 3: 305–9. Tykocinski, O.E. (2001) “I never had a chance: Using hindsight tactics to mitigate disappointments,” Personality and Social Psychology Bulletin 27: 376–82. van der Pligt, J., Zeelenberg, M., van Dijk, W.W., de Vries, N.K. and Richard, R. (1998) “Affect, attitudes, and decisions: Let’s be more specific,” European Review of Social Psychology 8: 34–66.

242

References

van Dijk, E. and Zeelenberg, M. (2003) “The discounting of ambiguous information in economic decision making,” Journal of Behavioral Decision Making 16: 341–52. —— (2005) “On the psychology of ‘if only’: Regret and the comparison between factual and counterfactual outcomes,” Organizational Behavior and Human Decision Processes 97: 152–60. Vasquez, J.A. (1997) “The realist paradigm and degenerative versus progressive research programs: An appraisal of neotraditional research on Waltz’s balancing proposition,” American Political Science Review 91: 899–913. Vertzberger, Y. (1991) The World in their Minds: Information Processing, Cognition, and Perception in Foreign Policy Decisionmaking, Stanford, CA: Stanford University Press. Vinokur, A. and Ajzen, I. (1982) “Relative importance of prior and immediate events: a causal primacy effect,” Journal of Personality and Social Psychology 42: 820–9. Vispoel, W.P. (1998) “Reviewing and changing answers on computer-adaptive and self-adaptive vocabulary tests,” Journal of Educational Measurement 35: 328–45. Wagenaar, W.A. and Keren, G.B. (1988) “Chance and luck are not the same,” Journal of Behavioral Decision Making 1: 65–75. Walsh, C.R. and Byrne, R.M.J. (2004) “Counterfactual thinking: The temporal order effect,” Memory and Cognition 32: 369–78. —— (2005) “Counterfactual thinking about reasons for acting,” manuscript submitted for publication. Waltz, K.N. (1979) Theory of International Politics, Reading, MA: Addison-Wesley. Wasserman, D., Lempert, R.O. and Hastie, R. (1991) “Hindsight and causality,” Personality and Social Psychology Bulletin 17: 30–5. Wegener, D.T. and Petty, R.E. (1997) “The flexible correction model: The role of naïve theories of bias in bias correction,” in M.P. Zanna (ed.) Advances in Experimental Social Psychology (vol. 29, pp. 141–208), Mahwah, NJ: Erlbaum. Wegner, D.M. and Bargh, J.A. (1998) “Control and automaticity in social life,” in D.T. Gilbert, S.T. Fiske and G. Lindzey (eds) The Handbook of Social Psychology (vol. 1, pp. 446–96), New York: McGraw-Hill. Weiner, B. (1985) “‘Spontaneous’ causal thinking,” Psychological Bulletin 109: 74–84. —— (1986) An Attribution Theory of Achievement, Motivation and Emotion, New York: Springer. Weinstein, N.D. and Klein, W.M. (1995) “Resistance of personal risk perceptions to debiasing interventions,” Health Psychology 14: 132–40. Wells, G.L. and Gavanski, I. (1989) “Mental simulation of causality,” Journal of Personality and Social Psychology 56: 161–9. Wells, G.L., Taylor, B.R. and Turtle, J.W. (1987) “The undoing of scenarios,” Journal of Personality and Social Psychology 53: 421–30. Westermann, R., Spies, K., Stahl, G. and Hesse, F.W. (1996) “Relative effectiveness and validity of mood induction,” European Journal of Social Psychology 26: 557–80. White, M. (1987) “The virtue of Japanese mothers: Cultural definitions of women’s lives,” Daedalus 116: 149–63. Wiener, R.L., Gaborit, M., Pritchard, C.C. and McDonough, E.M. (1994) “Counterfactual thinking in mock jurors’ assessments of negligence: A preliminary investigation,” Behavioral Sciences and the Law 12: 89–102.

References

243

Wilensky, R.C. (1983) Planning and Understanding: A Computational Approach to Human Reasoning, Reading, MA: Addison-Wesley. Williams, T., Goodwillie, S.M., Eden, C.L. and Ackermann, F.R. (1996) “Modelling the management of complex projects: Industry/university collaboration,” paper presented at the UNESCO International Conference on Technology Management (UnIG ’96), Istanbul, Turkey. Wilson, T.D., Wheatley, T., Meyers, J.M., Gilbert, D.T. and Axsom, D. (2000) “Focalism: A source of durability bias in affective forecasting,” Journal of Personality and Social Psychology 78: 821–36. Wiseman, R. (2003) The Luck Factor, London: Century. Wohl, M.J.A. and Enzle, M.E. (2003) “The effects of near wins and near losses on self-perceived personal luck and subsequent gambling behavior,” Journal of Experimental Social Psychology 39: 184–91. Wolff, P. and Song, G. (2003) “Models of causation and the semantics of causal verbs,” Cognitive Psychology 47: 276–332. Wrosch, C. and Heckhausen, J. (2002) “Perceived control of life regrets: Good for young and bad for old adults,” Psychology and Aging 17: 340–50. Zeelenberg, M. (1999) “Anticipated regret, expected feedback and behavioral decision-making,” Journal of Behavioral Decision Making 12: 93–106. Zeelenberg, M. and Beattie, J. (1997) “Consequences of regret aversion: 2. Additional evidence for effects of feedback on decision making,” Organizational Behavior and Human Decision Processes 72: 63–78. Zeelenberg, M. and Pieters, R. (2004) “Consequences of regret aversion in real life: The case of the Dutch postcode lottery,” Organizational Behavior and Human Decision Processes 93: 155–68. Zeelenberg, M., Beattie, J., van der Pligt, J. and de Vries, N.K. (1996) “Consequences of regret aversion: Effects of expected feedback on risky decisionmaking,” Organizational Behavior and Human Decision Processes 65: 148–58. Zeelenberg, M., Inman, J.J. and Pieters, R.G.M. (2001) “What we do when decisions go awry: Behavioral consequences of experienced regret,” in E.U. Weber, J. Baron and G. Loomes (eds.) Conflict and Tradeoffs in Decision-making (pp. 136–55), Cambridge, MA: Harvard University Press. Zeelenberg, M., van den Bos, K., van Dijk, E. and Pieters, R. (2002) “The inaction effect in the psychology of regret,” Journal of Personality and Social Psychology 82: 314–27. Zeelenberg, M., van der Pligt, J. and de Vries, N.K. (2000a) “Attributions of responsibility and affective reactions to decision outcomes,” Acta Psychologica 104: 303–15. Zeelenberg, M., van der Pligt, J. and Manstead, A.S.R. (1998a) “Undoing regret on Dutch television: Apologizing for interpersonal regrets involving actions and inactions,” Personality and Social Psychology Bulletin 24: 1113–19. Zeelenberg, M., van Dijk, W.W. and Manstead, A.S.R. (1998b) “Reconsidering the relation between regret and responsibility,” Organizational Behavior and Human Decision Processes 74: 254–72. —— (2000b) “Regret and responsibility resolved? Evaluating Ordóñez and Connolly’s (2000) conclusions,” Organizational Behavior and Human Decision Processes 81: 143–54. Zeelenberg, M., van Dijk, W.W., Manstead, A.S.R. and van der Pligt, J. (1998c)

244

References

“The experience of regret and disappointment,” Cognition and Emotion 12: 221–30. —— (2000c) “On bad decisions and disconfirmed expectancies: The psychology of regret and disappointment,” Cognition and Emotion 14: 521–41. Zeelenberg, M., van Dijk, W.W., van der Pligt, J., Manstead, A.S.R., van Empelen, P. and Reinderman, D. (1998d) “Emotional reactions to the outcomes of decisions: The role of counterfactual thought in the experience of regret and disappointment,” Organizational Behavior and Human Decision Processes 75: 117–41.

Author index

Adams, J.S. 175 Associated Press 77

Gould, S.J. 207 Gunasegaram, S. 69

Bagby, D.R. 104 Bartolone, J. 14–15 Bell, D.E. 149 Ben Ze’ev, A. 129 Bothwell, R.K. 166 Branscombe, N.R. 35, 36, 38, 166, 169, 175 Brickman, P. 53 Buck, M.L. 174, 175, 177 Burt, M.R. 191 Buunk, B.P. 157 Byrne, R.M.J. 5, 71, 72

Hart, H.L.A. 3–4, 29, 33, 47, 52, 55–6 Heckhausen, J. 149 Heider, F. 136 Henik, E. 4, 6–7, 21 Hilton, D.J. 4, 16, 52–3, 59, 98 Honoré, A.M. 3–4, 29, 33, 47, 52, 55–6

Catellani, P. 6, 190, 191–2, 193, 194–5, 196 Cheng, P.W. 48 Coelho, P. 130 Collingwood, R.G. 20 Craik, K. 5 Dawes, R.M. 26 Dhami, M.K. 4, 6, 169, 171, 173, 174 Duhon, K.W. 166 Fischhoff, B. 117 Fitzgerakd, L.F. 191 Frankl, V. 110 Frijda, N.H. 150, 155, 160 Galinsky, A.D. 5, 102, 123 Gavanski, I. 13, 33–4, 35, 36 Gibbons, F.X. 157 Gilovich, T. 155 Girotto, V. 48, 56, 58, 59 Goldinger, S.D. 26 Goldvarg, E. 17 Goodman, N. 2

Italian Supreme Court 183 James, W. 103 Jervis, R. 26 Johnson-Laird, P.N. 5, 17 Jones, E.E. 189 Kahneman, D. 2–3, 4, 12, 13, 14, 17, 21, 23, 26, 38–9, 44, 45, 85, 86, 88, 92, 114, 115, 129, 132, 147, 153, 168, 169–70, 186, 210 Kaplan 123 Kelley, H.H. 11, 44–5, 46, 172 Keren, G.B. 130 Kincannon, A.P. 4, 29 Krahé, B. 190 Kray, L.L.J. 5 Kruger, J. 123 Landman, J. 149 Larrick, R.P. 157 Lebow, R.N. 213, 216 Lehman, D.R. 14, 15–16, 17–20, 19–20 Leventhal, G.S. 180 Levy, L. 107 Lewis, D. 2, 115 Liljenquist, K.A. 5 Lockwood, P. 87, 91

246

Author index

Lonsway, K.A. 191 Loomes, G. 149 McClure, J.L. 4, 55 McConnell, A.R. 69 McGill, A.L. 188 McGillis, D. 189 Mackie, J.L. 3–4, 12, 15–16, 25, 28–9 McMullen, M.N. 5, 78, 83, 86, 91, 100, 146 Macrae, C.N. 187, 195 Mandel, D.R. 4, 6, 14, 15–16, 17–20, 23–5, 32–3, 52, 147–8, 152, 169, 171, 173, 174, 188 Mandler, G. 88 March, J.G. 103 Markman, K.D. 5, 78, 83, 84, 85, 86, 87, 90, 91, 92, 100, 146 Martin, L.L. 81 Medin, D.I. 156 Medvec, V.H. 100, 143 Milesi, P. 6, 190, 193, 194–5, 196 Mill, J.S. 98, 101–2 Miller, D.T. 3, 12, 69, 88, 112, 121–2, 124, 147, 153, 168, 169–70, 174, 175, 177, 210 Moore, P.C. 105 Morris, M.W. 5, 105, 122 Mussweiler, T. 79, 92, 93

Schwartz, P. 108 Schwarz, N. 80 Segura, S. 5, 69 Shafir, E. 154 Shakespeare, W. 44 Shapira, Z. 103 Sherman, S.J. 69 Shimanoff, S.B. 149 Sim, D.L.H. 122 Simon, H. 95 Slugoski, B.R. 4, 98 Smith, M. 144, 145 Smith, R.H. 170 Souza, K.A. 4, 6 Spellman, B.A. 4, 20, 29, 31, 36, 49–51, 54, 57, 69 Stalnaker, R. 2 Stose, S.J. 4 Strack, F. 79 Sugden, R. 149

Nario-Redmond, M.R. 166, 175 Newton, I. 1 N’gbala, A. 36, 38 Niedenthal, P.M. 152, 182n1 Novick, L.R. 48

Tamuz, M. 105 Taylor, S. E. 113 Teigen, K.H. 6, 131, 134, 135–6, 137, 139–40, 141 Tetlock, P.E. 4, 6–7, 21, 86, 100, 213, 216 Trabasso, T.R. 14–15 Tulving, E. 1, 27 Turley, K.J. 166, 186 Turnbull, W. 112 Tversky, A. 2–3, 4, 13, 14, 17, 23, 26, 38–9, 92, 114, 129, 147, 153, 154, 186, 213 Tykocinski, O.E. 155

Oettingen, G. 83, 85 Olson, J. 3, 4, 13, 91, 178, 182

van Dijk, E. 6, 154–5, 156 Varey, C.A. 44, 45, 85, 115, 132

Palich, L.E. 104 Pritchard, D. 144, 145 Rescher, N. 130–1, 132, 143 Roese, N.J. 3, 4, 5, 13, 88, 99–100, 101, 142, 178, 182

Wagenaar, W.A. 130 Walsh, C.R. 5, 71, 72 Wells, G.I. 13, 33–4, 35, 36, 51, 58 Wiener, B. 166 Wiseman, R. 145 Wrosch, C. 149

Salmon, W.C. 23, 49 Sanna, L.J. 88

Zeelenberg, M. 6, 39–40, 151–2, 154–5, 156, 159, 160, 170, 173

Subject index

Note: page numbers in italics refer to figures or tables abnormality criterion 47–8; antecedents 166; behavior 197; Gestalt 114; intentionality 52; mutability 183–4; scenarios 98–9; social context 189 accidents 54–5, 129–30, 132, 133; see also traffic accident action-inaction effect 38–41, 153, 191–2, 197 actions: behavioral models 95–6; controllability 59, 71–2; counterfactual alternative 68; implementation 103–4, 107–8; intentional 47–9; rule following 95; voluntary 55–7; see also outcomes actuality principle 23, 32–3 affect: accessibility mechanism 79–80; assimilation 78, 88–9; contrast 77–8, 129; downward counterfactuals 3, 110–11, 143; downward evaluation 81; future improvement 89; negative 81, 168; performance 91; positive 82; upward reflection 82; see also emotion alternatives 33–4, 38, 61, 67–73, 213–16 anger 174–5, 176 antideterminism 207, 216 Apparent Reality, Law of 155 assimilation 78, 80, 85–6, 88–9 attribution: ANOVA model 172; blame 165, 170; causality 51, 56; expectation 189; responsibility 165, 185 attribution theory 11, 44, 52–3, 136 behavior: abnormality 197; changes 95, 100; emotion 88; gender 190–1; guilt 182n1; norms 188, 190–1; social category 187; trivial 122, 123

behavioral models 95–8, 161 belief-system defense 6–7 bias 41–2, 43n6, 106, 117, 200, 216 blame: attribution 165, 170; counterfactual 112, 168, 171–2; factual thinking 171–2; guilt 167, 182; routine-violation 166, 169; temporality 111–12 bounded rationality 22 brain disorders 110–12 “but for” 38 butterfly effect 119 causal chains 47–8, 49–50, 58; coincidence 50–1; intuitive multiple regression analysis 48–9; justice system 174; opportunity 51, 55–7, 58, 59; pre-emptive 52; sufficiency 59; unfolding 51, 52–5, 58, 59 causal induction studies 16, 46 causal reasoning 22–3, 28, 30–3, 34–5, 39–41 causality 16, 21, 25, 45–7, 168; attribution 51, 56; complex 45; distal 52–3, 55–7; events 30–2, 44–5; inference 29, 80, 81; motivation 80–3; mutability 30–1, 36–8, 42; necessary judgment 54; proximal 53, 55–7; ratings 37–8; SPA 31–2; sufficiency 16–17, 29–30, 32, 57, 59; timing 36–8 closeness 115, 139; see also near misses closure needs 201, 202, 206–10, 211–13, 215 cognitive-conservatism bias 200 cognitive-consistency theory 201 cognitive dissonance theory 90 coincidence chains 50–1

248

Subject index

communication 194–6, 197 Comparative Feeling, Law of 150 comparisons: avoiding 158–60; counterfactual 129, 152–3; incomparable 156–7; individual differences 157–8; known/unknown 154–6; reflection and evaluation model 78–9; regret 156–7 compensation payments 166–7, 187 complacency 83, 90 computer models 106–7 conditionals 61, 63, 65–6 confidence 104 confusion matrices 17, 18, 19 conjunctions 63–4 consequences 28–9 conservatism 204, 205–6 conspiracy theories 120, 121–2 control 20, 59, 71–2, 111, 178 coping strategies 113 “could have been” 15; see also “mighthave-been” counterfactual cognition 202–10 counterfactual constraints 183–4, 194–6 counterfactual fallacy 112 counterfactual thinking 1–2, 12–13, 25–6, 27; automatic/controlled processing 88–9; causal explanations 21, 25, 30–3, 45–7; close 85–6, 135–6, 212–13; dysfunctional 111–12; emotion 147–8, 152, 169–70, 181; explicit 190–2; functions 5, 110–12; implicit 192–3; interdependence 94–5, 101–3; normality-restoring 2–3, 12–15, 26, 67, 88–9, 133; regulatory focus 87–8; self-focused 101, 188; sense-making 113, 114, 115, 116; see also downward counterfactual thinking; upward counterfactual thinking counterfactual thinking experiments 199–200, 201 counterfactual thinking research 166–8 covariational thinking 21, 22–3, 48 crediting causality model 49–51, 54, 57; see also Spellman’s probability updating account crime 181–2, 189 dangers 101, 137–8 decision making: motivation 155–6; regret 160; regret-averse 150, 158–9; social comparison 157; uncertainty 103–4

defying the odds 115–18, 119, 120 Delta-P 48, 54, 57 determinism 21–2, 117, 120 disappointment 101, 151, 168, 170 disaster 132–3, 135–6 disjunction effect 154 dissonant/consonant scenarios 206–10 distance factors 143–4 downward counterfactual thinking: affect 3, 110–11, 143; avoidance 101–3, 105–6; contrast 80; experiential learning 99–100; motivation 83; persistence 86–7; prisoners 168–80; and upward counterfactual thinking 77–8 downward evaluation 79, 80; causal inferences 81; complacency 83, 90; controlled procession 89; motivation 83; persistence, tasks 87; positive affect 81; prevention focus 82–3 downward reflection 80; automatic procession 89; causal inferences 81; motivation 83, 90; negative affect 81; persistence, tasks 87; prevention-focus 82–3, 87, 91–2; promotionorientation 87, 91–2 emotion 6; behavior 88; comparisons 129; counterfactual thinking 147–8, 152, 169–70, 181; lottery 147; luck 146; outcomes 169; prisoners 170; upward evaluation 80; see also affect emotional amplification hypothesis 3, 168, 169–70, 186 envy 141 evaluation: evidence 211–12; outcomes 96–8, 100–1, 104–6; psychology of 105–6; thinking modes 5, 78–9; upward counterfactual thinking 109; see also downward evaluation; upward evaluation “even if” 7, 13, 35–6, 105 events 129; antecedents 28–9; causality 30–2, 44–5; consequences 28–9; general/particular 46–7, 55; interpretation 185; positive 143; public/private 44; recent 68–70; repetition 90; single 31–2 evolutionary theory 215 exam performance 80, 83, 89, 99–100 exceptionality effect 115–18, 184, 186–7, 189 existentialism 112–13 expectations 189, 194–5, 197–8, 210

Subject index 249 experiential learning 95–100 experts 212, 213, 215 factual thinking 171–2 failure 129 fairness 165, 176–7, 178, 180 fantasies 85–6, 89, 90 fate: autonomy 120; avoidance 115; defying the odds 119; exceptionality 115–18; hindsight bias 117; “if only” 119–20; love 118, 119, 120; mutability 119–20; tragedy 119–20 feelings-as-information hypothesis 80–1 first-instinct fallacy 123 fortune 140–1 frustration 92 gambling 111–12, 130, 131–2, 142 games of chance 99, 155 gender differences 188, 189, 190–1 generalizations, universal 46 goal setting 86–7, 90–1, 149, 194–5 gratitude 141 guilt 180–1; behavior 182n1; blame 167, 182; counterfactual fallacy 112; prisoners 165, 168, 170, 172–3; selfblame 6, 101; shame 172–3 hindsight bias 41–2, 43n6, 106, 117, 216 historians 199–200, 215; see also experts history 200–2; Cold War 209–10; Eurocentrism 207–8; mental models theory 200–1; policy-making 199–200, 202; reactions to new discoveries 210–12; South Africa 205–6; Soviet Union 202–5; World War II 209 human error fallacy 103 ideology 208, 211 “if only”: control 71–2; fate 119–20; human error 103; inevitability 51; mutability 35, 59; rape case 183; simulation 105; traffic accident 186 “if-then” 13–14, 124–5 imprisonment 168, 181–2 inaction 191–2 incongruence hypothesis 177 indeterminism 21–2 individual differences 157–8 inferences 29, 62, 64–7, 80, 81 information processing 95, 96–7, 213 intentionality 48, 52–5, 180

interdependence 94–5, 100–3 interpretation 65–6, 184, 185 intuition 92, 131, 150 judgment dissociation theory (JDT) 22–5, 32, 169 jurors 70, 167, 180, 193 justice system 178, 180, 181; causal chains 174; counterfactual thinking 166–8, 183, 185; exceptionality effect 184; hypothetical cases 166, 167 laws, universal 47 learning: experiential 95–100; organizations 104–8; personal 101 life episodes 136, 165 likelihood 21–2; see also probabilities lottery 132, 147 love 118, 119, 120 luck 129–30, 136; carelessness 138–9; disaster 132–3, 135–6; emotion 146; formula for 131; fortune 140–1; gambling 130; good/bad 130–1, 134, 141–3, 145–6; narrative structure 139–40; near misses 118, 142; norm theory 142–3; perceptions 144–5; post-computed 133; risk 137–9, 144; unexpectedness 130–2, 139–40 meaning 110, 112–13, 120, 125 memory 1, 67, 70–1 mental models theory 5, 13–14, 61–3, 73, 200–1 mental rehearsal 103–4 “might-have-been” 105, 118, 152–4, 178, 179; see also “could have been” missed opportunities 147–9 mood-as-input perspective 81, 87, 88 motivation 87–8, 111, 180; causality 80–3; decision making 155–6; downward counterfactuals 83; downward reflection 90; trade-offs 89–91; upward evaluation 83, 90; upward reflection 83 multiple necessary conditions schema 46 mutability 37, 38; abnormality 48, 183–4; antecedents 14; causality 30–1, 36–8, 42, 56; conspiracy theories 122; ‘even if’ effect 35–6; fate 119–20; ideology 208; ‘if only’ effect 35, 59; intentionality 48; legislation 124–5; meaning 112, 125; mental simulation 114–15;

250

Subject index

mutability – contd. necessity judgments 57–8, 59; order effects 38; outcomes 14; reality 177–8; sense-making 114, 115; temporal 111–12 near-death experiences 116 near-histories 106–7 near misses 114–15; accidents 129–30, 133; air traffic 105–6; flights 98, 129; gambling 142; luck 118, 142; opportunities 135–6; political forecasters 100; regret 153; tragedy 116 necessary causes view 15–16 necessity judgments 54, 57–8, 59 negligence suit 43n6, 166, 188 negotiation 102, 123 nonconformity effect 184, 187–93, 196, 197 norm theory 3, 142–3, 153, 169, 183–4, 186–7 norms 186–7, 188, 190–1; see also social norms obligation 66 obsessive-compulsive behavior 124 Olympic medalists 100, 143 opportunities missed 135–6, 147–9 opportunity chains 51, 55–7, 58, 59 optimism 90 order effects 34–5, 38, 153 organizations: action implementation 107–8; learning 95, 104–8; rule induction 106–7; simulations 109 outcomes: counterfactual 133–4; emotional reactions 169; evaluation 96–8, 100–1, 104–6; factual 133–4; fairness 175; meaning 120; mood 88; mutability 14 peer ratings 134, 137–8 perpetrator/victim 166, 190 persistence in tasks 83–7 personal growth 114 pessimism 103, 200, 215–16 policy-making 199–200, 202–10 politics 100, 107–8 possibilities 28, 69; counterfactual conditionals 64, 66–7; forbidden 72; memory 70–1; mental models theory 61, 62–3; multiple 73; pre-/postaction 68; reality 28; true 62 possible worlds concept 63

post-traumatic stress disorder 77–8 prefactual scenarios 107, 108, 109 prejudice 124–5 prevention-focus 17–20, 28–30, 82–3, 87, 91–2 prisoners 6, 165–6, 170 –171; anger 176; downward counterfactual thinking 168–80; emotion 170; fairness 176–7; guilt 165, 168, 170, 172–3; “might-have-been” 178, 179; self-blame 172; shame 172–3; upward counterfactual thinking 168–80, 177–80 probabilities: after event 41–2; alternatives 38; attractiveness 133–4; conditional 22; distance 133–4, 143–4; manipulated 49–50; representativeness 21–2; subjective 144 project management 105 promotion-orientation 82, 83, 87–8, 91–2 psychosocial research 187, 196–7, 198 race 189 rape cases 175; acquaintance rape 183; action–inaction effect 191–2; behavioral norms 190–1; blame 166; date rape 35; “if only” 183; perpetrator 166, 187, 190; routine-violation 187; social norms 189–90; victim 183, 186, 191–2, 193, 195, 196 Rape Myth Acceptance Scale 191 rational models 95–6, 99 reality 182; alternatives 67–72, 213–15; causal reasoning 28; mutability 177–8; possibilities 28; preconceptions 215; self-esteem 90 rebound effect 195 recency effects 50 reflection 21–2, 78–9; see also downward reflection; upward reflection reflection and evaluation model (REM) 80–1, 93; empirical testing 83–8; negative affect 81; prevention focus 82–3; promotion focus 82; social comparisons 87–8 regret 39–41, 129, 147, 148–9, 151–2, 170; action/inaction 38–41; causal reasoning 39–41; comparative nature 150, 156–7; contrasts 100–1; decision making 158–9, 160; ‘mighthave-been’ 152–4; minimizing strategy 158–60; near miss 153;

Subject index 251 studies of 149–50; time 160–1; upward counterfactual thinking 168 responsibility 120, 165, 166, 184, 185 risk 99, 100, 137–9, 144 routine-violation 116–17, 166, 169, 186, 187 rules 95, 98–9, 101–3, 106–7 scenario analysis 108, 109 self-blame 6, 101, 172, 180–1 self-evaluation 79, 90 sense-making 113, 114, 115, 116 shame 101, 165, 168, 170, 172–3 similarities 21–2, 25, 92–3 simulations: counterfactual 45, 98–9; counterfactual conditional 13–14; elaborative/automatic 86; “even if” 105; heuristics 17; “if only” 105; learning from experience 94–5; organizations 109; outcomes 100–1; prefactual 108; upward/downward 78–9, 94, 99–104 social categories 187, 188–9 social comparisons 79, 87–8, 91–3, 157 Social Context Model of Counterfactual Constraints 184–5, 196–8 social norms 22, 72, 193; nonconformity effect 184, 187–93; rape case 189–90; social categories 188–9; violating 102–3 Spellman’s probability-updating account (SPA) 20, 24–5, 31–2; see also crediting causality model stereotypes 124, 189–90, 191, 193, 195 stock investment 39, 86, 158 subjunctive mood 66 substitution principle 23 success 129, 188 superstition 120, 122–4 suspicion 121–2 temporal chains 49–50 temporality 1, 45, 111–12, 118–19, 160–1

terrorist attacks, 11 September 2001 11, 77–8, 116, 124; FBI 19; World Trade Center 59, 121 thought experiments 26 time see temporality timing effects 36–8 traffic accident 36–8, 70–1, 186–7 tragedy 113, 114, 116, 119–20 uncertainty 103–4 unexpectedness 118–20, 130–2, 139–40 upward counterfactual thinking: affect 3; anger 174–5; and downward counterfactuals 77–8; evaluation 109; experiential learning 99–100; fairness 174–5; improvement 101–3; judgment dissociation theory 169; persistence 86–7; preparative 80, 110–11; prisoners 165–6, 168–80; regret 168; reoffending 182 upward evaluation 79; emotion 80; motivation 83, 90; negative affect 81; persistence in tasks 87; preventionfocus 87, 91–2; promotionorientation 83, 87, 91–2 upward reflection: complacency 83, 90; controlled procession 89; fantasies 85–6, 89; motivation 83; persistence in tasks 85, 87; positive affect 82; promotion-orientation 83 verb effects 44 victims: gender differences 189; inaction 191–2; innocent 116–17; perpetrator 166, 190; rape case 183, 186, 191–2, 193, 195, 196; routineviolation 186; stereotypes 190 vignette studies 4, 142 World Trade Center 59, 116, 121 World War I 199, 201, 203, 208, 209 World War II 132, 199, 209 “would-have-been” 105

eBooks – at www.eBookstore.tandf.co.uk

A library at your fingertips!

eBooks are electronic versions of printed books. You can store them on your PC/laptop or browse them online. They have advantages for anyone needing rapid access to a wide variety of published, copyright information. eBooks can help your research by enabling you to bookmark chapters, annotate text and use instant searches to find specific words or phrases. Several eBook files would fit on even a small laptop or PDA. NEW: Save money by eSubscribing: cheap, online access to any eBook for as long as you need it.

Annual subscription packages We now offer special low-cost bulk subscriptions to packages of eBooks in certain subject areas. These are available to libraries or to individuals. For more information please contact [email protected] We’re continually developing the eBook concept, so keep up to date by visiting the website.

www.eBookstore.tandf.co.uk