Social Comprehension and Judgment: The Role of Situation Models, Narratives, and Implicit Theories

Robert S. Wyer, Jr.

SOCIAL COMPREHENSION AND JUDGMENT The Role of Situation Models, Narratives, and Implicit Theories

Social Comprehension and Judgment The Role of Situation Models, Narratives, and Implicit Theories

Social Comprehension and Judgment The Role of Situation Models, Narratives, and Implicit Theories

Robert S. Wyer, Jr. Hong Kong University of Science and Technology

2004

LAWRENCE ERLBAUM ASSOCIATES, PUBLISHERS Mahwah, New Jersey London

Copyright © 2004 by Lawrence Erlbaum Associates, Inc. All rights reserved. No part of the book may be reproduced in Any form, by photostat, microform, retrieval system, or any other means, without the prior written permission of the publisher.

Lawrence Erlbaum Associates, Inc., Publishers 10 Industrial Avenue Mahwah, New Jersey 07430

Cover design by Kathryn Houghtaling Lacey

Library of Congress Cataloging-in-Publication Data Wyer, Robert S., Jr. Social comprehension and judgment : the role of situation models, narratives, and implicit theories / Robert S. Wyer. p. cm. Includes bibliographical references and index. ISBN 0-8058-4190-3 (alk. paper) 1. Social perception. 2. Human information processing—Social aspects. 3. Judgment. 4. Memory. I. Title. BF323.S63W94 2003 153—dc21 2003052865

Books published by Lawrence Erlbaum Associates are printed on acid-free paper, and their bindings are chosen for strength and durability. Printed in the United States of America 10 9 8 7 6 5 4 3 2 1

Dedicated to Rashmi, Kathy, and Natalie and The Social Cognition Group

Contents

Preface

xi

PART I: INTRODUCTION AND BASIC CONCEPTS 1.

INTRODUCTION Types of Social Knowledge 6 The Representation of Knowledge in Memory 10 Situation Models 16 Generalized Event Representations 17 Generalized Entity Representations 22

2.

THE ARCHITECTURE OF THE INFORMATION-PROCESSING SYSTEM Limitations of Wyer and Srull’s (1989) Theory 26 Structure of the Processing System 29 Storage and Retrieval Processes 37 Goal-Directed Versus Non-Goal-Directed Processing: Spontaneous Reminding Processes 41 Concluding Remarks 45

3.

DETERMINANTS AND EFFECTS OF KNOWLEDGE ACCESSIBILITY General Considerations 47 Frequency and Recency Effects of Knowledge Activation on Judgments and Decisions 49 Effects of Awareness 53 Effects of Knowledge Accessibility at Different Stages of Processing 57 The Effects of Prior Judgments on Subsequent Ones 63 Determinants and Consequences of the Accessibility of Goals and Motives 65 The Activation of Behavioral Dispositions 72 Concluding Remarks 75

3

26

47

vii

viii

CONTENTS

PART II: COMPREHENSION PROCESSES 4.

THE COMPREHENSION AND VALIDATION OF INFORMATION ABOUT FAMILIAR PEOPLE AND EVENTS: THE ROLE OF SITUATION MODELS The Content and Structure of Situation Models 80 The Role of Visual Imagery in Comprehending Social Events 83 Spontaneous Comprehension Processes 88 The Comprehension of Verbal Statements 92 Spontaneous Validity Judgments 98 Concluding Remarks 104

79

5.

THE CONSTRUCTION OF EPISODE MODELS AND GENERALIZED NARRATIVE REPRESENTATIONS The Construction of Episode Models 106 The Construction of Generalized Narrative Representations 116

106

6.

THE IMPACT OF GENERALIZED EVENT REPRESENTATIONS ON COMPREHENSION AND MEMORY The Role of Prototypic Event Representations in the Comprehension of One’s Own and Others’ Experiences 125 The Effects of Communicating About an Experience on Memory and Judgment 137 Concluding Remarks 149

124

7.

PRAGMATIC INFLUENCES ON THE INTERPRETATION OF STATEMENTS MADE IN A SOCIAL CONTEXT General Considerations 152 Responses to Uninformative Messages 155 Reactions to Communications That One Believes to Be Untrue 160 Pragmatic Influences on Impression Formation in Informal Conversations 165 The Effect of Nonverbal Behavior and Conversational Style on Impression Formation, Judgment, and Information Seeking 178 Pragmatic Communication in Close Relationships 186 Concluding Remarks 187

151

CONTENTS 8.

THE DYNAMICS OF HUMOR ELICITATION: THE EFFECTS OF INFORMATIONAL CONTEXT ON THE INTERPRETATION AND ELABORATION OF NARRATIVES Historical Background 192 A Comprehension–Elaboration Theory of Humor Elicitation 198 Cognitive Elaboration Processes 209 Reactions to Disparaging Humor 218 Concluding Remarks 227

ix 189

PART III: INFERENCE PROCESSES 9.

CRIMES, VACATIONS, AND POLITICAL CANDIDATES: THE CONSTRUCTION AND USE OF NARRATIVES IN SOCIAL JUDGMENTS The Impact of Narratives on Jury Decision Making 232 A Comparison of Narrative-Based Versus Nonnarrative-Based Information Processing 234 The Influence of Narratives on Political Impression Formation 240 The Role of Visual Imagery on Verbal Information Processing: Additional Considerations 249 General Effects of Narrative-Based Processing on Explanation and Prediction 255 Concluding Remarks 265

231

10.

THE IMPACT OF IMPLICATIONAL MOLECULES AND IMPLICIT THEORIES ON INFERENCES ABOUT ONESELF AND OTHERS General Considerations 267 Effects of Implicit Theories on Reconstructive Memory 274 Motivational Influences on Implicit Theory Construction and Use 278 The Role of Implicit Theories in Heuristic-Based Judgments 280 Individual Differences in Implicit Theories 286 Implicit Theories of Social Support: The Effect of Perspective on Theory Activation 288 Cultural Differences in Implicit Theories 290 The Role of Implicit Theories in Marital Satisfaction 295 Effects of Implicit Theories on Behavioral Decisions 299 Final Comment 305

267

x

CONTENTS

11.

THE ROLE OF AFFECT IN INFORMATION PROCESSING General Considerations 307 Affect and Concept Accessibility 312 The Influence of Affect on the Recall and Interpretation of Information 315 The Use of Affect as Information 322 Affective Versus Descriptive Bases for Judgment 325 Affect-Confirmation Processes 332 Adjustments for Bias 336 Automatic Influences of Affect on Information Processing 339 Motivational Influences of Affect on Information Processing 342 A Performance-Feedback Model of Affect as Information 344 Affect, Attitudes, and Behavior 358 Final Comments 363

306

12.

EPILOGUE: THE BOOK, THE AUTHOR, AND PHILOSOPHICAL RUMINATIONS

365

APPENDIX: SUMMARY OF POSTULATES REFERENCES

374 380

AUTHOR INDEX SUBJECT INDEX

405 415

Preface

Authors often become disenchanted with the books they have written almost as soon as they are published. Perhaps this is inevitable. Although a book reports the culmination of one’s theoretical and empirical work at the time it goes to press, it nevertheless contains the seeds of discontent. That is, the author in writing the book becomes painfully aware of the limitations of his research and the conceptualization underlying it, and consequently is already motivated to move beyond the confines of this conceptualization by the time it appears in print. At least, this has been true of me. A book I published in 1974, entitled Cognitive Organization and Change, provided a fairly rigorous analysis of cognitive consistency theories of the way beliefs and attitudes were organized in memory, individual and situational differences in the cognitive responses to new information, and the way in which different features of information were combined to form judgments. The book assumed that humans were analogous to computers, generating outputs based on operations that were specified in various “software” routines that were activated by the “user” (in the case of humans, the demands of the social environment). By the time the book was published, however, I had already begun to appreciate the need for a more general conceptualization of mental representation. A second book (Social Inference and Attribution, with Don Carlston, published in 1979) proposed a rather complex associative network model of social memory that took into account the role of knowledge accessibility in judgments and decisions. In this book, we recognized the impact that script-like representations of events could have on social judgment. However, we did not explore this possibility with any degree of rigor. More generally, the cumbersome nature of the network representation, coupled with the fact that it provided little insight into how the information represented was actually used, convinced me of the futility of using this approach to account for many phenomena that were being identified in research on person impression formation and the use of judgmental heuristics. xi

xii

PREFACE

The third book (Memory and Cognition in Its Social Context, with Thom Srull) appeared in 1989. This book remedied several deficiencies of the 1979 volume, providing a comprehensive description of social information processing that specified the activities at several different stages: the interpretation, organization, and storage of information in memory, its later retrieval and integration with other relevant knowledge to make a subjective judgment or decision, and the translation of this subjective response into an overt one. This rather grandiose formulation not only generated specific predictions that received empirical support but had the flexibility to integrate the implications of more specific theories of research that had been developed. By the time the book was published, however, I had become aware of several obvious limitations. For one thing, the title of the book was a misnomer. The theory I had proposed purported to describe the comprehension and use of information of the sort that one encounters in daily life, In fact, however, it was applied almost exclusively to the processing of linguistic information of the sort people receive in laboratory situations but are unlikely to encounter elsewhere (e.g., lists of unrelated behaviors of a fictitious person, verbal descriptions of routine sequences of events of little intrinsic interest, etc.) In short, although the book purported to be about cognition “in its social context,” the social context was typically absent in much of the research to which the theory was applied. Stimulated by our dissatisfaction with the applicability of the theory to phenomena outside the laboratory, my students and I began to examine the processing of information of the sort people actually encounter in the course of their daily lives. This research took many directions. For example, we investigated the impressions people formed on the basis of a person’s opinions about social policies with which they agreed or disagreed rather than abstract behaviors that had few implications for matters of personal concern to the recipients. Moreover, we studied the impressions people formed on the basis of information conveyed in conversations rather than written lists of unrelated behaviors described out of context. Moreover, we explored the way individuals respond to information that violated social norms of communication and that was conveyed for the purpose of eliciting amusement rather than to convey an impression of its referent or to espouse a particular point of view. In doing so, we began to take more seriously the fact that information conveyed in social situations was frequently in the form of a narrative, or story, and was often accompanied by visual images that were either based on pictures or direct observation or generated by the recipients themselves. Finally, we explored the role of affect in information processing.

PREFACE

xiii

This book provides a conceptual integration of this work. It proposes a general theoretical formulation of the way that the sort of information acquired in the course of daily life is comprehended and represented in memory, and how it is later used as a basis for judgments and behavioral decision. In doing so, it takes into account both the spontaneous comprehension of information about specific persons and events and the more deliberative, goal-directed interpretation of information that occurs when information is acquired in a social context. In addition, it considers not only the representation of this information in memory but also the way the information is later used as a basis for judgments and decisions. A major emphasis throughout the volume is on the construction and use of narrative representations of knowledge and the way visual images influence the formation of these representations and the judgments that are based on them. The role of affective reactions in this cognitive activity is also discussed. The book is divided into three sections. After a conceptual overview of the approach to be taken (chap.1), it outlines the general theoretical framework within which the discussion is housed, focusing on assumptions about the storage and retrieval of information (chap. 2) and reviews recent research on the impact of knowledge accessibility on judgments and decisions (chap. 3). Although the architecture of the information-processing system we propose is similar to that described in the 1989 volume, the conceptualization of memory processes provided here is quite different. The second section deals specifically with the comprehension of information. Chapter 4 discusses the mental simulations that are formed of situationally and temporally specific events (i.e., “event models”) in the course of comprehending descriptions of familiar persons and events, and specifies the conditions in which information that stimulates these simulations is spontaneously recognized as true or false. Chapter 5 discusses both the representation of situation-specific sequences of events that are formed spontaneously in the course of comprehending the events and more abstract representations that constitute implicit theories about the causal and temporal relations among events of the sort people encounter in daily life. Chapter 6 discusses the manner in which generalized narrative representations influence memory for the specific experiences that exemplify them. In this context, the influence of socialization practices on the construction of narratives about oneself, and on the self-concepts that result from these representations, are considered. The next two chapters discuss cognitive reactions to information that violates either normative expectations for the type of information that is conveyed in the social context at hand (chap. 7) or expectations that result

xiv

PREFACE

from previously acquired information about the information’s referents (chap. 8). In the latter context, we consider the effects of expectancy violations on the elicitation of humor and the extent to which the cognitive activities that occur in response to these violations can increase or decrease the humor that is elicited in a social context. The final section describes the inferences that are based on information of the sort that is conveyed in social situations. Chapter 9 outlines the way visual characterizations of events (either pictures or self-generated images) influence the impact of verbal descriptions of the events. Chapter 10 discusses the effects of people’s implicit theories about their social world on the inferences they draw from specific experiences. Finally, chapter 11 analyzes the role of affective reactions in information processing and the judgments and decisions that result from this processing. Many individuals have contributed both directly and indirectly to the content of this volume. My consideration of the role of narratives in information processing was stimulated in large part by a multisemester workshop conducted at Illinois by Julian Rappaport. G. A. Radvansky’s empirical and theoretical work on the nature of situation models was also a major stimulant. My conceptualization of the role of affect in information processing was profoundly influenced by my participation in an ongoing seminar on emotion and cognition conducted by Jerry Clore. Clore’s work and that of his students are well represented in this volume. However, I am particularly indebted to the University of Illinois Social Cognition group. This group, an informal consortium of faculty, postdoctoral students, and graduate students from several disciplines, met weekly over a 20-year period to discuss theoretical and experimental issues in the general area of information processing. Members of this group, both individually and collectively, have had a profound influence on the ideas represented in this volume and on my professional career more generally. I am particularly indebted to two subsets of individuals whose work is widely cited in this volume. Galen Bodenhausen, Lee Budesheim, Bob Fuhrman, Deborah Gruenfield, Alan Lambert, Victor Ottati, Jong Won Park, L. J. Shrum, and Dave Trafimow, participants in the Group during the period in which my views on social information processing were rapidly changing, were not only inspirational colleagues but close friends, and are largely responsible for many of the changes in my thinking that occurred. The stimulation provided in later years by Dolores Albarracin, Stan Colcombe, Carol Gohm, Linda Isbell, Eric Mankowski and Michaela Wänke was equally influential. The research and ideas of all of these individuals are readily apparent throughout this book. I should also acknowledge the contributions of Donnel Briley, Catherine Yeung, and

PREFACE

xv

Candy Fong, who worked with me at the Hong Kong University of Science and Technology, whose work is also represented. I owe a special note of appreciation to Rashmi Adaval. Both as a collaborator and as a wife, she has contributed in both tangible and intangible ways both to this book in particular and to my personal and professional life more generally. Her research on both the role of narratives in memory and judgment, and on affect and information processing, is prominent throughout this book. However, her personal support and encouragement has been equally invaluable. It is extremely difficult for anyone to be both a colleague and a spouse. Her willingness to play these two often conflicting roles, her tolerance for my idiosyncratic work habits, and her love and support throughout the past decade, has been inexpressibly gratifying. Finally, I want to acknowledge my deep appreciation to Larry Erlbaum and his staff at Lawrence Erlbaum Associates. The support he has given me over the past 30 years has been a constant source of gratitude. Larry’s commitment to excellence, and his sincere dedication to the enhancement of basic research in psychology, have won him the respect and admiration of all psychologists who have had the opportunity to work with him. I can only hope that this volume confirms his confidence in me and approximates the standards of excellence for which he strives. Robert S. Wyer, Jr.

P A R T

I INTRODUCTION AND BASIC CONCEPTS

C H A P T E R

1 Introduction

Human memory is a collection of thousands of stories we remember through experience, stories we remember by having heard them, and stories we remember by having composed them. Any story in memory could have gotten there in one of these three ways. The key point is that, once these stories are there in our memory, we rely upon them for all that we can say and understand. . . . —Schank and Abelson (1995, p. 3)

Schank and Abelson (1995) argued that virtually all the meaningful social knowledge we acquire is in the form of stories. This claim is probably overstated (Brewer, 1995; Rubin, 1995). Nevertheless, the role of narratives in the acquisition and transmission of information in daily life is undeniable. The content of everyday conversations almost invariably includes stories we tell about ourselves and personal acquaintances, descriptions of books we have read or movies we have seen, and jokes that we tell to one another. Narratives are also invoked in the course of understanding the causes or likely consequences of real and hypothetical social events (e.g., the outcome of a presidential election, or the India–Pakistan dispute over Kashmir). Persuasive messages and television commercials often stimulate us to imagine the sequence of events that might result from taking a particular course of action. Life itself is in the form of a narrative, consisting of a sequence of temporally related events that we experience as either participants or observers. In short, narratives are fundamental to an understanding of ourselves and of the world in which we live. The importance of narrative forms of knowledge is recognized in virtually every area of psychology. Research and theory on prose comprehen3

4

CHAPTER 1

sion and learning have long been concerned with the comprehension and mental representation of stories and temporally ordered sequences of events (Graesser, 1981; Mandler & Johnson, 1977; Stein & Glenn, 1979). Developmental psychologists (e.g., Miller, 1994; Nelson, 1993) have identified the role of stories in parent–child interaction and socialization more generally. Narrative forms of knowledge can often constitute implicit theories that people use both to explain their own and others’ past experiences and to predict the future (Dweck & Leggett, 1988; Ross, 1989). The influence of these theories on behavior and judgments has been identified in research on topics as diverse as personal achievement (Dweck, 1975), predictions of future life events (Oishi, Wyer, & Colcombe, 2000), marital relations (Holmes & Murray, 1995; Murray & Holmes, 1996), personality development (McAdams, 1988) and community–individual relations (Mankowski & Rappaport, 1995). Despite the consideration of narratives in these specific areas of research, few attempts have been made to develop a broader theoretical formulation of social information processing that permits their influence to be conceptualized in relation to that of other types of knowledge that people encounter in their daily lives. Several types of knowledge representations are likely to exist (Pennington & Hastie, 1993), and their role in information processing can differ. One must understand the factors that govern the retrieval of this knowledge, how it is used to comprehend and construe the implications of new information to which it is relevant, and the nature of its influence on judgments and behavioral decisions about its referents. This book attempts to provide this understanding. In particular, it is concerned with the comprehension and use of information of the sort people encounter in the course of their daily lives. This information can be obtained through direct experience, from movies and television, from reading newspapers, or from conversations with friends or strangers. The information can be transmitted in writing but also visually or acoustically. Moreover, it can be conveyed in several modalities simultaneously. Finally, although the information can sometimes concern hypothetical people and situations, it more often refers to actual persons and events about which recipients already have substantial knowledge. Finally, the information is often conveyed in a social context, and can elicit affective or emotional reactions toward either its referent or, in some cases, the communicator. Much of this information, particularly that which is acquired through direct experience, is necessarily transmitted in the form of a narrative, that is, a temporally related sequence of events. The conceptualization we propose addresses the way this information is comprehended and represented in memory, and how it is later recalled and used in making judgments and decisions. Our focus on “real-world” information processing does not imply that the research we bring to bear on it

INTRODUCTION

5

has not been conducted in controlled laboratory settings. In fact, most of the research reported is of this variety. However, the type of information of concern in this research, the referents of the information, the context in which it was presented, and the type of judgments and decisions for which it had implications, is similar to that encountered in daily life situations. The results of this research, and the formulation we bring to bear on it, are consequently more likely to generalize to nonlaboratory situations than work we have done in the past. In this introductory chapter, we review the different types of knowledge people acquire in their daily lives, and discuss briefly the various ways this knowledge is represented in memory. Chapter 2 describes the theoretical formulation of information processing within which the research and theory discussed later is conceptualized. This formulation has features in common with the earlier formulation proposed by Wyer and Srull (1989). However, it is modified in several important ways that permit it to account for the comprehension and use of information about familiar persons and events of the sort we acquire in daily life. Chapter 3, another introductory chapter, discusses the determinants and effects of knowledge accessibility, which is central to much of the theory and research presented in later chapters. Part II of the book is devoted to the representation of knowledge about social events and its role in comprehension. Chapters 4 and 5 examine the sorts of representations that people form specifically in the course of comprehending social events that they either read about or directly experience. Chapter 5 also discusses the more abstract event representations that are formed in the course of deliberative goal-directed activity. Chapter 6 addresses the extent to which generalized event representations can influence the representations that are formed of situation-specific occurrences, and the impact of forming abstract representations of an experience on later memory for the experience. Chapter 7 considers the effects of the situational context in which information is presented on its comprehension. Finally, chapter 8 applies the conceptualization developed in previous chapters to a phenomenon that has received surprisingly little attention in research on social information processing but is a central ingredient of communication outside the laboratory: namely, the reactions to humor. Part III of this volume examines the use of social information to make judgments and behavioral decisions. Chapter 9 focuses specifically on the way verbal and visual information combine to affect judgments under conditions in which narrative representations of information are and are not likely to be constructed. Chapter 10 considers more generally the impact of narrative representations of knowledge on judgment, focusing largely on the impact of implicit theories that people construct and use as a basis for inferences

6

CHAPTER 1

about specific events. Finally, chapter 11 addresses an issue largely ignored in previous chapters but nevertheless of central importance in understanding the effects of information in situations outside the laboratory, namely, the impact of affective reactions on judgments of the persons, objects, and events to which these reactions are (or, in some cases, are not) relevant. The conceptualization proposed in this volume has been largely stimulated by research conducted in our own laboratory during the past 15 years. However, its implications for phenomena identified in other areas are considered as well. We therefore hope that the conceptualization provides a perspective on the dynamics of information processing in everyday life situations that is not only useful in integrating a variety of known phenomena but suggests new directions for research and theorizing.

TYPES OF SOCIAL KNOWLEDGE Social knowledge can vary in terms of its referent (what the knowledge is about), its modality (verbal, visual, or affective), the generality of its implications, and the way in which it is acquired (through direct experience, or through cognitive operations that are performed on other cognitive material). Two types of knowledge should be distinguished at the outset. Declarative knowledge concerns the referents of our everyday life experience (e.g., persons, objects, events, social issues that we read about, or oneself). In contrast, procedural knowledge concerns the sequence of actions that one performs in pursuit of a particular goal (driving a car, using a word processor, etc.). Whereas declarative knowledge is reflected in the information we can recall about an entity or that we implicitly draw on in the course of attaining a particular objective, procedural knowledge is reflected in the sequence of cognitive or motor acts that are actually performed in the pursuit of this objective. People can of course have declarative knowledge about how to attain a particular objective, and might sometimes consult this knowledge for use as a behavioral guide. Once the procedure is well learned, however, it may often be applied automatically, with little if any conscious cognitive mediation. These automated procedures can be conceptualized as “productions” of the sort suggested by J. R. Anderson (1982, 1983; see also E. R. Smith, 1990). Thus, they may metaphorically have the form of “If [X], then [Y]” rules in which [X] is a configuration of perceptual or cognitive stimulus features and [Y] is a sequence of cognitive or motor acts that are elicited automatically when the eliciting conditions are met. These productions, which are acquired through learning, are strengthened by repetition, and can ultimately be activated and applied with minimal cognitive mediation. The rou-

INTRODUCTION

7

tines involved in driving a car (e.g., putting in the clutch, turning on the ignition, putting the car in gear, gradually releasing the clutch while stepping on the gas, etc.) initially require conscious thought. However, they ultimately come to be performed without consultation of one’s declarative knowledge about the sequence of steps involved, and require few if any cognitive resources (Schneider & Shiffrin, 1977). As Bargh (1997) asserted, a very large amount of our social behavior is likely to involve the use of these automatically activated productions. Procedural knowledge plays a very important role in social information processing, as I indicate frequently throughout this volume. For the present, however, I will concentrate my discussion on the content and structure of declarative knowledge. I first consider the referents of this knowledge, and the modality in which it is acquired. I then turn briefly to the different ways in which this knowledge is represented in memory, and discuss the need to understand these matters in conceptualizing the processing of information of the sort we receive in the course of daily life. Referents of Social Knowledge The stimulus information that impinges on us in a given social situation can often far exceed our capacity to assimilate its implications. In a conversation, for example, several persons typically communicate to one another both verbally and nonverbally, and in doing so, are likely to convey information about not only the topic under discussion but also about themselves. Moreover, this information is often transmitted in the context of many other situational stimuli that are largely irrelevant to either the topic at hand or the individuals who are discussing it. We are usually not passive recipients of this information. Rather, we actively try to determine its meaning. Furthermore, we might construe its implications for the attainment of a particular goal that we have in mind. In doing so, however, we necessarily pay more attention to some features than to others. The features that receive the greatest attention depend in part on our perception of the referent (i.e., what the information is supposed to be about). The referent of a piece of information can be either an entity (a person or object, a place, oneself, etc.) or an event. However, most events involve persons and objects and occur in a specific (although perhaps unspecified) location. Furthermore, one’s knowledge about an entity often concerns events in which the entity has been involved. As a consequence, the referent of information is often subjective, depending on not only the context in which the information is received but also the interests and objectives of the recipient. For example, consider the statement, “Muhammad Ali knocked out George Foreman in Zaire.” This statement could be about Ali, Foreman, or the event itself. In some cases, of course, perceptions of the

8

CHAPTER 1

referent of information are determined by the context in which the information is received. Thus, someone who hears the aforementioned statement during a discussion of Muhammad Ali is likely to identify Ali as its referent, whereas someone who reads the statement in a magazine article about boxing might consider the event itself to be the referent. In addition, however, individual differences in prior knowledge and motivation (e.g., whether one is a fan of Muhammad Ali or George Foreman) can also influence perceptions of a statement’s referent. These situational and individual differences in perception of the referent can have an impact on how the knowledge is represented in memory and the likelihood that the knowledge comes to mind later. For example, a person who learns that Ali knocked out Foreman in a conversation about Ali, and consequently perceives Ali to be the referent, may not recall the information later when information about George Foreman is being discussed. (For further discussion of this possibility, see Wyer & Srull, 1989.) Furthermore, when the information conveyed in a particular situation is detailed and complex, recipients are likely to think more extensively about features that pertain to its referent (as they perceive it) than other features. Thus, the referent-related features are likely to be retained, whereas more peripheral features seem to be forgotten. For example, a man who learns that V. S. Naipaul won the Nobel Prize for Literature might remember the specific context in which he acquired the information a short time after it was conveyed. Over time, however, the information becomes part of his general knowledge and the conditions that led to its acquisition are not recalled.1 Affect and Emotion. Although the referents of knowledge are usually external to the organism, internally generated stimulation can also be a source of information. (See Strack and Deutsch, 2002, for a theoretical analysis of these alternative sources of information.) For example, external stimuli can elicit affective or emotional reactions, and these reactions can be used as bases for both thoughts and overt responses to the stimuli (Schwarz & Clore, 1983, 1996; Wyer & Carlston, 1979). This does not mean 1 The dissociation of knowledge from the context in which it is learned does not mean that contextual features are erased from memory. In the previous example, I learned about Naipaul’s award from my wife, who stopped into my office to tell me about it one day at work, and I wrote the preceding paragraph shortly afterwards. I would normally not be able to recall where and when I acquired knowledge of the award. However, being reminded of the example while going over this manuscript several months later was sufficient to stimulate a relatively detailed mental image of the event, including my own and my wife’s physical locations in the room. Thus, although these contextual features became “lost” in the interim, they continued to exist in memory and could be retrieved by a subset of cues that identified the event and its use as an example in this book. The theory to be proposed in this volume attempts to account for this and related memory phenomena.

1

INTRODUCTION

9

that affective reactions per se are part of stored knowledge. Rather, they are conditioned or unconditioned responses to cognitions about their referent, and can sometimes be the referents of cognitions, but they are not themselves part of the cognitive system (Wyer, Clore, & Isbell, 1999). This view, which contrasts with that proposed by Wyer and Srull (1989; see also Bower, 1981), is discussed more fully in chapter 11. Relational Knowledge. Much of the knowledge we require about entities and events is relational. These relations may be physical or conceptual. For example, we know not only that a nose and eyes are parts of a face but also their positions in relation to one another. Furthermore, we know that our bedroom is down the hall from the living room, that Chicago is west of New York, that lions are larger than rabbits, and that the earth goes around the sun. We also know that people graduate from high school before going to college, that smoking is a cause of lung cancer, and that people have to order a meal in a restaurant before they eat it. This knowledge requires not only understanding of the entities being related but the relations themselves (“larger than,” “causes,” “above or below,” etc.). Note that in many instances, the mental representation of these relations requires visual imagery. Modalities of Knowledge Information is conveyed in many different modalities and formats. For example, it can be transmitted verbally in the form of propositions. It can also be conveyed visually as in a picture or movie, or acoustically, as in music. Our knowledge can likewise be in different modalities. That is, our knowledge of a person could consist in part of verbal descriptions of the person’s traits, and our knowledge of restaurants could include abstract characterizations of the events that usually take place there. However, we also have mental images of people we meet and places we visit, and might have acoustic representations of a person’s voice, or of the Beatles’ rendition of “Hey Jude.” Social experiences are often a configuration of verbal, visual, and acoustic stimulation that combines to form our knowledge of what occurred. There has been some controversy about whether information is actually coded into memory in the modality in which it is conveyed or is represented in a common format that can be decoded into different modalities, depending on situational demands (J. Anderson, 1978; Kosslyn, 1980; Pylyshyn, 1973). Certainly we can easily translate many visual stimuli into linguistic terms, and can form visual images of people and events on the basis of verbal descriptions of them. Images can be acoustic as well as visual. (This is apparent from the fact that people who have seen a foreign movie

10

CHAPTER 1

are later unable to remember whether actors’ statements to one another were conveyed in subtitles or were dubbed.) Linguistic representations of the actors’ utterances become integrated with the visual representation of the movie and the quality of the speakers’ voices in a way that provides a memory of the events very similar to the memory one would have if the words had actually been spoken. Be that as it may, it is most convenient in the present volume to assume that codings of information in different modalities do exist in memory and, in some cases, can be stored independently. Several studies we describe in this volume attest to the importance of visual and auditory information processing and the impact of visually coded knowledge on memory and judgment. The way in which visual and verbal codings of knowledge theoretically combine to influence the representations that are formed, and how these representations are later used, are detailed in chapters 4 and 5.

THE REPRESENTATION OF KNOWLEDGE IN MEMORY

General Considerations As the preceding discussion testifies, it is virtually impossible to discuss the type of knowledge we acquire without making implicit assumptions about the representation of this knowledge in memory. However, the mental representation of information can differ in several ways from the information itself. Some features of the information we receive are not stored in memory at all, whereas other features that were not specified in the information may be added (Barclay, 1973; Loftus, 2000). Furthermore, detailed descriptions of an event can often be interpreted in terms of abstract (e.g., attribute) concepts that are stored in memory independently of these descriptions (Carlston, 1980; Higgins, 1996; Wyer & Srull, 1989). In addition, information might be mentally reorganized into a form that is relevant to the goal one happens to be pursuing (cf. Hamilton, Katz, & Leirer, 1980; Wyer & Bodenhausen, 1985). Moreover, information that is transmitted visually may be coded into memory linguistically, and verbal information may sometimes elicit visual images. It might nevertheless be argued that people extract similar implications from information regardless of the mental representation they form of it. If this were so, an understanding of the mental representations that are formed of information would be of little importance in predicting its impact. We would simply need to characterize clearly the stimulus information to which people are exposed, and to obtain a reliable indication of which as-

INTRODUCTION

11

pects of this information they retain (as evidenced by the aspects of the information they can recall). This would be sufficient to determine its impact. In fact, however, research in social and cognitive psychology since the 1970s suggests this contention is naïve. That is, a conceptualization of the cognitive operations that are performed on information, and of the mental representations that result from these operations, is essential for understanding its influence. Because the comprehension and mental representation of experiential knowledge is a central focus of this volume, some empirical support for this conclusion may be worth summarizing. 1. Communication and Persuasion One of the first demonstrations in social psychology of the need to take into account the cognitive activities that people perform in response to information was constructed by Greenwald (1968). He showed that the influence of a persuasive message was often unrelated to the implications of those aspects of the message that participants could recall. Rather, its impact was governed by the implications of the thoughts they had about the message at the time it was received. These cognitive responses, which included elaborations of the message content, counterarguments, and thoughts that were unrelated to the message itself, presumably reflected recipients’ comprehension of the message and their evaluation of its validity as well as the implications of other previously acquired knowledge that they brought to bear on the position advocated. In a study by Loken (1984), participants read a list of information items pertaining to a college professor with instructions either to form an impression of the individual or to decide whether they wanted to take a course from him. The information consisted of either favorable or unfavorable trait descriptions, and either favorable or unfavorable consequences of taking the course. In some cases, individual pieces of information had implications for both attitudes toward taking the course and attitudes toward the professor (e.g., “Dolan is interesting,” “Taking Dolan’s course is time consuming,” etc.). In other cases, they had implications for only one of these attitudes (e.g., “Dolan is clumsy,” “Taking Dolan’s class fulfills a graduation requirement,” etc.). Later, participants reported their attitudes toward both the person and the behavior, and then recalled the information they had received. The type of information that participants recalled was influenced by the task objectives they were given. Nevertheless, the effects of experimental variables on attitude judgments did not mirror their effects on recall. Moreover, correlations between attitudes and the evaluative implications of attitude-relevant information that participants recalled were very low, ranging from .05 to .24. Loken concluded that participants spontaneously formed attitudes toward both the professor and taking the course at the

12

CHAPTER 1

time the information was received, and later reported these attitudes independently of the information they could remember at the time of judgment. 2. Person Impression Formation In a study by Carlston (1980), participants read descriptions of a target person’s behavior that exemplified two traits with evaluatively different implications. In one condition, for example, behaviors (e.g., “told his girlfriend that her hair-do was ugly”) could be interpreted as both honest and unkind. Participants judged the target with respect to one of the two traits and then, either a few minutes or several days later, evaluated him with respect to the other trait. Participants’ second judgments were evaluatively biased by the first. In other words, participants who initially evaluated the target’s honesty (and thus judged him to be honest) later reported him as being more kind than participants who had not made this initial judgment. Correspondingly, participants who made kindness judgments at the outset (and thus rated the target as unkind) judged him to be less honest than other participants did. Moreover, these effects increased over time. Participants apparently interpreted the information in a way that was relevant to the first judgment they made and formed a mental representation of the target as the type of person who had the trait being judged. Then, having done so, they based their second judgment on the implications of this representation rather than reconsidering the original information they had received. In a series of studies by Srull and Wyer (1983), some participants read descriptions of a person’s behavior that were either predominantly favorable or predominantly unfavorable with instructions to form an impression of the actor. Then, after either a short or long delay (5 min or 45 min), they were asked to report their liking for him. Other participants, who were asked to recall the behaviors they had learned, remembered fewer behaviors after a long delay than after a short delay. However, participants rated the target more extremely in the long delay condition (e.g., as more likable or more dislikable, depending on the predominant implications of the behaviors). Apparently, these participants formed a generally favorable or generally unfavorable impression of the target on the basis of the predominant theme of the behaviors they had read. After a delay, their recall of the specific behaviors diminished, but their overall impression remained. Therefore, their judgment, based on this impression alone, was more polarized. 3. Assimilation and Contrast In a study by Higgins and Lurie (1983), subjects rated the harshness of sentence imposed on a criminal by a “target” judge (Jones) along with the sentences recommended by three “context” judges. Participants rated Jones’s sentence as more lenient when sentences recommended by the

INTRODUCTION

13

context judges were long than when they were short. The critical conditions of the study, however, occurred a week later. At this time, participants were given a sample of the sentences imposed by a different group of context judges that were again either long or short. Then, after receiving the information, they were asked to recall the actual sentence that Jones had imposed, as conveyed in the first session. Instead of recalling this sentence directly, however, participants apparently recalled their rating of him as harsh or lenient, and then reinterpreted the implications of this rating with reference to the sentence imposed by the context judges in Session 2. Therefore, they “recalled” the target’s sentence to be longer when the sentences imposed by the second set of judges were long than when they were short, and this was true regardless of the sentences imposed by the context judges in Session 1. 4. Impact of Information That One Is Told to Disregard In a study by Wyer and Budesheim (1987; see also Wyer & Unverzagt, 1985), participants read a series of behaviors of a target person with instructions to form an impression of him. In some cases, the first behaviors were favorable and the last behaviors were unfavorable. In other conditions, the order was reversed. On the pretense that an error had been made, some subjects were told after reading the first set of behaviors that these behaviors actually pertained to a different person than the one they were supposed to consider and to disregard them. Other participants read the entire set of behaviors and then were told to disregard the last set they had considered. Finally, participants were asked both to judge the target with respect to the traits to which the information pertained and to recall the information they received. When participants had been told to disregard the first behaviors in the series, their judgments were not influenced by these behaviors although they could recall the behaviors quite well. When participants had been told to disregard the last behaviors in the series, these behaviors had an impact on their judgments although their recall of the behaviors was poor. Thus, instructions to disregard the information had opposite effects on recall and judgments in the two conditions. These studies suggest that the mental representations that participants form of information can exert a major influence on their judgments independently of the information itself, and that the magnitude of this influence can increase over time. Greenwald and Banaji (1995) described several additional situations in which people are influenced by information they cannot recall and, in some cases, are unaware that they ever received (see also Bargh, 1997). Thus, neither the implications of the information that people receive nor the implications of the information they can remember provide

14

CHAPTER 1

much insight into the impact of this information on judgments. Rather, the impact is mediated by the cognitive operations that people perform on the information and the mental representation that results from these operations. A Conceptual and Methodological Approach to Investigating the Influence of Mental Representations To the extent that mental representations of information have an impact on judgments, a question arises as to how their content and structure can be determined. Any characterization of the mental representations that people form from information is obviously metaphorical and does not reflect the physical representation of the information in the brain. For one thing, information is not stored in particular locations, but rather, is distributed throughout the cerebral cortex. Thus, all theoretical models of mental representation must be viewed in terms of their utility and not whether they are true or false. The general logic underlying theory and research on mental representations can be conveyed with reference to Fig. 1.1. This figure distinguishes between input processes, recall processes, and judgment processes. Both the type of stimulus information presented and the recipients’ processing objectives at the time they encounter it can influence the cognitive operations they perform on the information and, therefore, the mental representation they construct. This representation is then stored in memory. If participants are later asked to recall the information, they may activate a procedure for extracting this information from the representation they formed earlier. If, on the other hand, they are called upon to make a judgment or behavioral decision, they may engage in search and retrieval operations that are relevant to this objective, and may use the material they extract from the mental representation as a result of these operations to compute their response. Two things are worth noting, however. First, the objectives that exist at the time the information is received may differ from those that occur later, at the time the information is retrieved for use in attaining a particular objective. Second, the search, retrieval, and computational operations that are performed at this second stage depend on the particular judgment or decision to be made and, therefore, are likely to differ from the operations that are performed when people’s objective is to recall the information they received. This means there may be little relationship between the implications of the information people recall and the judgments they make, as the research cited earlier suggests. This leads to the logic underlying research on social information processing in general, and much of the research to be summarized in this vol-

15

FIG. 1.1. Relations among various components of social judgment. Observable (independent and dependent) variables are enclosed by solid lines, and mediating variables by dashed lines. Rectangles denote “states” and ovals denote “processes.”

16

CHAPTER 1

ume more specifically. That is, the mental representation that is formed from information is a hypothetical construct and cannot be observed directly. However, given a specific set of assumptions about the retrieval operations that occur when people are asked to recall the information, and a metaphor for describing the associations among features of the information that exist in memory, one can use the amount, type, and order of information to conceptualize the content and structure of the mental representations that was formed at the time the information was received. Then, once this is established, one can hypothesize and investigate empirically the way this mental representation is used to attain specific objectives to which it is relevant. Much of the research and theory summarized in this volume is based, implicitly or explicitly, on this general approach. The social knowledge we acquire can be represented in memory in several ways. The nature of these representations, when they are constructed, and how they are used, are a major focus of this book. Two types of mental representations that occupy our greatest attention are worth noting briefly. These representations, to be denoted situation models and generalized representations, are distinguished in terms of both their content and structure and the point at which they typically come into play in social information processing. Although we consider both types of representations in detail in subsequent chapters, a general overview of the representations is perhaps desirable.

SITUATION MODELS We assume that people who receive information about situation-specific events, either verbally or through direct experience, form mental simulations of these events in the course of comprehending them that indicate not only the people and objects that are involved in the events but their spatial and temporal relatedness. These mental simulations, or situation models (Johnson-Laird, 1983, 1989; Kintsch, 1998) are often represented in memory in terms of mental images of the events and, in some cases, propositionally as well. To the latter extent, the representations are somewhat analogous to a “picture plus caption.” Moreover, a temporally and thematically related sequence of events might be represented by a multiple-segment representation somewhat analogous to a “comic strip.” (For a similar metaphor in conceptualizing the mental representation of event sequences, see Abelson, 1976.) We assume that situation models are often formed spontaneously in the course of comprehending information (see chap. 4 for a discussion of these processes). This information could be either obtained through direct experience or conveyed verbally. However, a necessary condition for the con-

INTRODUCTION

17

struction of these models is that the events or states to which they refer are temporally and spatially specific. That is, the phenomena must occur at a specific (although not necessarily specified) place and point in time. This constraint is implicit in the assumption that a mental image is formed. Thus, the statements, “The boy kicked the ball” and “The book is on the table” describe events or states of affairs that existed at a particular time and place, and a mental picture can be formed of them. In contrast, “the boy owns a ball,” or “the book is interesting,” are not spatially or temporally constrained, and a mental picture of the situations is not constructed (Radvansky, Wyer, Curiel, & Lutz, 1997). Rather, a generalized representation is formed on the basis of these statements that does not have an image component.

GENERALIZED EVENT REPRESENTATIONS More complex generalized representations can also be formed in the course of higher level goal-directed processing that occurs once information has initially been comprehended. For example, people who wish to tell someone about a movie they have seen might describe the things that happened in more abstract terms than they were shown in the movie and, in doing so, might construct a more general, linguistic representation of the sequence of events that occurred. Generalized event representations can also be formed in the course of evaluating the implications of a persuasive message, describing someone’s personality, or generating a conclusion on the basis of logical reasoning. Although these goal-directed representations might also be based on previously formed situation models, they are stored in memory independently of these models.

The Dimensionality of Event Representations Generalized event representations can vary along several dimensions. Generalizability. A generalized event representation might refer to a particular experience that involves specific individuals and occurs at a given time and place. Alternatively, it could pertain to a general class of individuals, situations, or events, instances of which occur frequently at different types. Thus, for example, I might have a representation of going to dinner at Zorba’s with Rashmi at 7 p.m. last Thursday. In addition, I might have a representation of going out to eat at restaurants that generalizes over times and places.

18

CHAPTER 1

Abstractness. A narrative representation could be encoded in terms of specific concepts that apply to single entities. Thus, the representation of my dinner at Zorba’s might specify where I sat, what I ate, and the details of the conversation I had over dinner. Alternatively, it could be coded in very abstract terms similar to those used to convey the example provided in the preceding paragraph. Personal Relevance. Event representations can be about oneself and one’s personal experiences, about acquaintances, or about unknown or fictitious individuals. Similarly, they can describe either events one has personally encountered (e.g., one’s observation of a teacher’s interchange with an obnoxious student) or those that occur in a remote and unfamiliar clime (e.g., the border dispute between India and Pakistan). Time Frame. A representation could refer to people and events in either the past, the present, or the future, or could extend over two or more of these time periods. Types of Generalized Event Representations Several different types of event representations can be distinguished in terms of values along these dimensions. In this volume, we refer to all representations as narratives. Certain types of narratives are nevertheless worth distinguishing, including scripts, stories, histories, and implicit theories. 1. Scripts and Plans Scripts are prototypic sequences of events that occur in a particular situational context (Schank & Abelson, 1977). They can refer to either specific persons (e.g., my wife’s ritual of getting ready for a dinner party) or people in general (a visit to a restaurant). However, although scripts can refer to particular individuals and situations, they apply to routinized sequences of actions that involve these persons and situations that typically occur under the circumstances at hand. A plan is a special type of script in which the events described produce a particular (typically favorable) end state. These constructions are typically applied to new situations, often serving as behavioral guides. For example, an individual who has formed a script of changing a flat tire as a result of observing others engage in the activity could use this script as a plan in determining how to change a tire when personally required to do so. Most research on the representation of scripts in memory (Bower, Black, & Turner, 1979; Graesser, 1981; Nottenburg & Shoben, 1980; Trafimow & Wyer, 1993) has focused on mundane event sequences that are very familiar to members of a Western culture. As noted earlier, however, individuals

INTRODUCTION

19

can form routinized sequences events that are specific to themselves, or their family (e.g., I might have a representation of the sequence of acts my spouse routinely performs when getting ready for a party). As we see in chapter 5, there is some question about how often these representations are actually constructed (cf. Colcombe & Wyer, 2002). Nevertheless, several such constructions undoubtedly exist. The fact that the events that compose scripts are routinized and generally mundane distinguishes them from other types of narratives in two ways. First, they are likely to have little evaluative significance for the persons involved. (A visit to a restaurant, for example, which terminates with paying the bill and leaving, has relatively neutral implications.) Second, scripts are unlikely to be the topic of conversation. That is, the events that compose them are expected to occur, and so mentioning them has little information value. 2. Stories Stories, unlike scripts, concern sequences of events at least one of which is unique. That is, it deviates from expectations based on prior knowledge about the types of people and situations involved. In addition, the events have some evaluative significance for the protagonists (Mandler & Johnson, 1977; Stein & Glenn, 1979). This conceptualization of a story is more restricted than Schank and Abelson’s (1995). In fact, a substantial literature in cognitive psychology has been concerned with both the appropriate definition of stories and their representation in memory (cf. Black, Galambos, & Read, 1984; Brewer, 1995; Brewer & Lichtenstein, 1981; Mandler & Johnson, 1977; Stein, 1982). In this work, stories are postulated to have certain common features that are captured in large part by a “story grammar” (Mandler & Johnson, 1977; Stein & Glenn, 1979). Although these grammars differ in detail (see Stein, 1982, for a comparison of alternative story grammars), they share the assumption that stories have a definable beginning and end along with an event that provides a transition between these states. Other features that are sometimes used to distinguish stories from other representations are of less utility. For example, Brewer and Lichtenstein (1981) argued that a story must also elicit some degree of affect or emotion in the recipient. However, it seems undesirable to define a story in terms of responses of the recipient rather than attributes of the story itself. (Brewer and Lichtenstein’s conceptualization implies that the same narrative might constitute a story to one person but not to another, depending on the affect it happens to elicit.) Brewer and Lichtenstein (1981) also assumed that the function of stories is to entertain. However, the same narrative might be communicated for the purpose of being entertaining in one context but for the purpose of making

20

CHAPTER 1

a point or exemplifying a principle in another context. It would be conceptually confusing to define the same narrative differently, depending on the conditions in which it is conveyed. It therefore seems more reasonable to define a story as simply a narrative that (a) describes events involving specific individuals or groups that are localized in time and place, (b) contains at least one event that deviates from expectations based on prior knowledge of the type of individuals and situations involved, and (c) whose end state has emotional significance for one or more of the protagonists. 3. Histories A history, like a story, consists of a sequence of events that are specific to a particular person or group and are localized in time and place. Unlike stories, however, a history does not necessarily contain unique or unexpected events. Nor does it necessarily have emotional significance for the protagonists. Histories can concern real-world events in which one is not involved. However, it can also refer to oneself or an acquaintance. The mental representation of a personal history is clearly not a single sequence of events that occurs over the course of a lifetime. Rather, an individual is likely to construct several histories, each of which is restricted to life events that are relevant to the purpose for which it is formed (communicating about oneself to another, comprehending one’s own or another’s personal experience, etc.). Once constructed, these sets of segments are likely to be stored independently of one another in memory. As these observations imply, histories can contain overlapping segments. Moreover, a history does not usually cover all the events that occur during the time period to which it pertains. Thus, one might construct different histories that cover the same or overlapping time period. For example, a personal history of how one became a psychologist could temporally overlap a history of one’s romance with an ex-call girl. However, the events that compose these mental constructions might be totally different. According to this conceptualization, stories and histories are not mutually exclusive constructs. A history would be a story to the extent it contains unexpected events and is emotionally significant for one or more of the parties involved. However, not all histories are stories, and not all stories are histories. 4. Implicit Theories and Implicational Molecules A theory, like a script, is composed of a prototypic sequence of events. However, the events are causally as well as temporally related. Moreover, the events are often defined more broadly. A theory is not localized in time and, therefore, can be applied in many different circumstances. It might

INTRODUCTION

21

nevertheless concern a particular individual (e.g., oneself) or general category of individuals (African Americans, fraternity members, etc.) as well as people in general. Similarly, it could be specific to a particular type of situation (living in public housing, student–faculty interactions) or could apply more generally. In all cases, however, the events that compose a theory are likely to recur at different times. For example, one might have an implicit theory that (a) people with alcoholic fathers feel insecure, (b) these feelings lead them to pursue unrealistically high goals, (c) their failure to attain these goals leads to depression and ultimately to drinking, and (d) they wind up becoming alcoholic like their fathers. Each segment encompasses a large number of more specific events and functions as a general concept that can be used to interpret and organize new information. To this extent, a theory can function as an event stereotype. Theories play a very important role in processing new information about the world in which we live and making inferences on the basis of it. They are used both to explain events that have occurred in the past and to predict the future. Moreover, they can be used to fill in “gaps” in a sequence of events about which one has incomplete information. In this regard, some theories can consist of only two or three segments that, in combination, constitute a general principle. In these instances, the principle could function as a linguistically coded “title,” or “header” of the theory to which it refers. For example, the principle “people get what they deserve” could be the title of a two-segment theory composed of the events “A person does something bad” and “The person gets punished.” These theories can be conceptualized as implicational molecules of the sort suggested by Abelson and Reich (1969; see also Wyer & Carlston, 1979, 1994). That is, they consist of a set of general propositions that have become associated through learning. For example, the propositions associated with the principle that people get what they deserve could compose the molecule [P does something good (bad); Something good (bad) will occur to P].

The use of an implicational molecule is theoretically governed by a completion principle. That is, if the information one receives about a specific situation instantiates all but one of the propositions that compose a molecule, an instantiation of the remaining proposition will be inferred. Thus, people who have formed the implicational molecule described in the preceding example might predict that a particular person who has done a bad thing will be punished. However, they might also infer that someone who has experienced misfortune has done a bad thing or for other reasons deserves his or her fate. Several other examples of the use of implicational molecules, and implicit theories more generally, are discussed in detail in chapter 10 of this volume.

22

CHAPTER 1

GENERALIZED ENTITY REPRESENTATIONS Generalized representations are obviously not restricted to events. They can also pertain to persons, to groups of persons, to objects, or to places. In fact, most of the mental representations that have typically been investigated in social information processing research (for reviews, see Carlston & Smith, 1996; Wyer & Carlston, 1994) are of this type. For example, people who receive descriptions of someone’s traits and behaviors with the goal of forming an impression of them may think about the individual pieces of information in relation to one another in order to understand their relatedness or, perhaps, to reconcile apparent inconsistencies among them. In the course of this cognitive activity, associations are presumably formed among the pieces of information that influences the ease of recalling the information later. The entity representations that are constructed in the course of goaldirected processing can often be conveyed metaphorically in terms of an associative network. That is, concepts and units of knowledge (e.g., situation models) are represented by nodes in memory. When a person thinks about one concept or knowledge unit in relation to another, an association is formed between them that is represented by a pathway. Thus, suppose people are asked to form an impression of a person, John, on the basis of set of behavior descriptions. If they interpret a particular behavior of the person (e.g., “returned a lost wallet”) as honest, this establishes an association between a concept denoting the behavior and the trait concept “honest.” If they interpret several other behaviors (“refused an opportunity to cheat on an examination,” “told his girlfriend that her hair-do was ugly,” etc.) in terms of the same trait, an association is formed between these behaviors and the trait concept as well, leading to a trait–behavior cluster as shown in Fig. 1.2a. Alternatively, suppose both these and other behaviors (e.g., “honks at slow drivers,” “reads poetry,” etc.) are thought about with reference to John without construing their trait implications. Then, associative pathways are formed between the behavior concepts and the person concept, “John,” as shown in Fig. 1.2b. General traits that are attributed to John as a result of processing these behaviors can also become associated with this concept. If people happen to think about some of the behaviors in relation to others, associative links may be formed between the behaviors as also shown in the figure. Finally, if they think extensively about John’s engaging in some behaviors but not others, the former behaviors may become more strongly associated with John than the latter ones do. This difference is conveyed by the thickness of the associative links between the behaviors and John (cf. Wyer & Carlston, 1979), as also shown in Fig. 1.2b.

INTRODUCTION

23

FIG. 1.2. (a) Metaphorical trait-based representations formed as a result of interpreting behaviors in terms of trait concepts. (b) Metaphorical representation of a person, “John,” described by behaviors and traits. Thicker lines denote stronger associations.

The representations of the sort shown in Fig. 1.2 are theoretically the result of conscious cognitive activity (but see Chartrand & Bargh, 1996, 2002, for a different conclusion). Moreover, the associations shown in the figure reflect the implications of specific assumptions about the nature of this activity. Given assumptions about the processes that underlie the extraction of specific features from the network, these implications can be evaluated

24

CHAPTER 1

empirically. One set of assumptions, espoused by Hastie and Kumar (1979) and others (Srull, 1981; Srull, Lichtenstein, & Rothbart, 1985; Wyer & Srull, 1989), which captures the processes involved in person impression formation on the basis of linguistically coded traits and behaviors, is discussed briefly in chapter 7. However, an associative network metaphor has been applied successfully in other domains as well (cf. Park, Yoon, Kim, & Wyer, 2001; Park & Wyer, 1993; Wyer & Carlston, 1979). The conceptualization described previously has typically been applied to representations that are formed of fictitious entities. Representations of actual people and objects may be similar with two exceptions. First, the representation of a familiar person might often include the referent’s name and also a mental image of his or her physical appearance. Second, the features that are associated with the individual might be added to the representation over a period of time, as new information about the person is acquired. These additional features simply become linked to the central concept of the referent, much as the ones that were conveyed at the outset. On the other hand, once features have become part of a representation, they cannot be modified or erased. (See chap. 2 for an elaboration of this assumption.) When an existing representation appears to be inappropriate, a new representation is constructed whose features differ from the original. Thus, people might form different representations of the same person or object that are stored independently in memory, each with a different (although perhaps overlapping) set of associated attributes. Finally, representations might pertain to a group or category as well as a single entity. For example, people might form a representation of the persons in a particular social group, of U.S. Presidents, or of African Americans. These representations might consist not only of features that are typically associated with members of the group or category as a whole, but also names of the members that belong to the group and have been thought about in relation to it. In this regard, the name of a particular person who belongs to several groups could be contained in several different representations. However, these representations would not be interconnected but would exist independently in memory. This assumption distinguishes the present conceptualization from more general associative network models (e.g., Collins & Loftus, 1975; Wyer & Carlston, 1979), which assume that all representations with common features are interconnected. An assumption underlying the conceptualization is that once a mental representation is formed (either spontaneously in the course of comprehension or as a result of goal-directed processing), this representation is stored in memory as a single unit of knowledge and is later retrieved as a whole. Thus, in our example, a representation of John would presumably be stored in memory as a unit and might later be retrieved as a whole when information about “John” is sought. Other representations of the individual,

INTRODUCTION

25

based on information obtained at different times, might also be formed and stored independently. These latter representations might include situation models of the sort described earlier. This means than when information about a person or object is sought, the type and implications of the information retrieved depends on which of several potentially applicable knowledge representations come to mind. This, in turn, depends on its accessibility in memory. This matter is discussed more fully in chapter 2 within the framework of the general information-processing formulation we propose.

C H A P T E R

2 The Architecture of the Information-Processing System

A complete theoretical formulation of information processing must specify the content and structure of memory, the storage and retrieval processes that are involved in the transmission of knowledge throughout the cognitive system, and the way this knowledge comes into play in the pursuit of specific goals to which it is relevant. Wyer and Srull’s (1989) theory attains many of these objectives, and several features of this earlier theory are retained in the conceptualization I now propose. However, several assumptions of the earlier formulation must be modified in order to account for the processing of information about familiar people and events that is acquired through direct experience. Before describing these modifications, certain limitations of the earlier model are worth noting. (These limitations are shared by other, more restricted conceptualizations of social information processing as well; see Brewer, 1998; Carlston, 1994; Smith, 1990; Wyer & Carlston, 1979.) LIMITATIONS OF WYER AND SRULL’S (1989) THEORY The Comprehension and Validation of Information About Familiar People and Events Perhaps the most general deficiency of the original model is its failure to specify the operations that govern the comprehension of information at the time the information is first encountered, before goal-directed cognitive ac26

ARCHITECTURE OF THE INFORMATION-PROCESSING SYSTEM

27

tivity is initiated. Wyer and Srull simply assumed that input information is spontaneously encoded in terms of “low level” semantic concepts that are applied independently of the modality in which the information is received. They provided no indication of how this is done. Furthermore, their assumption that all incoming information is encoded semantically constrains the model’s ability to address the comprehension of information that pertains to people and events about which one already has substantial knowledge. The model assumed that referent-specific knowledge does not come into play until a later, goal-directed stage of processing. One implication of this assumption is that the validity of information about well-known people and events is not assessed until some time after the information is comprehended, and may not occur at all unless people have a specific goal to which this assessment is relevant. This seems intuitively unlikely. Most Americans would almost immediately recognize that the assertion “Abraham Lincoln was President of the United States” is true in the course of comprehending it. By the same token, they are equally quick to recognize that “Mao Tse Tung was President of the United States” is false. These validity assessments appear to occur spontaneously in the absence of any specific goal that requires them. Not all information is spontaneously verified at the time it is comprehended, however. The statement “Abraham Lincoln visited Boston” is likely to be comprehended without any conscious recognition that it is either true or false. Statements that refer to unknown persons may be similarly comprehended without spontaneously validating them. The conditions in which spontaneous validity assessments do and do not occur in the course of comprehending information is not specified by either the Wyer and Srull (1989) formulation or others. Information is likely to be processed quite differently when it is comprehended with reference to previously acquired knowledge about its referents than when it is not. For example, suppose people receive information that a person is kind, followed by information that he both (a) bought dinner for a street person who did not have enough money to eat and (b) berated his secretary for coming 10 minutes late to work. Recipients are likely to think more extensively about the second behavior, which is ostensibly inconsistent with the trait description than about the first one, leading this behavior to be better remembered later (Hastie & Kumar, 1979; Srull, 1981; Wyer & Gordon, 1982). However, suppose people receive information that a conservative Republican has both organized a sit-in at a nuclear power plant and picketed an abortion clinic. In this case, recipients are less likely to think about the behaviors in terms of their consistency with the general characteristics of the actor. Rather, they think about them with reference to their knowledge about the social implications of the behavior and their perception that these implications are desirable. Consequently, they have

28

CHAPTER 2

better recall of the behaviors that reflect policies they consider to be incorrect or undesirable (Wänke & Wyer, 1996; Wyer, Lambert, Budesheim, & Gruenfeld, 1992). In short, information that has little relevance to one’s prior knowledge about its referents is likely to be thought about in terms of the semantic consistency of its implications. However, information about persons and situations about which one has prior knowledge is usually thought about with reference to this knowledge, and so its semantic consistency has less impact. The aforementioned studies provide only few examples of the role of referent-specific knowledge in comprehending and validating new information. The effects of spontaneously recognizing information as true or false in the course of comprehending it, and the reactions to the implications of these assessments, are central to an understanding of humor elicitation (Wyer & Collins, 1992), emotional communication (Gaelick, Bodenhausen, & Wyer, 1985; Scott, Fuhrman, & Wyer, 1991) and persuasion (Gruenfeld & Wyer, 1992; Wyer & Gruenfeld, 1995). This becomes clear in later chapters of this volume.

The Nonverbal Coding of Social Events A second limitation of Wyer and Srull’s (1989) theory is its focus on linguistically coded information processing. The information we acquire in daily life is often obtained nonverbally, through direct observation. Although linguistically coded information (e.g., verbal utterances) can accompany our observations, this information is also conveyed in the context of visual and acoustic features that are important for comprehending it. (One’s interpretation of a colleague’s comment “You idiot!” is clearly different, depending on the speaker’s facial expression and tone of voice.) The interpretation of nonverbal behavior, both in isolation and in the context of verbal behavior, has received much attention (for a recent review, see DePaulo & Friedman, 1998). However, although the multiple modalities in which information can be represented in memory have been recognized in cognitive psychology (A. Anderson, Garrod, & Sanford, 1983; Graesser, Singer, & Trabasso, 1994; Schank & Abelson, 1977, 1995; Zwaan, Magliano, & Graesser, 1995), the implications of this fact has rarely been considered in social psychological theory and research. Carlston (1994) noted that visual and linguistic processing systems can interact in the course of attaining a specific processing objective (i.e., impression formation). However, the nature of this interaction is left ambiguous in his model. Wyer and Srull (1989) allowed for the construction of mental representations in multiple sense modalities, but they also fail to articulate the nature of these representations and the conditions in which they are formed.

ARCHITECTURE OF THE INFORMATION-PROCESSING SYSTEM

29

In this regard, the modality in which information is presented is not necessarily the modality in which it is reported. Visual information can sometimes be encoded into memory linguistically. Moreover, linguistically coded information can elicit mental images (Bransford, Barclay, & Franks, 1972; Garnham, 1981; Radvansky, Gerard, Zacks, & Hasher, 1990; Reyes, Thompson, & Bower, 1980; Wyer, Adaval, & Colcombe, 2002). These processes, however, are not articulated in the theory proposed by Wyer and Srull (1989). Other, more specific problems with the original theory will be noted presently in the context in which they become relevant. The conceptualization we propose in this and later chapters is intended to address the earlier deficiencies while at the same time preserving the general architecture of the information-processing system of the original model and, therefore, many of its implications. In the next section, we outline the general structure of the proposed system, noting the assumptions that differ from those of the original model. We then summarize the assumptions we now make concerning information storage and retrieval, which differ significantly from those of the original formulation.

STRUCTURE OF THE PROCESSING SYSTEM The architecture of the social information-processing system we propose is described in Fig. 2.1. The system, which is structurally very similar to that postulated by Wyer and Srull (1989), consists metaphorically of four information storage units (the Sensory Store, the Work Space, the Permanent Storage Unit, and a Goal Specification Box), four special-purpose processing units (the Comprehender, Encoder/Organizer, Inference Maker and Response Selector), and an Executor that directs the transmission of cognitive material from one unit to another. The original theory differs from the model to be proposed in terms of (a) the organization of memory in Permanent Storage, (b) the functions of the Comprehender, and (c) the processes of information retrieval. Storage Units Of the four storage units the theory postulates, the Sensory Store is least important for the issues of concern in this volume. This unit simply records the configuration of sensory input material that impinges on the cognitive system at any given movement, in fairly veridical form. However, the trace of this information decays rapidly (in a manner of seconds), and if it is not involved in further processing, it is lost (cf. Sperling, 1960). The postulation of this storage unit recognizes that information in multiple sense modalities

30

FIG. 2.1. Schematic representation of the information processing system. Circles and ovals refer to processing units. Rectangles denote storage units. Arrows indicate the direction of transmission of cognitive material between these units.

ARCHITECTURE OF THE INFORMATION-PROCESSING SYSTEM

31

can simultaneously impinge on the processing system and can serve as input to the Comprehender (see Fig. 2.1). However, its importance for the issues of concern here is limited, and the unit will generally not be discussed further. 1. The Work Space The structure and function of the Work Space is similar to that of the original model. This unit functions as a temporary store of all information that has been encoded by the Comprehender along with other material that is or has recently been involved in higher level goal-directed processing. As shown in Fig. 2.1, input to the Work Space can include material that has been interpreted and transmitted to it by the Comprehender, concepts and knowledge representations that have been retrieved from Permanent Storage for use in attaining specific processing objectives, and the results of this processing. The capacity of the Work Space is large, and as long as its contents are relevant to the attainment of a processing objective that is active, it can be retained indefinitely. However, if a processing goal has been attained, or if information has not been used for some time, the Work Space may be cleared to permit easier access to material that is relevant to other, more immediate objectives. This may be done either automatically, as new material is deposited, or volitionally (for evidence, see Srull & Wyer, 1983). Phenomenally, this simply means that people can intentionally stop thinking about material of relevance to one objective and concentrate on a second.2 These considerations become important in the context of an additional assumption of both the Wyer and Srull model and the present one. That is, only the output of higher order goal-directed processing is transmitted to Permanent Storage. Therefore, once the Work Space is cleared, any input information from the Work Space that has not been involved in goaldirected cognitive activity is irretrievably lost. 2. Permanent Storage A major modification of the present model in relation to that proposed by Wyer and Srull surrounds the conceptualization of long-term memory or, in other words, Permanent Storage. We first describe the assumptions of the original model, and then discuss the modifications to be made. 2

On the surface, this implication appears inconsistent with the result of research on thought suppression (Bodenhausen & Macrae, 1998; Wegner, 1994), which suggests that conscious attempts not to think about a concept increases its accessibility in memory. However, this inconsistency is illusory, as we point out later in this volume.

2

32

CHAPTER 2

The Original Model. Wyer and Srull conceptualized long-term memory metaphorically as a number of content-addressable storage bins. One, semantic bin, was likened to a “cognitive dictionary” that contained semantic concepts necessary to interpret incoming stimulus information. Referent bins, which in combination functioned as a cognitive “encyclopedia,” contained representations of a particular referent (entity, event, etc.) that are formed in the course of goal-directed cognitive activity. Each bin was identified by a “header,” or set of features that defined or circumscribed its contents. Different representations could be formed of a referent at different points in time, being stored in a bin as separate units of knowledge in the order they are constructed. Thus, more recently formed representations were on top. These representations, which might differ in form and modality, could theoretically be retrieved independently of one another in the course of a top-down search of the bin contents. This means that the most recently formed and deposited representations are most likely to be identified later. Different bins were theoretically formed to contain information about referents at different levels of specificity. Thus, an individual might form a “George W. Bush” bin containing information about Bush’s actions and policies while in office, but also a “U.S. President” bin containing more general information about U.S. Presidents, the names of specific exemplars, and so on. Note from this example that pieces of knowledge in one bin (e.g., the name of a specific U.S. President) can include features in the header of another bin. Moreover, not all information about a referent is necessarily stored in the bin that pertains to this referent. Thus, suppose one’s objective at the time a piece of information is received about George W. Bush pertains to presidents in general rather than to Bush in particular. Then, it would be stored in a “U.S. Presidents” bin. Consequently, it would not be retrieved if one is later searching in the “Bush” bin for relevant knowledge about George W. Bush. The Proposed Modification. The “bin” construct permitted the Wyer and Srull model to predict and explain a number of phenomena that are difficult for alternative formulations to capture (e.g., see Srull & Wyer, 1980, 1983; Wyer & Bodenhausen, 1985; Wyer & Budesheim, 1987; Wyer & Unverzagt, 1985). At the same time, the conceptualization is inadequate to account for the comprehension of information about known people and events. One problem derives from the aforementioned assumptions that (a) the information contained in a referent bin is only accessed in the course of goal-directed cognitive activity, and (b) a top-down search of the bin is required in order to identify it. Because of these assumptions, the conceptualization has difficulty accounting for the spontaneous comprehension of information about known people and events, or for the spontaneous recognition that the information one receives is true or false.

ARCHITECTURE OF THE INFORMATION-PROCESSING SYSTEM

33

In the revised theory, therefore, we drop the construct of bins. Instead, following multiple-trace theories of long-term memory (cf. Hintzman, 1986), we assume that each mental representation that is formed of a given body of information (either at the time the information is first received or later, in the course of higher order, goal-directed processing) has its own memory trace. Each of these representations can become involved in the processing of information at the initial comprehension stage as well as at later stages, in ways that we specify presently. This modification of the theory permits new information about a known referent (e.g., George W. Bush) to be comprehended immediately in terms of a specific representation of this referent without requiring goal-directed cognitive activity. 3. Goal Specification Box The Goal Specification Box is a temporary store of the procedures that are currently being used to attain the particular goals that are actively being pursued at the time. These procedures, or goal schemas, are drawn from Permanent Storage, and are consulted by the Executor for instructions about the cognitive steps required to attain these goals. More than one goal schema can be stored in the Goal Specification Box at one time. This means that the system has the capability of pursuing more than one goal simultaneously. However, the size of the Goal Specification Box is limited. Consequently, if a particular procedure is complex, the number of other procedures that can be stored in the Goal Specification Box, and, therefore, the number and complexity of other ongoing cognitive activities that can be pursued simultaneously, decreases. Special Purpose Processing Units The Wyer and Srull conceptualization postulates four special purpose processing units, each of which performs a particular function. The first, Comprehender, is an initial lower level encoding device that interprets stimulus input information that enters the processing system from the Sensory Store. Unlike other processing units, the Comprehender is not controlled by the Executor. That is, it receives stimulus information from the Sensory Store and automatically interprets it in terms of concepts and knowledge structures that are drawn from Permanent Storage (Fig. 2.1). This is done independently of any more specific processing objective that might exist. In the original theory, the activities of the Comprehender were restricted to the encoding of information in terms of semantic concepts. In the present conceptualization, however, the functions and capabilities of the Comprehender are expanded in ways that permit it to interpret information in terms of previously acquired referent-specific knowledge as well. These processes are articulated in detail in chapter 4.

34

CHAPTER 2

The other three processing units come into play in the course of specific goal-directed activity, and are activated on the basis of instructions in a goal schema that is used by the Executor to monitor this activity. The characteristics of these units are not appreciably different from those described by Wyer and Srull (1989). The Encoder/Organizer interprets stimulus information transmitted to it from the Comprehender in terms of more abstract concepts that are relevant to the goal being pursued, or organizes the information with reference to a pre-existing configuration of features that are specified in the information. The Inference Maker generates specific subjective inferences on the basis of new or pre-existing information using logical, algebraic or other combinatorial rules, or, in some cases, heuristics. The Response Selector then translates subjective inferences into an overt response, which could be a behavioral decision, a judgment along a category response scale, or a verbal utterance. Each processing unit is equipped with a library of routines (conceptually equivalent to cognitive productions, as mentioned in chap. 1) that it draws on to perform its functions. The nature of these productions can depend on both the goal being pursued and the type of information being processed. Thus, for example, the procedures performed by the Inference Maker depend on whether the goal is to form a general impression or to make a comparative judgment. On the other hand, they can also depend on whether the information is conveyed visually or verbally, or whether it is conveyed in a list or a narrative (cf. Adaval & Wyer, 1998). The Executor The Executor monitors the flow of cognitive material between memory storage units and special purpose processing units. It has no “mind” of its own, but instead, takes instructions from a pre-existing goal schema that is active in the Goal Specification Box. These schemas presumably provide stepby-step descriptions of the sequence of steps that must be performed in order to attain the objective at hand. For example, suppose people receive descriptions of a person for the purpose of forming an impression. The goal schema that guides the attainment of this objective might indicate that the individual behaviors should be interpreted in terms of trait concepts, that the evaluative implications of these traits should be assessed, and the implications of these assessments should then be combined to form an inference of the person’s general likeableness, and that this evaluation should be communicated to another verbally. The Executor would presumably recognize the first two (trait and evaluative) encoding tasks to be the responsibility of the Encoder/Organizer and would direct the information to this unit with instructions to perform the encoding. Once the results of the evaluative encoding are avail-

ARCHITECTURE OF THE INFORMATION-PROCESSING SYSTEM

35

able, the Executor would transmit these encodings to the Inference Maker with instructions to combine their implications into an overall inference. Then, once this inference is computed, the Executor would transmit the inference to the Response Selector with a request to generate an overt response in a language that was applicable (a verbal utterance or, if the inference is to be reported in an experiment, a numerical value along a response scale). Note that the routines contained in the libraries of the various processing units, as well as the goal schemata used by the Executor to govern the flow of information in pursuit of specific processing objectives, are generally not specified by the conceptualization we propose. These routines and procedures are the subject of research on specific types of cognitive activity of the sort described by Wyer and Srull (1989) as well as later chapters of the present volume. However, many social psychological theories derived from this research can incorporated into the more general conceptualization we have proposed. For example, the model of person impression formation and memory proposed by Srull and Wyer (1989; for earlier versions, see Hastie & Kumar, 1979; Srull, 1981), which specifies the cognitive processes involved in forming an impression and the cognitive representations that are formed as a result of these processes, can be viewed as a specification of an “impression formation” goal schema along with a specification of the organizational processes performed by the Encoder/Organizer and stored in the Encoder/Organizer’s library. The algebraic integration processes postulated by Norman Anderson (1971; 1981), several of the cognitive heuristics identified by Tversky and Kahneman (1974), and the attribution processes postulated by Bem (1972) and Jones and Davis (1965), can be viewed as theories about the routines performed by the Inference Maker and the conditions in which they are applied. The impact of contextual stimuli on the use of category response scales (Lynch, Chakravarti, & Mitra, 1991; Ostrom & Upshaw, 1968; Parducci, 1965; Wyer, 1974) and opinion survey responses (Schwarz, 1994; Strack, 1994) are essentially theories about the processes performed by the Response Selector. These and other theoretical formulations can in principle be incorporated into the general framework we propose. Effects of Multiple Goal Schemata Two more general aspects of the conceptualization are worth noting at this point. First, more than one goal schema might be relevant to the attainment of a given objective. An alternative to the sequence described in our example might involve a comparison of the configuration of trait encodings to the pre-existing knowledge representation of a known person, or alternatively, a prototype, and, if there is sufficient match, an inference of the per-

36

CHAPTER 2

son’s likeableness from a pre-existing evaluation of the person to which this representation refers (Fiske & Pavelchak, 1986). (Thus, for example, the configuration of features might match those of one’s roommate or, alternatively, the “typical college professor,” leading the evaluation to be similar to that associated with this individual.) In this case, which alternative schema is applied could depend on its relative accessibility in memory and the ease of performing the cognitive operations it specifies (Logan, 1988). In this regard, several recent formulations of information processing have been viewed as “dual processing” models (e.g., see Chaiken & Trope, 1999). These models recognize that different processes are likely to govern memory and judgment, depending on a number of situational and motivational factors that happen to exist. These alternative processes are generally implicit in the conceptualization we propose and also the Wyer and Srull theory, both of which postulate that several different cognitive processes may be specified in the alternative goal schemata that can potentially be brought to bear on the attainment of a particular processing objective as well as in the libraries of various processing units. Although the present conceptualization does always not specify the nature of these processes, it clearly allows for their existence. Moreover, the conceptualization takes into account the possibility that different processes can operate simultaneously. Several alternative processes of encoding, organization, and inference are described in detail throughout this volume. Conscious Versus Nonconscious Processes Second, Wyer and Srull assume that “consciousness” resides in the Executor. Thus, the Executor is “aware” of the general processing stages to be performed, as specified in the goal schema. Furthermore, it is aware of the inputs to processing units, the material that is drawn from Permanent Storage to use in operating on these inputs, and the output of processing at each stage. On the other hand, the cognitive operations performed by the various processing units, which are governed by procedures stored in the libraries of these units, are not subject to awareness. For example, suppose a man is asked his impression of a woman who reads Tolstoy and pickets abortion clinics. He might judge the woman to be both “intellectual” and “feminist” on the basis of these behavioral descriptions. Furthermore, he might be aware that the characterization reminded him of a college professor he knew in the 1960s, and that he based his liking for the woman on his feelings about this professor. This is because the output of processing at each stage is retained in the Work Space and, as such, is accessible to the Executor. On the other hand, the man would have little insight into why he interpreted the behavior in terms of these particular traits rather than others, or into the processes that led to his inference that

ARCHITECTURE OF THE INFORMATION-PROCESSING SYSTEM

37

the woman was similar to the professor. We consider the results of automatic and deliberative processing later in this volume. For the time being, it is sufficient to note that the theory can account in principle for differences in conscious and nonconscious processing and can specify the conditions in which each type of processing may operate.

STORAGE AND RETRIEVAL PROCESSES The assumptions surrounding the storage and retrieval of information are intimately connected to assumptions about the structure of memory. In the original theory, information in Permanent Storage was identified by (a) compiling a set of features (“probe cues”) that specify the type of information desired, (b) identifying a referent bin whose header contains these features, and (c) conducting a probabilistic, top-down search of memory for a knowledge representation that contains these features. These assumptions imply that the likelihood of identifying a particular knowledge representation will increase with both the recency of its acquisition and use (and, therefore, its proximity to the top of the bin) and the frequency of its acquisition and use (i.e., the number of times it is represented). In the present theory, we retain the construct of “probe cues.” However, the bin construct is eliminated and with it, the sequential search processes that Wyer and Srull (1989) postulated. Consequently, the effects of recency and frequency on accessibility are accounted for in other ways.

Storage Our assumptions about the storage of information can be stated simply in terms of a postulate: Postulate 2.1. Each knowledge representation that is formed, either in the course of comprehension or in the course of higher level, goal-directed processing, constitutes a separate unit of knowledge that is stored independently of other knowledge units.

This assumption has broad implications. For example, it suggests that if two conceptually related representations of a referent are formed at different times, they will be stored separately in memory. Consequently, one representation can later be retrieved independently of the other. Thus, for example, suppose people who have formed a detailed representation of a personal experience are later asked to describe this experience to others. They may do so by retrieving the representation and encoding it in terms of

38

CHAPTER 2

more general, semantic concepts that can be communicated verbally. This more general, verbally coded representation is then stored separately from the more detailed representation on which it is based, and can correspondingly be retrieved independently of the latter representation for use in later processing. This latter processing could lead still other representations to be constructed. In the preceding example, people might later recall their abstract representation of the experience and use it to make a trait inference about the protagonists. This inference, in turn, could constitute yet another representation that is stored independently of either of the others. Note that according to Postulate 2.1, a higher order representation does not replace the original. That is, all of the knowledge representations that have been formed coexist in memory. Whether or not a given representation is retrieved, however, depends on factors described shortly. Direct evidence for the independent-storage postulate is somewhat limited. However, a study by Klein and Loftus (1993) is suggestive. Participants were asked both to report whether they had each of several traits and, in addition, to recall a specific behavior that exemplified the traits. However, the order of responding to the two requests was varied. The time that participants required to verify that they had a trait did not depend on whether they had previously recalled a behavior that exemplified it. By the same token, the time to report a trait-related behavior did not depend on whether or not participants had previously reported having the trait implied by this behavior. Thus, participants did not base their trait judgments on the relevant behavior that was salient to them at the time. Furthermore, making a trait judgment that a behavior exemplified did not facilitate their later recall of this behavior. These findings suggest the independence of traits and behavior in memory. The independence of trait representations and behavioral representations may be more characteristic of self-knowledge than of knowledge about others. However, other evidence of the independence of conceptually related knowledge representations are discussed in chapter 6 as well as elsewhere in this volume. This conceptualization obviously requires the assumption that Permanent Storage has unlimited capacity. That is, once a representation has been encoded into memory, it is never erased. Given the virtually infinite number of experiences that individuals have over the course of a lifetime, this assumption may seem nonsensical. An alternative assumption is suggested by distributed memory or connectionist models, which assume that memories are not localized but rather, are represented by an array of activated features that, in different strengths and combinations, can capture the sum total of experiences to which the system has been exposed. In principle, a very large number of unique memories can theoretically be captured by a small number of feature nodes. (E.g., 10 nodes, each of which can

ARCHITECTURE OF THE INFORMATION-PROCESSING SYSTEM

39

be assigned 10 different strengths, could potentially represent 1010 unique memories.) Several distributed memory models of social information processing have been proposed (e.g., Humphreys & Kashima, 2002; Smith & DeCoster, 1998, 1999), and it is conceivable that such a conceptualization of Permanent Storage would be viable in the present context. However, all theories of mental representation are inherently metaphorical, as noted in chapter 1, and should be evaluated on the basis of their utility and not their validity. The assumptions underlying the conceptualization we propose provide a more parsimonious account of the phenomena being considered in this volume than the distributed memory models proposed to date. Retrieval Processes The retrieval processes we postulate are borrowed in part from a previous formulation by Ratcliff (1978) and make use of a “tuning fork” metaphor (for related conceptualizations, see Albrecht & Myers, 1995; Kintsch, 1998). Many implications of this conceptualization are somewhat similar to those of a spreading activation, associative network model (cf. J. Anderson, 1983; Collins & Loftus, 1975; Wyer & Carlston, 1979). In contrast to these latter models, however, our formulation requires no assumptions about the organization of these knowledge representations in memory. We assume that when a configuration of stimulus features is activated, they “vibrate” and this causes the configuration to resonate with knowledge representations (memory traces) that contain some or all of these features. If the resonance of a knowledge representation exceeds a threshold value, it is also activated. When a knowledge representation is no longer involved in processing, it becomes deactivated. Like a tuning fork, however, the resonance does not dissipate instantly, but decreases gradually over time. These processes can be stated more formally in terms of five additional postulates.Ç Postulate 2.2. The activation of a set of stimulus features causes all knowledge representations that contain these features to resonate. The resonance of a given representation, k, increases with (a) the similarity of its features to those in the stimulus set and (b) the length of time the stimulus set has been activated. The subjective similarity of the stimulus features to the representation, and thus the rate at which its resonance increases, is estimated by the quantity: Res(k) =

n(pI k ) , n(k )

where n(k) is the number of features in the representation and n( p I k ) is the number of features that are common to both the representation and the stim-

40

CHAPTER 2

ulus set. If the resonance reaches a threshold value, T (k), the representation is activated.

One implication of Postulate 2.2 is worth noting at this point. That is, when features of the probe set are contained in more than one representation, the resonance at the representation with the fewest number of noncommon features will reach threshold faster. Thus, the probe “football game” is more likely to activate an abstract representation of football games than a specific representation of the 2002 Super Bowl. Postulate 2.3. (a) If two or more representations resonate with a given set of probe cues, the one that reaches threshold most quickly is identified. (b) If several representations reach threshold simultaneously, a composite of the features is activated. The weight of each feature in the composite is a positive function of the proportion of times it occurs in the set of representations involved.

Postulate 2.3a is consistent with the assumptions of the independent trace model proposed by Logan (1988). Postulate 2.3b is suggested by Hintzman’s (1986) conception of the process of abstracting general concepts from exemplars. For example, it implies that when several relevant knowledge representations are activated by a given set of stimulus features, the features that are typical of these representations as a whole are most likely to come to mind and other features, which characterize only a few of these representations, tend to be filtered out. Thus, suppose a baseball fan who encounters the statement “Nolan Ryan pitched a no-hitter” has a single instance of this event represented in memory. In this case, the features “Nolan Ryan” and “no-hitter” in combination should activate a detailed representation of the game in question. In contrast, a fan who has observed all six of Ryan’s no-hit games is likely to activate a more general set of features that are common to these games without thinking about any particular one of them. But suppose the same two individuals encounter the general statement “a ball player pitched a no-hitter.” In this case, both fans might activate a general set of features that do not specify any given baseball pitcher or behavior. The preceding analysis suggests that if baseball fans encounter the statement “Nolan Ryan pitched a no-hitter,” they will typically comprehend it in terms of a previously formed representation of Nolan Ryan engaging in this activity rather than in terms of more general semantic concepts associated with baseball players and no-hit games. That is, they will comprehend the statement in terms of different concepts than they would if they encountered the statement “a baseball player pitched a no-hitter.” Although these observations may seem intuitively obvious, Wyer and Srull’s (1989) model would not make these distinctions.

ARCHITECTURE OF THE INFORMATION-PROCESSING SYSTEM

41

Determinants of Knowledge Accessibility The next two postulates, which are implied by numerous studies of the determinants and effects of knowledge accessibility (for reviews, see Bargh, 1997; Higgins, 1996; Wyer & Srull, 1989), become important in conceptualizing the effect of previously using a knowledge representation on their subsequent reactivation and use at a later time. Postulate 2.4. If a knowledge representation has resonated with a set of probe cues, this resonance does not dissipate immediately but rather, decays gradually over time. Postulate 2.5. The activation threshold of a knowledge representation in memory, T(k), is an increasing function of the number of times the representation has been activated in the past.

In effect, Postulate 2.4 permits the theory to account for the effect of recency of activation on concept and knowledge accessibility that is consistently observed in research on priming effects (Higgins, Bargh, & Lombardi, 1985; Srull & Wyer, 1979). Postulate 2.5 permits the theory to account for the effect of frequency of activation that is also observed in these studies. One implication of these postulates is worth noting. Specifically, the effect of recency of activation, which is a function of the residual resonance that exists at a memory location after the knowledge representation at the location has been deactivated, is transitory. That is, the effect decreases over time as the residual resonance dissipates. In contrast, the effect of activation frequency, which is localized in the level of the activation threshold, is relatively permanent. These implications of the theory are discussed more fully in chapter 3.

GOAL-DIRECTED VERSUS NON-GOAL-DIRECTED PROCESSING: SPONTANEOUS REMINDING PROCESSES Postulate 2.2 specifies the rules that govern the retrieval of a particular knowledge representation from Permanent Storage, given that a set of features are identified for use as “probe cues.” However, it does not indicate the processes that underlie the selection of these features. These processes typically come into play once information has been comprehended and transmitted to the Work Space and, therefore, are governed by the Executor. The nature of this cognitive activity depends in part on whether or not the information is processed in pursuit of a particular goal.

42

CHAPTER 2

Goal-Directed Retrieval Processes If previously acquired knowledge is required for a particular purpose, the following steps are assumed: 1. The Executor compiles a set of features that exist in the Work Space, giving greater weight to those that have been activated and used more recently. These features (“probe cues”) are likely to be ones that (a) are relevant to the objectives one is pursuing and (b) have recently been transmitted to the Work Space by the Comprehender. However, the number of features that are compiled may depend on the specificity of the objective being pursued. When this objective is relatively nonspecific, features may be fortuitously included in the set that are not relevant to it. 2. The probe cues identified in Step 1 vibrate, leading them to resonate with knowledge representations in either the Work Space or Permanent Storage that contain them. The representation(s) that reach activation threshold most quickly, based on criteria specified in Postulate 2.2, are identified, and copies of them are transmitted to the Work Space for use in goal-directed activity. 3. If no representation is identified in Step 2, the set of probe features is recompiled and the search is repeated. These processes are somewhat similar to those proposed by Norman and Bobrow (1979) in accounting for the retrieval of information from “very long-term memory.” The fact that goal-irrelevant features are sometimes included in the probe set has two noteworthy implications. First, the inclusion of these features can sometimes prevent the identification of a goalrelevant representation that exists in memory. This may be experienced subjectively as a “tip of the tongue” phenomenon, where one is temporarily unable to identify the name of a well-known acquaintance or a movie one has seen. In these instances, extraneous probe cues may interfere with the retrieval of a representation containing this information, and so the name may not come to mind unless the set of probe cues is recompiled. Second, the configuration of probe cues that results from the presence of a goal-irrelevant feature can sometimes lead to the retrieval of knowledge that would not normally come to mind. This can account for the spontaneous recall of a person or event that had not been thought about for many years. To borrow an example from Wyer and Srull (1989), a person who is asked to describe his hometown might normally use the name of the town (e.g., “Delhi, New York”) as a probe cue, activating a general representation of its physical characteristics and the type of people who live there. However, suppose the individual had been discussing religion just before being asked to provide this description. Then, a feature activated by this

ARCHITECTURE OF THE INFORMATION-PROCESSING SYSTEM

43

discussion (e.g., “church”) might be fortuitously included in the probe set. This feature, in combination with the town’s name, might stimulate a memory of adjourning to the basement of the church while the collection was being taken in order to avoid putting money in the plate. This behavior may not have been thought about for 40 years, and might have been buried in memory forever if the two probe features had not been fortuitously activated in combination.3 Information Processing During the Free Flow of Thought Much of our cognitive activity in daily life appears to occur without any particular objective in mind. We frequently engage in daydreaming, or let our minds wander aimlessly from one topic to another in an ostensibly haphazard manner. In the course of this free flow of thought, however, we may be reminded of something we have to do, or of a problem we need to solve, and we suddenly find ourselves immersed in thoughts about how to accomplish these objectives. At other times, our cognitive musings can stimulate a memory of a past experience or acquaintance that we have not thought about in years. A complete formulation of information processing in situations outside the laboratory must be able to specify the processes that surround the free flow of thought and the transition of these processes into more deliberative, goal-directed cognitive activities. To account for these processes within the present conceptualization, we assume that when individuals do not have a specific goal in mind, the processing system enters into a continuous feedback loop that continues indefinitely until a goal is identified, either as a result of external demands (i.e., transmitted by the Comprehender) or internally generated cognitive activity. These processes can be described easily with reference to Fig. 2.2. The 3

In comparing the retrieval processes described here with the processes postulated by Wyer and Srull (1989), one further implication should be noted. The earlier model assumed that in the pursuit of goal-relevant information, the Work Space was searched first, before information is retrieved from Permanent Storage. In the present conceptualization, this assumption is an implicit by-product of Postulate 2.5. That is, a set of probe cues is likely to resonate to the greatest extent with existing knowledge representations that have been activated and used most recently. Because the knowledge representations contained in the Work Space have typically been used recently in the pursuit of processing objectives, they would normally be reactivated more quickly than representations in Permanent Storage for use in attaining goals to which they are relevant. To this extent, the two conceptualizations are quite compatible. On the other hand, the content of the Work Space typically consists of representations that have either been (a) recently recalled from Permanent Storage, or (b) recently formed and transmitted to Permanent Storage by the Executor. These representations would be activated in Permanent Storage according to criteria assumed by Postulate 2.2. To this extent, the implications of the assumption that the Work Space is searched first are largely redundant with those of the retrieval postulates that we have already stated.

3

FIG. 2.2. Flow diagram of information processing in the absence of a specific objective (e.g., during the free flow of thought). Circles denote the processes involved. Diamonds indicate conditions that give rise to different types of processing.

44

ARCHITECTURE OF THE INFORMATION-PROCESSING SYSTEM

45

first two stages of processing are similar to Steps 1 and 2 of the process assumed to underlie goal-directed retrieval. That is, the Executor identifies a set of features that happen to exist in the Work Space (either transmitted from the Comprehender or resulting from prior cognitive activity). These features in combination activate a representation in Permanent Storage, a copy of which is sent to the Work Space. The Executor scans this representation for a goal specification and, if none is found, repeats the process just described, recompiling a set of probe cues and using them to activate another representation in Permanent Storage. This continues until a goal specification is identified as a result of the content of the material retrieved from Permanent Storage or transmitted from the Comprehender. Phenomenologically, this simply means that when people do not have a particular goal in mind, the configuration of cognitions they happen to be thinking about cues the retrieval of a previously formed representation. Features of this representation, once retrieved, may combine with other activated concepts to cue the retrieval of yet another representation, and so on until a goal is identified. Thus, for example, suppose a student overhears a conversation in which one person comments that “John gave Peter an answer during the exam.” A representation of the statement that is constructed in the course of comprehending it is transmitted to the Work Space. Features of this representation (e.g., “exam”) along with other features that happen to be in the Work Space at the time (e.g., “psychology”) are sampled, and a set of features is compiled (e.g., “psychology, exam”). These features may resonate with a representation in Permanent Storage that contains them (e.g., the thought that the final exam in psychology is only 3 days before Christmas). As this knowledge does not have goal implications, a new set of probe features is compiled which might now include “Christmas.” These features then resonate with representations in Permanent Storage that contain this feature, perhaps leading a representation to be retrieved that concerns buying a Christmas present for one’s boyfriend. In this case, however, the Executor recognizes this as a goal. Consequently, the system leaves the feedback loop. That is, the Executor identifies a goal schema pertaining to the objective and deposits it in the Goal Specification Box, and then uses the content of the schema to direct processing relevant to attainment of the goal (e.g., the construction of a plan about what to buy and where to purchase it).

CONCLUDING REMARKS The information-processing model outlined in this chapter provides a framework for conceptualizing much if not all of the theory and research discussed in the remainder of this volume. This will be particularly clear in

46

CHAPTER 2

our analysis of the comprehension of personal experiences, and of information about actual persons and events of the sort we encounter in the course of daily life. Several important implications of the conceptualization are localized in Postulates 2.4 and 2.5, which govern the factors that influence the accessibility of knowledge in memory and, therefore, the likelihood that this knowledge is brought to bear on the processing of new information. Indeed, knowledge accessibility is a primary consideration in all phases of information processing from the initial comprehension of information to the generation of an overt response. Before embarking on a more detailed analysis of comprehension and judgment, therefore, a more general discussion of the determinants and effects of knowledge accessibility will be useful. The next chapter provides this discussion.

C H A P T E R

3 Determinants and Effects of Knowledge Accessibility

The retrieval postulates outlined in chapter 2 come into play at many stages of processing and therefore play a role in our discussion throughout this volume. However, they have particular implications for the factors that influence the accessibility of concepts and knowledge in memory. The important role of knowledge accessibility in information processing is recognized in numerous areas of social psychology, including the comprehension of information in terms of semantic concepts (Higgins, Rholes, & Jones, 1977; Srull & Wyer, 1979, 1980), stereotyping (Bodenhausen & Macrae, 1998), consumer judgment (Adaval & Monroe, 2002), cultural influences on judgment (Briley & Wyer, 2002; Hong, Morris, Chiu, & Benet-Martinez, 2000), and overt behavior (Bargh, Chen, & Burrows, 1996; Dijksterhuis & van Knippenberg, 1998). Higgins (1996) provides an extensive review of this research, and a duplication of his efforts is unnecessary here. In this chapter, we restrict our attention to representative phenomena for which the proposed conceptualization has implications.

GENERAL CONSIDERATIONS The general effects of concept and knowledge accessibility are intuitively obvious. That is, people who are called upon to make a judgment or decision seldom consider the implications of all of the knowledge they have acquired that bears upon it. Instead, they use the first relevant judgment or decision criterion that comes to mind, ignoring other criteria that might be equally or more applicable. Taylor and Fiske (1978) were among the first to 47

48

CHAPTER 3

recognize this possibility, arguing that many judgments and decisions are made “off the top of the head.” A more formal statement of this phenomenon was proposed by Chaiken (1980, 1987). She argued that people who have to make a judgment or decision first apply the criterion that comes to mind most quickly and easily and then assess their confidence that the implications of using this criterion are sufficient to attain the goal at hand. Only if this confidence is below some minimal “confidence threshold” do they search for additional criteria. Thus, they may search further if they are highly motivated to make a “correct” decision, or if for some reason they believe that the criterion they initially applied may be biased. Normally, however, they base their judgment or decision on the first relevant criterion they happen to think about. The implications of Chaiken’s conceptualization can be formalized in the following, sufficiency postulate: Postulate 3.1. People retrieve and use only the amount of information that they consider to be sufficient for attaining the processing objective they are pursuing at the time. The information they use is likely to be the first goalrelevant information that comes to mind or, alternatively, permits their objective to be attained most quickly and easily.

Chaiken’s formulation has been applied primarily in conceptualizing responses to persuasive communications (Chaiken, 1987; Chaiken, Wood, & Eagly, 1996). However, it has more general applicability to information processing (Chaiken, Liberman, & Eagly, 1989), and many of the phenomena reviewed later in this volume can be viewed from the perspective she suggests. The discussion in this chapter is restricted to representative research that bears on three main points to which the formulation outlined in chapter 2 is particularly relevant: 1. The primary determinants of knowledge accessibility are (a) the recency with which knowledge has been acquired and/or used in the past and (b) the frequency with which it has been applied. In the latter regard, knowledge that has been employed in many situations over a long period of time can become “chronically” accessible. 2. The accessibility of knowledge representations for use in attaining a particular processing objective can be activated by extraneous factors that are objectively irrelevant to this objective and of which persons may not be consciously aware. In fact, increasing awareness of these extraneous factors is likely to decrease their influence. 3. The effects of differences in knowledge accessibility are apparent at several stages of processing, including the spontaneous comprehension of information at the time it is first received, the more deliberative interpreta-

DETERMINANTS AND EFFECTS OF KNOWLEDGE ACCESSIBILITY

49

tion of information that occurs in the course of attaining a particular goal, inference processes, and behavioral decision making. Differences in knowledge accessibility can also influence the goals that individuals pursue under conditions in which more than one objective is potentially relevant, and the procedures that are spontaneously activated and used to attain a particular goal (e.g., productions of the sort postulated by Anderson, 1983).

FREQUENCY AND RECENCY EFFECTS OF KNOWLEDGE ACTIVATION ON JUDGMENTS AND DECISIONS The earliest research in social cognition to examine the impact of knowledge accessibility was conducted in the area of person impression formation. When people are asked to form an impression of someone on the basis of the person’s behaviors, they often interpret these behaviors in terms of traits. This activity is presumably performed by the Encoder/Organizer based on concepts that are transmitted to it by the Executor. Many behaviors are ambiguous. That is, they have implications for values along several different trait dimensions or, alternatively, for different values along the same dimension. Under these conditions, the interpretation of the behavior can depend on the particular concept that is brought to bear on it. This, in turn, can depend on both the recency and the frequency of its use in the past and, therefore, its accessibility in memory (Postulates 2.4 and 2.5). Once the behavior of a person is interpreted, this interpretation provides the basis for a more general characterization of the person and an evaluation of his or her general likeableness. Early Demonstrations Higgins et al.’s (1977) well-known study was among the first to demonstrate the effects of recency. Participants were asked to form an impression of a target person who ostensibly wanted to cross the Atlantic in a sailboat, was well aware of his ability to do things well, felt that he didn’t need to rely on others, and rarely changed his mind once he had made a decision. Before reading a description of these characteristics, however, participants were unobtrusively exposed to a number of trait adjectives in the course of performing a color-naming (Stroop) task. These adjectives could all be used to interpret the target’s behaviors. In some cases, however, the adjectives were favorable (“adventurous,” “confident,” “independent,” and “persistent”) and in other cases, they were unfavorable (“reckless,” “conceited,” “aloof,” and “stubborn”). Participants apparently interpreted the information in terms of concepts activated by the trait adjectives to which they had

50

CHAPTER 3

been exposed in the color-naming task, and these interpretations influenced their evaluations of the target. That is, they evaluated the target more favorably when they had been exposed to favorable trait adjectives than when they had been exposed to unfavorable ones. In a similar series of studies (Srull & Wyer, 1979, 1980), participants performed a sentence construction task that required the use of concepts associated with hostility. Later, they were asked to form impressions of a target person whose behaviors were ambiguous with respect to this trait (e.g., “refused to pay the rent until the landlord painted his apartment,” “demanded his money back from a salesperson,” etc.). Both the number of hostilityrelated items presented in the sentence-construction task and the time interval between this task and the impression formation task (immediately, 1 hour or 1 day) were manipulated. Both of these factors influenced the likelihood that participants interpreted the target’s behavior as hostile, indicating that both frequency and recency played a role. Moreover, a supplementary study indicated that the effects only occurred when participants were exposed to trait concepts before they read the target information (Srull & Wyer, 1980). When the concepts were activated after the information was presented, they had no effect at all. This indicates that the effects of primed concepts on judgments were mediated by their impact on the interpretation of the trait information at the time the information was received and did not have a direct influence on judgments independently of this information. The Relative Persistence of Frequency and Recency Effects Although the frequency and recency of prior activation both have an impact on the use of trait concepts to interpret information, the underlying basis for these effects theoretically differ. As noted earlier, the effects of recency of activation implied by Postulate 2.4 are transitory, whereas the effects of activation frequency, which impacts on the activation threshold (Postulate 2.5) are more enduring. This was demonstrated empirically by Higgins, Bargh, and Lombardi (1985). Participants were initially exposed to adjectives exemplifying both favorable and unfavorable trait concepts (“adventurous” vs. “reckless”) that could each be used to interpret the same behavior (e.g., “crossing the Atlantic in a sailboat”). However, the relative frequency and recency of exposure to the adjectives were varied. When participants were asked to interpret the behaviors a very short time after exposure to the trait adjective, they tended to use the more recency primed trait concept rather than the frequently activated one. When a longer period of time had elapsed, however, the effects of activation frequency predominated over recency, as im-

DETERMINANTS AND EFFECTS OF KNOWLEDGE ACCESSIBILITY

51

plied by Postulates 2.4 and 2.5. (For an alternative conceptualization of these effects, see Higgins et al., 1985.) The enduring influence of activation frequency has important implications. That is, concepts that have been frequency used in the past can become chronically accessible and, therefore, can influence comprehension and judgments in the absence of any immediate situational factors that activate them. A study by Anderson, Reynolds, Schallert, and Goetz (1977) provides an example. Participants were asked to read the following story: Every Saturday night, four good friends get together. When Jerry, Mike, and Pat arrived, Karen was sitting in her living room writing some notes. She quickly gathered the cards and stood up to greet her friends at the door. They followed her into the living room but as usual, they couldn’t agree on exactly what to play. Jerry eventually took a stand and set things up. Finally, they began to play. Karen’s recorder filled the room with soft and pleasant music. Early in the evening, Mike noticed Pat’s hand and the many diamonds. As the night progressed, the tempo of play increased. Finally, a lull in the activities occurred. Taking advantage of this, Jerry pondered the arrangement in front of him. Mike interrupted Jerry’s reverie and said, “Let’s hear the score.” They listened carefully and commented on their performance. When the comments were all heard, exhausted but happy, Karen’s friends went home.

After reading the story, participants were asked what the protagonists had commented about. The story is normally interpreted as the description of a card game. However, it could also describe the rehearsal of a woodwind ensemble. Some participants, who were students in a weight-lifting class, typically indicated that the protagonists had commented on how well they were playing cards. In contrast, students who were planning a career in music education indicated that the protagonists had commented on the sound of their music. Thus, concepts and knowledge that were chronically accessible to participants as a result of their curriculum and interests apparently affected their interpretation of the story and this interpretation influenced their answers to questions about it. The Additivity of Situationally Induced and Chronic Knowledge Accessibility A further implication of Postulates 2.4 and 2.5 is of particular interest in this regard. That is, the effect of activation frequency (which are mediated by changes in activation threshold) and the recency of activation (which are mediated by the amount of residual resonance that remains at a knowledge representation once it is no longer activated) are independent. This independence was demonstrated by Bargh, Bond, Lombardi, and Tota (1986).

52

CHAPTER 3

They showed that the chronic accessibility of a trait concept (inferred from individual differences in the a priori likelihood of using the trait to describe acquaintances in a previous experiment) and situationally induced accessibility of the trait (resulting from exposure to the trait in a laboratory task) had independent and additive effects on the use of the concept to describe a target person on the basis of information about the target’s behavior. Although the independence of these interpolated effects are implied by Postulates 2.4 and 2.5, they are difficult to explain on the basis of the “bin” conceptualization proposed by Wyer and Srull (1989). According to the latter conceptualization, priming trait concepts should make them equally accessible (either in the Work Space or in the “semantic” bin) regardless of the frequency of their prior activation, and consequently should lead these concepts to be used to an equal extent regardless of their chronic accessibility. In other words, recency should override the effects of chronic accessibility. Therefore, Bargh et al.’s (1986) findings are embarrassing to the Wyer and Srull model. On the other hand, they are quite consistent with the retrieval postulates of the present conceptualization. Generalizability of Concept Activation Over Attribute Dimensions In the study by Higgins et al. (1977), activating favorable and unfavorable trait concepts that were descriptively inapplicable for interpreting the target’s behavior had no influence on participants’ evaluations of him. This finding suggests that the effects of activating trait concepts do not have a direct impact on judgments of the target, but rather, influence judgments only through their mediating influence on how the information about the person’s behaviors is interpreted. As we noted earlier, however, impression formation is a two-stage process. Once a target person’s behaviors are interpreted and a general impression is formed of him, the person may be inferred to have other attributes that are characteristic of the sort of individual implied by this impression. Thus, as Srull and Wyer (1979, 1980) found, activating concepts associated with hostility led the target person not only to be described as hostile but also to have other attributes that are typical of a hostile individual. (For a more recent explication of these two processing stages, see Trope, 1986.) The conclusion that activated trait concepts influence the interpretation of only those behaviors to which they are descriptively applicable may nevertheless need to be qualified. Stapel and Koomen (2000) found that when stimulus attributes with extreme evaluative implications (“malevolent,” “warm,” etc.) are primed, they do appear to influence the interpretation of behaviors that are descriptively unrelated to them. Apparently, adjectives with extreme implications spontaneously activate more general evaluative

DETERMINANTS AND EFFECTS OF KNOWLEDGE ACCESSIBILITY

53

concepts (“good” or “bad”) that, once activated, influence the interpretation of information along evaluative as well as descriptive dimensions.

EFFECTS OF AWARENESS Many effects of previously acquired concepts and knowledge on the interpretations of information can occur without awareness of the factors that led them to become accessible in memory. This possibility is implicit in the conceptualization we propose. That is, consciousness theoretically resides in the Executor. The processes performed by the Comprehender, which are not controlled by the Executor, occur without awareness. Correspondingly, the knowledge representations that are involved in these spontaneous comprehension processes can later be activated and applied without consciousness of why they came to mind or, for that matter, without awareness that these representations had recently been used. For similar reasons, the representations that are later identified by the Executor for use in goaldirected processing are unlikely to contain cognitions about the recency or frequency with which the representations have been formed or used in the past, as these cognitions are typically irrelevant to the objectives for which the representations were formed. Therefore, these representations are likely to be used as bases for judgments and decisions without considering the reasons for their accessibility. Consequently, as studies cited in the preceding section testify, factors that fortuitously make knowledge accessible in memory can increase the use of this knowledge. Moreover, this can occur without awareness of the factors that influence its accessibility. Nonconscious Effects of Knowledge Accessibility The first compelling demonstration of the nonconscious influence of concepts and knowledge was reported by Bargh and Pietromonaco (1982). In this research, participants were exposed to a computerized text of “perceptual vigilance” that required them to press a button as soon as a light appeared on the computer screen. In actuality, the points of light were words that were presented subliminally. In some conditions, a number of words conveyed hostility (“hostile,” “kill,” etc.). After performing the task, participants as part of a different experiment were asked to form an impression of a person on the basis of behaviors that were ambiguous in terms of the hostility they conveyed. Participants were unable to distinguish between words that were presented and those that were not presented in a recognition memory task after the experiment. Nevertheless, they judged the target to be more hostile if they had been subliminally exposed to hostilityrelated words than if they had not.

54

CHAPTER 3

Even when people are conscious of the conditions in which knowledge has been acquired, they may not include a specification of these conditions in the representation of this knowledge that they store in memory. Consequently, they may later retrieve and use the representation as a basis for judgment without considering the conditions that led the knowledge to be acquired. Two studies are interesting to view in this light. In a study by Hasher, Goldstein, and Toppin (1977; see also Kelley & Lindsey, 1993), participants completed a general knowledge questionnaire in which they judged the validity of several statements about unfamiliar events and situations. Then, in a second session of the experiment, they completed a similar questionnaire containing some of the same statements they had considered earlier. Participants reported stronger beliefs in the validity of these statements than they had in the first session. Representations of the belief statements that participants had considered in the first session were apparently stored in memory without any indication of the conditions in which they were encountered. Consequently, these representations were retrieved when participants encountered the statements in the second session, leading the statements to seem familiar. However, participants attributed the statements’ familiarity to information they had received outside the laboratory rather than to the presence of the statements in the original questionnaire. As a result, they inferred that the statements were likely to be true. In a conceptually similar study, Jacoby, Kelley, Brown, and Jasechko (1989) exposed participants to a number of names of persons, some of whom were familiar and others of whom were fictitious. Then, 24 hours later, they were given a second list that included a subset of the fictitious names they had seen earlier and were asked to indicate which names in the list referred to well-known persons. Participants were more likely to misidentify fictitious persons as well known if they had seen their names in the earlier list than if they had not. Thus, both this study and Hasher et al.’s research suggest that knowledge that is activated in one context can later be recalled and used as a basis for judgments without considering the reasons that led it to be familiar. The general tendency for representations of knowledge to be stored in memory independently of the conditions in which the knowledge was acquired has particularly important implications for the impact of daily life experiences on perceptions of social reality. As Shrum and his colleagues (O’Guinn & Shrum, 1997; Shrum, Wyer, & O’Guinn, 1998) have found, people often confuse the persons and events they encounter in the real world with those they have seen on television. Consequently, heavy television viewers tend to overestimate the incidence of events that occur frequently in soap operas and crime shows. We elaborate this research more fully in chapter 10 in the context of discussing the role of implicit theories on judgments and decisions.

DETERMINANTS AND EFFECTS OF KNOWLEDGE ACCESSIBILITY

55

Effects of Awareness on Avoidance of Bias In fact, when people are aware that the concepts they used when performing an ostensibly unrelated task might influence their interpretation of the information they receive later, they may consciously avoid the use of these concepts. In a study by Lombardi, Higgins, and Bargh (1987), for example, participants were exposed to trait concepts in a priming task that were associated with earlier (a) adventurousness and self-confidence or (b) recklessness and conceitedness. They were then asked to report their impressions of a person whose behaviors could exemplify both sets of traits and, finally, were asked to recall the priming words to which they had been exposed. Participants who could not remember the priming words were positively influenced by them in much the same way observed by Higgins et al. (1977) in the study described earlier. However, participants who could remember the words showed a boomerang effect; that is, they were less inclined to interpret the target’s behavior in terms of the primed concepts than in terms of the alternatives. Similar results were obtained by Strack, Schwarz, Bless, and Kuebler (1993) in a different research paradigm. Even when people are aware of the reasons why concepts and knowledge come to mind, they may not always avoid using them. This avoidance should only occur when the individuals are both motivated and able to search for alternatives. Martin, Seta, and Crelia (1990) demonstrated these contingencies under conditions in which trait concepts that were applicable for interpreting a target’s behavior were blatantly primed. When participants were both able and motivated to search for alternative concepts to use, boomerang effects of priming occurred; that is, participants were less inclined to use the primed concept than they were in the absence of priming. However, when participants were distracted while performing the evaluation task and were less able to search for alternative concepts to apply, primed concepts had a positive impact on judgments similar to those observed by Higgins et al. (1977) and others. A second study by Martin et al. (1990) showed that primed concepts only had a positive influence on target judgments when participants had little intrinsic motivation to think about the information received (i.e., low need for cognition; see Cacioppo & Petty, 1982) but had boomerang effects on judgments by participants whose motivation (need for cognition) was high. Particularly interesting are the results of a third study in which participants were led to believe they were forming impressions of the target person either as individuals or as members of a group. Primed concepts had boomerang effects on judgments by individuals who believed they were participating as individuals. However, participants who thought they were participating as a group engaged in “social loafing” (Harkins, Latane, & Wil-

56

CHAPTER 3

liams, 1980). That is, they used the primed concepts as a basis for interpreting the information without bothering to seek alternatives. The Effect of Thought Suppression on Concept Accessibility It therefore seems reasonable to conclude that when people are aware that the concepts and knowledge that come to mind at the time they make a judgment, they often actively avoid the biasing influence of these cognitions and, in doing so, are actually less likely to use them than they might otherwise be. However, the effects of this cognitive activity can be somewhat ironic. Suppose individuals consciously avoid the use of a concept or knowledge representation in responding to a stimulus. This conscious suppression takes cognitive effort. Moreover, it is impossible to consciously avoid using a concept without thinking about the concept one is trying to avoid. Consequently, active attempts to avoid using a concept can actually increase its accessibility in memory, making the concept more likely to be used later under conditions in which attempts to suppress it are no longer made. Thus, as Wegner (1994) noted, instructing people to avoid thinking about a white bear may stimulate conscious thoughts about the bear, thus making it more accessible in memory than it would have been in the absence of this cognitive activity and, therefore, more likely to come to mind later.4 An intriguing series of studies by Bodenhausen, Macrae, and their colleagues demonstrated the effects of conscious thought suppression in the use of stereotypes (for a review, see Bodenhausen & Macrae, 1998). In a representative set of studies (Macrae, Bodenhausen, Milne, & Jetten, 1994), participants were given information about a skinhead with instructions to form an impression of him. However, some participants were told not to rely on stereotypes in arriving at their impression, whereas others were not given this instruction. The former participants, relative to those who had not been told to avoid using the stereotype, (a) responded more quickly to stereotype-related words in a subsequent lexical decision task, (b) described a second skinhead more stereotypically once the request to avoid using stereotypes was no longer imposed, and (c) sat further away from a skinhead while waiting to participate in a later part of the experiment. Thus, 4

A game we used to play as children anticipated Wegner’s insight. That is, we would tell naïve playmates to try not to think of a white horse and then, after ensuring that they were doing so, would ask them to report the direction in which the horse was facing. They invariably found this question quite meaningful and responded to it quickly. Attempts to suppress thoughts about the horse typically elicited an image of the horse that individuals found easy to describe despite their assurances that they were trying not to think about it.

4

DETERMINANTS AND EFFECTS OF KNOWLEDGE ACCESSIBILITY

57

the attempt to suppress the use of the stereotype in the initial task increased rather than decreased the accessibility of concepts associated with the stereotype. Consequently, it increased the use of the stereotype as a basis for judgments and behavioral decisions in situations in which the sanctions against using it were removed.

EFFECTS OF KNOWLEDGE ACCESSIBILITY AT DIFFERENT STAGES OF PROCESSING The retrieval processes implied by Postulates 2.1 through 2.5 apply at all stages of cognitive activity, from the initial interpretation of information to the generation of an overt response. The role of knowledge accessibility in these processes is central to our discussion throughout this volume, being particularly evident in conceptualizing the impact of generalized representations on memory for exemplars (chap. 5), the role of implicit theories in explanation and prediction (chap. 10), and the impact of affective reactions on judgments and decisions (chap. 11). In anticipation of this discussion, a brief review of the effects of knowledge accessibility at different stages of processing may be helpful. This review is not intended to be exhaustive (for more extensive reviews, see Bargh, 1997; Higgins, 1996). Instead, we provide representative examples of the effects that occur, several of which raise considerations that become important in our later discussion. Input Processes: Selective Attention and the Activation of Bipolar Concepts As Bargh and Pietromonaco’s (1982) findings indicate (see also Srull & Wyer, 1979, 1980), information that is ambiguous in terms of its implications for a given attribute (e.g., hostility) is more likely to be interpreted as exemplifying this attribute when concepts associated with this attribute are easily accessible in memory than when they are not. However, many attribute concepts are bipolar (hot vs. cold, hostile vs. kind, etc.). Moreover, these concepts may become associatively linked in memory by virtue of having been thought about in relation to one another. If this is so, exposure to one bipolar concept (e.g., “honest”) is likely to activate the other (“dishonest”) as well and, therefore, to increase the likelihood that the latter concept is used to interpret information to which it applies. To see the implications of this, suppose two pairs of bipolar concepts, “honest/dishonest” and “kind/unkind,” are associatively linked as shown in Fig. 3.1. Suppose further that the behavior, “refuses to give an answer to a friend during an exam” is associatively linked to both “honest” and “un-

58

CHAPTER 3

FIG. 3.1. Hypothetical associative network showing the relations among behaviors, bipolar trait concepts (honest, dishonest, kind, and unkind), and the priming stimuli that activate these concepts (Ph, Pd, Pk, and Pu, respectively).

kind.” Then, priming “honest” should increase the likelihood of interpreting the behavior as “honest,” whereas priming “unkind” should increase the likelihood of interpreting it as “unkind.” However, because “dishonest” and “kind” are associatively linked to “honest” and “unkind,” respectively, priming these concepts should have similar effects. In other words, priming either bipolar concept along a trait dimension should increase the likelihood of using a value along this dimension to interpret information to which the dimension is relevant. Park, Yoon, Kim, and Wyer (2001) confirmed these hypotheses. In one experiment (Park et al., 2001, Experiment 1), participants performed a scrambled sentence task that required the use of one of four concepts: kind (e.g., “offered carry to help elder and luggage”), unkind (“refused carry to help elder and luggage”), honest (e.g., “homework to copy refused another’s”) or dishonest (e.g., “homework to copy tried another’s”). Later, they read a story containing descriptions of a target person’s behavior that could be interpreted as both honest and unkind (“told a girlfriend that her new hair style looked terrible,” “insisted that a friend leave his address on a car he had barely scratched,” etc.). Participants read the behaviors and then both described the target in their own words and, after doing so, evaluated the person. The proportions of participants who used the trait exemplified by the behavior (i.e., either honest or unkind) to describe the target are shown in Table 3.1 as a function of priming conditions. When primed concept was directly applicable for interpreting the target’s behavior, participants were more likely to use this concept (pooled over the two relevant priming conditions, M = .294) than to use the equally applicable but unprimed concept

59

DETERMINANTS AND EFFECTS OF KNOWLEDGE ACCESSIBILITY TABLE 3.1 The Proportion of Times the Target Was Described as Honest and Unkind as a Function of Priming Conditions Primed Concept Applicable for Describing Target

Proportion of times the target was described as Honest Unkind

Primed Concept Inapplicable for Describing Target

Honest

Unkind

Dishonest

Kind

.280 .160

.115 .308

.407 .074

.231 .231

Note. Based on data from Park et al. (2001).

(M = .137). However, this was also true when the bipolar opposite (inapplicable) concepts were primed (.315 vs. .152, respectively).5 Finally, these interpretations influenced participants’ overall evaluations of the target. That is, participants evaluated the target more favorably when they had been primed with a trait pertaining to hostility (M = 4.55) than when they had been primed with a trait pertaining to kindness (M = 3.76), and this was true regardless of whether the primed trait was directly applicable for interpreting the behavior (4.08 vs. 3.58 when “honest” and “unkind” were primed, respectively) or when the bipolar opposites were primed (5.02 vs. 3.94, when “dishonest” and “kind” were primed, respectively). The behavioral information that participants received in the preceding study pertained to two different dimensions, but its implications along each dimension were very clear. An additional experiment (Park et al., 2001, Experiment 2) showed that when the implications of information along a dimension are ambiguous, activating concepts can influence not only the dimension along which the information is evaluated but also the value it is assigned along this dimension. In this study, participants first performed a task that required the use of concepts associated with either “good health,” “bad health,” “good taste,” or “bad taste.” Later, they were given the description of a milk product as having “more artificial sweeteners.” Although this description clearly implied that the product was bad for the health, its implications for tastiness were ambiguous. That is, it could imply that the 5

This latter effect appears more pronounced when “dishonest” was primed than when “kind” was primed. However, this difference results from a more general tendency to judge the target as honest rather than as unkind, which inflates the effect of priming “dishonest” but decreases the effect of priming “kind.” Pooling over the two trait replications, the overall effect of priming was similar in magnitude regardless of whether the primed concept was applicable for describing the target or its bipolar opposite.

5

60

CHAPTER 3

product was either good tasting (sweet) or bad tasting (too sweet). Participants who had been exposed to health-related concepts described the product as unhealthy regardless of whether “good health” or “bad health” was primed. In contrast, participants who were exposed to taste-related concepts described it as having either good taste or bad taste, depending on which taste-related concept had been activated. Thus, primed concepts influenced not only the dimension along which the information was interpreted but also the value it was assigned along this dimension. Inference Processes: Belief Formation and Change People who are asked to report their belief that a statement concerning it is true are unlikely to search memory for all of the belief-related knowledge they have acquired. Rather, they are apt to rely on the first belief-relevant knowledge that comes to mind (Postulate 3.1). That is, they search memory for a possible reason why the statement might or might not be true and, if they find such a reason, use its implications as a basis for their judgment without further ado. To this extent, the reason they identify and, therefore, the belief they report, may depend on the recency with which the reason had been thought about. We discuss this possibility in more detail in chapter 10 of this volume. However, two examples are worth mentioning in the present context. Participants in a study by Henninger and Wyer (1976) were asked to report their beliefs in a series of propositions. Some propositions were syllogistically related. That is, one, “informational” proposition, A (e.g., “The army is recruiting people of below average intelligence”), had implications for the validity of a second, target proposition, B (e.g., “The quality of the peace-time army is deteriorating”). In some cases, beliefs in the informational proposition were reported early in the questionnaire, and the target proposition was not considered until later. In other conditions, the target proposition was considered at the outset. Participants’ use of their belief in the informational proposition as a basis for inferring the validity of the target was inferred on the basis of a quantitative model of syllogistic inference proposed by Wyer and Goldberg (1970; see also McGuire, 1960, 1981).6 Par6

6

If beliefs are defined in units of probability (along a scale from 0 to 1), the model asserts that P(B) = P(A)*P(B/A) + P(~A)*P(B/~A)

Where, P(B) is the belief that Proposition B is true, P(A) and P(~A) are the beliefs that A is and is not true, and P(B/A) and P(B/~A) are beliefs that B is true if A is and is not true, respectively. This model provides an accurate quantitative description of the relations among these beliefs, and also the effect of changing beliefs in A on beliefs in B (Wyer, 1970, 1974; for summaries of evidence, see Wyer, 1974; Wyer & Hartwick, 1980). The implications of this model are discussed more fully in chapter 10.

DETERMINANTS AND EFFECTS OF KNOWLEDGE ACCESSIBILITY

61

ticipants based their beliefs in the target on their beliefs in the informational proposition when they had evaluated this proposition earlier in the questionnaire. When they evaluated the target proposition at the outset, however, they typically identified a different subset of previously acquired knowledge to use as a basis for reporting their belief, and so this belief was inconsistent with their belief in the informational proposition they considered subsequently. Interestingly, when participants reported their beliefs again a week later, their beliefs in the target were consistent with their beliefs in the informational proposition in both order conditions. Apparently, the informational proposition, having been reported in Session 1, was accessible to both groups of participants in Session 2, and so both groups of participants used it as a basis for inferring the validity of the target. This interpretation obviously assumes that no other relevant information bearing on the validity of the target was received or thought about during the 1-week interval between sessions. The target propositions pertained issues that participants were unlikely to think about spontaneously, and so this assumption seems reasonable. In an unpublished study by Wyer and Hartwick (see Wyer & Srull, 1989), however, some participants were asked to report their beliefs that drinking coffee was desirable after having been exposed to a proposition that either implied that this was true (“Coffee keeps you alert”) or implied that it was false (“Caffeine destroys nerve cells”). Others were asked to report their beliefs that student use of the university health center would increase after exposure to a proposition suggesting that the event was likely to occur (“Dental services are scheduled to be provided at the health center”) or unlikely (“Some doctors at the health center are about to lose their licenses”). In both cases, participants based their beliefs in the target proposition on the implications of the particular information proposition to which they had been exposed. However, whereas the effects of informational propositions on participants’ beliefs about the use of the health center were still evident in a second experimental session 1 week later, their effects on beliefs in the desirability of drinking coffee were not. The reason for this difference is rather obvious. That is, participants in the 1-week interval between experimental sessions were likely to have numerous coffee-drinking experiences with implications for the desirability of this activity, and these more salient implications overrode the effects of the knowledge they had selectively activated and used in the first experimental session. In contrast, participants had few if any experiences with implications for the use of the health center during the 1-week interim, and so the knowledge they had activated in the first session continued to have an impact. The studies described in this section demonstrate the effects of activating prior knowledge about an issue on beliefs and opinions. However, this is

62

CHAPTER 3

not the only type of knowledge that people bring to bear on these judgments. An equally important source of information is provided by one’s own past behavior toward the persons and objects to which the issue pertains. Bem (1972), for example, noted that people often use their past behavior (e.g., a decision to advocate a position publicly) to infer their belief that the position is valid (see also Albarracin & Wyer, 2000). These effects are likely to be mediated by an implicit theory that people have acquired about the causal relatedness of their beliefs and their behavior under the particular situational conditions at hand. We explore this possibility more fully in chapter 10. Response Processes: The Use of Activated Concepts as Standards of Comparison The effects of accessible concepts on the interpretation of information typically occur at the time the information is first received. At the time a judgment is made, the same concepts can have other effects. In particular, judgments are typically made with reference to a standard of comparison. Moreover, these standards may be applied spontaneously, without awareness. Thus, we may judge a baby to be “big” and a house to be “small” without feeling any inconsistency in these judgments at all. The effects of knowledge on standards of comparison can sometimes offset or override the effects that occur at the comprehension stage. A study by Herr (1986) is of interest in this context. Participants were initially exposed to the names of persons who were known to be either moderately hostile (e.g., Muhammad Ali or Howard Stern) or extremely hostile (e.g., Adolf Hitler or Genghis Khan). Later, they were asked to form impressions of a person whose behaviors were ambiguous with respect to this trait. It seems reasonable to suppose that the priming stimuli employed in this study had two effects. First, they activated trait concepts that were associated with the individuals being primed, leading participants to interpret the target’s behaviors in terms of these concepts. At the same time, they activated concepts of the individuals themselves that participants later used as standards of comparison in evaluating the target. These effects could potentially offset one another. Herr’s results were consistent with this possibility. Participants who were exposed to moderate exemplars of hostility judged the target as more hostile than participants in a control condition, confirming the findings obtained by Srull and Wyer (1979) and by Bargh and Pietromonaco (1982). However, participants who were exposed to extreme exemplars judged the target to be less hostile than control participants did. The exemplars to which participants were exposed may have been used as standards of comparison in both conditions. When the exemplars were only moderately ex-

DETERMINANTS AND EFFECTS OF KNOWLEDGE ACCESSIBILITY

63

treme, however, the effect of using them as standards was not sufficient to override the effect of the activated trait concepts on interpretation of the information at the time it was received. When the exemplars provided a more extreme standard of comparison, the effect of using them in this capacity predominated, producing a contrast effect. In this regard, standards of comparison for evaluating a person are generally less likely to be activated by trait concepts per se than by the exposure to persons who possess the traits. Thus, as Stapel, Koomen, and van der Pligt (1997) found (see also Moskowitz & Skurnik, 1999), contrast effects of priming are more likely to be evident in the latter case than the former. The activation and use of stimuli as standards of comparison can occur without awareness. This was demonstrated in a series of studies by Adaval and Monroe (2002). In some of their studies, participants were exposed subliminally to either high or low numbers before being asked to judge a target product whose price fell between the two sets of priming stimuli. Participants’ judgments of the target product were lower when high numbers had been subliminally primed than when low numbers had been primed. That is, the target was judged as less expensive, but also less desirable, in the former condition than in the latter. Thus, priming stimuli appeared to induce a disposition to use either a high or low value as a standard of comparison regardless of the dimension along which judgments were made.

THE EFFECTS OF PRIOR JUDGMENTS ON SUBSEQUENT ONES In most research on the effects of knowledge accessibility, the concepts and knowledge that participants bring to bear on their judgments have been activated by situational and individual differences factors that are ostensibly irrelevant to the judgment or decision they are called upon to make. However, concepts and knowledge are often activated in the course of cognitive activity that is directly relevant to the persons or objects being judged. For example, people in the course of judging an object may retrieve concepts and knowledge that are specifically relevant to this judgment. If these cognitions remain accessible at the time a second judgment or decision is made, they may influence the latter response as well. Studies described earlier can be interpreted in this light. For example, participants in an early study by Carlston (1980) were given behaviors that exemplified by kindness and honesty (e.g., “told his girlfriend that her new hair style was ugly”), and were asked to judge the person with respect to one of these traits. In the course of making this judgment, participants presumably formed a general impression of the person, based on their prior knowledge about the sort of individual who typically possesses the trait

64

CHAPTER 3

they are asked to judge. Once this impression was formed, it may have influenced their later judgments independently of the behavioral information they had received earlier. Thus, in our example, participants who were asked to judge the target’s honesty judged him as honest. Having done so, they later judged the target as more kind than they would have done if they had not made the initial judgment. A quite different series of studies by Thomas Mussweiler and Fritz Strack (for a review, see Mussweiler, 2003) provide a further example of these effects and confirm assumptions underlying their interpretation. In a typical study (Mussweiler & Strack, 2000b), participants were first asked to indicate whether Mt. Everest was greater or less than a specific anchor value (e.g., “Is Mt. Everest higher or lower than feet?”). The anchor value was either high or low. In all cases, these values were sufficiently extreme that participants would normally consider the target to fall in between them. In some cases, however, the values were fairly plausible (e.g., 45,000 ft. vs. 2,000 ft.), whereas in other cases, the values were totally implausible (e.g., 300,000 ft. vs. 65 ft.). Participants after making the comparative judgment were asked to estimate the actual height of Mt. Everest. Participants typically judged the actual height of Mt. Everest to be greater when they had compared it to a high value than when they had compared it to a low one. Moreover, this tendency was even greater when the anchor values were implausible (115,728 ft. vs. 9,271 ft., respectively) than when they were plausible (36,106 ft. vs. 27,783 ft., respectively). On the surface, these results appear to resemble the anchoring-andadjustment phenomenon identified by Tversky and Kahneman (1974). That is, people in judging a stimulus arbitrarily identify a high or low anchor value along the judgment dimension and then adjust upward or downward in relation to this standard. However, they do not adjust enough, with the result that their judgment is displaced toward the anchor they have arbitrarily used (for examples of this effect in social judgment, see Schwarz & Wyer, 1985). Mussweiler and Strack (1999a) raise an alternative possibility, however. That is, people who estimate a stimulus in relation to a standard may activate a body of general knowledge that is consistent with it. Although they may ultimately reject the anchor value as implausible, the subset of knowledge and its implications, having been made accessible, is then used as a basis for their own estimates. To distinguish between the two alternative interpretations, Mussweiler and Strack (1999a) asked participants after making their comparative judgments to perform a lexical decision task requiring them to identify words that were semantically associated with either high anchor values, low anchor values, or irrelevant. Participants were quicker to identify words that were associated with the anchor to which they had been exposed than the opposite anchor, confirming the assumption that exposure to the anchor in-

DETERMINANTS AND EFFECTS OF KNOWLEDGE ACCESSIBILITY

65

creased the accessibility of anchor-consistent knowledge in memory. The anchoring-and-adjustment processes assumed by Tversky and Kahneman (1974) would not predict this difference. The phenomena identified by Mussweiler and Strack generalize over a variety of content domains (Mussweiler & Strack, 1999a, 1999b, 2000b). A particularly intriguing implication of their conceptualization surrounds the effects of comparing oneself to a standard on self-judgments. In one study (Mussweiler & Strack, 2000b), participants were first asked whether the number of simple mathematical computations (e.g., 2 + 5 = ?) they could perform in 1 minute was greater or less than either a high value (62) or a low one (10). After doing so, they estimated the actual number they could do and also rated their intelligence along a scale from 1 to 7. Consistent with results of earlier studies, participants estimated that they could do more problems when they had been exposed to the high anchor than when they had been exposed to the low one. On the other hand, they rated their general intelligence to be lower in the first condition. This combination of results makes salient two different effects that standards of comparison can have. On one hand, a standard can activate a selective subset of general knowledge that is consistent with it, thus biasing the criteria that people bring to bear on their numerical estimates. At the same time, the standard can influence the range of subjective values with which these numerical estimates are associated, as suggested by Parducci (1965) and Upshaw (1969), leading persons to assign lower values to these estimates when the anchor is high than when it is low (for an analysis of these effects and the processes underlying them, see Ostrom & Upshaw, 1968; Wyer & Srull, 1989).

DETERMINANTS AND CONSEQUENCES OF THE ACCESSIBILITY OF GOALS AND MOTIVES The preceding discussion focused on the accessibility of concepts and knowledge that are used to interpret and make inferences about an external referent. As noted in chapters 1 and 2, however, goals and motives are also components of the cognitive system. As such, the activation of these motives and their influence on judgments and decisions should be subject to similar processes. People are obviously motivated by a wide variety of general and specific goals, and the behaviors that are required to attain these goals in any given situation may not be compatible. In this case, people’s behavioral decisions may depend in part on which goal happens to predominate at the time the decisions are made. In some cases, the relative accessibility of a particular goal or motive is likely to be activated by features of the stimuli to which the goal is relevant.

66

CHAPTER 3

Shiv and Fedorikhin (1999), for example, found that people were inclined to choose to eat chocolate cake rather than fruit salad when they did not have the opportunity to think about their decision, whereas the reverse was true when they were able to deliberate. Presumably, the visual stimulus features of the cake led hedonic goals to be activated very quickly, whereas the competing goal of good health took more time to be activated. In many instances, however, people may acquire more general motivational orientations that, once activated, can influence their judgments and behavior in a variety of specific situations. Such general motivational dispositions are often activated by features of a stimulus situation to which they are relevant. However, their accessibility can also be influenced by extraneous factors that are not directly relevant to the situation at hand. Two quite different bodies of literature provide evidence of this possibility. One concerns the desire to maintain beliefs in a just world (Lerner, Miller, & Holmes, 1976). A second surrounds the relative emphasis placed on positive versus negative consequences in making a behavioral decision (cf. Higgins, 1997). The Motivation to Believe in a Just World: Implications for Reactions to Rape Motivational influences on attitudes and opinions, which were called to the attention of social psychologists by Festinger (1957), are well established. The nature of these effects obviously depends on the particular motive involved (e.g., see Arkin, Gleason, & Johnson, 1976; Baumeister & Newman, 1994; Giner-Sorolla & Chaiken, 1987; Psyzczynski & Greenberg, 1987; Zanna & Cooper, 1976). One motivational determinant of judgments, proposed by Lerner et al. (1976; Lerner & Miller, 1978), is the desire to believe that the world is just. Lerner et al. assume that people have a need to maintain beliefs in a just world in order to reassure themselves that they (who are righteous) will personally not fall victim to misfortune for reasons beyond their control. One manifestation of this motivational disposition is that people are not only disposed to believe that individuals who do something wrong will be punished but also, that people who encounter misfortune are responsible for their fate. In other words, individuals not only get what they deserve but deserve what they get. Lerner et al. (1976) provided several examples of this tendency. Walster (1966), for example, found that people judged victims of an automobile accident more negatively if the victims had been seriously injured than if they had not. Similarly, Lerner and Simmons (1966) found that observers of a “learning” experiment in which a confederate was ostensibly shocked for making errors disparaged the learner more when the shocks were severe than when they were mild.

DETERMINANTS AND EFFECTS OF KNOWLEDGE ACCESSIBILITY

67

Beliefs in a just world are not the only possible basis for judgments and decisions, of course. To this extent, the use of this criterion should be increased by situational factors that increase the accessibility of these beliefs and the motive to maintain them. This possibility was demonstrated by Wyer, Bodenhausen, and Gorman (1985) in a study of the cognitive mediators of reactions to rape. We were interested in whether activating aggression-related concepts in one context might influence people’s reactions to rape situations that they encountered in a different, ostensibly unrelated context. To do so, we asked participants to engage in two ostensibly unrelated experiments. The first study was ostensibly concerned with the things shown in the media that college students find objectionable. On this pretense, we exposed participants to slides of 12 pictures. Nine of these pictures showed objects and events that participants were unlikely to consider offensive. The other three pictures varied. In one condition, these slides showed aggressive acts of the sort that occurred frequently (e.g., police subduing a criminal, a boxing match, etc.) and, therefore, activated concepts that aggression is normal and socially sanctioned. In a second condition, the pictures portrayed severely negative outcomes of aggression that activated the concept that human beings were cruel and inhumane (e.g., a lynching episode, a dead soldier with a hole in his head, etc.). The third, “control” set portrayed stimuli that might be considered unpleasant but were unrelated to aggression (e.g., deformed babies, a smoking advertisement, etc.). Participants rated each of the 12 pictures in terms of how objectionable it was. Then, they were told that the experiment (which took about 10 min) was over, but that because there was time remaining, we would like them to help out another faculty member who was conducting a study in a different room down the hall. The “new” study was introduced as an investigation of the factors that people consider important in judging criminal cases. On this pretense, participants were asked to read descriptions of four rape cases, and in each case, to report several reactions. Two questions concerned (a) whether the defendant should be convicted and (b) whether he actually was convicted. Three others concerned the victim’s responsibility for the incident (e.g., whether she provoked the rape, whether she could have avoided it, etc.). We assumed that the pictures presented in the first experiment would activate concepts that participants would use to construe the implications of the rape scenarios they encountered in the second experiment. On a priori grounds, several hypotheses seemed plausible. First, if pictures of socially sanctioned acts of aggression activated concepts that aggression is normal, they might decrease beliefs that the defendant should be convicted. On the other hand, if pictures of extremely negative consequences of aggression activate concepts that people are cruel and inhumane, they

68

CHAPTER 3

might increase beliefs that the defendant should be convicted. Moreover, they might increase sympathy for the victim, decreasing beliefs that she was responsible for the incident. An alternative possibility, however, is suggested by the assumption that exposure to instances of aggression threaten people’s beliefs that the world is just and, therefore, activate a motive to reaffirm this belief. This motive, once activated, may influence participants’ reactions to the rape incidents they encounter subsequently. If this is so, exposure to severe consequences of aggression should increase beliefs that the defendant was punished (i.e., that he got what he deserved). However, it should also increase beliefs that the victim was responsible for the incident (that she deserved what she got). Results summarized in Table 3.2 confirm the latter hypothesis. Exposure to acts of aggression had little influence on judgments of either the defendant or the victim. (If anything, they increased beliefs that the defendant should be convicted, contrary to the assumption that these beliefs would be mediated by perceptions that aggression was normal and socially sanctioned.) Moreover, exposure to severe consequences of aggression also had no effect on participants’ beliefs that the defendant should be convicted. However, exposure to these stimuli increased beliefs that the defendant was convicted and, at the same time, increased beliefs that the victim was responsible for the incident. Thus, these results are consistent with the hypothesis that exposure to extreme consequences of aggression activated concepts that threatened participants’ perceptions that the world is just. Therefore, it motivated them to reaffirm these perceptions by believing not only that the defendant got what he deserved, but also that the victim deserved what she got. These results could have implications for the impact of the media on perceptions of rape and reactions to aggression more generally. The effects TABLE 3.2 Judgments of the Defendant and Victim as a Function of Concepts Activated by Priming Stimuli Priming Stimuli

Belief that defendant should be convicted Belief that defendant was convicted Belief that victim was responsible

Control

Socially Sanctioned Aggressive Acts

Severe Outcomes of Aggression

8.97 3.95 2.97

9.62 3.47 3.07

8.70 5.10 4.20

Note. Judgments are reported along a scale from 0 (not at all likely) to 10 (very likely). Based on data from Wyer, Bodenhausen, and Gorman (1985).

DETERMINANTS AND EFFECTS OF KNOWLEDGE ACCESSIBILITY

69

of situationally primed concepts on judgments are often of short duration (but see Srull & Wyer, 1979, for evidence of priming effects over a period of 24 hours). However, the effects of frequent exposure to stimuli on concept accessibility are much more enduring (Higgins, 1996; Higgins, Bargh, & Lombardi, 1985). It therefore seems reasonable to assume that frequent exposure to extreme violence could induce a chronic tendency to maintain beliefs in a just world that is manifested in a variety of contexts. To this extent, it could have effects similar to those that Wyer, Bodenhausen, and Gorman (1985) observed. The Motivation to Avoid Negative Outcomes People who are confronted with a decision are likely to base their judgments on their perception of its possible consequences. In doing so, they might potentially consider both the positive consequences that might result from their decision and the negative consequences it could have. However, both individual and situational differences may exist in the relative emphasis that people place on these consequences. This possibility is recognized by Higgins’ (1997) conceptualization of regulatory focus. Higgins postulates that individuals can have either a promotion focus (which is manifested by a concern with positive consequences of one’s behavior) or a prevention focus (characterized by a concern with avoiding negative decision outcomes). These motivational orientations may reflect general dispositions that, once activated, generalize over situations. Although chronic individual differences can exist in these orientations, their relative influence can also be determined by situational factors that influence their relative accessibility. 1. Chronic Differences in Promotion Versus Prevention Focus Chronic differences in the motivation to attain positive outcomes or avoid negative ones are suggested in a study by Briley, Morris, and Simonson (2000). At the same time, these motives may not be activated spontaneously. Briley et al. speculated that European Americans were disposed to be promotion focused, whereas Asians were more inclined to be prevention focused. To examine this possibility, representatives of both cultural groups were asked to choose between (a) an alternative with a very favorable value along one dimension (A) and a very unfavorable value along a second dimension (B), (b) an alternative with a very unfavorable value along A and a very favorable value along B, and (c) an alternative with moderate values along both dimensions. Thus, the first two alternatives had both more favorable and more unfavorable features than the third, “compromise” alternative. Briley et al. reasoned that participants

70

CHAPTER 3

who were primarily motivated by the desire to attain positive outcomes would prefer either of the first two alternatives to the third. However, individuals who were motivated to avoid negative outcomes would prefer the compromise alternative. They found that when participants were asked to make decisions without having to justify their choices, cultural differences in these choices did not emerge. However, asking European Americans to give a reason for their decision increased their preferences for the alternatives with extremely favorable values, whereas asking Asians to give a reason increased their preferences for the compromise alternative. Thus, chronic cultural differences in promotion and prevention focus appear to exist. However, these differences are only evident if participants are stimulated to think about the basis for their decisions. 2. Situational Influences on Promotion Versus Prevention Focus

Feelings of Membership in an Ad Hoc Group. Situational factors that induce concerns about avoidance of negative consequences can override chronic differences in this tendency. A series of studies conducted by Donnel Briley (Briley & Wyer, 2002) investigated this possibility and, in doing so, confirmed the generalizeability of the motivational orientations that Higgins postulates. Aaker and Lee (2001) found evidence that stimulating people to think of themselves as part of a group increases their focus to negative features of a situation relative to positive ones. Based on these findings, we assumed that stimulating people to feel part of a group would induce a prevention focus, the effects of which would generalize over a variety of decision situations. In two studies, participants’ feelings of group membership were induced by having them ostensibly participate in either a group or an individual achievement task. After doing so, participants in Experiment 1 reported their agreement with implications of a number of proverbs, some of which emphasized the desirability of balance and equality (e.g., “it takes two to make a quarrel,” “when the shoulder pole is not secure at both ends, the load will slip off,” “the pole is easy to carry if the load is balanced,” etc.). (For evidence that responses in this task are correlated with preferences for equality in actual decision situations, see Briley et al., 2000; Weber, Hsee, & Sokolowska, 1998.) We expected that participants with a prevention focus would prefer situations that minimized the likelihood that one individual would receive negative outcomes at others’ expense, and, therefore, would be more likely to endorse proverbs that advocated outcome equality than would participants without this orientation. This was in fact the case, as shown in the first column of Table 3.3 (top); people reported stronger

71

DETERMINANTS AND EFFECTS OF KNOWLEDGE ACCESSIBILITY TABLE 3.3 Mean Evaluations of Equality-Oriented Proverbs, and Proportion of Participants Who Chose Candies of Different Types, as a Function of Participating as Individuals or in a Group and Regulatory Focus

Endorsement of equality-oriented proverbs Individual participation Group participation M Proportion of participants choosing candy of different types Individual participation Group participation M

No Instructions (Experiments 1 and 2)

Promotion Focus (Experiment 3)

Prevention Focus (Experiment 3)

5.07 5.60 5.34

5.08 5.07 5.07

5.49 5.60 5.55

.45 .73 .61

.50 .55 .43

.59 .81 .71

Note. Based on data from Briley and Wyer (2002).

agreement with these proverbs when they had participated as part of a group than when they had taken part as individuals. In Experiment 2, participants after performing the group task were dismissed. Upon leaving the experiment, however, they were given the opportunity to take two candies as an additional reward for participating. Two types of attractive candy were available, and so participants could choose either two candies of the same type or one candy of each type. We expected that individuals with a desire to avoid negative consequences of their choice would be inclined to choose one candy of each type, thereby minimizing the risk of experiencing postdecisional regret if their momentary preference for the two types of candy happened to change. Consistent with this reasoning, participants were more likely to choose candies of different types after participating in the experiment as members of a group (.73) than after participating as individuals (.45). To confirm the assumption that the effects of group membership on behavior in these studies were mediated by its impact on prevention focus, we conduced an additional experiment in which this focus was manipulated more directly (Briley & Wyer, 2002, Experiment 3). This study was virtually identical to the first two experiments except that participants were told either that (a) the individual/group who performed best on the achievement test would receive a monetary reward (promotion-focus conditions) or, alternatively, (b) individuals/groups would receive a reward except for those who did poorer than average (prevention-focus conditions). Proverb endorsement and candy choices under these two conditions are shown in the last two columns of Table 3.3. Participants were more inclined to endorse

72

CHAPTER 3

equality-related proverbs, and also more likely to choose candies of different types, under prevention-focus conditions than under promotion-focus conditions. Note also that under promotion-focus conditions, proverb endorsement and candy choice were each very similar to that observed under individualparticipation conditions of Experiments 1 and 2, whereas under preventionfocus conditions, they were similar in magnitude to that observed under group-participation conditions of the first two experiments. Thus, these data provide strong confirmation of the assumption that inducing feelings of group membership activated a prevention focus, and therefore had the same impact as a more direct manipulation of this motivational orientation. Feelings of Cultural Identity. A second series of studies (Briley & Wyer, 2002, Experiments 5 and 6) confirmed these conclusions using a quite different means of inducing feelings of group membership, and also different indices of prevention focus. Specifically, participants were both European American and Hong Kong Chinese college students whose feelings of group membership were activated by making them aware of their cultural identity. This was done by exposing representatives of each culture to symbols of their own or a different culture as part of an ostensibly unrelated test of general knowledge. (Specifically, European Americans were exposed to pictures of Marilyn Monroe, a Dixieland jazz band, the American flag, etc. Chinese were exposed to pictures of the Great Wall, a Chinese musical instrument, people doing calligraphy, etc.). Following this task, some participants were asked to state their preferences for different combinations of outcomes to themselves and another in a hypothetical resource allocation task. Others were administered the product choice task employed by Briley et al. (2000) and described earlier. Exposing participants to symbols of their own culture not only increased their preferences for equality in the resource allocation task but also increased their likelihood of choosing the compromise alternative in the product choice task, and this was true of both European Americans and Hong Kong Chinese. Thus, making participants aware of their cultural identity activated a disposition to avoid negative outcomes that overrode the chronic cultural differences in prevention focus that Briley et al. (2000) observed.

THE ACTIVATION OF BEHAVIORAL DISPOSITIONS Behavior that results from promotion and prevention focus is likely to be mediated by conscious decisions to seek positive outcomes or avoid negative ones. In some cases, however, behavior may be mediated by “produc-

DETERMINANTS AND EFFECTS OF KNOWLEDGE ACCESSIBILITY

73

tions” of the sort postulated by Anderson (1983) and described in chapter 1, which are activated and applied spontaneously, with a minimum of conscious deliberation. These productions, like declarative knowledge representations, could depend on their accessibility in memory at the time. A series of studies by Bargh and his colleagues are particularly intriguing. In one study (Bargh, Chen, & Burrows, 1996, Experiment 1), participants completed a sentence construction task that required the use of concepts associated with either politeness or rudeness. Upon completing the task in the experimenter’s absence, they went to another room to turn in their work. Upon arriving, however, they found the experimenter engaged in a discussion with a graduate student. The time that participants waited before they interrupted the conversation was recorded. Participants interrupted more quickly if they had previously been primed with rudenessrelated concepts than if they had been primed with politeness-related concepts. On the other hand, participants’ post-experiment judgments of the experimenter’s own rudeness was not influenced by the priming manipulation. Thus, the primed concepts appeared to influence participants’ behavior independently of their interpretation of the experimenter’s actions. Effects of Stereotype Activation on Own Behavior Two additional studies by Bargh et al. (1996) demonstrated that the activation of stereotypes can induce stereotype-consistent behavior, and that this is true even when individuals are not themselves members of the stereotyped group. In a particularly provocative study (Bargh et al., 1996, Experiment 2), participants completed a similar sentence construction that in this case required the use of concepts associated with the elderly (“aged,” “bingo,” etc.). After completing the task, participants were dismissed from the experiment, and the time they took to walk to the elevator was recorded. Participants who were primed with elderly related concepts walked more slowly to the elevator than control participants. This was true even though the participants themselves were not elderly but rather were college students. Participants in a third study (Bargh et al., 1996, Experiment 3) were subliminally exposed to pictures of either African-American or EuropeanAmerican faces while they performed a tedious perceptual task on the computer. Upon completing the task, however, they were told that the computer had malfunctioned and they would have to perform the task a second time. Participants’ nonverbal behavior (facial expressions, etc.) were recorded and coded for manifestations of irritation and hostility. Participants manifested significantly more hostility when they had been exposed to African-American faces than when they had been exposed to European-Americans.

74

CHAPTER 3

Qualifications and Extensions Bargh et al.’s studies combine to suggest that unobtrusively exposing participants to stimuli can activate sequences of behavior under conditions in which they are minimally aware of either the priming stimuli or the behavior itself. Moreover, these effects are not mediated by their effects on priming on the interpretation of the situation or the person toward whom the behavior is directed. To this extent, they are consistent with the assumption that features of the priming stimuli activated a production that elicited a behavioral sequence automatically, with minimum awareness of the conditions that gave rise to it. It should be noted, however, that the priming stimuli may not by themselves be sufficient to activate the production. That is, the eliciting conditions of a production are likely to include features of the situation as well. It seems unlikely that people who are primed with rudeness-related concepts in Bargh et al.’s first experiment would have been rude if the experimenter’s own behavior had not provoked it. Nor would the exposure to African-American faces activate hostile behavior in the absence of situational conditions in which the behavior was potentially appropriate. Rather, both primed features and situational features in combination may be necessary to form the eliciting conditions for a production. In fact, the same primed features might elicit quite different behavior, depending on the situational features that accompany them. This possibility is made salient in a series of studies by Stanley Colcombe (Colcombe & Wyer, 2001). In some conditions of this study, Caucasian participants were first subliminally primed with either African American, Asian, or Caucasian faces using procedures very similar to those employed by Bargh et al. (1996). Then, they completed a shortened version of the mathematics section of the Graduate Record Examination. African Americans are stereotyped as unmotivated to perform well in intellectual activity, whereas Asians are stereotyped as highly motivated to achieve. To this extent, priming African-American faces should decrease White participants’ attempts to perform well on the mathematics test, whereas priming Asian faces should increase the effort they expend. Results shown in the first column of Table 3.4 confirm this hypothesis. That is, participants attained lower scores on the test when African-American faces had been primed, and higher scores when Asian faces had been primed, than they achieved when White faces had been primed. The fact that the effects of priming stimuli on participants’ behavior occurred without awareness was made salient by additional conditions in which the priming stimuli were presented above awareness threshold. (To justify this, participants were told that the perceptual task had been used in another, unrelated experiment and that we had been unable to modify the program to eliminate the stimuli from being shown.) As shown in the righthand column of Table 3.4, participants under these conditions performed

75

DETERMINANTS AND EFFECTS OF KNOWLEDGE ACCESSIBILITY TABLE 3.4 Mathematics Test Performance as a Function of Priming Stimulus and Prime Type

African American faces Caucasian faces (control) Asian faces

Subliminal Prime

Overt Prime

420 452 512

502 446 524

Note. Scores are converted to percentiles based on GRE norms. Adapted from Colcombe and Wyer (2001).

better when they had been exposed to either African-American or Asian faces than under control conditions. In these conditions, participants who were exposed to African-American faces apparently worked hard in order to distance themselves from this stereotypically low achieving group, whereas those who were exposed to Asian faces were stimulated to work hard in order to emulate this stereotypically high achieving group. Thus, participants performed better in both conditions, albeit for different reasons. The effects observed under subliminal priming conditions require further consideration, however. It is not completely clear why activating a stereotype would have a positive influence on the behavior of individuals to whom the stereotype does not apply. A conceptualization by Prinz (1990) suggests one possible explanation. He postulated that in order to comprehend another person’s behavior, one must often imagine oneself performing it. This mental simulation could establish an association between a representation of others’ behavior and a representation of one’s own. As a result of this association, factors that activate concepts of another’s behavior might activate a representation of one’s own behavior in similar situations, producing the effects that both Bargh et al. (1996) and Colcombe and Wyer (2001) observed. CONCLUDING REMARKS As the research in this chapter testifies, cognitive responses to situations are determined to a large extent by the subset of concepts and knowledge that happen to be easily accessible in memory at the time the responses are generated. The research we have described does not begin to cover the numerous ways in which differences in knowledge accessibility can have an impact. It nevertheless suffices to indicate the importance of taking these differences into account in conceptualizing information processing of the sort we discuss in the chapters that follow. This becomes immediately important in the next section, where we analyze the comprehension of social information of the sort we acquire through direct experience.

P A R T

II COMPREHENSION PROCESSES

C H A P T E R

4 The Comprehension and Validation of Information About Familiar People and Events: The Role of Situation Models

The comprehension of an event can occur in two stages. The first stage is spontaneous and automatic, and occurs whenever the event can be easily understood in terms of concepts and knowledge that comes easily to mind at the time. It is difficult to observe a boy pounding a nail into a piece of wood, or to see a book lying on a table, without comprehending this experience. It is equally difficult to read descriptions of these events without comprehending them. For example, try not to understand “The boy pounded a nail into a piece of wood,” or “the book is on the table.” For a native English speaker, the comprehension of such statements is spontaneous and virtually uncontrollable. In many cases, however, comprehension requires deliberative cognitive activity. Just as we comprehend some verbal statements spontaneously, we immediately recognize that others cannot be easily understood. For example, “The nail pounded a tree into the boy” is structurally identical to “The boy pounded a nail into a tree.” To understand the first statement, however, one must consciously construct a cartoon-like mental image of an animated nail driving the tree into a screaming juvenile. This construction is deliberative and requires cognitive effort. Moreover, the comprehension of some events requires an understanding of other events that precede or follow them. This information is particularly necessary when the event is inconsistent with a preexisting conception of the things that typically occur in the situation at hand. For example, a nuclear physics professor might have difficulty computing a 15% tip at a restaurant. Or, a person’s comment at a party might violate a conversational norm to be polite, informative, or truthful (Grice, 1975). These events 79

80

CHAPTER 4

are likely to stimulate conscious cognitive activity in an attempt to understand why they occurred. However, this deliberative cognitive activity is only likely to occur when the comprehension processes that are spontaneously activated by exposure to information are insufficient. Before discussing this activity and the conditions that give rise to it, one must first understand the processes that underlie the spontaneous comprehension of information. The present chapter focuses on these processes. We first describe the mental representations that are formed in the course of comprehending events of the sort we often encounter or read about in daily life. Then, we discuss the processes that underlie the construction of these representations. Finally, we consider the implications of our conceptualization for the spontaneous recognition of statements about known persons and events as either true or false. These latter implications are of considerable importance in understanding responses to information outside the laboratory, as our discussion in later chapters testifies.

THE CONTENT AND STRUCTURE OF SITUATION MODELS Most social experiences are dynamic. That is, they take place over a period of time. Such an experience can be conceptualized as a sequence of states of affairs along with events (e.g., behaviors) that transform one state into another. When an experience is described verbally, however, the entire sequence of events that occur is unlikely to be specified. For example, our observation of a boy throwing a ball necessarily includes both the action itself and features of the situation in which it takes place. It might also include other events that precede or follow the ball-throwing event per se (e.g., the ball sailing through the air and being caught by another person). In contrast, a verbal description of the event (e.g., “The boy threw the ball”) does not specify its antecedents or consequences. Nor for that matter, does it indicate the nature of the ball, physical characteristics of the thrower, or the situational context in which the event occurred. To infer these characteristics, therefore, we must relate the verbal description to a past experience we have had whose features match those of the description but contain others as well. These latter features may then be added to the mental representation we form. The comprehension process just described essentially involves a mental simulation of the event to which the information refers. We conceptualize this mental simulation as a situation model. The situation model construct is not new (Johnson-Laird, 1983, 1989). In cognitive psychology, it has been used to conceptualize the processes that underlie language comprehension

THE ROLE OF SITUATION MODELS

81

(Johnson-Laird, 1983; Kintsch, 1998; Zwaan & Radvansky, 1998), spatial memory (Taylor & Tversky, 1992) and logical reasoning (Schaeken, JohnsonLaird, & d’Ydewalle, 1996). Other research has examined their role in both comprehending observed events (e.g., a movie; Magliano, Dijkstra, & Zwaan, 1996) and imagining experiences that have not yet occurred (cf. Glenberg, 1970; Glenberg, Kruley, & Langston, 1994; Graesser, Singer, & Trabasso, 1994). Perhaps surprisingly, however, their role in social information processing is not widely recognized. The mental simulation of a situation typically requires assumptions about the relations among the entities (persons and objects) involved in it. These relations are most easily represented in a mental “image.” An image is often coded visually, and thus can be likened to a “picture.”7 However, it should not be equated with a photograph. For one thing, not all features that are found in a photograph are likely to be mentioned. (Thus, the mental image that is formed in response to the statement “the boy kicked the ball” might specify the nature of the ball and portray the boy in the act of kicking it, but might not include a specification of the boy’s hair color or what he was wearing.) On the other hand, a mental image, as we conceptualize it, may convey information in other modalities. Thus, a representation of “The man shouted, ‘Get out of my way!’ ” may contain an acoustically coded representation of the man’s tone of voice as well as the thing he said. Two types of situation models are of particular relevance in conceptualizing social experiences. Event models are simulations of specific actions and, in isolation, are similar to states of affairs. However, a number of temporally and thematically related event models might be combined to form a multiple-segment episode model that conveys the temporal and causal relatedness of the events as well as the events themselves. As the preceding discussion implies, all event and episode models are assumed to have an image component. This is true regardless of whether the information that leads the model to be constructed is conveyed nonverbally, orally, or in writing. When information is conveyed verbally, the representation of it in memory is also likely to have a propositional component that provides a linguistic description of the event or situation being represented. To this extent, an event model might be analogous to a “picture plus caption,” and an episode model to a mental “comic strip.” Note, however, that although the verbal description of an event may spontaneously lead to the construction of a mental image in the course of comprehending it, an observation or a picture of the event can often be compre7 The representation of visual images in memory has been somewhat controversial (cf. Kosslyn, 1980; Pylyshyn, 1973; Shepard & Metzler, 1971). However, we do not intend to enter into this debate. The construct of a visual image as used in this volume, like mental representations more generally, should be evaluated in terms of their utility as metaphors and not in terms of their validity as depictions of the physiology of the brain.

7

82

CHAPTER 4

hended without recoding it linguistically. In this case, the recoding may not normally be performed (see Postulate 3.1). The preceding discussion can be summarized in two postulates: Postulate 4.1. Event and episode models are mental simulations of an event, or sequence of events, that are temporally and spatially constrained. These models are constructed spontaneously in the course of comprehending events that occur at a specifiable (although not necessarily specified) time and place. Postulate 4.2. An event model can consist of both a mental image and a linguistically coded description of the event or state of affairs to which it refers. Although the image component of the model is obligatory, the linguistic component is optional.

A critical feature of our conceptualization is indicated in Postulate 4.1. That is, not all of the representations that people form of social information are situation models. Situation models are restricted to events or states of affairs that are localized in space and time. This restriction is important. Although the time and place of an event is not always specified, it must be implicit. Thus, statements such as “The book is on the table,” or “The boy kicked the ball,” refer to states and events that necessarily occurred at a particular place and point in time. However, many verbal statements that people encounter are not temporally or situationally constrained. For example, “The book is interesting,” or “The boy owns a ball,” are unrestricted as to time and place. Therefore, the comprehension of these statements would not involve the construction of a situation model. For example, they might be represented linguistically, but a mental image of their referents might not be formed. Note that although most visual images are situationally and temporally specific, this is not necessarily the case. For example, a mental picture of an acquaintance could contain nonverbally coded features of the person’s face and other physical attributes but might be independent of any particular situation in which the individual is involved. Moreover, a schematic representation of a prototypic situation could provide expectations for the type of events that are likely to occur without its features being verbally articulated. These schematic representations can influence the spontaneous initial appraisals of a new situation that one encounters in a manner similar to that postulated by Lazarus (1982, 1991), eliciting feelings that the situation is benevolent or hostile without a clear understanding of why this is so. These schematic representations, or “situation prototypes,” can also be part of the eliciting conditions of a production of the sort noted in chapter 1, which automatically stimulate behavior under conditions in which they are activated. Thus, people often behave quite differently in a classroom sit-

THE ROLE OF SITUATION MODELS

83

uation than at a party, and differently with friends than with casual acquaintances. These behavioral dispositions are activated and applied without conscious analyses of the situational factors that give rise to them. Headers Episode models are often assigned a header, or title, that constitutes a generalization of the sequences of events to which the model refers and that essentially tells what the model is about. The header of a model may be coded propositionally. Thus, “the fiasco at Lydia’s dinner party” might be the title of an episode model whose segments depict my stumbling into a table and spilling a bloody Mary on a guest’s new dress, causing her to drop a dish of guacamole on the hostess’s white shag rug. Features of a model’s header can serve as retrieval cues for the model itself, activating the model as a result of the processes described in chapter 2. In this regard, many general statements that convey attitudes and opinions could be headers of more specific episode models that exemplify them. For example, “Religious fanaticism is dangerous” could be the header of an episode model of the events surrounding the attack on the World Trade Center. Not all propositions are model headers, of course. On the other hand, a given proposition might be the header of more than one model. It is interesting to speculate that the strength of one’s belief in a given proposition is a positive function of the number of episode models for which it serves as a header. Because episode models are constructed from specific experiences, this proposition would be consistent with Fazio’s (1990) proposal that the strength of an attitude (or opinion) is typically greater when it is derived from personal experience than when it is not. In this and the following chapter, we describe the construction of event and episode models and their role in the comprehension of social information. Before doing so, however, it will be useful to discuss more generally a fundamental assumption that underlies the postulation of these models, namely, that verbal event descriptions are often spontaneously encoded in terms of mental images.

THE ROLE OF VISUAL IMAGERY IN COMPREHENDING SOCIAL EVENTS According to Postulates 4.1 and 4.2, verbal images are fundamental components of situation models and, therefore, are formed spontaneously in the course of comprehending events that are temporally and situationally constrained. This is true even when the events are described verbally. On the other hand, the linguistic coding of the events that compose a situation

84

CHAPTER 4

model is optional. Therefore, this coding is not always performed in the course of comprehending events that are conveyed visually (e.g., in terms of pictures).

The Construction of Images in Text Comprehension Some of the earliest and best-known indications that mental images are constructed in the course of comprehending text were provided by John Bransford and his colleagues (cf. Bransford, Barclay, & Franks, 1972; Bransford & Johnson, 1972). In an early study (Bransford et al., 1972), some participants were asked to learn sentences that on the surface appeared anomalous (e.g., “The haystack was important because the cloth would rip,” or “The notes were sour because the seam was split”). Not surprisingly, their memory for these sentences was typically poor. In another condition, however, the same sentences were each preceded by a single word (in our examples, “parachute” and “bagpipes,” respectively). In the context of these words, memory for the sentences improved substantially. The most plausible explanation of these results is that the cue word stimulated the construction of a mental image of an event or sequence of events in which the statements made sense (e.g., an image of a parachutist landing in a pile of hay, or a frustrated piper trying to get music out of his defective instrument). Although these results could perhaps be interpreted without recourse to the assumption that mental images were formed of the events, an explanation in terms of this assumption is the most parsimonious. A study of recognition memory by Garnham (1981) provided circumstantial evidence of a different sort. In this study, participants after being asked to learn a number of sentences were given a recognition memory test in which the test sentences differed slightly from those they had seen earlier. In some cases, the events described in the two sentences presumably occurred in the same situation; for example, participants were initially exposed to “The hostess bought a mink coat from the furrier” but were later asked to verify “The hostess bought a mink coat at the furrier’s.” In other cases, however, the events described by the two sentences occurred in different situations; for example, “The hostess received a telegram from the furrier” and “The hostess received a telegram at the furrier’s.” In each case, the learning and test sentences were identical in structure. However, the events described in the first pair of sentences presumably occurred in the same situation and, therefore, were likely to elicit similar mental images. In contrast, the events described in the second pair occurred in different situations and were likely to elicit quite different mental images. Therefore, if participants constructed these images spontaneously in the course of com-

THE ROLE OF SITUATION MODELS

85

prehending the sentences, they should be more likely to confuse the first pair of sentences than to confuse the second pair, as reflected in their recognition responses. This was in fact the case. A study by Glenberg, Meyer, and Lindem (1987) is particularly compelling. Participants read a story in which a sweatshirt either was associated with the protagonist at the outset (“John put on his sweatshirt and went jogging”) or was separated from him (e.g., “John took off his sweatshirt and went jogging”). Their recognition of specific features mentioned in the passage (including “sweatshirt”) was then assessed. The time that elapsed between the mention of the feature in the story and its occurrence in the recognition list was the same in all cases. Nevertheless, participants recognized the target object more quickly in the first condition than in the second. Participants in the first condition apparently formed a mental image of John wearing the sweatshirt that persisted throughout the mental representation they formed of the events that occurred later, and this image was salient at the time of recognition. In the second condition, however, this was not the case. Finally, a study by Black, Turner, and Bower (1979) suggested that when people form mental images on the basis of event descriptions, they construct these images from a specific visual perspective. Participants read passages describing pairs of events that were either likely to be imagined from the same perspective (e.g., “Mary was reading a book in her room. John came in to talk to her.”) or from different perspectives (“Mary was reading a book in her room. John went in to talk to her.”). Participants typically took longer to read and comprehend the second type of passage than the first. In the first case, readers presumably formed an image of Mary from someone in the room that was maintained while comprehending the second. In the second case, however, the second sentence stimulated a mental image from the perspective of someone outside the room, and so readers were required to shift their perspective in order to comprehend it in the context of the first sentence. As a consequence, reading and comprehension time increased. Differences in the Comprehension of Picture and Text Information Using quite different methodologies, all four studies described earlier converge on the conclusion that mental images are constructed spontaneously in the course of comprehending information about situation-specific events. In all cases, however, the evidence is circumstantial. More direct support for Postulate 4.2, requires a comparison of the comprehension of verbal descriptions of events with comprehension of the same events when they are

86

CHAPTER 4

conveyed in pictures. Specifically, suppose participants read a verbal description of an event. If this description spontaneously stimulates a mental image of the event, the event should later be identified just as easily on the basis of a picture as it would be if a picture of the event had been conveyed in the first place. Wyer, Adaval, and Colcombe (2002) examined this possibility. In doing so, we also evaluated the assumption that although verbal descriptions of an event are spontaneously encoded in terms of mental images, pictures of events are not spontaneously encoded linguistically. We constructed two scenarios of about 14 events each. One scenario, which concerned a day at school, described a male student getting up in the morning, getting dressed, eating breakfast, leaving his apartment, finding that his bicycle had a flat, walking to school, getting to class late, leaving, meeting a friend, going to the library, and falling asleep at the table. The second, a dating scenario, portrayed the student getting ready for a date, meeting the date, buying tickets for a movie, buying some popcorn, eating it while watching the show, leaving, going to a pub, observing his date flirt with a friend, sulking while walking home, but kissing his date goodnight at the door. Four sets of stimuli were constructed for each scenario, two composed of “pictures” and two composed of “captions.” The two “picture” scenarios were composed of photographs of the events, with one set taken from different perspectives than the other. The two “caption” scenarios were also constructed composed of verbal descriptions that were similar in meaning but different in wording (e.g., “discovered his bicycle tire was flat” vs. “saw that his bike had a flat tire”). Participants were shown the events composing one of the two scenarios on a computer screen with instructions that we were interested in how people comprehend everyday life events. This was done in three conditions. In picture-only conditions, only pictures of the events were presented. In caption-only conditions, only verbal descriptions of the event were conveyed, and in a third, picture-plus-caption condition, both pictures and verbal descriptions were shown. After receiving the event descriptions and a short delay, participants were presented stimuli portraying a number of events with instructions to decide whether the event had occurred in the sequence they had seen earlier. Some of the events were described verbally and others were described in pictures. Moreover, some stimuli referred to the same events that occurred in the original scenario but were from a different perspective (in the case of photographs) or were worded differently (in the case of verbal descriptions). Other stimuli, however, described different events entirely. Participants were told that the stimuli they would see would not be identical to those they had seen earlier, and to base their judgments on whether the stimulus event was “similar in meaning” to the one they had

87

THE ROLE OF SITUATION MODELS

seen earlier. In each case, the time that participants took to judge the stimuli was recorded. 1. Results Participants under all conditions presumably formed a multiple-segment episode model on the basis of the events they encountered regardless of the modality in which they were conveyed. According to Postulate 4.2, however, participants who received written descriptions of the events should spontaneously form mental images of them and, therefore, should have these images in memory at the time of judgment. Therefore, they should later be able to identify pictures of the events as easily as they would if pictures of the events had been presented in the first place. Data shown in the top half of Table 4.1 support this hypothesis. That is, when participants were asked to verify an event on the basis of a picture, they responded just as quickly if the event had been described verbally (M = 1.90 s) as they did if the event had been portrayed in a similar but not identical picture (M = 1.97 s). However, Postulate 4.2 also implies that people do not spontaneously code pictures verbally when these codings are not necessary to comprehend the events they portray. If this is the case, however, participants who are later asked to verify an event on the basis of a verbal description of it must recode the picture linguistically at the time of judgment in order to perform the verification. This recoding presumably takes time. Consequently, participants should take longer to verify the event than they would if a verbal description had initially been presented. This was also the case, as Table 4.1 indicates. That is, when participants were told to identify an event on the basis of a verbal description of it, they took longer to do so when the event had initially been conveyed in a picture (M = 2.53 s) than when it had been described in words (M = 1.71 s). TABLE 4.1 Mean Time to Verify Pictures and Captions as a Function of Presentation Order and the Type of Information Initially Presented

Narrative order Response time to Response time to Scrambled order Response time to Response time to

Picture

Verbal Description

Picture Plus Verbal Description

pictures verbal descriptions

1.97 (19) 2.53

1.90 (14) 1.71

2.05 (20) 1.96

pictures verbal descriptions

2.16 (19) 2.26

2.50 (18) 2.23

1.91 (16) 2.09

Note. Adapted from Wyer et al. (2002).

88

CHAPTER 4

2. A Qualification The results just described are quite consistent with Postulate 4.2. There are nevertheless some qualifications on the validity of this postulate. In other conditions of Wyer et al.’s study, the events were presented in scrambled order rather than in the order they occurred. In this condition, participants could not spontaneously form a multiple-segment episode model of the sequence as a whole. Rather, comprehension of this sequence required deliberative cognitive activity. Results shown in the bottom half of Table 4.1 indicate that in this condition, participants could verify an event on the basis of a verbal description of it just as quickly when the event had originally been conveyed in a picture (M = 2.26 s) as when the event had been described verbally in the first place (M = 2.23 s). However, they took longer to verify an event on the basis of a picture when the event had been described verbally (M = 2.50 s) than when it had been described pictorially (M = 2.16). Although this pattern of response times was unexpected, it has a plausible explanation. Verbal symbols are easier to manipulate mentally than visual images are. Consequently, participants who received pictures of the events in scrambled order may have subjectively translated them into linguistic terms in order to construct a mental representation of the sequence of events as a whole. As a result, they could later verify a verbal description of the events just as easily as participants who had been given verbal descriptions of the events in the first place. When participants received verbal descriptions of the events in scrambled order, however, they relied on linguistic codings alone to construct a representation of the sequence without including visual images in this representation. Therefore, when they were later asked to verify a picture, they were required to recode it linguistically in order to compare it with the linguistic representation they had constructed earlier, and this took time. As a consequence, they took longer to make this comparison than participants who had received a picture at the outset.

SPONTANEOUS COMPREHENSION PROCESSES To reiterate, much of the knowledge we acquire in daily life consists of event and episode models of specific events that we observe or read about. This knowledge may often pertain to the activities of actual people and situations that we learn about over a period of time. To this extent, new information about the person and situations is likely to be comprehended in terms of this referent-specific knowledge rather than in terms of more abstract concepts of persons and events in general. Thus, new information about Bill Clinton’s affair with Monica Lewinsky is likely to be comprehended spontaneously in terms of prior knowledge about these specific protagonists rather than more general concepts pertaining to U.S. Presi-

THE ROLE OF SITUATION MODELS

89

dents and staff associates, or to men and women more generally. As we noted earlier, the conceptualization proposed by Wyer and Srull (1989) could not account for this possibility. Earlier in this chapter, we distinguished three different types of representations that might be constructed spontaneously in the course of comprehending new information. The representations differ in both the type of input information that stimulates their construction and the terms in which the representation is encoded. Specifically: 1. Episode and event models may be constructed from either verbal descriptions of situationally and temporally specific events or direct observations of these events. These representations consist of mental images and, in some cases, linguistically coded descriptions of the events being depicted. 2. Nonspecific representations of persons and events can be constructed from verbal descriptions that are not temporally or situationally specific (e.g., “the boy owns a baseball,” “John went to college at Yale,” “Peter is dishonest”). These representations are coded linguistically but do not contain mental images. 3. Situational appraisals are constructed of the context in which more specific information is conveyed. These appraisals, and the representations formed of them, are nonverbal. All three types of representations obviously come into play in social information processing. They are not only formed from new information but, once constructed, can be used as bases for comprehending experiences and information that are acquired subsequently. In this chapter, we focus primarily on the construction of event models in the course of comprehending events involving familiar persons of the kind we are likely to observe or read about, and will defer our discussion of the construction of episode models to the chapters that follow. Much of our discussion, however, is applicable to the construction of nonspecific person and event representations as well. Functions of the Comprehender The comprehension processes we propose are theoretically localized in the Comprehender (see Fig. 2.1). Our conceptualization of this unit differs from that of the corresponding unit in the Wyer and Srull (1989) model in terms of its accessibility to knowledge in Permanent Storage and, therefore, its ability to comprehend information about specific persons and events with which one is already familiar. The Comprehender is assumed to have two components: a Parser and a Simulator. These components and when they come into play are shown schematically in Fig. 4.1.

90

CHAPTER 4

FIG. 4.1. Theoretical sequence of processing steps involving the Parser and Simulator in the course of comprehending visual and verbal information.

1. The Parser The Parser only comes into play when the input information is coded linguistically. In this case, the Parser transforms verbal information into the equivalent of a Subject–Predicate proposition. The output of this Parser then resonates with previously constructed mental representations in Permanent Storage. In this regard, a major difference between the conceptual-

THE ROLE OF SITUATION MODELS

91

ization of the Simulator and those of the Comprehender postulated by Wyer and Srull lies in the range of material in Permanent Storage to which it has access. In the original theory, the Comprehender had access to only semantic concepts that were stored in a semantic bin. In contrast, the Simulator has access to all previously formed knowledge representations that exist. Thus, information about a specific person or event about which one already has knowledge is comprehended with reference to aspects of this referentspecific knowledge without requiring any interpolated cognitive activity. This process is elaborated presently. If the input information is verbal and a previously constructed representation exemplifies the subject–predicate proposition that is formed of it, the Parser retrieves this representation from Permanent Storage and uses it to form a linguistically coded representation of the new information as well. If the predicate describes an event or stimulus that is situationally and temporally constrained, however, it is also likely to resonate with a preexisting nonverbally coded representation (e.g., an event or episode model). In this case, the Simulator is activated. 2. The Simulator The basic operations performed by the Simulator are theoretically similar regardless of whether the original stimulus information is conveyed visually (as in a picture or movie) or verbally. It has the ability to (a) identify referential knowledge in Permanent Storage, (b) evaluate the compatibility of new information with previously formed event and episode models, (c) substitute features of the new information for those of preexisting models to form a new model, and (d) combine individual segments extracted from previously formed models to form a new episode model that depicts a unique sequence of events. The output of the Simulator is largely nonverbal (visual or acoustic). If both a verbally coded representation has been formed by the Parser and a nonverbally coded representation has been formed by the Simulator, these representations are then integrated into a single representation, as indicated in Fig. 4.1. The general processes outlined in Fig. 4.1 are potentially applicable regardless of the type of input information, and theoretically underlie the formation of all three types of representations described earlier. However, the most complex comprehension processes occur when verbal information is presented about situationally and temporally specific events. In the pages that follow, therefore, we first discuss these processes and present some empirical evidence bearing on them. In doing so, we note the implications of the conceptualization for not only the comprehension of information but also the spontaneous recognition of its validity or invalidity. Finally, we

92

CHAPTER 4

turn briefly to the processes that underlie the comprehension of events that are directly observed. These latter considerations raise some additional issues that become important in later chapters of this volume.

THE COMPREHENSION OF VERBAL STATEMENTS Theoretical Considerations To reiterate, a verbal statement that enters the Comprehender is theoretically transformed by the Parser into the equivalent of a proposition consisting of a subject and a predicate. The subject can refer to either a specific person (“Michael Jordan”), an object (“Gone with the Wind”), or a general category (“basketball players,” “book,” etc.). The predicate can convey a specific or general act (“sank a three-pointer,” “played basketball”), or can specify a relation between the statement’s subject and an attribute (e.g., “is an athlete,” “has brown hair,” “is on the table”). Although our examples typically refer to predicates that describe actions, they apply equally well to predicates of other types. Once the Parser identifies the subject and predicate, it uses the linguistic codings of them as retrieval cues to access concepts in Permanent Storage that they exemplify and forms a linguistic representation of the statement in terms of these concepts. In addition, it transmits a specification of the subject and predicate to the Simulator. The functions of the Simulator are captured in four postulates: Postulate 4.3. The subject and predicate identified by the Parser are independently used as probe cues of memory, resonating with all of the mental representations that contain them (see Postulate 2.2). In each case, the representation whose resonance level reaches threshold most quickly is most likely to be identified.

One implication of this postulate should be noted at the outset. That is, the type of representations activated by the subject and predicate are likely to differ. In particular, the subject’s features are likely to activate a generalized person representations of the sort described in chapter 1 (Fig. 1.2). That is, this representation may consist of a set of general features that are independently associated with the individual to whom the subject refers. Although some of these features can be specific behaviors the individual has performed, they are also likely to include more general descriptions of the person’s personality and behavior. On the other hand, the features of the predicate, which pertain to behaviors or actions, are likely to activate previously formed event and episode models of these actions. The actors depicted in these representations could

THE ROLE OF SITUATION MODELS

93

sometimes be the same person that is specified in the subject of the statement to be comprehended. However, they could include many other individuals as well. In each case, the speed with which a previously formed knowledge representation reaches threshold depends on factors specified in the retrieval postulates described in chapter 2. That is, it increases with the number of features of the representation that are also contained in the probe set (Postulate 2.2) and also the degree to which the representation is already “vibrating” as a result of prior cognitive activity (Postulate 2.4). Note that if a preexisting knowledge representation exists that contains both the probe cues pertaining to the subject and the probe cues pertaining to the predicate, it is likely to be activated much more quickly than other representations that contain only one set of cues. Moreover: Postulate 4.4. If a memory search activates a previously formed knowledge representation whose features include both the subject probe cues and the predicate probe cues, the new information is spontaneously recognized as redundant with prior knowledge and a new representation is not formed.

When the two sets of probe cues activate different knowledge representations, it means that a previously formed knowledge representation of the information does not exist. Then, the process we assume is somewhat more complicated. In this case, the probe cues associated with the subject are likely to activate a number of knowledge representations that contain them, some of whose features are more likely to overlap than others. In addition, the predicate probe cues activate all of the models that explicate this action or state of affairs. (If the stimulus is “Michael Jordan sank a threepointer,” for example, it would activate all preexisting situation models whose predicate was “sank a three-pointer” or the equivalent.) The subjects of these situation models could vary, but they are likely to have a subset of features in common. In the composite set of subject features that are activated, the features that are represented most often are weighted most heavily. Then, the following postulate applies: Postulate 4.5. If no previously formed model (i.e., knowledge unit) refers to both the subject and predicate of a target statement, the features that are activated by the target’s subject are compared to a weighted composite of the features associated with the target’s predicate. This comparison yields a global estimate of similarity, S. A new situation model is formed if either (a) S is greater than a minimum threshold value, TC, or (b) the target’s subject contains a subset of features that are common to the subject of all models activated by the probe. In the latter case, the new model is constructed by combining the composite of features activated by the subject and the composite of features activated by the predicate. In all cases, however, only the features that are weighted heavily in the composite are included in the new model that is constructed.

94

CHAPTER 4

This postulate obviously requires unpacking. Suppose the target statement is “Michael Jordan sank a three-pointer.” According to Postulate 4.3, features of the Subject resonate with all previously stored representations containing these cues, whereas “sank a three-pointer” resonates with all previously formed event models of this activity. However, several of these models are likely to have Michael Jordan as the subject. Then, according to Postulate 4.4, these representations should be activated by resonance from both the target’s subject and the target’s predicate, and the statement should be regarded as redundant with prior knowledge. However, suppose the target statement is “Michael Jordan rode a motorcycle.” In this case, the predicate may not activate models whose subject is Michael Jordan. However, features that are frequently associated with motorcycle riders are “big,” “muscular,” and “athletic.” Because these features are likely to be conveyed in the generalized person representations that are activated by “Michael Jordan,” the similarity, S, should be above threshold TC . Therefore, a new model is likely to be constructed. In contrast, suppose the target statement were “Woody Allen rode a motorcycle.” In this case, the person representations activated by “Woody Allen” are less likely to include “big,” “muscular” and “athletic.” Consequently, the similarity index computed on the basis of a comparison features and the composite set activated by “rides a motorcycle” may be below threshold. Nevertheless, the features of “Woody Allen” are likely to include those that are common to all previously formed models of motorcycle riders (e.g., “human”). Thus, a new model of this statement might still be formed, but it would not be constructed as quickly as it would if statement had referred to Mr. Jordan. Finally, suppose the target statement was “A hippopotamus rode a motorcycle.” In this case, the similarity index would not only be below threshold, but the features of the subject would not be shared by all previously formed models of “rode a motorcycle.” In this case, therefore, a situation model of the statement would not be constructed spontaneously. The implications of this analysis for comprehension become clear in the context of the next postulate: Postulate 4.6. A statement is comprehended spontaneously if it is redundant with prior knowledge or if a new model is constructed of it. If neither is the case, comprehension of the statement is not spontaneous, but requires (Executor-controlled) goal-directed processing.

Thus, to summarize our example, “Michael Jordan sank a three-pointer” and “Michael Jordan rode a motorcycle” would both be comprehended very quickly. “Woody Allen rode a motorcycle” would also be comprehended spontaneously, but would take more time to comprehend than the

THE ROLE OF SITUATION MODELS

95

first two statements. In contrast, “A hippopotamus rode a motorcycle” would not be comprehended spontaneously, but would require conscious Executor-controlled processing. An interesting implication of the preceding analysis derives from Postulate 4.5. When a new situation model is formed, only those features that occur frequently in the representations that are activated by the subject and predicate are depicted in this model. This means that the greater the amount of information one has acquired about a referent in the past, the less detailed the new information is likely to be. Thus, the representation of Michael Jordan riding a motorcycle is likely to be less detailed if one has encountered numerous instances of people riding motorcycles, or has had exposure to Mr. Jordan in many different situations, than if one’s previous experience is limited in these respects. More generally, one’s initial encounter with an exemplar is likely to be represented more vividly than later encounters that have been comprehended in terms of more general sets of features. A related implication surrounds the comprehension of a statement about members of a general category (e.g., “basketball players”). According to Postulate 4.5, comprehending such a statement should depend on the diversity of exemplars who are known to have engaged in the action described. For example, consider the statement “A basketball player wore Nikes.” In this case, a generalized person representation of “basketball players” may be activated that refers to Michael Jordan as an exemplar (see chap. 1). Moreover, previously formed situation models of “wears Nikes” are likely to include Michael Jordan as the subject as a result of television commercials that portray him as such. Consequently, the resonance elicited by “basketball player” and “wears Nikes” could lead a previously formed model of Michael Jordan in Nike sports shoes to be activated very quickly, and so the statement might be comprehended almost as quickly as a statement that refers to Michael Jordan directly. In contrast, suppose the category does not have a known exemplar that has engaged in the activity in question. Then, different predictions would be made. For example, “The tennis player wore Nikes” should be comprehended less quickly than “The basketball player wore Nikes.” This should be true even though Nikes are associated just as strongly with playing tennis as with playing basketball. The reason is that wearing Nikes and playing basketball are both strongly associated with “Michael Jordan,” whereas wearing Nikes is less likely to be associated with a previously formed model of a specific tennis player. Empirical Evidence Wyer and Radvansky (1999) reported evidence in support of this conceptualization. In one study, participants read a series of statements on a computer screen with instructions to press one of two designated keys on the

96

CHAPTER 4

keyboard to indicate whether the statement was comprehensible or not. Some of the statements were anomalous (e.g., “A show played poker”). Others were meaningful and of four different types: a. true exemplar–referent statements (e.g., “Jane Fonda does aerobics”)8 b. false exemplar–referent statements (e.g., “Jane Fonda plays professional hockey”); c. exemplar–referent statements of unknown validity (e.g., “Jane Fonda rides a motorcycle”) d. category–referent statements that are likely to be true but for which there is no known exemplar (e.g., “A steel worker does aerobics”). All four of these statements were judged as comprehensible (mean likelihood = .92). However, they differed significantly in the time required to make this assessment. Of particular interest is a comparison of the time to comprehend true exemplar–referent statements and the time to comprehend category–referent statements. According to the earlier Wyer and Srull (1989) theory, the subject and predicate of the exemplar–referent statements would have to be instantiated in terms of more general semantic concepts in order to be comprehended. According to this theory, therefore, it should take longer to comprehend these statements than to comprehend statements that were already conveyed in terms of these concepts. In contrast, the present theory implies that true exemplar–referent statements are interpreted in terms of preexisting situation models, and therefore should be comprehended very quickly. This was in fact the case. The first column of Table 4.2 (Experiment 1) shows the mean time to comprehend the four types of statements described earlier. True exemplar–referent statements were comprehended more quickly (M = 1.35 s) than either category–referent statements (M = 1.68 s), false exemplar–referent statements (M = 1.64 s), or exemplar–referent statements of unknown validity (M = 1.55 s). Other differences in comprehension time are also consistent with expectations. For example, exemplar–referent statements for which a previously formed model exists should take less time to comprehend than statements for which a new model must be constructed. Moreover, false exemplar–referent statements (e.g., “Jane Fonda plays professional basketball”) should take longer to comprehend than statements of uncertain validity. This is because the similarity index that is computed on the basis of a global comparison of features activated by their subject (“Jane Fonda”) with features that are common to the models activated by the predicate (“plays professional 8 At the time of this study, Jane Fonda was a well-known promoter of aerobics on television and magazines.

8

97

THE ROLE OF SITUATION MODELS TABLE 4.2 A Comparison of the Time Required to Comprehend Statements and the Time to Judge Their Validity Comprehension Time

Validation Time

Difference

1.35 1.64

1.55 1.86

0.20 0.22

1.55

2.07

0.52

1.68

1.88

0.20

1.84

1.99

0.15

2.17

2.23

0.06

Experiment 1 Exemplar-referent statements Definitely true (e.g., “Mozart wrote a Symphony”) Definitely false (e.g., “Mozart played rock and roll”) Unknown validity (e.g., “Jane Fonda “read her horoscope”) Category-referent statements with no known exemplar (e.g., “a Musician did aerobics”) Experiment 2 True exemplar-referent statements (e.g., “Jane Fonda did aerobics”) Category-referent statements with known exemplar (e.g., “An actress does aerobics”)

Note. Response times are in seconds. Adapted from Wyer and Radvansky (1999).

basketball”) is below comprehension threshold, leading to an additional stage of processing (Postulate 4.5). Although the difference in time to comprehend the two types of statements (1.64 s vs. 1.55 s) was not statistically significant, it directionally confirmed this expectation. In a second study, participants were asked to comprehend either true exemplar–referent statements (“Jane Fonda does aerobics”) or statements that referred to the category to which the exemplar belonged (e.g., “an actress does aerobics”). According to our assumptions, the second type of statement should activate the same preexisting event model as the first type and, therefore, should be comprehended nearly as quickly. This was in fact the case, as also shown in Table 4.2 (Experiment 2). Although comprehension generally took longer in this study than in the first one, the times to comprehend true exemplar–referent statements and category–referent statements with a known exemplar did not appreciably differ (1.84 s vs. 2.00 s, respectively). Priming Effects Wyer and Radvansky’s (1999) second experiment evaluated a more subtle implication of the theory we propose. These implications are suggested in part by Postulate 2.4, that the resonance of a previously accessed knowledge representation decreases gradually over time. This means that if the

98

CHAPTER 4

residual resonance has not dissipated at the time resonance is emitted from other sources, the representation may be activated more quickly than it otherwise would. This possibility has unique implications for the effects of exposure to an exemplar-referent statement on the comprehension of a category-referent statement that refers to the same or a different event. For example, suppose participants are exposed to an exemplar-referent statement (e.g., “Jane Fonda did aerobics”). This statement may increase the accessibility of other event models that refer to the same subject (e.g., “Jane Fonda played in a movie”). This, in turn, should increase the speed of recognizing a new statement that refers to this activity, making it easier to comprehend. Moreover, this should be true even if the new statement refers to someone else. Thus, exposure to “Jane Fonda did aerobics” should increase the speed of comprehending not only “Jane Fonda played in a movie” but statements about other persons in the same activity (e.g., “Dustin Hoffman played in a movie”). In contrast, the subject of a categoryreferent statement (e.g., “An actress did aerobics”) should activate a generalized person representation (cf. Fig. 1.2), but is less likely to activate event models. Consequently, exposure to this type of statement should not influence the comprehension of statements that refer to different behaviors. The second experiment reported by Wyer and Radvansky (1999) provided some support for this prediction. Participants were first asked to comprehend either an exemplar-referent statement about a person’s behavior (e.g., “Jane Fonda did aerobics”) or a statement that referred to a category to which the exemplar belonged (e.g., “An actress did aerobics”). Then, they were asked to comprehend a target statement that referred to either the same behavior or a different one with which the exemplar was associated (e.g., “played in a movie”). Exposure to a category-referent statement decreased the time required to comprehend a target statement only if the two statements referred to the same behavior. In contrast, exposure to an exemplar-referent statement decreased the time to comprehend a target statement regardless of whether it referred to the same behavior or to a different behavior with which the exemplar was also associated.

SPONTANEOUS VALIDITY JUDGMENTS The most unique implications of Wyer and Radvansky’s (1999) formulation have yet to be considered. As noted earlier, people who receive information about a known person or event may not only comprehend the event but may judge it to be true or false. Moreover, this determination is often made automatically in the course of comprehending the statement, in the absence of a more specific goal that requires its assessment. Thus, the

THE ROLE OF SITUATION MODELS

99

statement “Abraham Lincoln wore a beard” is comprehended and immediately recognized a true. By the same token, “Abraham Lincoln shot John F. Kennedy” is immediately recognized as false. These assessments occur spontaneously, in the absence of any goal that requires them. These phenomena and the conditions in which they occur are taken into account in the formulation we propose. According to the theory, the recognition that a statement is true or false is often a necessary by-product of the comprehension processes embodied in Postulates 4.4 and 4.5. This is most obvious when people’s comprehension of a statement activates a preexisting situation model of the event to which it refers. In this case, awareness of the prior existence of the model is typically sufficient to affirm its validity. There may be exceptions to this. Situation models can obviously be constructed of fictitious events (e.g., “Santa Claus came down the chimney on Christmas Eve”). However, the representation of these events is likely to contain a “tag” denoting their fictional character. In the absence of this tag, statements that instantiate an existing situation model may be spontaneously recognized as true in the course of comprehending them. The following postulate formalizes this assumption. Postulate 4.7. A statement is spontaneously recognized as true in the course of comprehending it if (a) a situation model of the event or state described already exists in memory and (b) no tag has been attached to the model that designates it as false.

Spontaneous judgments of a statement’s validity can also occur when a previously formed model of the situation does not exist. As Postulate 4.5 implies, comprehension may often be a two-stage process. That is, persons who encounter a statement first compute a global estimate of the similarity (S) between the statement and the referents of previously formed situation models of the predicate, and comprehend the statement if this similarity is above threshold. In the course of this determination, however, they theoretically become aware of the value of S in relation to two other threshold values that are necessary to confirm or disconfirm its validity. If S exceeds or falls short of these values, the statement may be identified spontaneously as true or false. To formalize: Postulate 4.8. A statement about an event is recognized as true in the course of comprehending it if the similarity of its subject to those of previously formed models of the event (S) exceeds an upper threshold, TT . The statement is identified as false if S is below a lower threshold, TF . If S falls between TT and TF , the statement is comprehended but its validity is not spontaneously evaluated.

100

CHAPTER 4

Postulate 4.9. Spontaneous inferences that a statement is true or false are transferred to the Work Space along with a representation of the statement that has been formed in the course of comprehending it.

Thus, as an example, “An Israeli donated money to a synagogue” would be quickly judged as plausible because the categorical representation of “Israeli” is likely to share many features with the subject of these models (S > TT). In contrast, “Hitler donated money to a synagogue” is likely to be immediately recognized as false, because few features of Hitler are likely to characterize the subject of previously formed models of donating money to a synagogue (i.e., S < TF). Finally, “Madonna donated money to a synagogue” would not be immediately evaluated as either true or false, because Madonna’s known features are neither highly similar nor highly dissimilar to those of the relevant models subjects (TF < S < TT). That is, the latter statement would be comprehended, but its validity would only be assessed if more specific goal-directed processing required it. Empirical Evidence Wyer and Radvansky (1999) obtained support for the implications of these postulates in the two experiments described earlier. Rather than being asked to comprehend the statements they were presented in these conditions, some participants in these studies were asked to indicate whether the statements were true or false. If statements were verified spontaneously in the course of comprehending them, there would be little difference between the time required to make these verifications and the time required to comprehend them. If verification requires additional processing subsequent to comprehension, the time to perform the verification should be greater than comprehension time. According to Postulate 4.8, however, this latter condition arises only when the subject of a statement is neither similar nor dissimilar to the subject of preexisting models associated the predicate (e.g., “Madonna donated money to a synagogue”). Data bearing on these possibilities are summarized in Table 4.2. The first column of data refers to the time required to comprehend different types of target statements, as indicated in our earlier discussion. The second column summarizes the time to judge whether the same statements were true or false. Although the statements in Experiment 1 were verified more slowly than they were comprehended, this difference was appreciable only for exemplar–referent statements of unclear validity. The difference in all other cases was small and did not differ over the three types of statement, suggesting a general response bias that was independent of the cognitive processes underlying judgment. In Experiment 2, the time to verify statements

THE ROLE OF SITUATION MODELS

101

and the time to comprehend them were even more similar. Thus, these results are quite consistent with the formulation we propose. Further Implications The results we have brought to bear on the comprehension and validation of propositions were obtained in a rather artificial research situation. However, results of research in other areas are consistent with this conceptualization. Of particular interest in this regard are the implications of Postulate 4.7. To reiterate, this postulate states that if a pre-existing situation model of a statement exists in memory, and if no tag has been attached to the model that denotes it a false, the statement will become spontaneously as true. Thus, suppose participants construct a new situation model of a statement in one context in the course of comprehending it but do not spontaneously assess its validity. Later, however, they encounter the same statement in a different context. To comprehend the statement, they presumably retrieve the situation model they constructed of it earlier (Postulate 4.4). However, because the statement is comprehended in terms of this model, it should be spontaneously identified as true (Postulate 4.7). Put more simply, the exposure to and comprehension of a statement at one point in time may increase beliefs in the statement’s validity at a later time. The study by Hasher et al. (1977) described in chapter 3 provides direct support for this prediction. That is, people who had reported their beliefs in statements about unfamiliar events and situations in an initial experimental session increased their beliefs in the statements when they encountered them in a second experimental session 1 week later. Presumably, participants’ comprehension of the statements in the initial session led them to construct situation models of the events described by the statements. Consequently, they retrieved and used these models to comprehend the same statements when they encountered them later. However, the pre-existence of these models in memory should increase inferences of the statements’ validity, as Hasher et al.’s findings indicate. These phenomena may pertain not only to the verification of a single event, but also to sequences of events of which multiple-segment episode models are constructed. In a study by Green and Brock (2000), participants read a story about a violent crime. They were clearly aware that the story was factual. Nevertheless, they later estimated the incidence of story-related events in the real world to be higher than participants who had not read the story. In a similar vein, Shrum and his colleagues (O’Guinn & Shrum, 1997; Shrum, Wyer, & O’Guinn, 1998) found that heavy television viewers were more likely than light viewers to overestimate the incidence of crimerelated events in the real world. These quite different sets of studies con-

102

CHAPTER 4

verge on the possibility that people, either in the laboratory or when watching television, construct episode models of the events they encounter and store these models in memory. Later, they are likely to retrieve the models when contemplating events that are redundant with their implications, and therefore estimate the events to be more likely to occur. Similar conclusions can be drawn from research on the effects of generating an explanation for hypothetical events on later beliefs that the event will occur (Ross, Lepper, Strack, and Steinmetz, 1977; Sherman, Skov, Hervitz, & Stock, 1981). We discuss these studies in more detail in chapter 9. An Alternative Conceptualization The conceptualization proposed in this chapter should be considered in the context of a formulation proposed by Gilbert (1991) on the role of comprehension on beliefs. Gilbert hypothesized that belief formation is a twostep process, which first involves comprehension and only later involves verification. Moreover, in order to comprehend a statement, one must necessarily entertain the possibility that it is true. Thus, for example, comprehension of the statement “George W. Bush donated money to the Taliban” entails a tentative acceptance of this state of affairs, and only later is the statement’s validity rejected. One implication of this conceptualization is that people who have been asked to learn true and false statements under conditions of high information load are more likely to recall false statements as true than to recall true statements as false. This is presumably because the heavy processing demands prevent the second (verification) stage of processing from occurring. Thus, the statements are likely to be judged as true, based on results of the first stage of processing alone. As a consequence, objectively true statements are evaluated correctly but objectively false statements are not. The processes that Gilbert postulates to occur at the comprehension stage are compatible with those assumed by the conceptualization proposed here. That is, the comprehension of a statement about an event involves the construction of a mental simulation of it (i.e., a situation model). Moreover, the storage of this representation in memory increases its likelihood that a statement asserting the event it describes will later be judged as true (Postulate 4.8). On the other hand, this postulate distinguishes between the comprehension of a statement and the criteria that are used to estimate its validity. Moreover, it specifies the conditions in which the latter assessment is made spontaneously rather than at a later, postcomprehension stage of processing. Our conceptualization and Gilbert’s are therefore compatible with one qualification. That is, according to the proposed model, statements that differ substantially from the subset of previously formed models of the events they describe are likely to be below falsifica-

THE ROLE OF SITUATION MODELS

103

tion threshold (TF). These statements should be judged as invalid spontaneously at the time the statements are comprehended and should later be reported as false regardless of information processing load. Thus, for example, the proposed theory predicts that the statement “Abraham Lincoln landed on the moon” would be spontaneously judged as false regardless of processing load. Gilbert’s formulation does not seem to make this prediction. With this exception, however, the two formulations are consistent.

The Comprehension of Observed Events and Situations The comprehension of observed events is in some respects simpler and in other respects more complex than the comprehension of verbal descriptions of the events. For one thing, observations typically consist of not just one but a continuous stream of actions. For another, observations typically include features of not only the event itself but also the situational context in which it occurs. In the first regard, it seems unlikely that people store a continuous stream of experiences in memory. Rather, as Newtson (1973, 1976) argued, we are likely to extract visual frames from an ongoing stream of events that identify the actions that occur. The specific nature of these frames has been somewhat controversial. Newtson (1973, 1976), for example, assumed that the frames extracted are located at breakpoints between the end of one event and the onset of another. As Ebbesen (1980; see also Wyer & Srull, 1989) pointed out, however, this assumption seems rather tenuous, and the empirical support for it is equivocal. Following Ebbesen (1980), we assume that the representation formed in the course of comprehending an ongoing sequence of behavioral events consists of static frames, each of which corresponds to a meaningful act, and therefore, constitutes a mental “picture” of the event that occurred. Thus, suppose a person enters his house, takes off his coat, goes into the kitchen, opens the refrigerator and gets a can of beer. Observers do not need to store a mental “movie” of these actions in order to comprehend and remember what went on. Rather, they need only to extract a frame that is typical of each event and that permits its essential features to be identified. Each of these frames could normally be represented in a mental image of the sort we assume to compose an event model, and the sequence of frames as a whole could constitute an episode model. The operations of the Simulator in this process must nevertheless be clarified. We assume that as an individual observes a sequence of events, features of these events resonate with previously constructed event models in memory. Although the features of the observation are constantly changing, the features that occur within the time interval associated with a

104

CHAPTER 4

meaningful event are all likely to be sufficiently similar to those of a previously formed event model to activate it. Thus, for example, the features extracted from the observation of someone taking off his coat may all resonate with a static event model of this behavior, leading the model to be accessed and used in much the same manner it is used to comprehend a verbal description of the event. However, the features of an observed event sequence also include aspects of the situational context in which the sequence occurred. These features can also resonate with situational features of preexisting event and episode models in Permanent Storage. The subset of situational features that are most typical of these models could constitute a situational appraisal of the sort described earlier in this chapter. However, because situational features are common to all of the specific events that are observed, these features, along with others that occur frequently in the models that have been activated, may be transmitted to the Work Space as a separate representation that is independent of the more specific event models that are constructed. This situational representation could operate in several ways. First, if the events that are specified in the activated models are relatively homogenous, the situational appraisal defined by these features could give rise to expectations for the sorts of events that are likely to occur in the situation at hand. Second, the appraisal, once it is transmitted to the Work Space, might be included among the eliciting conditions of a production of the sort described in chapter 1, and thus could stimulate the observer’s own behavior in the situation. This possibility, which could help to account for the different effects of concept activation on behaviors described in chapter 3 (cf. Bargh et al., 1996; Colcombe & Wyer, 2001), is considered further in chapter 7, where we discuss the situational norms that govern expectations for the communications that are exchanged in social interaction and the impact of deviations from these expectations on the interpretation of these communications.

CONCLUDING REMARKS To summarize, the proposed formulation specifies the conditions in which a statement about a situation-specific event or state of affairs is spontaneously identified as true or false in the course of comprehending it. Our analysis has largely been restricted to statements about events and states that are temporally and situationally constrained (i.e., those that are comprehended by constructing a situation model). However, similar processes should guide the comprehension and verification of other types of statements as well. Be that as it may, the implication that many statements are verified spontaneously in the course of comprehending them has further

THE ROLE OF SITUATION MODELS

105

implications. As implied by Postulate 4.9, statements that are spontaneously verified as true or false are transmitted to the Work Space along with a designation of their validity, and this spontaneous validity assessment may be subject to further processing. For example, communications that are obviously true or obviously false may violate normative principles of conversation, and therefore may stimulate a conscious effort to understand why the communication was uttered. This additional processing may sometimes lead to a reconstrual of the statement’s meaning that differs from the meaning implied by the situation model that was constructed in the course of comprehending it. We elaborate this possibility and its implications in chapter 6. However, the discussion of comprehension processes in this chapter has focused on the representations formed of statements about single events or, alternatively, on event models. Equally important considerations surround the way in which event models are combined to form a multiplesegment episode model of a thematically related sequence of events. The next chapter is devoted to these matters.

C H A P T E R

5 The Construction of Episode Models and Generalized Narrative Representations

Chapter 4 focused largely on the comprehension of statements about single events of the sort that are represented in memory as event models. Most of the information we acquire in daily life, however, consists of sequences of events that theoretically give rise to the construction of multiple-segment episode models. To understand the conditions in which these models are constructed, several questions must be answered. These questions concern the number of episode models that are formed from a given sequence of events, the role of thematic relatedness in the construction of these models, and the way in which the temporal order of events is reconstructed when the events are not contained in the same model. We attempt to provide tentative answers to these questions in this chapter. Not all mental representations of event sequences are episode models, of course. In the last section of this chapter, we discuss briefly the construction of generalized narrative representations. Then, in chapter 6, we consider the conditions in which these generalized representations are used to comprehend specific experiences, and also the extent to which abstract representations of an experience, once constructed, can influence memory for this experience. THE CONSTRUCTION OF EPISODE MODELS How Many Episode Models Are Formed? People obviously do not construct a single representation of their experiences over the course of a lifetime, or even over the course of a day. Nor do they construct a single representation of the events they read about in a 106

EPISODE MODELS AND GENERALIZED NARRATIVE REPRESENTATIONS

107

novel or see in a movie. Rather, they are likely to construct several different episode models, each pertaining to a given sequence of events that are thematically as well as temporally related. Thus, the statements “John packed his bags. John left for the airport.” may be represented in a two-segment episode model because they refer to the same person and are both related to going on a trip. In contrast, the sentences “John packed his bags. Mary watered the flowers.” are likely to be represented in different models because they appear to have little to do with one another. These considerations are captured by a more general postulate: Postulate 5.1. Once a situation model has been formed, subsequent information is integrated into the model if and only if this information concerns the same situation as the model or a thematically related one.

This integration is presumably performed by the Simulator. Thus, the Simulator presumably forms an event model of the first piece of information it receives. When it encounters a second piece, it forms an event model of this as well and then compares its features with those of the first model. If the two sets of known features are compatible, features of one model that are either unspecified or absent are instantiated in terms of specified features of the other. An episode model is then formed composed of both event models in combination. On the other hand, suppose the features of the second model are incompatible with those of the first. Then, the integration is not performed. Thus, suppose people who have formed an event model of “George packed his bags” encounter the statement “He left for the airport.” The referent of “he” in the second statement is compatible with the representation of “George” in the first statement, and so a model of the second statement is likely to be integrated with the first model, producing an episode model of the events described two statements in combination. However, suppose the second statement is “She left for the airport.” In this case, “she” cannot be instantiated in terms of the features of the first representation, and so a separate model of the event is constructed. In some cases, there might be an a priori reason to suppose that statements are thematically related despite appearances to the contrary. In this case, an integration of the two statements might be performed. However, this would not occur spontaneously by the Simulator, but rather, would involve higher order, Executor-controlled processing. Thus, in our example, people might speculate that “George” is actually a woman. Or, they might infer that “she” refers to an individual who was mentioned in a previous statement to which they are not privy. However, these speculations would not occur automatically.

108

CHAPTER 5

Note that these latter, more deliberative integration processes require a modification of the representation that was initially formed of the first statement. As we indicated in chapter 2, new features can be added to a representation, but existing features can never be erased. Thus, the goaldirected processing described in the previous paragraph would result in two representations: (a) the initial event model, which contains only the original information, and (b) a second representation that contains implications of both the original information (as modified to accommodate the new material) and the new. This assumption, which becomes of particular importance in our discussion of higher order comprehension processes (see chap. 7), is captured by the following postulate: Postulate 5.2. When new information requires a modification of an event or episode model that had been constructed on the basis of prior information, the original model is retained in memory. However, a second representation is constructed consisting of a modified version of the first model along with the implications of the new information.

In this regard, the thematic relatedness of information is not an inherent property of the information itself. That is, it can depend to a large extent on the prior knowledge of the recipient. For example, the statements “Mary got a coke. John went to the bathroom.” may appear to be more thematically related to people who know that John and Mary are preparing to see a movie at a local theater than to people who do not have this knowledge. A more graphic example is provided by the following passage written by Nancy McCarrell (see Bransford & Stein, 1984): Remember Sally, the person I mentioned in my last letter? You’ll never guess what she did this week. First, she let loose a team of gophers. The plan backfired when a dog chased them away. She then threw a party but the guests failed to bring their motorcycles. Furthermore, her stereo system was not loud enough. Sally spent the next day looking for a “Peeping Tom” but was unable to find one in the yellow pages. Obscene phone calls gave her some hope until the number was changed. It was the installation of blinking neon lights across the street that finally did the trick. Sally framed the ad from the classified section and now has it hanging on her wall. (p. 51)

To the naive reader, the events described in the passage appear to have little to do with one another. However, a reader who has prior knowledge that the woman is trying to get rid of her noisy neighbors is much better able to understand the causal and thematic relatedness of the events. The question is how to determine empirically whether episode models are formed of information and, if so, how many. A procedure for determining this was developed by Radvansky and Zacks (1991) in investigating the

EPISODE MODELS AND GENERALIZED NARRATIVE REPRESENTATIONS

109

construction of situation models of states of affairs. In a typical study (Radvansky & Zacks, 1991), participants were asked to learn statements concerning an object and its location. In multiple-object, single-location conditions, the statements described 1, 2, or 3 objects in a given location (e.g., “The book is on the table.” “The cup is on the table.” “The pen is on the table.” In single-object, multiple-location conditions, the sentences described a given object in 1, 2, or 3 locations (“The book is on the table.” “The book is on the chair.” “The book is on the floor.”). Radvansky and Zacks (1991) reasoned that people can easily imagine several objects in a given location and, therefore, would construct a single situation model under multipleobject, single-location conditions regardless of the number of objects involved. However, an object cannot occupy different locations at the same time. Therefore, in single-object, multiple-location conditions, participants would have to form a different model on the basis of each statement. Participants after learning the statements to criterion were given a recognition memory test, and the time to verify the statement was observed. Radvansky and Zacks argued that if participants formed a different model on the basis of each statement, the models would interfere with one another at the time of retrieval, and the magnitude of this interference would increase with the number of models involved. To this extent, the time to recognize a given statement should increase with the number of statements in the set. However, suppose the statements were integrated into a single situation model. Then, there should be no interference, and so the verification time should not depend on the number of statements. Results were consistent with expectations based on this assumption. That is, response times increased with the number of statements presented under singleobject, multiple-location conditions, but were independent of the number of statements presented in multiple-object, single-location conditions. Radvanksy and Zacks’ study not only provided insight into the construction of situation models but also suggested a general methodology for determining the conditions in which they are constructed (for other demonstrations of the effectiveness of the procedure, see Radvansky, Spieler, & Zacks, 1993). This methodology was therefore applied in investigating the validity of Postulate 5.1. In this study (Radvansky, Wyer, Curiel, & Lutz, 1997), participants were asked to learn sentences each describing the relationship of a person (e.g., a lawyer) to an object (e.g., a toothbrush). In some cases, the person and object were related in terms of a specific behavior (e.g., “the lawyer is buying a toothbrush”), whereas in other cases, they were related in terms of ownership (e.g., “the lawyer owns a toothbrush”). Finally, the objects presented in some conditions (e.g., a toothbrush, shaving cream, and aspirin) could all be purchased in a single location (e.g., a drugstore) but in other cases, the objects (e.g., a toothbrush, a bicycle, and a diamond ring) were typically purchased in different loca-

110

CHAPTER 5

tions. As in Radvansky and Zacks’ (1991) studies, participants after learning the statements were asked to indicate whether or not these statements (along with distracters) had been among those they had been asked to learn. Response times to presented items were recorded. Although the design of the study was complex, the predictions were straightforward. That is, participants should form a single episode model from statements that describe a single person buying objects in a single location. In this case, therefore, the time to verify such statements should not depend on the number that were presented. However, separate event models should be formed of a person who buys objects in different locations, or of different persons who buy the same object regardless of location. Finally, when the relationship involves ownership, which is not temporally constrained, event models should not be formed at all. Therefore, in all cases but the first, verification times should increase with the number of statements presented. Results shown in Fig. 5.1 are quite consistent with these expectations. The figure shows mean verification times to statements in each experimental condition as a function of the number of statements presented. The top left panel shows that when the statements described a single person’s purchase of objects in a single location, verification time was virtually identical regardless of the number of statements presented. In all other conditions, however, verification time increased with the number of statements presented. Thus, these findings confirm the implications of Postulate 5.1. First, event models are only likely to be formed in the course of comprehending statements about events that are situationally and temporally constrained. Second, event models are likely to be integrated into a single multiple-segment episode model only if the events occur in a single situation. A qualification on the latter conclusion, however, is indicated in the next section. Effects of Thematic Relatedness Radvansky et al.’s (1997) study indicated that events that occur in different locations may be represented separately in memory rather than integrated into a single episode model. However, the construction of episode models may be influenced by other factors, the effects of which could override the impact of situational differences. If events are causally related, for example, they may be incorporated into a single model even if they occur in different locations. By the same token, thematically unrelated events are likely to be represented in different models even if they occur in the same situation. An early study by Wyer and Bodenhausen (1985) provided indirect evidence of these differences. Participants read a story about events that occurred at a cocktail party that was written from the perspective of a person who attended it. Two target episodes were included among several others

111

FIG. 5.1. Mean time to verify a statement as a function of (a) the number of persons associated with a given object (denoted by squares) or (b) the number of objects associated with a given person (denoted by circles). The two top panels show responses to items pertaining to buying objects that could be found in either a single location (a drugstore) or different locations. The two bottom panels show responses to items pertaining to owning these objects. RT = response time. Based on data reported by Radvansky et al. (1997).

112

CHAPTER 5

that took place in the course of the party. One episode consisted of events that were causally related and occurred in the same situation. Specifically, someone bumped a guest’s arm as he was reaching for an hors d’oeuvre, causing him to spill his drink on a woman’s dress. The woman thereupon called the offender a disparaging name and stalked off to the bathroom to clean up. In contrast, the events composing a second episode were causally related but occurred in different physical locations. That is, it concerned a guest’s telling a story about Willa, who learned that her father was dying, packed her bags and got on a plane to San Francisco, had three drinks on the plane to calm herself, but felt tipsy by the time the plane landed. Consequently, she forgot the name of the hospital where her father was staying and wound up crying on the streets of San Francisco. One of the episodes was conveyed nearer the end of the party scenario than the other. However, the order in which the events composing each episode were described was varied, sometimes being in chronological order and in other cases being in the reverse order (e.g., “Willa had cried on the streets of San Francisco. She had forgotten the name of the hospital where her father was staying. She was feeling tipsy as a result of having three drinks on the plane . . .”). Moreover, the events were all mentioned in sequence or were separated by other, unrelated events that occurred at the party and temporarily distracted the visitor from those involved in the episode being described. Participants after reading the entire cocktail party narrative and a short delay were asked to recall what they had read. Participants typically recalled the events composing each target scenario in chronological order regardless of the order in which they had been presented and regardless of whether the descriptions of them were separated by unrelated events. This suggests that they integrated the events in each scenario into a multiplesegment episode model that they later used as a basis for recalling what had occurred, and this was true regardless of whether the events occurred in the same physical situation or in different situations. However, they recalled the last story mentioned in the narrative before recalling the one that occurred earlier. Thus, participants formed separate episode models of the two event sequences despite their occurrence in the same general situation (the party), perhaps assigning them a common header (e.g., “the cocktail party”). Thus, the residual resonance of the more recently constructed episode model led it to be the first one retrieved and reported (Postulate 2.4). Temporal Coding of Episode Models The conclusion that thematically unrelated sequences of events are organized into different episode models and are stored separately in memory is quite plausible. But if this is the case, how can the temporal order of the se-

EPISODE MODELS AND GENERALIZED NARRATIVE REPRESENTATIONS

113

quences of events be reconstructed? We can often remember going to Bali on vacation before George Bush was elected President despite the fact that the events are totally unrelated. How is this accomplished? The order of such events may in some situations be computed rather than remembered. For example, the events may each be compared to a “landmark” event (e.g., my wife’s having a baby) that is personally relevant and whose temporal position is fixed (Loftus & Marburger, 1983). Or, people may simply apply a “heuristic,” reasoning that the more they can remember about an event, the more recently it probably occurred (Brown, Rips, & Shevell, 1985). Even within the same situation, however, the order of events that are ostensibly unrelated to one another can often be constructed fairly accurately. It is conceivable that in the latter situations, people may assign temporal “tags” to events that denote the order of their occurrence and permit their order to be reconstructed later. This possibility was examined by Wyer, Shoben, Fuhrman, and Bodenhausen (1985). People read a version of the “Willa” story employed by Wyer and Bodenhausen (1985) and mentioned earlier. In some cases, however, the first and last scenes were modified to make them ostensibly unrelated to those that occurred in the middle one. These events, which were presented in a paragraph, were as follows: Willa was awakened by the alarm. She got out of bed. She had breakfast. She got on a plane to San Francisco. She had three drinks. She felt dizzy. She was glad when the plane landed. She went shopping in the city. She bought a sweater.

It seemed reasonable to suppose that participants would construct three episode models in the course of comprehending the events (pertaining to getting up in the morning, the plane ride, and the shopping spree), designating each by a header that refers to what the model is about. They might also assign a temporal tag to each model (e.g., “early,” “middle,” “late,” etc.) to indicate when the episode occurred in relation to others. Thus, suppose participants are later asked which of two events occurred earlier. They might first identify the episode model(s) in which the events were contained and compare the tags associated with them. If the tags differ, they can make a quick decision. If the tags are the same, however, participants must compute the order of the events on the basis of their location in the episode model itself, and so they take more time to respond. Thus, the overall decision time, Ttotal , might be given by the equation:

114

CHAPTER 5

Ttotal = t0 + tI + tC1 + tC2 , Where tI is the time to identify the episode model in which the events are contained, tC1 is the time to compare the temporal tags associated with the models, and tC2 is the time to locate the relative positions of the events within a given episode model should this prove necessary. To evaluate this conceptualization, participants after reading the passage were given several different pairs of events and asked in each case to decide which event in the pair occurred sooner. Participants took more time to make this decision when one or both events occurred at the boundary of an episode model than when they occurred in the middle, suggesting that more time was required to identify the model in which an event was contained (tI ) when its location was relatively ambiguous. Second, participants took more time to compare events if they occurred in the same model than if they occurred in different ones, and this was true independently of the positions of the events in the story as a whole. This finding is consistent with the assumption that when the events are contained in the same model, their order cannot be determined on the basis of the temporal tags assigned to the models. Therefore, an additional amount of time (tC2 ) is required to make this determination.9 In additional conditions of Wyer, Shoben, Fuhrman, and Bodenhausen (1985) study, the three scenes of the story were similar to those employed by Wyer and Bodenhausen (1985) and, therefore, were thematically related. In this case, decision times were generally longer than they were when the scenes were unrelated. Wyer et al. attributed this difference to an increase in tI (the time to identify the model in which the events were contained). However, other results were comparable to those that occurred when the scenes were unrelated, suggesting that different episode models were formed in this condition as well. This conclusion might seem to contradict the implications of Wyer and Bodenhausen’s findings. However, a sequence of thematically related events may be more likely to be integrated into a single episode model when the events are described in the context of several other, unrelated episodes, as was the case in the cocktail party story that Wyer and Bodenhausen’s participants read. When a sequence of 9 As an aside, it might be noted that the conceptualization predicts that in general, the time to compare two events will be less when they are far apart in the sequence than when they are close together. This suggests a “symbolic temporal distance” effect of the sort observed by Nottenburg and Shoben (1980) and similar to that obtained in other judgment domains (cf. Banks, 1977). According to this conceptualization, however, the effect results from the fact that the likelihood that the events being compared are in different models (and, therefore, tC2 = 0) increases with the distance between them. In fact, although a symbolic distance effect was observed in Wyer, Shoben, and Bodenhausen’s (1985) study, it was eliminated when the relative numbers of between-model and within-model comparisons was controlled.

9

EPISODE MODELS AND GENERALIZED NARRATIVE REPRESENTATIONS

115

events is conveyed in isolation, the different situations appear more distinct, and so different models may be constructed. Effects of Familiarity An additional factor that might determine the extent to which a sequence of events is divided into different episode models is the similarity of the sequence to ones that participants have been exposed in the past and are already represented in memory. For example, it seems intuitively reasonable to suppose that familiar sequences are encoded into broader conceptual units than novel ones (Newtson, 1973, 1976). Thus, multiple-segment episode models may be formed more frequently in the first case than the second. Stan Colcombe and I (reported in Wyer et al., 2002) examined this possibility using procedures similar to those employed by Radvansky and his colleagues and described earlier. We constructed four stories, each composed of events that occurred in three related situations. The events in two stories were fairly novel; one described Willa’s ill-fated trip to visit her sick father and a second described a woman’s encounter with a bear while camping. The events composing the other two stories were routine, describing a visit to a restaurant (with scenes pertaining to entering, eating, and paying) and going to a movie (with scenes pertaining to buying tickets, eating some popcorn, and seeing the show). Three versions of each story were prepared. Each version described six events, with the number of events taken from each scene (1, 2, or 3) varying in a counterbalanced design. (Thus, one version of the restaurant story described one event that occurred while entering, two events that occurred while ordering, and three events that occurred while paying; another described three events that occurred while entering, one while eating, and two while paying, etc.) Participants read the stories under one of two conditions. In other-referent conditions, the protagonist in the story was a hypothetical person. In self-referent conditions, participants were told to imagine themselves experiencing the situations described, and the protagonist was referred to as “you.” After learning the stories to criterion, participants were asked to verify the events that were mentioned in each story. We assumed that the extent to which verification time increased with the number of events described in each scene of the stories indicated the extent to which participants formed a single episode model of the scene or separate models. Pooled over otherreferent and self-referent conditions, the time to verify novel events increased with the number presented in the scene to which they pertained (1.23 s vs. 1.35 s when 1 vs. 3 events were presented, respectively), suggesting that a different model was formed of each scene. In contrast, the time to verify familiar events did not depend on the number presented (1.48 s vs.

116

CHAPTER 5

1.40 s, respectively). These data therefore confirm the expectation that participants are more likely to construct a single episode model from a series of familiar events than from a sequence of novel ones. In addition, the effects of set size were contingent on the nature of the protagonist. Specifically, verification times increased with the number of events presented in a scene when the protagonist was described as “you” (1.32 s vs. 1.41 s, when 1 vs. 3 events were presented, respectively). This was not the case, however, when hypothetical others were involved (1.40 s vs. 1.33 s, respectively). Thus, participants were apparently less likely to construct a single multiple-segment episode model of events that they imagined themselves experiencing than events that they imagined an unfamiliar other experiencing, and this was true even when the situation described was very familiar. These results suggest that the criteria that people employ in comprehending self-referent events may differ from those they apply when comprehending information about unfamiliar others. More direct evidence of this difference (Colcombe & Wyer, 2002) is described in chapter 6.

THE CONSTRUCTION OF GENERALIZED NARRATIVE REPRESENTATIONS Episode models of social experiences are presumably formed spontaneously in the course of comprehending them. However, more abstract narrative representations of temporally related events exist in memory as well. Unlike episode models, these generalized event representations are unlikely to be constructed automatically. Rather, they are formed in the course of conscious goal-directed activity that requires them. This activity can potentially be of several types. First, people who wish to tell others about an experience they have had or a movie they have seen may describe the events that occurred in fairly general terms. In pursuit of this objective, they may form a linguistically coded representation of the events that is more abstract than the episode model they formed of these events in the course of comprehending them, and may store the representation it in memory independently of this model. Second, people who have had a number of very similar experiences in a given type of situation might form a generalized representation of the events containing features that the experiences have in common but are not specific as to time and place. Thus, for example, people might form a general representation of the events that occur in a restaurant (looking at the menu, ordering, eating, etc.) based on a number of specific experiences in which the events have occurred. These generalized representations, which have been referred to elsewhere as scripts (Schank & Abelson, 1977),

EPISODE MODELS AND GENERALIZED NARRATIVE REPRESENTATIONS

117

event prototypes (Colcombe & Wyer, 2002) or event schemas (Graesser & Nakamura, 1982; Wyer & Srull, 1989), can function as implicit theories about the events that occur in the type of situation at hand. As we note in the next chapter, however, the abstraction of generalized representations from specific exemplars may occur less frequently than is often assumed. Colcombe and Wyer (2002) showed that people typically do not form generalized narrative representations of events on the basis of exemplars unless they are exposed to several different exemplars of the generalized narrative in temporal proximity. Even then, the abstraction may not occur unless people are stimulated to think about the exemplars in relation to one another. Thus, although generalized event representations undoubtedly exist in memory, they may be constructed in other ways. In fact, many generalized event representations might not be based on personal experience at all. Rather, people might acquire a general understanding of the causal relations among event concepts through social learning, without directly experiencing the individual events that exemplify these concepts. Thus, a child may be told that people who study hard get good grades in school, and that these grades help them to get a good job upon graduation. This sequence of events could constitute an implicit theory that the child accepts without first-hand knowledge of any specific instance in which the theory holds. Finally, people may construct generalized event representations in the course of making a judgment to which it is relevant. For example, people often wish to explain why a past event occurred, or to predict a future event. Alternatively, they might wish to assess the desirability of a particular course of action. These events could either be situation specific or more general. For example, one might wish to explain either why a particular friend got AIDs or, alternatively, why the incidence of AIDs in underdeveloped countries is increasing. Because the processes of constructing these generalized representations are somewhat less obvious than the process of constructing other types, these processes and their implications are worth discussing in somewhat more detail. General Considerations To explain or predict a specific event, people might attempt to identify a pre-existing episode model in which the event is embedded and use segments that precede or follow this event as bases for their inference. When such a model does not exist, however, a two-stage process may be required. First, people identify a generalized propositional representation of the event to be explained (i.e., “people get AIDs”). Then, having done so, they search memory for propositions that describe possible antecedents of it and, if they find one, they base their explanation on this proposition.

118

CHAPTER 5

The assumption that people spontaneously think about the antecedents and consequences of an event is suggested by McGuire and McGuire (1991) in conceptualizing the content and structure of thought systems. This conceptualization, which was stimulated in part by McGuire’s (1960, 1981) earlier work on the cognitive organization of propositional knowledge, assumes that individuals who are intrinsically or extrinsically motivated to construe the likelihood that an event will occur may search memory for antecedents of the event (i.e., factors that will either facilitate or interfere with the event’s occurrence). In contrast, individuals who are stimulated to consider an event’s desirability are likely to ponder its consequences. In support of these assumptions, the McGuires found that when people were asked to report their spontaneous thoughts about a target event (e.g., “a vaccine to prevent AIDS will be developed”), 78% of their thoughts pertained to either antecedents of the event or consequences of it. Moreover, providing information about others’ beliefs that an event would occur stimulated participants to generate more thoughts about antecedents of the event than about its consequences (56% vs. 22%), whereas providing information about others’ estimates of the event’s desirability stimulated more thoughts about its consequences than about its antecedents (55% vs. 31%). The McGuires’ research does not indicate which of several antecedents or consequences happens to be identified in the course of this cognitive activity. However, when several alternative propositions (or, in some cases, sets of propositions) could potentially describe a cause or effect of a target event, people are likely to identify and use the one that is most accessible in memory at the time (see chap. 3). The accessibility of such a proposition presumably depends on the strength of its previously formed association with the target or, if no prior association exists, on the frequency or recency with which it has previously been identified and used for other purposes (see Postulates 2.4 and 2.5). In either case, the identification of the proposition establishes (or strengthens) an association between the proposition and the target, and so the two propositions in combination may be stored in memory as a unit, thus functioning as an implicit theory of the sort described in chapter 1. The construction of this theory should have at least two consequences. First, people who recall one proposition are likely to recall the other as well. Second, they should be inclined to use the theory that is constructed from the propositions as a basis for judgments to which it is applicable. Empirical Evidence A study with Jon Hartwick (Wyer & Hartwick, 1984) confirmed these predictions. Participants were first given a list of randomly ordered propositions and instructed to indicate whether the propositions were understandable.

EPISODE MODELS AND GENERALIZED NARRATIVE REPRESENTATIONS

119

Thirty-two of the propositions comprised 16 pairs of assertions that occupied the positions of A and B in a causal statement of the form, “If A, then B.” Moreover, the likelihood that A was true, and the clarity of its implications for B, were systematically varied over the 16 pairs. (Thus, for example, a pair in which A was likely to be true and that had clear implications for B consisted of the statements, “Trucks conveying heavy cargo destroy highways,” and “The weight limit on truck cargo will be decreased.” In contrast, a pair in which A was likely to be true and had unclear implications for B was composed of the statements, “There will be no national borders by the end of the year 2000” and “War will be a thing of the past by the end of the [20th] century.”) This task presumably increased the accessibility of the propositions in memory and, therefore, increased the likelihood that participants would later identify and use them for purposes for which they were applicable. Participants then completed a two-part questionnaire in which they were asked to report their beliefs in half of the propositions they had encountered in the comprehension task (i.e., the likelihood that the events described would occur) and to estimate the desirability of the events described in the remaining ones. This was done in such a way that for a given A–B pair, participants reported either (a) their beliefs in both A and B, (b) the desirability of both propositions, (c) their belief in A and the desirability of B, or (d) their belief in B and the desirability of A. They were then dismissed but returned for a later session 1 week later, at which time they were asked to recall as many of these propositions as they could. We expected that participants who were asked to report their belief in B would search memory for an antecedent of the event. Moreover, they should use the first relevant antecedent that comes to mind (Postulate 3.1; see Henninger & Wyer, 1976). Therefore, if A had been encountered in the comprehension task and had clear implications for B, participants should identify and use it as a basis for their judgment, thereby establishing an association between A and B that might otherwise not have existed. Similarly, participants who were asked to report the desirability of A should search for a consequence and identify B, and this should also establish an association between the two propositions. In contrast, participants who are asked to report their belief in A should not identify B. Correspondingly, people who are asked to estimate the desirability of B should not consider A. In these conditions, therefore, an association between A and B should not be formed. The strength of association that participants formed between A and B in the first session of the experiment should be reflected in their recall of the propositions later. Specifically, suppose participants had established an association between A and B. Then, if they recall B during the second session, they should think of A as well. In other words, their likelihood of recalling A

120

CHAPTER 5 TABLE 5.1 Strength of Association Between Syllogistically Related Propositions as a Function of the Plausibility of the Antecedent (A), the Clarity of its Implications for the Consequence (B) and the Type of Ratings Made of A and B Rating Made of Consequence (B)

Characteristics of Antecedent (A) Clear implications for consequence Plausible Implausible Mean Unclear implications for consequence Plausible Implausible Mean

Rating Made of Antecedent (A)

Likelihood

Desirability

Likelihood

Desirability

.734 .462 .598

.505 .156 .331

.442 .274 .358

.798 .345 .572

.320 .375 .347

.439 .402 .420

.366 .482 .424

.393 .295 .344

Note. Numbers refer to the difference between the probability of recalling A given that B was recalled and the probability of recalling A given that B was not recalled. Adapted from Wyer and Hartwick (1984).

if they recall B should be greater if they have recalled B than if they have not. The difference between these two conditional probabilities therefore provided an index of the strength of association that people had formed between A and B in the first session of the experiment. This index is shown in Table 5.1 as a function of (a) the a priori likelihood that A was true, (b) the clarity of A’s implications for B, and (c) the type of rating that participants had made of the propositions in the first session of the experiment (likelihood vs. desirability). When A had clear implications for B, it was more strongly associated with B when participants had reported either the likelihood that B was true (.598) or the desirability of A (.572) than when they had reported either the likelihood that A was true (.358) or the desirability of B (.331). When A’s implications for B were unclear, however, these contingencies were not apparent. Thus, when A’s implications for B were not obvious, participants were less likely to recognize their relatedness during the first session of the experiment, and so they based their judgments on other criteria. As a result, an association between the two propositions was not formed, and so their recall of B had no impact on the likelihood of recalling A. Other Determinants of Association Wyer and Hartwick’s (1984) findings obviously did not provide a complete picture of the associations that are formed among general propositions of the sort that compose an implicit theory or implicational molecule. McGuire

EPISODE MODELS AND GENERALIZED NARRATIVE REPRESENTATIONS

121

and McGuire (1991) identified several other factors that can exert an influence on these associations. For example, they postulated that people are likely to think about antecedents of events over which they feel they have control, but are more inclined to think about consequences of events over which they have little influence. Therefore, because people feel they have more control over the events in their personal lives than over events in the society at large, they tend to think about antecedents of personal life events (e.g., having a paper rejected for publication) but are more likely to think about consequences of social events (e.g., a terrorist bombing). For possibly similar reasons, people are disposed to think about the antecedents of events that are likely to occur in the distant future but to think about the consequences of immediately impending events over which they have less control. Other motivational factors could also influence the type of associations that people form. For example, people may be motivated to think about desirable rather than undesirable consequences of events that are likely to occur, and to think about factors that facilitate rather than inhibit the occurrence of events they consider desirable. Although the effects of these factors were not consistently supported by the McGuires’ data, their potential importance should not be entirely dismissed. Indeed, a study by Albarracin and Wyer (2001) provided evidence of these effects. Briefly, participants who had been induced to feel either happy or unhappy by writing about a past experience read a persuasive message containing either strong or weak arguments in favor of performing a particular behavior. This was done in either the presence or absence of situational distraction. After reading the message, participants reported their attitudes toward the behavior being advocated, followed by estimates of the likelihood and desirability of specific behavioral consequences. Participants who were not distracted based their beliefs and evaluations of the consequences on the content of the message, and these beliefs and evaluations determined their overall attitude toward the behavior. In contrast, participants who were distracted from thinking about the communication based their attitudes on the extraneous affect they were experiencing and misattributed to their feelings about the behavior. Once these attitudes were formed, however, they had reciprocal effects on outcome-related beliefs and evaluations. For example, participants with favorable attitudes reported stronger beliefs that desirable consequences would occur, and reported that likely-to-occur outcomes were more desirable. The Construction of Implicational Molecules Propositions that become associated as a result of the processes described in this chapter constitute an implicit theory about the causal relatedness of the events they describe that can be used to comprehend and make infer-

122

CHAPTER 5

ences about more specific experiences. A conceptualization of these inference processes was mentioned briefly in chapter 1. To reiterate, sets of general propositions that become associatively linked in memory can form implicational molecules that are stored in memory as conceptual units and, therefore, can be brought to bear on the comprehension of new information that exemplifies them (Abelson & Reich, 1969; Wyer & Carlston, 1979, 1994). In combination, the propositions that compose a molecule can exemplify a general principle that one uses to comprehend and make inferences about the world in which one lives. Many types of molecules can potentially exist.10 The molecules of primary concern in the present context consist of two or more verbally coded propositions, each describing a different event or state of affairs. In combination, the segments of a molecule exemplify a generalization about the world in which we live. For example, the generalization that smoking causes lung cancer might be represented in the twosegment molecule: [P smokes; P has (will get) lung cancer].

Other generalizations may have moral overtones. For example, the generalization that people get what they deserve might be conveyed in a “justdeserts” molecule; i.e.: [P does a bad thing; P encounters misfortune]

or, alternatively, [P does a good thing; P has good fortune].

Once implicational molecules are formed, they can be used to interpret specific events to which they are applicable. This possibility is particularly important when not all of the information that is necessary to instantiate the propositions in a molecule is available. For example, people might use the “smoking” molecule described earlier to infer that a person who smokes will get lung cancer. Alternatively, they might infer that a person they hear has lung cancer is a smoker. Similarly, they might use the justdeserts molecule to infer that a person who stole money will be punished, or that a person who was involved in a serious accident was likely to have 10

Not all implicational molecules pertain to events. Some, implied by cognitive balance theory (Heider, 1957), may concern the sentiment relations among a group of individuals (Picek, Sherman, & Shiffrin, 1975; Sentis & Burnstein, 1979). Other, “syllogism molecules” may specify the logical relations among propositions (Loken & Wyer, 1983). These possibilities, and empirical evidence bearing on them, have been summarized elsewhere (cf. Wyer & Carlston, 1979, 1994).

10

EPISODE MODELS AND GENERALIZED NARRATIVE REPRESENTATIONS

123

done something wrong. A more formal statement of these inference processes is provided in chapter 10, where we discuss the impact of implicit theories on judgments and decisions in some detail. The associations that give rise to the construction of an implicational molecule can often be formed as a result of the cooccurrence of its constituents in the course of comprehending information and making judgments. However, they can also be motivationally determined. As Kunda (1990) pointed out, people can often be motivated to believe that they have desirable attributes and that these attributes will allow them to have a happy and successful life. Therefore, they may often construct implicit theories and implicational molecules that confirm this belief. In doing so, they may selectively attend to new information, and selectively seek and retrieve previously acquired knowledge, that confirms their desire to maintain a favorable view of themselves and their world. This possibility is elaborated in chapter 10, where we consider the motivational factors that underlie the construction and use of implicit theories in more detail.

C H A P T E R

6 The Impact of Generalized Event Representations on Comprehension and Memory

People construct at least two types of narrative representations of the experiences they encounter in daily life. On one hand, they spontaneously form episode models in the course of comprehending specific events they personally experience, observe, or read about. In addition, they may intentionally form generalized representations of events in the pursuit of a more specific objective that requires them. These latter representations can sometimes be abstract codings of events that that occur frequently in certain types of situations and, therefore, constitute event prototypes (Colcombe & Wyer, 2002) or schemata (Graesser & Nakamura, 1982). They can also be linguistically coded descriptions of a specific experience that people form in the course of communicating about the experience to others. Although generalized event representations and episode models are theoretically stored in memory as separate units of knowledge (Postulate 2.1), they can obviously exert an influence on one another. For example, event prototypes can be used as a basis for comprehending new experiences that exemplify them and, therefore, can influence the mental representation that is formed of these experiences. Moreover, an abstract narrative representation that is formed in the process of communicating about an experience might later be recalled and used to attain other objectives to which it is relevant, and this could be done without consulting the episode models on which the representation was originally based. The present chapter discusses these possibilities in detail and presents data bearing on them. We first examine the role of prototypic event repre-

124

IMPACT OF GENERALIZED EVENT REPRESENTATIONS

125

sentations on the comprehension and representation of new experiences that exemplify them. In doing so, we consider the conditions in which these prototypes are formed and the extent to which they are actually used to comprehend one’s own experiences and those of familiar others. Second, we discuss the abstract representations that people form of a personal experience in the course of communicating about the experience to others, and consider the way in which these representations can influence memory for details of the experience being represented. Finally, we discuss the implications of these communication phenomena for the impact of telling stories about oneself on self-perceptions. In this context, we consider differences in the storytelling that characterizes parent–child interactions in different societies and discuss their implications for cultural differences in self-perception.

THE ROLE OF PROTOTYPIC EVENT REPRESENTATIONS IN THE COMPREHENSION OF ONE’S OWN AND OTHERS’ EXPERIENCES Many of the events that compose our daily life experiences are fairly routine. That is, they occur repeatedly, with minor variations, at many different times and in many places. The events that typically occur in the course of visiting a restaurant, cashing a check at the bank, or changing a flat tire, and the sequence in which these activities take place, are fairly commonplace, and there is widespread agreement within a particular culture concerning the events and when they occur. People may also engage in sequences of behaviors that, although idiosyncratic, they personally perform routinely on a daily basis. For example, I typically get up at 5 a.m., brush my teeth, do 30 sit-ups, and read a novel while drinking a cup of coffee. I then walk the dog, return home and take a shower before beginning work at about 8 a.m. This series of behaviors is unlikely to describe anyone but myself. Nevertheless, it is fairly invariant. It might therefore seem intuitively likely that people construct generalized representations of sequences of events that they personally experience or observe others experience. However, these representations would not have the character of an episode model. In particular, the events are likely to be coded propositionally rather than in terms of mental images. To the extent these event prototypes are formed, they seem likely to influence the comprehension of new experiences that exemplify them. In fact, however, the conditions in which this influence occurs are more restricted than one might expect. In the pages that follow, we consider the nature of these restrictions.

126

CHAPTER 6

Preliminary Considerations The assumption that people construct event prototypes is quite common in research and theory on prose comprehension (e.g., Bower, Black, & Turner, 1979; Graesser, 1981; Graesser, Gordon, & Sawyer, 1979; Schank & Abelson, 1977). These representations are unlikely to be formed spontaneously, however. Rather, they are constructed in the pursuit of specific processing objectives that require them. This might be done in a manner similar to that suggested by Hintzman’s (1986) conceptualization of the construction of schematic representations from exemplars. Once such a representation is formed, however, people store it in memory, and can later recall and use it to comprehend new experiences to which it is applicable. Several theories have been proposed to address these matters (Bower et al., 1979; Graesser et al., 1979; Graesser & Nakamura, 1982; Trafimow & Wyer, 1993). Most of these theories employ a principle of cognitive economy similar to that implied by Postulate 4.4. That is, they assume that when people encounter a sequence of events that is redundant with the events contained in an event prototype, they usually do not construct a detailed representation of these events. Rather, they retain only enough features that permit them to reconstruct the events that occurred on the basis of the prototype. However, the theories make different assumptions about the nature of the features that are retained. For example, Graesser et al. (1979; see also Bower et al., 1979) assume that when a sequence of events can be predicted on the basis of a more general event prototype, people typically do not store a representation of these events in memory at all. Rather, they store only a “pointer” to the prototype (i.e., a set of retrieval cues that identify it) along with a set of translation rules that instantiate features of the prototype in terms of specific features of the exemplar. Thus, suppose people read that Joe went to Dom’s Patio Villa for dinner, ordered lasagna and a bottle of Chianti, and paid $15.95 for the meal. They might not construct an episode model of the events. Instead, they might store a pointer to a prototypic restaurant scenario along with the translation rules, customer = “Joe,” restaurant = “Dom’s,” meal = “lasagna,” drink = “Chianti,” etc. These translation rules, along with the content of the prototype, would be sufficient to reconstruct what had occurred. They could also be used to infer the occurrence of other events that were not specified in the information but typically occur in restaurants (e.g., that John looked at the menu or actually ate the meal). Note that these inferred events cannot be distinguished in the representation from the events that were actually mentioned. As a result, errors may occur in recalling the information later (Graesser et al., 1979). A related conceptualization was proposed by Graesser and Nakamura (1982). They suggest that when people read descriptions of events the events that compose a prototype, they comprehend these events in terms

IMPACT OF GENERALIZED EVENT REPRESENTATIONS

127

of the prototype and copy them into the representation they form, instantiating the generic features of the prototype in terms of specific ones that are mentioned in the information. However, they do not include segments of the prototype to which the information does not refer. Thus, the events that were specified in the information and those that were not mentioned can be more easily distinguished. A limitation of both conceptualizations, however, surrounds the representation of events that are not instantiations of the prototype. The occurrence of these events in daily life is less often the exception than the rule. For example, a man who visits a restaurant is likely to look at the menu, order the meal, eat and pay the bill. However, he might also ponder a painting on the wall, say hello to a colleague who is seated at a nearby table, and spill some of his wine on the tablecloth. The way in which these latter events are integrated into the representation of the experience as a whole is not clearly specified. In fact, the theories simply assume that the events are appended to the representation as “tags.” To this extent, there would be no way to reconstruct the point in the sequence that the events occurred. In many cases, however, this reconstruction is obviously possible. A conceptualization developed by David Trafimow (Trafimow & Wyer, 1993) eliminates this ambiguity. Trafimow, like Graesser et al. (1979), assumes that people who encounter a sequence of events that exemplify a prototypic sequence do not normally retain these events in the representation they form. However, there is an important exception to this rule. That is, when a nonprototypic event occurs, people retain a prototypic event that occurred in temporal proximity to it in order to localize its position in the sequence. Thus, in our restaurant example, people who learn about John’s visit to Dom’s might not normally retain the events that occurred. However, suppose that while John was ordering his meal, he said hello to a colleague at the next table. In this case, the person may retain a depiction of John’s ordering the meal in the representation, thus permitting the point at which the nonprototypic event took place to be reconstructed. 1. An Empirical Test Trafimow designed an intriguing series of experiments to evaluate these possibilities. Suppose that when participants interpret a sequence of events in terms of a prototype, they only retain prototype-relevant events in the representation they construct if these events are necessary to localize nonprototypic ones. Then, the addition of nonprototypic events to a sequence should increase the number of prototypic ones that are retained and, therefore, should increase the ability to recall these events at a later point in time. On the other hand, suppose participants do not comprehend the sequence in terms of a prototype, but rather, construct an episode

128

CHAPTER 6

model of the events. In this case, additional event descriptions are likely to increase comprehension difficulty and, therefore, to decrease the likelihood of retaining the presented events in the representation that is formed. In other words, the addition of the nonprototypic events might decrease the recall of the other, prototype-related ones. (Evidence that the likelihood of recalling a given event decreases as the total number of events increases is consistent with the set size effects observed in other research paradigms; see Anderson & Bower, 1973; Srull & Brand, 1983.) To evaluate this possibility, Trafimow and Wyer (1993) identified several situations with which college students were familiar (e.g., cashing a check, photocopying a piece of paper, making tea, etc.) and constructed six events that typically occurred in each. The photocopying sequence, for example, included “found the article,” “got some change,” “found a machine,” “aligned the copy,” “put in the coins,” and “pressed the button.” However, several of the events composing each sequence were not unique to the situation at hand, but could occur in other situations as well. Therefore, without a prior indication of the situation being described, the nature of the situation was unlikely to be immediately identified. In addition to the prototypic events, six unrelated events were identified that could plausibly occur in the situation of concern (e.g., “remembered he had to return a phone call,” “ate a piece of candy,” “saw a person he knew,” etc.). Stimulus stories were then constructed by combining the prototyperelated events and the prototype-unrelated ones. In one experiment (Trafimow & Wyer, 1993, Experiment 2), the stories contained four prototyperelated events and either two or six unrelated ones. In some cases, the story was introduced with a statement that identified the situation (e.g., “John needed to photocopy a paper”) and in other cases, the situation was not indicated until the end (“After he had photocopied the paper, John left”). Participants read four stories that differed in terms of (a) the number of prototype-unrelated events presented, and (b) whether the prototype was identified at the outset or not until the end. Then, after a short delay, they were asked to recall all of the events they had read about. Results were quite consistent with expectations. When the situation was not identified at the outset, participants were unlikely to identify and use a prototype to comprehend the information. In this case, increasing the number of nonprototypic events presented decreased the proportion of prototypic events that were recalled (from .218 to .195, when 6 vs. 2 unrelated events were presented, respectively). When the situation was identified at the outset, however (thus stimulating participants to comprehend in the information in terms of a prototype), increasing the number of nonprototypic events presented increased the proportion of prototypic events that participants recalled (from .179 to .279 when 2 vs. 6 nonprototypic events were presented).

IMPACT OF GENERALIZED EVENT REPRESENTATIONS

129

More subtle implications of the theory were also confirmed. For example, if participants retain prototypic events in the representation in order to localize the nonprototypic events, the associations they form between these events should lead one event to cue the recall of the other. To this extent, participants’ recall of a prototypic event should be preceded by the recall of the nonprototypic event that came before it in the sequence. When the prototypic situation was identified at the outset, this occurred with a greater probability than would be expected by chance. When the situation was not specified until the end of the story, however, this was not the case. A more general prediction of the theory results from its implication that when no prototypic-unrelated events are included in the sequence, no prototypic events at all should be retained in the representation that is formed. Under these conditions, participants theoretically have no way of distinguishing between the prototypic events that were presented and those that were not. Consequently, participants should not only have poor recall of the presented events but also should tend to recall events as presented when they actually were not mentioned. On the other hand, suppose several unrelated events are added to the representation, leading the presented prototype-related events to be retained. Then, these events can later be distinguished from the prototypic events that were not presented, and so intrusion errors should decrease. An additional study (Trafimow & Wyer, 1993, Experiment 3) confirmed these predictions. Participants read stories composed of four prototype-related events and either four unrelated events or no such events. When the situation was identified at the outset, participants recalled a greater proportion of prototypic events when four unrelated events had been presented than when none had been presented (.21 vs. .13), confirming the results of the earlier study. At the same time, the number of intrusion errors was significantly lower in the former condition than the latter (0.25 vs. 0.96). 2. Methodological Implications Trafimow and Wyer’s (1993) findings, which were replicated by Colcombe and Wyer (2002, Experiment 1), provide evidence of the impact of using event prototypes to comprehend situation-specific sequences of events. However, the conditions in which this occurs may be more limited than one might expect. For one thing, prototypic event representations are usually not formed spontaneously. Rather, they are only constructed in order to attain a specific objective that requires them. Moreover, even if these representations have been formed and stored in memory, they may not always be retrieved and used. As we assumed in chapter 4, people who encounter a situation-specific sequence of events are likely to comprehend it with reference to a pre-existing episode model of the person and/or type of event in

130

CHAPTER 6

question. To this extent, they might interpret one exemplar of a routinized sequence of behavior in terms of another exemplar that happens to come to mind quickly rather than a more abstract representation that is less easily accessible in memory. (Thus, I might interpret a description of someone’s visit to McDonald’s in terms of an episode model of a personal experience I have recently had at McDonald’s rather than a more general representation of fast-food restaurant visits.) It can sometimes be difficult to distinguish between the use of an exemplar and the use of a more general prototype (Barsalou, 1990). If the conclusions drawn from Trafimow’s studies are valid, however, the procedure he employed can be used for this purpose. Specifically, the use of an exemplar or an event prototype to interpret a sequence of thematically related events can be inferred from the effect of introducing theme-unrelated events on the recall of theme-related ones. That is, suppose the introduction of these events increases the likelihood of recalling theme-related events (as Trafimow and Wyer found when a prototype was not made salient at the outset). This would indicate that an abstract representation of the sequence was used to interpret. If the introduction of these events decreases the recall of theme-related ones, however, it would indicate that a more general event representation was not used. The studies described in the following pages are based on this reasoning. When Are Prototypes Formed? As we have speculated, the conditions in which people spontaneously abstract generalized event representations from a number of exemplars could be quite limited. A large number of exemplars may be necessary to stimulate the construction of a generalized event sequence. A study with Stan Colcombe (Colcombe & Wyer, 2002, Experiment 2) demonstrated this possibility. Materials for the study were similar to those employed by Gick and Holyoak (1983) in research in analogical reasoning. Specifically, we constructed six different exemplars of a problem situation that could be resolved by applying the same “convergence” principle (i.e., several small entities can do the work of one if they converge simultaneously on the task at hand). One story described a problem of fixing some laboratory equipment: A woman walked into her physics lab to find that a very rare and expensive type of light bulb had been left on overnight, fusing the filaments inside the bulb together. The woman knew that a sufficiently intense blast from an ultrasonic wave generator would be able to separate the filaments and thus render the bulb useful again. Several of these wave generators were available. Unfortunately, a blast from the generator at an intensity high enough to separate the filaments would shatter the glass bulb surrounding the filaments. However, the physicist took several generators and set them to a relatively low in-

IMPACT OF GENERALIZED EVENT REPRESENTATIONS

131

tensity. After carefully positioning the wave generators in a circle around the bulb, she simultaneously discharged all of them. This short blast from many different sources at low intensity separated the filaments and preserved the glass bulb. Thus, the light bulb was repaired, and the woman’s physics experiments continued successfully.

A second scenario described the dismantling of a bomb: Security police discovered that a terrorist had planted a bomb at a major American airport. A bomb squad was called in to deal with the problem. Unfortunately, the type of bomb that the terrorists had planted was so sensitive that it was almost impossible to defuse. The bomb squad knew that if they could cool the bomb down to a very low temperature using liquid nitrogen, it might be possible to defuse. However, if they cooled the bomb down unevenly or too quickly, it would definitely explode. The only cooling device they had at the airport used a single large hose, and would deliver too much liquid nitrogen at a single point, thus setting off the bomb. However, the bomb squad used an attachment to split the single hose into several smaller tubes. They positioned these tubes in a circle around the bomb. Then, they slowly released the nitrogen through the many different tubes. Focusing all of the small tubes on the bomb simultaneously cooled the bomb down slowly and evenly, and so the bomb was able to be defused.

Three other stories concerned (a) putting out the fire in a burning woodshed, (b) taking over a corporation, and (c) extinguishing an oil well fire. Participants were exposed to either one or five of these context stories followed by a target story. The target, which concerned an attack on the stronghold of an evil dictator, had four versions. Two versions described four theme-related events and either two or six unrelated ones. Two other versions described four theme-unrelated events and either two or six related ones. For example, a story containing six theme-related events and four theme-unrelated events was: A small country was under the rule of an evil dictator who was despised by the people. The dictator lived in a fortress with many roads leading to it like spokes in a wheel. The dictator was cruel. Rebels realized that they had enough forces to overrun the fortress if they could attack all at once. The rebel general noted that the weather had been turning colder lately. A rebel spy reported back that the dictator had planted mines that would allow only a few men to pass at once along the many roads leading to the fortress. Given that all of the rebel forces must attack the fortress at once, it seemed as though the rebels’ plans were foiled. Birds flocked in a nearby tree. However, the rebel general was very smart and instructed his men to split up into several groups, each taking a different road to the fortress. A dog howled in the distance. In these small groups, the men could pass over the mines without setting them off and

132

CHAPTER 6

meet up again at the fortress. Thus, the rebels overthrew the dictator, and the people in the kingdom lived happily ever after.

Participants were introduced to the study with instructions that we were interested in how people understand the sort of information they read in stories and novels outside the laboratory, and that they should read the stories as if they encountered them in a book. Participants read the stories in sequence, with the target story being at the end. Finally, after a 5-min filler task, participants recalled the target story in as much detail as possible. If participants constructed an episode model of the target story without referring to the stories they had read earlier, the addition of theme-unrelated events should decrease their recall of theme-related ones. This was the case when only one exemplar of the prototype preceded the target story. That is, participants recalled a greater proportion of theme-related events when six unrelated ones were presented (M = .50) than when only two were presented (M = .80).When participants had been exposed to five exemplars before reading the target story, however, they recalled a greater proportion of theme-related events in the first condition than in the second (.83 vs. .60, when two vs. six unrelated events were presented, respectively). Thus, exposure to five exemplars of the convergence problem appeared to stimulate the extraction of a prototype that, once it was formed, was used to comprehend subsequent stories. However, a single exemplar was insufficient to accomplish this. Colcombe and Wyer’s study provides one of the first empirical demonstrations that new prototypic representations are actually formed on the basis of accumulated past experience. Research in other domains (e.g., Klein & Loftus, 1993; Klein, Loftus, Trafton, & Fuhrman, 1992; Sherman & Klein, 1994) suggests that the likelihood of using trait concepts rather than specific behaviors to describe individuals in a given life domain increases with the amount of experience one has had in this domain. In only one of these studies, however (Sherman & Klein, 1994), was the amount of experience experimentally manipulated, and in this study, participants were given an implicit impression formation goal before they received information about the person to be described. Therefore, these results do not bear on the extent to which people spontaneously extract new prototypic representations in the course of comprehending information that exemplifies them. On the other hand, the generalizeability of the conclusions drawn from Colcombe and Wyer’s study may also be limited. For one thing, the exemplars to which participants were exposed were presented in temporal contiguity. Although participants were not explicitly told to think about the stories in relation to one another, their juxtaposition in the same situation may have stimulated them to believe they were supposed to do so. In everyday life, these situational demands do not arise. Visits to a restaurant, or obser-

IMPACT OF GENERALIZED EVENT REPRESENTATIONS

133

vations of a spouse’s actions when getting up in the morning, occur very frequently. However, these experiences are temporally separated, and may seldom be thought about in relation to one another. Consequently, although Colcombe and Wyer’s (2002, Experiment 2) findings suggest that prototypic event sequences can sometimes be formed from exemplars when they are encountered within a short period of time, the conditions in which this occurs in the absence of a specific goal that requires it may nevertheless be more the exception than the rule. This is particularly true in the case of personal experiences, to which we now turn. The Role of Episode Models and Prototypes in Comprehending Personal Experiences Many of the behaviors we perform in daily life (e.g., getting up in the morning, washing the dinner dishes, etc.) occur repeatedly with minor variations. The behaviors we see others perform (e.g., my wife’s getting ready for a party, or a roommate’s studying behavior) may also be fairly invariant. It might seem reasonable to suppose that people spontaneously extract a prototypic representation of the routinized behavior that they later apply in explaining a particular instance of the behavior and predicting its consequences. People might also use prototypic representations of their own behavior to predict what others will do (Ross, Greene, & House, 1977). However, this is not necessarily the case. For one thing, the conditions in which people form prototypic representations of their own and others’ behavior may be limited. The construction of these representations is presumably stimulated by the need to abstract features that are common to several different exemplars. In daily life, however, people are seldom called upon to consider their own behavior at one point in time in relation to their behavior at a different time. This is particularly true if the behavior is routine. As we suggested earlier, many routinized behaviors may be governed by a production (Anderson, 1983) that, once activated, is applied with little if any conscious cognitive mediation. To this extent, the behaviors that occur at one point in time may seldom need to be thought about in relation to similar behavior that occurred earlier. Moreover, even when this need arises, a single prior instance of these behaviors may be sufficient. To know what to do at a McDonald’s restaurant, for example, I only have to recall my most recent visit to the restaurant. I do not need to retrieve a prototypic representations of fast-food restaurant visits in general. A familiar other’s routinized behavior may also not be comprehended with reference to an abstract representation of the behavior. That is, an episode model of the behavior may be constructed on the basis of a previously formed episode model of the sequence that has been stored in memory, as suggested by Postulate 4.5. (Moreover, if the sequence is redundant with a

134

CHAPTER 6

previously constructed model, a new representation might not be formed of it at all; see Postulate 4.4.) An episode model of the other’s past behavior may also be sufficient for both explaining the individual’s present actions and predicting his or her future ones. Therefore, the conditions that stimulate the construction of a generalized event representation of a familiar other’s routinized behavior could also be very limited. 1. The Comprehension of Routinized Events Involving Self and Others Two studies by Colcombe and Wyer (2002, Experiments 3 and 4) confirm these conjectures. In one study, participants in an initial session of the experiment were asked to list both (a) sequences of activities that they performed on a daily basis and (b) sequences that they observed their mother and their father perform regularly. Then, in a second session 1 week later, we asked them to read four stories. One story was based on a routinized sequence of behaviors they had personally reported in Session 1, and a second was based on the routinized activities that a different participant had reported. A third story described activities that either their mother or their father routinely performed, and the fourth was based on the activities of another participant’s parent. Each story consisted of four routinized events and either two or six events that were not part of the routine. Participants in reading the stories were told to imagine that they (or one of their parents) were performing the activities. Then, after a short delay, they were asked to recall the stories they had read. Results summarized in Table 6.1 are very clear. Participants consistently recalled fewer routinized events when six novel behaviors were contained in the story (M = .49) than when only two were contained in it (M = .60). Moreover, this was true regardless of whether the stories pertained to TABLE 6.1 Proportion of Routinized Events Recalled as a Function of the Source of the Story, the Referent of the Story, and the Number of Nonroutinized Events Presented Source of Story

Stories about self Two unrelated events Six unrelated events Stories about parent Two unrelated events Six unrelated events

Self

Different Person

.63 .49

.51 .45

.70 .57

.55 .45

Note. Based on data from Colcombe and Wyer (2002).

135

IMPACT OF GENERALIZED EVENT REPRESENTATIONS

themselves (.47 vs. .57) or a parent (.51 vs. .62), and regardless of whether the stories were based on routines that participants themselves performed (.53 vs. .67) or those that another had engaged in (.45 vs. .53). Thus, these data suggest that participants did not use a generalized representation to interpret either stories about themselves or stories about their parent. 2. The Use of Previously Formed Prototypes to Comprehend Another’s Behavior There are two possible reasons why participants did not use prototypic representations in the preceding study. First, as noted earlier, people might not form prototypic representations of their own and well-known others’ behaviors. Another possibility, however, is that these generalized representations may exist in memory, but people may simply not draw upon them for use in comprehending new experiences involving themselves and familiar others. A second series of studies support the latter possibility. Participants in these studies read stories similar to those employed by Trafimow and Wyer (1993). That is, each story consisted of four behaviors that exemplified a prototypic script (cashing a check, changing a tire, etc.) and either two or six prototype-unrelated behaviors. In one study (Colcombe & Wyer, 2002, Experiment 1), participants were told that the protagonist in the stories was a hypothetical person (“Joe”), as in Trafimow and Wyer’s original studies. In a second experiment, however (Colcombe & Wyer, 2002, Experiment 4), they were told to imagine that the story was about their roommate. (In this study, the protagonist was referred to as “he” or “she,” depending on whether participants were male or female, respectively). In all cases, the situation to which the sequence pertained was identified at the outset, thus ensuring that participants aware of the applicability of the prototype for comprehending it. Results of the study are summarized in Table 6.2. When the stories referred to a hypothetical person, Trafimow and Wyer’s (1993) findings were

TABLE 6.2 Mean Proportion of Prototype-Related and Prototype-Unrelated Events Recalled as a Function of the Number of Unrelated Events Mentioned in the Story and the Type of Protagonist Protagonist Unfamiliar Other

Roommate

.30 .42

.50 .39

Number of prototype-unrelated events mentioned in story Two events Six events Note. Based on Colcombe and Wyer (2002, Experiments 1 and 4).

136

CHAPTER 6

replicated. That is, participants recalled a greater proportion of prototyperelated events when six unrelated events were presented than when two were presented (.42 vs. .30, respectively), confirming the fact that participants used a prototype to comprehend the story. When participants imagined the events occurring to their roommate, however, their recall of the prototype-related events was poorer when six unrelated events were presented than when two were presented (.39 vs. .50, respectively). In this case, therefore, a prototypic event representation was not employed. Concluding Remarks Colcombe and Wyer’s findings suggest that even when a generalized event representation exists in memory, people do not always use it to comprehend information about themselves or familiar others. This conclusion is consistent with the retrieval postulates described earlier, and the results reported by Wyer and Radvansky (1999). That is, people who receive information about the behavior of a well-known person can often construct a situation-specific event model involving this person very quickly. This is particularly true if a previously formed event model of the person’s behavior already exists in memory. In this case, they are likely to use this model as a basis for comprehending the information rather than relying on more general knowledge that comes less quickly to mind. In retrospect, this conclusion is not surprising. Unless people have a reason to think about different instances of their own or familiar others’ behavior in relation to one another, they are unlikely to engage in this cognitive activity. In making behavioral decisions, a previously formed episode model of the behavior and its consequences can often serve just as well as an abstract prototype. Moreover, as noted earlier, routinized behavior may be governed by productions and performed with minimal cognitive mediation. The influence of these productions is particularly evident when one is distracted by thinking about something else. (For example, a person who is thinking about how to revise a paper while driving to the grocery store may suddenly find himself about to enter his office parking lot. Similarly, a person may often forget whether he has locked the door to the house before leaving.) As Bargh (1997) contended, a very large proportion of our behavior may be governed by automatic processes, and there is little need to use a generalized representation of this behavior that we have stored as part of declarative knowledge. Circumstances can nevertheless arise in which a prototypic event for example representation is personally useful. When a person’s behavior has undesirable consequences, the individual might be motivated to avoid these consequences in the future and, therefore, may attempt to explain it. To generate this explanation, the person might retrieve a number of epi-

IMPACT OF GENERALIZED EVENT REPRESENTATIONS

137

sode models of past behavior in situations in which a similar outcome has occurred, and might abstract commonalities in the behavior’s antecedents. Similar attempts might be made when a behavioral outcome is unexpected. This cognitive activity is unlikely to occur spontaneously, however, but instead, is driven by Executor-governed processing objectives.

THE EFFECTS OF COMMUNICATING ABOUT AN EXPERIENCE ON MEMORY AND JUDGMENT Many abstract representations of experiences are formed in the course of social communication. People are often called upon to describe an experience to others. These communications can sometimes describe a procedure (e.g., instructions about how to change a tire or use a photocopier). More generally, however, they concern a situation-specific experience (e.g., a personal encounter one has had, or a movie one has recently seen) that one considers of sufficient interest to relate. In order to generate these communications, people presumably construct a linguistically coded representation of the experience and, once it is formed, store this representation in memory. The representations that people form in the process of communicating about an experience are presumably based on episode models that they formed of the experience at the time it occurred. However, the representations differ in several respects from the models on which they are based. Features that are tangentially relevant to the communication goals may be omitted entirely. Others, which were represented visually in the episode models, are characterized in more abstract, linguistic terms. The specific content and structure of the representations that are formed can depend in part on the communication objective at hand. When the goal is to describe the experience itself, the communication is likely to consist of verbal descriptions of the events that occurred (e.g., “captions” of the sort that sometimes accompany the images contained in an event or episode model). Suppose, however, that people want to convey their impressions of the events or the individuals involved in them. Then, their communications might not describe the events themselves, but rather, might consist of more general trait descriptions along with characterizations of their own attitudes or opinions. In each case, the representation that people form may reflect the content and structure of the communication they generate, and this representation may be stored in memory separately from the more specific episode models of the original experiences that were used to construct it (Postulate 2.1). This possibility has further implications. As noted in chapter 3, people who are called upon to make a judgment or decision are unlikely to retrieve and use all of the information they have available in memory that is poten-

138

CHAPTER 6

tially relevant to this decision. Rather, they retrieve only a subset of this information and do not search further unless their confidence in the validity of the information or its implications is below threshold (Postulate 3.1; see also Chaiken, 1987). This “sufficiency” postulate was applied earlier in conceptualizing the effects of knowledge accessibility on judgments. However, it has implications for memory as well. For example, if a generalized representation of information has recently been formed in the course of attaining a specific goal, this representation is likely to be used as a basis for not only making judgments but also reconstructing the events that occurred. That is, the original representation may only be consulted if the implications of the more recently formed goal-specific representation are below confidence threshold. A classic study by Loftus and Palmer (1974) provides an example. Some participants who had been shown a picture of an automobile accident were later asked to indicate how fast they thought the car was going when it “hit” the tree. Others were asked how fast it was going when it “crashed into” the tree. Participants estimated that the car was going faster in the second condition than the first. Moreover, when later asked to remember details of the picture they had seen, participants in the second condition were also more likely to report (incorrectly) seeing broken glass at the scene of the accident. Participants apparently reconstructed the scene of the accident in a way that was consistent with implications of the question they had been asked, forming a new mental representation that contained features in addition to those that were conveyed in the picture. Their later memory was then based on this newly formed image independently of the one they had constructed at the time they saw the picture. In Loftus and Palmer’s study, however, participants’ processing objective stimulated them to add features to the representation they had formed initially. When people wish to describe an experience to others, the representation they form is more likely to omit features that are of peripheral relevance to the description they are providing. This could also decrease the accuracy of recalling information under conditions in which this representation is consulted. In a study by Schooler and Engstler-Schooler (1990), participants were shown a series of human faces and asked to describe some of these events verbally. Later, they were given a memory test for the faces. One might expect that the increased amount of processing involved in describing the picture would make it more memorable. In fact, however, participants were less accurate in identifying the faces they had described than faces they had seen but not described. Participants who wrote an abstract description of a face apparently formed a new representation of it in which certain specific features that were useful in distinguishing it from other faces were omitted, and then used this new representation, which was less detailed than the original representation, as a basis for their later recognition responses.

IMPACT OF GENERALIZED EVENT REPRESENTATIONS

139

Note that the abstract representation that is formed in the course of communicating verbally about an event does not replace the original (Postulate 5.2). However, because it was formed more recently, it is typically more accessible in memory, and might often be considered a sufficient basis for recognition responses without consulting the more detailed representation on which it is based. If a goal-specific representation is considered irrelevant, or if participants’ confidence in the judgments implied by the representation is below threshold, the more detailed representation would presumably be retrieved and used. The Impact of Communication Objectives on Memory for Observed Experiences Adaval and Wyer (2003) investigated the conditions in which episode models and goal-specific representations of social experiences were likely to be activated and used as a basis for reconstructing these experiences. In doing so, we considered the effects of two communication objectives that people often have in daily life outside the laboratory: communicating a description of the events to another, and conveying an impression of the persons involved in the events. We compared the effect of introducing these goals at the time the event were observed with the effect of introducing them later, after episode models of the events had already been formed. 1. Procedure Participants in two studies watched the opening 12 minutes of Edward Albee’s Who’s Afraid of Virginia Woolf. The segment concerns a married couple who return home from a party and prepare for the visit of some lateevening guests. The protagonists engage in an animated conversation in which both positive and antagonistic comments are exchanged, thus leaving the couple’s actual feelings toward one another unclear. Although the protagonists’ nonverbal behaviors during the interaction (working on a crossword puzzle, washing dishes, fixing drinks, cleaning up papers in the living room, etc.) were relevant to a description of the sequence of events that occurred, they were relatively incidental to the couple’s personalities. Participants in some conditions watched the movie with no specific goal in mind, much as if they saw it in a theatre. After doing so, however, participants in impression-description conditions were asked unexpectedly to write a 5-min description of their impressions of the protagonists’ personalities, whereas participants in event-description conditions were asked to spend 5 min describing the sequence of events that occurred. Two other groups of participants were told they would be asked to perform these tasks before they viewed the movie. Finally, participants under comprehension-only conditions watched the movie without being given specific objectives either be-

140

CHAPTER 6

fore or afterwards. (These participants spent 5 min after watching the movie writing a description of the university campus.) After performing the writing task, participants were given a recognition memory test, the nature of which differed in the two experiments. In Experiment 1, participants were given verbal descriptions of statements and behaviors and asked in each case to indicate whether the statement or behavior described had occurred in the movie. In Experiment 2, we extracted both pictures of events that had occurred and acoustically recorded statements that protagonists made and presented these stimuli (along with pictures and utterances from other parts of the movie) on a computer with instructions to indicate whether the stimuli had occurred in the movie or not. 2. Predictions Participants were expected to construct a multiple-segment episode model of the movie in the course of comprehending it. In addition, participants with a communication objective should construct a more general, linguistically coded representation in the process of generating their communication, the content of which depends on the particular goal being pursued. Specifically, persons who described the sequence of events that occurred were expected to include characterizations of not only things the protagonists said but also things they did, as both types of events were relevant to this description. However, only protagonists’ verbal statements were particularly relevant to an understanding of their personalities. Thus, participants who were told to convey their impressions of the protagonists were likely to include trait descriptions in their representations along with descriptions of protagonists’ statements that exemplified these descriptions, but were unlikely to include nonverbal behaviors. When participants in these conditions are asked to verify a recognition item, they should first retrieve the goal-specific representation they have formed, which is most accessible in memory. If they consider the content of this representation to be a sufficient basis for evaluating the item, they should use it without further ado (Postulate 3.1). If it is not, relevant, however, they should default to the episode models they formed at the time they watched the movie. Thus, suppose participants who have written a description of the sequence of events that occurred are asked to verify a recognition item. They should consider the representation they formed in the course of generating this description to be a sufficient basis for verifying both protagonists’ statements and their nonverbal behaviors. However, because this representation is less detailed than the episode models they had formed, they should be less able to recognize both types of items than they would if they had not used this representation. In contrast, participants who conveyed their impressions of the protagonist should consider the representation

IMPACT OF GENERALIZED EVENT REPRESENTATIONS

141

they formed in the course of this activity to be sufficient for evaluating protagonists’ statements but not their nonverbal behaviors. Therefore, they should resort to the use of episode models to verify these behaviors. Consequently, although their recognition of protagonists’ statements should be adversely affected by the goal-specific representation they formed, their recognition of the protagonists’ behaviors should not be. 3. Results The results of both studies were very similar and were generally consistent with expectations. Table 6.3 shows the difference between recognition accuracy under each communication objective condition and accuracy under comprehension-only conditions.11 Writing a description of the sequence of events that occurred decreased recognition of both protagonists’ statements and their nonverbal behaviors, and this was true regardless of whether participants had watched the movie with this objective in mind or were not told of the objective until afterwards. In contrast, communicating an impression of the protagonists only decreased participants’ recognition of things the protagonists said and did not influence their memory for things the protagonists did. Moreover, this decrease was only evident when participants were not given an impression-formation objective until after they had seen the movie. The effects of objectives that participants had in mind at the time they watched the movie deserve further attention. All of the statements and behavior that participants consider interesting enough to communicate to others are likely to be of interest to themselves as well and, therefore, to be represented in the episode models they would normally form spontaneously while watching the movie. Therefore, the content of the goal-specific representation formed by participants with an event-description objective should be similar regardless of whether they were aware of this objective at the time they watched the movie or not. To this extent, the decrement in accuracy that results from their use of this representation as a basis for recognition should not depend on when this objective was imposed. This was the case, as Table 6.3 indicates.

11 The index of recognition accuracy used in Experiment 1, which corrects for guessing, is given by the equation:

11

P(Acc) =

P(hit) - P(False Alarm) , 1 - P(False Alarm)

where P(Hit) and P(False Alarm) are the probabilities of identifying an item as having been presented when it was or was not presented, respectively (Hilgard, 1951). This index could not be applied to Experiment 2, as the number of nonpresented items was too small to obtain a reliable index of P(False Alarm).

142

CHAPTER 6 TABLE 6.3 Effects of Communication Objectives, and the Point at Which These Objectives Were Induced on Recognition of Protagonists’ Statements and Nonverbal Behaviors Experiment 1

Event-description objectives Induced before watching movie Induced after watching movie Impression-description objectives Induced before watching movie Induced after watching movie

Experiment 2

Statements

Behaviors

Statements

Behaviors

–.189a –.065

–.360 –.297

–.070 –.044

–.046 –.095

.130 –.108

.000 .021

.045 –.022

.046 .046

Note. Based on Adaval and Wyer (2002). aCell entries refer to differences between the accuracy obtained under each task-objective condition and accuracy observed in comprehension-only conditions. In Experiment 1, accuracy was based on the index proposed by Hilgard (1951); see Footnote 12. In Experiment 2, accuracy was based on the proportion of correct responses to recognition items.

In contrast, participants who are told at the outset that they would be asked to describe their impressions may include features in the representation they form while watching the movie that are relevant to their impressions but are not depicted in the episode models they would form if they were only trying to comprehend what is going on. The implications of these additional features may then be included in the communication they generate later, and their recognition may benefit, as results also suggest. Effects of Reminders on Recognition Suppose people after seeing a movie or observing an interaction are later reminded of the events that occurred. They are likely to comprehend this reminder with reference to a previously formed representation of the original experience. In the course of doing so, they presumably construct yet another representation of what went on. According to the conceptualization we propose, the content of this new representation, and its consequent effects on memory, may depend on the nature of the previously formed representation that was used to interpret the reminder. Adaval and Wyer (2003) investigated these possibilities in a study that provides further support for the implications of Experiments 1 and 2. In this study, participants who had watched the movie with instructions to comprehend it were then asked to describe their impressions of the protagonists. After doing so, however, they were dismissed with instructions to return the next day for a second session of the experiment. In this session,

IMPACT OF GENERALIZED EVENT REPRESENTATIONS

143

some participants were re-exposed to the audio portion of a 4-min portion of the movie they had seen without the accompanying video. Others were re-exposed to the video portion without the accompanying audio. Still others received no reminder at all. Finally, all participants were given the verbal recognition task administered in Experiment 1. Based on the reasoning underlying our interpretation of the first two experiments, we assumed that participants who received a video reminder of the protagonists’ nonverbal behavior would find the goal-specific representation they had formed while reporting their impressions to be inapplicable for comprehending it and, therefore, would retrieve and use the episode models they had constructed at the time they first watched the movie. Because this model contains representations of things the protagonists said as well as things they did, both types of events should be included in the new representation they form. Therefore, if participants use this recently formed representation as a basis for recognition, their memory for both statements and behaviors should be facilitated. However, suppose participants receive an auditory reminder. These participants should consider the goal-specific representation they formed to be sufficient for interpreting it, and should use it without bothering to consult their episode models of the original movie. In this case, the new representation they form should be similar in content to the goal-specific representation they had formed earlier, containing only abstract depictions of the protagonists’ statements. Consequently, this representation should not benefit their later recognition of these statements relative to conditions in which the reminder was not given. Results are consistent with these conjectures. Relative to conditions in which no reminder was given, a visual reminder increased the likelihood of recognizing not only protagonists’ nonverbal behaviors (mean increase = .110) but their statements as well (M = .056). However, a visual reminder had relatively little effect on participants’ recognition of protagonists’ statements (M = .035) and actually decreased their likelihood of recognizing protagonists’ behaviors (M = -.126). General Conclusions Adaval and Wyer’s studies provide evidence of the influence of goal-directed processing of a social experience on later memory for this experience. Some caution should be taken in overgeneralizing our specific findings. Impressions of the protagonists in the movie we presented were conveyed largely through their statements to one another, and their nonverbal behaviors were relatively incidental to these impressions. In many social situations, individuals’ nonverbal behavior is likely to have a greater impact on the impressions that people form of them. In these cases, com-

144

CHAPTER 6

municating impressions of the individuals might interfere with observers’ memory for their behavior as well as the statements they make. The Effects of Storytelling on Self-Representations and Social Identity The effects of goal-relevant event representations on memory of the sort that Adaval and Wyer identified have potentially much broader implications. People are often called upon to recount a past experience they have had. In doing so, they may tailor their descriptions of the experience to the values and expectations of their audience, omitting some aspects they believe the audience may consider unpalatable while elaborating on others that the recipient may like (Higgins & Rholes, 1978). The representation they form as a result of this communication may be stored in memory and, therefore, is likely to be called upon again when they tell about the experience a second time. After repeated iterations of this process, a number of copies of the more abstract, goal-directed representation are formed that are more accessible in memory than the original. As a result, their likelihood of recalling this original representation may be low. In some cases, this process could have psychological benefits. For example, a man who has experienced a traumatic event might have a vivid episode model of the experience that elicits considerable negative affect when it is activated. In communicating about the event to others, however, he presumably forms a more abstract representation of the event. A second description of the event may be based on this recently formed representation, leading a still more general representation to be constructed. As the representations that are formed in the course of repeated communications about the event become more abstract, fewer emotion-eliciting features may be retained in them, and so the man can ultimately recall and discuss the episode without re-experiencing the emotions that occurred earlier. On the other hand, the original episode model of the experience is retained in memory, and can potentially be reactivated by a set of retrieval cues that are unique to a configuration of the features contained in it. In such an event, the man might find himself re-experiencing the emotions elicited by the original events even after a period of time has elapsed. This process has more general implications for the development of individuals’ perceptions of themselves and others, and for the norms and values that guide their behavioral decisions. McAdams (1988) argued that the stories people tell about themselves provide the main source of their personal and social identity (see Mankowski & Rappaport, 1995, for a similar view). However, suppose people modify the content of the stories they tell in order to comply with certain expectations or desires of the recipient, and the representations they form in the course of telling these sto-

IMPACT OF GENERALIZED EVENT REPRESENTATIONS

145

ries are stored in memory. These stories may then become the basis for their self-perceptions. This means that the self-perceptions that people develop are likely to depend on the type of situations in which they are typically called upon to tell stories about themselves and the expectations that others have for the content of these stories. Two bodies of research and theory bear on these processes. One concerns the impact of parent–child interaction in the construction of self-stories. A second area of research provides insight into cultural and situational influences on individuals’ personal and social identity.

The Construction of Narratives in Parent–Child Interaction The role of narratives in the acquisition of social knowledge is particularly evident at early stages of social development. Much of the knowledge that children acquire about themselves derives from the stories they tell about themselves and others. The role of storytelling in early childhood has been investigated intensively by both Katherine Nelson (1989, 1993; Nelson & Greundel, 1981) and Peggy Miller (1994, 1995; Miller, Fung, & Mintz, 1996) in the context of parent–child interaction. Nelson’s research has been concerned with the storytelling process itself and its impact on the construction of social reality. Miller’s research is based on similar assumptions but has focused on cultural differences in the stories that are exchanged between parents and children and the consequent impact of these differences on the norms and values that children acquire. 1. Basic Processes Both Nelson and Miller note that children’s generalized representations of their personal experiences are often constructed in the course of interactions with their parents. Young children typically do not have a built-in mechanism for organizing their memories. Therefore, this organization is often acquired through parent–child interactions in which the parent stimulates the child to recount aspects of an experience by questioning the child and providing prompts that fill in gaps in the sequence of events that took place. In other words, the child and the parent “collaborate” in constructing the story the child tells about his or her experience. The co-narrative that emerges may be retained in memory and thus may constitute the child’s “reality.” In short, the child’s “memory” for the experience may be largely a construction of the parent, based on the parent’s perception of the experience, rather than a true reflection of the experience from the child’s own point of view.

146

CHAPTER 6

Certain consequences of this process could be insidious. As Schank and Abelson (1995) pointed out, children may learn in the course of collaborative storytelling that certain types of events should be recounted but others should not. To the extent these “edited” stories become part of the child’s “reality,” they may be hard to modify later. The extent to which this is true is often difficult to disentangle. For example, children may fail to include aspects of a parent’s abusive behavior when describing the interaction because of the fear of repercussions. To this extent, these instances of child abuse might later be difficult to retrieve. At the same time, as Ceci and Brucks (1993) pointed out, a prosecutor who wishes to make a conviction might harass a child into reconstructing an incident of child abuse that may in fact not have occurred, leading this “false memory” to become part of the story that the child “remembers” as part of the original experience. (Similar false memories may result from therapy; see Hyman & Loftus, 2002.) Although our description of these processes focuses on the construction of an abstract representation of a single event, prototypic event representations could be formed similarly. A parent, for example, may rehearse the behavior a child should perform in a given social context in order to teach the child “proper” behavior. Furthermore, this could be done without reference to specific instances that exemplify the behavior. The children’s perceptions of causality as well as of morality (what behavior are generally appropriate) could be learned through these processes. Miller’s work provides examples of this learning. 2. Cultural Differences in Storytelling Peggy Miller (1995; Miller et al., 1996; Miller, Sandel, Liang, & Fung, 2001) has analyzed cultural differences in self-perceptions in terms of differences in the type of collaborative storytelling that characterized parent–child interaction. In an analysis of storytelling involving both European American and Taiwanese families (Miller et al., 1996; Miller et al., 2000; Miller, Wiley, Fung, & Liang, 1997), she found that quite different attention was given to positive and negative behaviors in the stories that children were encouraged to construct and retain about themselves. These differences are particularly evident in stories about misdeeds. For example, Chinese parents and their children are more likely than their American counterparts to make the child’s present misbehavior a central part of the story the child tells, but to relate this behavior to past transgressions as well. Thus, the self-stories that are likely to emerge from these parent–child co-narrations were likely to be self-critical and to focus on the child’s responsibility for the misdeed. Furthermore, by relating the child’s current misdeed to earlier ones, the stories encouraged the development of a more general conceptualization of the child’s undesirable behavior that had implications for his or her moral character.

IMPACT OF GENERALIZED EVENT REPRESENTATIONS

147

However, American parent–child co-narratives were more often characterized by a “self-favorability” bias (Miller et al., 2001). These stories focused much less frequently on a child’s past transgression, and when they did, the transgression was treated humorously, portraying the child in a positive light despite his or her misdeed. Thus, for example, one story concerned a child who had written on the wall and lied about it. Although the mother acknowledged the seriousness of the child’s negative act, she at the same time made light of it in helping the child communicate about the act to others. Another example: Tommy and his older brother remembered being punished for some misdeed committed a few days earlier, but none of the participants—Tommy’s brother or mother—could remember what they had done wrong, a baffling if not inconceivable state of affairs from a Taiwanese perspective. This practice of down playing transgressions in the narrative medium seemed to be part of a wider set of practices that [European American] caregivers used to protect their children’s self esteem. (p. 168)

An equally or more important difference in Eastern and Asian story telling was evident in the stories that parents told about themselves. Taipai mothers’ self-stories typically provided positive examples for the child to emulate. As Miller et al. (2001) suggested: . . . Taipai mothers’ goal seems to be to encourage their children to live up to their own high standards by citing concrete examples of admirable behavior from their own or other family members’ lives . . . [T]eaching by example is one of several traditional Chinese socializing strategies . . . (p. 170)

In contrast, European-American mothers’ stories about themselves were more likely to preserve the children’s self-esteem by reassuring them that something they had done was not really all that bad, and that even admirable persons were prone to error. Therefore, their stories typically conveyed their own misdeeds as a way of indicating that they were also not perfect and to err is human, rather than setting themselves up as standards to emulate. In this regard, many European-American parents were prone to tell “hell rising” stories about times in their youth in which they had behaved in willed and irresponsible ways. In contrast, there was not a single instance of this type of story by Taiwanese parents. In summary, the stories that European-American parents encouraged children to tell about themselves were focused on positive events or, when they concerned misdeeds, tended to treat them as humorous rather than a potentially serious defect in the child’s general personality. Moreover, American parents’ stories about themselves were intended to suggest that they, like their children, were also not perfect. The implications of these

148

CHAPTER 6

stories, which emphasized positive outcomes while minimizing the seriousness of misdeeds, seem likely to induce a general “promotion focus” (Higgins, 1997), that is, a tendency to attach importance to positive outcomes with little concern about negative consequences of behavior. Unlike Americans, Chinese parents encouraged their children to focus their stories on their misdeeds and to acknowledge responsibility for them. Moreover, parents’ stories about themselves established a high standard of excellence that their children were encouraged to emulate. In combination, these socialization practices seem likely to induce a “prevention focus,” that is, a concern with negative consequences of behavior. These general differences seem likely to stimulate the development of different implicit theories about oneself as responsible for negative outcomes and, moreover, in the desire to avoid these outcomes. In fact, this difference exists between European and Asian adults, and can have a profound impact on judgments and behavior (cf. Aaker & Lee, 2001; Briley et al., 2000; see chap. 3). Cultural Influences on Socialization The narrative representations that individuals acquire about themselves and others are transmitted directly through parent–child interaction. As Miller et al.’s research testifies, the character of these individual narratives can vary over cultures, and also over communities and subcultures within a society. However, cultural differences can also exist in more general narrative representations that individuals construct. Mankowski and Rappaport (1995) noted that the norms and values that pervade societies are governed by a shared cultural narrative that not only serves to guide behavior in specific situations but influences individuals’ feelings of personal and social identity. As they indicated: . . . the phenomena of identity development and change may be understood in terms of the appropriation of shared narratives into one’s personal life story on the one hand, and the creation of new narratives or modification of existing narratives on the other. (p. 213)

For example, a narrative that pervades American culture is exemplified by the “Horatio Alger” myth of individuals who persevere and behave righteously in the face of poverty and adversity ultimately overcome their hardships and become affluent and respected citizens. This narrative implies a standard of morality and behavior that influences many Americans’ values and behavioral decisions. Stories with similar implications surround the original settlers in America (e.g., the “Mayflower” stories) and the settlement of the “Wild West.” (Interestingly, these stories have been modified to such an extent over the years that they typically bear little resemblance to the actual events that occurred; see Zinn, 1980.) The life stories of a reli-

IMPACT OF GENERALIZED EVENT REPRESENTATIONS

149

gious figure can also provide the basis for shared norms and values in societies in which the religion predominates. The suggestion that individuals in a society share a dominant cultural narrative does not imply that the entire narrative exists as such in the mind of a single individual. Moreover, the influence of such a narrative on the construction of self-narratives may not be completely conscious. Just as features of the content of children’s self-stories are modified to conform to the values of their parents or others to whom they are communicating, a cultural narrative may imply general norms and values that individual’s consider socially desirable and, therefore, influence the content of the stories that individuals construct about themselves, leading some things to be emphasized and others to be omitted. This may be done without a clear understanding of the roots of these norms and values. Moreover, the functional autonomy of cultural norms and values from their original content is well established in Boorstin’s (1973) sociological analysis of the development of American thought. (Thus, e.g., the conviction that individuals have the right to own firearms is a remnant of a period in the development of the American West in which law enforcement was either nonexistent or ineffective, and individuals were required to “take law into their own hands” in order to survive.) It seems reasonable to suppose that once self-narratives become tailored to cultural norms and values, they are later retrieved without a conscious consideration of the personal experiences on which they are based. Thus, they ultimately become regarded as “true” by the storytellers themselves. As a result, they become part of individuals’ self-identity (McAdams, 1988) that are told to one’s children and others independently of the original narrative representations of the events that were formed at the time they occurred.

CONCLUDING REMARKS The influence of generalized narrative representations on memory clearly depends on the nature of the representation and how it is used. When event prototypes of the sort referred to elsewhere as scripts (Shank & Abelson, 1977) or event schemas (Graesser, 1981) are brought to bear on the comprehension of verbal descriptions of events involving fictitious individuals, they can influence the content of the representations formed of these events and stored in memory. However, these prototypes are much less likely to come into play in comprehending events that people imagine themselves or familiar others experiencing. Generalized representations of single experiences that people form in the course of telling or writing about these experiences may have a more

150

CHAPTER 6

profound influence on the memory for these experiences. Moreover, to the extent that these representations are later retrieved and used as a basis for judgments of oneself and others, they may come to function as “reality,” and thus can have a pervasive influence on self-perceptions, personal norms, and values. We return to this possibility in chapter 10, where we discuss the impact of generalized narrative representations on judgments and behavior in some detail.

C H A P T E R

7 Pragmatic Influences on the Interpretation of Statements Made in a Social Context

Event and episode models are constructed spontaneously in the course of comprehending information about social events. These processes are automatic. Try not to understand the statement, “The boy kicked the ball.” For a native English speaker, this is very difficult. However, the processes involved in comprehending such a statement, described in chapter 4, are obviously not sufficient to extract meaning from the information one receives. This is particularly true when the information is conveyed verbally. Many statements make use of terms with which the recipient is unfamiliar. Others appear anomalous upon first consideration. For example, statements such as “The haystack was important because the cloth would rip” can only be assigned meaning in the context of a larger body of knowledge (e.g., a discussion of the hazards of parachuting; cf. Bransford et al., 1972). When such statements are encountered, they are theoretically transmitted by the Comprehender to the Work Space with an indication that they cannot be understood, thus stimulating Executor-monitored comprehension processes. Still other considerations arise when information is conveyed in a social context. A person’s comments to one another are often made with a particular goal in mind: to convey new information, to entertain, to persuade, to criticize, or to self-promote. Recipients’ interpretation of these comments can depend on their perception of the communicator’s objective. More generally, recipients must not only comprehend a statement’s literal meaning but also identify its pragmatic implications. These implications can concern not only the meaning that the communicator intends to transmit but also more general feelings, attitudes, and personality characteristics that the speaker does not always wish to reveal. 151

152

CHAPTER 7

The meaning that communicators intend to transmit and the meaning that recipients extract from their messages can often differ. For example, a person who comments, “George W. Bush’s foreign policy is certainly making America popular overseas” might intend to be sarcastic, and to disparage Bush’s policy and not to praise it. However, a recipient might not perceive this intention. Alternatively, a professor who describes his graduate assistant as “about as helpful as a well-trained chimpanzee” might intend to convey that the assistant’s competence leaves something to be desired. However, a listener might infer that the professor is rather nasty and an unpleasant person to work for. In short, the pragmatic implications of a communication do not always correspond to either the literal meaning of the message or the meaning that the communicator would like to transmit. A conceptualization of communication and comprehension in daily life situations must take into account both (a) the factors that influence communicators’ attempts to tailor their comments in a way that will attain their objectives and (b) the factors that govern recipients’ comprehension of these communications (see Wyer & Adaval, 2003). In this chapter, however, we focus primarily on the second set of factors. In doing so, we consider two questions. First, what factors stimulate a recipient to think about the pragmatic implications of a message rather than taking it literally? Second, how does a recipient identify the pragmatic implications of a communicator’s message, and how do these implications, once identified, affect the recipient’s reactions to it? As we will see, these reactions can be reflected not only in judgments that are made on the basis of a message but also the search for additional information that explains why the message was transmitted.

GENERAL CONSIDERATIONS People’s perceptions of the pragmatic implications of a message can often be stimulated by their own goals in the situation at hand. A woman who must decide whether to accept a date might want to know whether the invitation reflects a genuine desire for her company or only a desire to have sex. More generally, people who wish to form an impression of a person or product on the basis of a communicator’s description of the referent may need to assess the likelihood that this description is biased. They may attempt to identify the pragmatic implications of the communicator’s message in order to make this assessment. On the other hand, people often spontaneously recognize that a statement’s literal meaning is not the meaning the communicator actually intends to convey. This occurs when the statement’s literal meaning violates normative expectations for the content and style of communications that

PRAGMATIC INFLUENCES

153

are typically exchanged in the situation at hand. As Grice (1975) first pointed out and others have since elaborated (e.g., Green, 1989; Higgins, 1981; Sperber & Wilson, 1986), social communications are governed by a set of implicit principles that are applied by both the communicator in generating a message and the recipient in construing its implications. In informal situations, for example, speakers are normally expected to be 1. informative—to convey new knowledge that the recipient does not already have and is likely to find interesting; 2. truthful—to convey correct information as the communicator sees it; 3. relevant—to say things that are related to the topic under discussion; 4. polite—to avoid offending the individuals to whom one is communicating; and 5. modest—to avoid unduly promoting oneself. These normative principles are obviously not universal. A modesty norm, for example, is applied less often in job interviews than in informal conversations, and truthfulness is less likely to guide exchanges between a customer and a used car salesman than a student’s discussions with a professor. Moreover, certain types of situations may activate specific expectations for communicators to ingratiate, to try to persuade, or to be amusing.12 It seems reasonable to suppose that when people receive a communication whose literal implications appear to deviate from expectancies for the content and style of messages that are typically conveyed in the situation in which they occur, they attempt to reconcile this discrepancy. In doing so, they often reinterpret the message, or the communicator’s purpose in transmitting it, in a way that is more consistent with their expectations. This, in turn, may require the recipients to consider the communication in relation to other available information in the situation at hand. As a result, they may have better recall of this information later, and may be more inclined to use it as a basis for judgment. The assumption that people spontaneously recognize that a communication violates a normative principle is based in part on Postulate 4.6. According to this postulate, many statements are spontaneously recognized as either redundant with prior knowledge (and thus in potential violation of a norm to be informative) or false (and thus in potential violation of a norm 12

The expectations for communications in a social situation can pertain not only to the content of the messages that are exchanged but the style in which they are transmitted. The extent to which parties to a conversation elaborate answers to one another’s questions, or ask questions in return, may influence perceptions of the communicators’ attitudes toward both the conversation and one another (Wyer, Swan, & Gruenfeld, 1995). Nonverbal behaviors can obviously play a role as well, as we note later in this chapter.

12

154

CHAPTER 7

to convey valid information) in the course of comprehending them. In the first two sections of this chapter, we discuss the identification of these two types of norm violations in some detail and describe research bearing on responses to these violations. The cognitive dynamics that underlie the recognition of other norm violations are discussed in later sections. Before embarking on this discussion, however, the basic processes we assume to underlie the identification are worth reviewing. As we noted in chapter 4, situation models are likely to be formed of the states or events described in a verbal communication. In addition, features of the situational context in which a communication is made can activate representations of previously encountered situations that share these features. This configuration of common features may be responded to schematically, thus constituting a situational appraisal of the sort suggested by Lazarus (1982, 1991). The situation-specific representations that give rise to this appraisal, however, contain other features that are not common to all of these representations but nevertheless fall within a circumscribed range of values. These features, which could pertain to characteristics of statements and behaviors that have occurred in the situation, give rise to expectations for the features that are likely to exist in new situations. Thus, suppose a specific communication (statement or verbal utterance) is encountered in a given situation. The Comprehender presumably constructs a representation of the statement’s literal meaning with reference to previously formed knowledge representations and computes an estimate of it similarity to these representations (S; see chap. 4). If this similarity is either very high (S > TT ; see Postulate 4.6) or very low (S < TF), the Comprehender appends a “tag” that denotes the statement as true (i.e., as redundant with prior knowledge) or false (i.e., inconsistent with prior knowledge) and transmits it to the Work Space along with a representation of the message itself. When the Executor encounters a tag that denotes the statement as true or false, it instructs a special-purpose processing unit (e.g., the Encoder/Organizer) to compare the statement’s similarity index (S) to the range of values that are specified in the situational appraisal and, if it falls within this range, accepts the statement’s literal meaning. If, however, the similarity index falls outside this range, the Executor directs the Encoder/Organizer to construe other interpretations of the statement that are more compatible with situation-based expectations. Stripped of the cumbersome terminology imposed by the theory, this conceptualization simply means that people may spontaneously recognize that a statement violates normative principles that govern communications of the sort that are usually transmitted in the situation at hand. When this occurs, they attempt to determine the reason for this violation. In doing so, they may reinterpret the statement, or the purpose for which it was con-

PRAGMATIC INFLUENCES

155

veyed, in a way that is more consistent with the principle that is violated— that is, whether the statement is interpreted as redundant with prior knowledge or is interpreted as false. We consider these two possibilities in turn.

RESPONSES TO UNINFORMATIVE MESSAGES Suppose the representation of a statement that is transmitted to the Work Space is tagged as “true” (Postulate 4.6) and, therefore, as redundant with prior knowledge. If the representation is inconsistent with expectations for communications that are encountered in the situation at hand, it may stimulate attempts to explain its occurrence. This may involve two steps, as shown in the right half of Fig. 7.1: 1. Recipients consider why the source of the statement might consider it to be informative. For example, they might speculate that the source is unaware of their knowledge. Or, they might infer that the source knows of a reason why the statement might not be true, thus making a reaffirmation of its validity necessary. 2. If recipients conclude that the statement cannot possibly be false, they may conjecture that it is not intended to convey information about its referent per se, but rather, is to convey the source’s implicit attitude toward the event or state described in the statement. Thus, for example, suppose a man hears someone comment, “George W. Bush is not a card-carrying member of the Ku Klux Klan.” He might never before have considered this possibility, and might normally regard the statement as going without saying. Therefore, if the statement is made under conditions in which communications are expected to be informative, the recipient might speculate that Bush might actually have been accused of belonging to the Klan, thus making a denial of his membership informative. This inference, however, could plant a seed of doubt in the recipient’s mind about Mr. Bush’s status that did not exist before the denial was made. To this extent, the denial might actually increase the recipient’s belief that Bush belongs to the organization or, at least is in sympathy with the organization’s principles. On the other hand, suppose people overhear the comment “Americans are not allowed to vote until they reach the age of 18.” To Americans, this is common knowledge and, furthermore, could not possibly be false. Consequently, recipients who consider the statement to be uninformative are unlikely to speculate that it might actually be invalid. Instead, they are likely to infer that the communicator is intending to convey an implicit attitude toward the state of affairs described (“Americans are not allowed to vote

156

FIG. 7.1. Theoretical flow diagram of processing steps involved in identifying the implications of statements that are recognized as true or false. The source and recipient of the statement are denoted S and R, respectively.

PRAGMATIC INFLUENCES

157

until they reach the age of 18, and this is certainly a good thing,” or “Americans are not allowed to vote until they reach the age of 18, whereas citizens of other countries can vote much earlier.”). To this extent, recipients would not change their belief in the validity of the statement, but instead might change their attitude toward the state of affairs that the statement describes. Note, however, that these effects should only occur under conditions in which uninformative communications are counternormative. When the messages are not intended to convey new information, they should theoretically be taken at face value, and attempts to make them informative should not be evident. Effects of Uninformative Statements on Recipients’ Beliefs Gruenfeld and Wyer (1992) investigated implications of the preceding analysis. Participants read a series of statements about social events. In some cases, the statements asserted the validity of target propositions that college students were likely on a priori grounds to believe to be false (e.g., “The CIA is engaged in illegal drug trafficking,” “Lyndon Johnson was responsible for the assassination of President John F. Kennedy,” etc.). In other cases, however, the statements denied the validity of these propositions (e.g., “The CIA is not engaged in illicit drug trafficking,” etc.) and, therefore, were uninformative. Finally, some participants believed that that the statements were taken from newspapers (whose purpose is to convey new information), whereas others believed they were taken from an encyclopedia (whose purpose is to preserve archival knowledge). Participants first judged the importance of the information conveyed in the statements. Then, after doing so, they estimated their beliefs in both the target propositions and related ones (e.g., “The CIA engages in other illegal activities,” “Lyndon Johnson was an enemy of John F. Kennedy,” etc.). Statements that affirmed the validity of the propositions constituted new information, whereas statements that denied their validity were redundant with participants’ prior knowledge about the persons and events in question. However, we expected that when participants were told that the statements came from encyclopedias, they would accept the statements without further consideration. In contrast, participants who were told that the statements came from newspapers should assume that they were intended to be informative and, therefore, might infer that there was some reason to suppose that the propositions being denied were in fact true. To this extent, they might actually increase their belief in these propositions. This was in fact the case. Table 7.1 shows the difference between beliefs in propositions under each experimental condition (reported along a 0–10 scale) and beliefs reported by control participants who were not exposed

158

CHAPTER 7 TABLE 7.1 Effects of Assertions on Beliefs in Target Propositions and Related Propositions Assertion Type

Effect on beliefs in target proposition Newspaper source Encyclopedia source Effect on beliefs in related propositions Newspaper source Encyclopedia source

Affirmation

Denial

1.16* 1.81*

1.06* 0.48

1.15* 1.62*

1.20* 0.91*

Note. Cell entries refer to differences between beliefs reported after exposure to an assertion and context-free beliefs reported in the absence of the assertion. Differences denoted by asterisks are significantly greater than 0, p < .05. Adapted from Gruenfeld and Wyer (1992).

to either denials or affirmations. Affirmations that were taken from encyclopedias increased participants’ beliefs that target propositions were true, whereas denials had little effect. These effects confirm the assumption that affirmations constituted new information but denials did not. News headlines that affirmed the validity of the target propositions also increased beliefs that the propositions were true. (This increase was somewhat less than it was when the statements came from encyclopedias, suggesting that participants perceived newspapers to be relatively less credible.) On the other hand, denials of the propositions’ validity that were taken from newspapers also increased participants’ beliefs that the propositions were true. In fact, this increase was just as great as the increase produced by affirmations! Thus, participants attempted to interpret denials in a way that made them informative, leading them to infer that there might be some reason to suppose that the propositions were true, and this led the statements to have a boomerang effect. Effects of Uninformative Statements on Attitudes As noted earlier, however, some uninformative assertions cannot possibly be false. In this case, participants’ attempts to make them informative could lead them to infer that they are implicit expressions of opinion. To this extent, the assertions might influence participants’ estimates of the desirability of the state of affairs being denied by the proposition rather than their beliefs in their validity. Gruenfeld and Wyer (1992) reported evidence of this as well. For example, exposure to the news headline, “Wearing seatbelts is required by law in Illinois” significantly decreased participants’ opinion that wearing seatbelts should be required. Similarly, exposure to the state-

PRAGMATIC INFLUENCES

159

ment, “Many food products are sold in plastic and styrofoam containers” decreased subjects’ beliefs that there should be restrictions on the packaging of food product, and exposure to “Not all college students drive American automobiles” increased beliefs that college students should drive these automobiles. General Implications 1. Communication and Persuasion The conditions constructed by Gruenfeld and Wyer (1992) provide examples of innuendo effects of the sort that are used intentionally by propagandists. Their use is particularly common in the political arena. For example, a candidate might attempt to plant a seed of doubt about an opponent’s integrity, not by asserting that the opponent has engaged in illegal activities but rather, by actually denying it. For example, the assertions “I personally have no evidence that my opponent embezzled funds while serving as an executive of General Motors,” or “I do not for one minute believe that my opponent is a card-carrying member of the Ku Klux Klan,” which might normally go without saying, could stimulate people to suppose that the opponent if not an embezzler or a Klan member, is nonetheless likely to have engaged in activities that give rise to these suspicions. A perhaps more familiar instance of the intentional application of the informativeness principle occurs in advertising. Suppose an advertisement for Brand X asserts, “X contains no hydratropine.” Consumers may have no idea what hydratropine is. In an attempt to make the statement informative, however, they may infer that there is some reason why promoters of Brand X made the statement. In particular, they may infer that (a) hydratropine is undesirable and (b) other products have it. Correspondingly, the equally meaningless assertion, “X contains hydratropine” might stimulate them to infer that hydratropine is desirable and that brands other than X do not have it. 2. Communication of Emotions Similar effects occur in interpersonal communication. For example, a woman might comment to her spouse, “It’s cold in here.” If she has rarely complained about the room temperature in the past, her husband may regard this as new information and take steps to remedy the situation. However, suppose the wife’s remark is one of many similar comments she has made over a period of time. In this case, the husband would consider the statement to convey little new information about his wife’s feelings. Rather, he might infer that the statement is a veiled criticism of his compulsion to save energy by keeping the thermostat at 65°F. As this example points out,

160

CHAPTER 7

the recognition of a statement as uninformative, and the construal of its intended meaning, often requires a substantial amount of shared knowledge. Watzlawick, Beavin, and Jackson (1967) noted that people in close relationships often acquire a private meaning system that permits them to convey thoughts and feelings of which other listeners are unaware. This shared knowledge presumably has its impact through its mediating influence on perceptions of one another’s statements as uninformative and the reinterpretation of their literal meaning that results from it.

REACTIONS TO COMMUNICATIONS THAT ONE BELIEVES TO BE UNTRUE Statements that are transmitted to the Work Space by the Comprehender may often contain tags that identify them as false or, at least, inconsistent with the implications of prior knowledge (Postulate 4.6). These statements, like statements that are redundant with prior knowledge, are likely to stimulate cognitive activity in an attempt to understand why they were made. These processes may be similar to those that underlie responses to violations of the informativeness principle. However, the inferences that result from the activity are somewhat different. The processes that seem likely to underlie reactions to untrue communications can be summarized with reference to the left half of Fig. 7.1. Specifically, individuals who hear a statement they recognize as false (S < TF , as defined in Postulate 4.6) are likely to assess whether the communicator knows that this is the case. If the source is likely to believe it is true, the recipients may take the statement’s literal implications at face value. They might also do so if they think that the source knows the statement is false but is unaware that they share this knowledge. In this case, they might conclude that the source is trying to deceive them. However, suppose recipients believe that the source is aware of their knowledge. Then, they are likely to infer that the source does not mean the statement to be taken literally, but rather, intends it to be ironic. Reactions to ironic statements can depend on both their topic and whether their literal implications are favorable or unfavorable. In some cases, ironic statements are intended to be funny. In other cases, however, they can be indirect expressions of hostility. These motives are sometimes difficult to distinguish. Suppose a woman whose husband has had three traffic accidents in the past 2 years comments to her guests, “My husband is a wonderful driver.” The comment could be a true expression of opinion, and an attempt to defend her spouse’s abilities despite his poor driving record. To the extent it is intended to be sarcastic, however, it could either convey hostility or be meant as a tease. Which interpretation the guests

PRAGMATIC INFLUENCES

161

make is likely to depend on their view of the woman’s more general feelings toward her spouse. If they believe that the couple’s relationship is basically harmonious, they are likely to infer that the woman is teasing. If the couple is known to have a history of marital conflict, however, they might interpret the statement as hostile. These observations are rather self-evident. The conceptualization outlined in Fig. 7.1b has greater interest, however, when a statement’s validity is clearly a matter of opinion. Suppose a person comments, “What this country needs is another Ronald Reagan.” This comment is more likely to be ironic if it is made by a liberal Democrat than if it is made by a conservative Republican. On the other hand, conservative Republicans are less likely than liberal Democrats to interpret the statement as ironic. That is, they are more likely to believe that the statement is true, and to accept its literal meaning as the intended meaning regardless of its source. Accuracy Violations and Amusement: The Identification of Witticisms To reiterate, people often interpret a statement spontaneously as false in the course of comprehending it (Postulate 4.6). When it is obvious to recipients that the source also believes the statement to be false and, therefore interpret the statement as ironic, they are often amused. This was demonstrated empirically by Isbell, Wyer, and Collins (2002). Participants were exposed to a number of vignettes in which one person made a comment to another. In some cases, the comment was likely to be true. In other cases, the comment was literally false, but if interpreted as ironic, was similar in meaning to the first statement. Finally, in some cases, the intended meaning of the statement was more favorable than its literal meaning, and in other cases, it was less so. For example, one vignette described a discussion between two students about where they had spent the summer. In some versions, one student made a true statement that conveyed either a favorable description of its referent (“Switzerland is certainly a great place to spend the summer—all those lakes and high mountains”) or an unfavorable description (“Champaign, Illinois, is certainly an awful place to spend the summer—all that corn and high humidity”). In other cases, however, the student’s statement was literally false (“Switzerland is certainly an awful place to spend the summer . . . ,” and “Champaign, Illinois is certainly a great place to spend the summer . . .”), but its intended meaning was similar to that of the corresponding true statement. In another scenario, two professors discussed who was the most (or least) intelligent public figure in the 20th century. In this case, the literally true statements were “Einstein gets my vote (for most intelligent)” and “Dan Quayle gets my vote (for least intelligent)” and the corresponding false statements were “Einstein gets my

162

CHAPTER 7

vote (for least intelligent)” and “Dan Quayle gets my vote (for most intelligent).” Participants read one version of each vignette with instructions to estimate how amused they would be if they overheard the statement being made under the conditions described. Pooled over vignettes, participants regarded ironic statements as more amusing than literal ones (4.9 vs. 2.7), and this was true regardless of whether the intended meaning of the statements was favorable (4.6 vs. 1.9) or unfavorable (6.3 vs. 3.5). It would obviously be incorrect to conclude that all ironic statements are amusing. Sarcasm can often not be amusing at all. A person who has just been turned in by a classmate for cheating on an exam might respond to the other, “You are certainly a fine friend,” which is intended to convey quite the opposite. This comment is unlikely to elicit amusement. The necessary conditions for humor elicitation are discussed in chapter 8 (see also Apter, 1982; Wyer & Collins, 1992). Nevertheless, the evidence that ironic statements are more likely to elicit amusement than their true counterparts is worth noting. Effects of Invalid Communications on Information Search The sources of the ironic statements presented in Isbell and Wyer’s study obviously believed that their statements were false, and so recipients were likely to believe that the statements were intended to amuse. In many instances, however, people are uncertain of a source’s beliefs in the statements that he or she makes. This is particularly true when the statements express opinions that are somewhat controversial. People who encounter statements of opinion that they personally believe to be invalid may often attempt to seek additional information that helps them to understand why the statements were made. In addition, they may review their prior knowledge about the issues to which the statements pertain in order to confirm their belief that the opinion expressed is incorrect. This cognitive activity is likely to be reflected in recipients’ ability to recall the information later. Information that is thought about more extensively with reference to prior knowledge is generally easier to recall (Craik & Lockhart, 1972; Wyer & Hartwick, 1980). If this is so, however, people should be more likely to recall opinion statements they spontaneously recognize as false than either statements they consider to be true or statements whose validity they have not spontaneously evaluated. This possibility has not been investigated under conditions in which information is conveyed in conversations. If the proposed conceptualization is valid, however, the cognitive activities that underlie people’s recall of opinion statements they recognize as false should occur even when the

PRAGMATIC INFLUENCES

163

statements are conveyed out of their social context. Several studies bear on this possibility. In a study of impression formation reported by Wyer, Lambert, Budesheim, and Gruenfeld (1992), participants received information about a target person who was characterized as either a conservative Republican or a liberal Democrat and, in addition, was described by either favorable or unfavorable trait adjectives (e.g., “kind,” “hostile,” etc.). This general description was followed by a list of both (a) behaviors that were both evaluatively consistent and evaluatively inconsistent with the trait descriptions (e.g., “invited a charity collector for lunch,” “shouted and honked at slow drivers,” etc.), and (b) opinion statements that reflected either a liberal or conservative political orientation. Participants after receiving the information were asked to recall the information they had read. Participants were more likely to recall behaviors that were consistent with the target person’s initial trait description than behaviors that were inconsistent with it, replicating the results of many other studies (Hastie & Kumar, 1979; for a review, see Srull & Wyer, 1989). However, they were more likely to recall opinions with which they disagreed than opinions with which they agreed, and this was true regardless of the consistency of these opinions with the target’s political ideology. Thus, when participants encountered opinion statements about matters with which they were familiar, they comprehended these statements with reference to their previously acquired knowledge, and devoted a particular amount of time trying to understand the basis for opinions they regarded as incorrect. Therefore, they had better recall of these opinions regardless of the opinions’ consistency with the general attitudes of the individual who expressed them.13 1. Individual Differences in the Processing of Invalid Information People obviously differ in the amount of cognitive effort they perform in an attempt to reconcile information they consider to be invalid. Wänke and Wyer (1996) speculated that differences in this motivation might vary with participants’ own social and political ideology. Participants with a liberal social and political orientation are inclined to consider the underlying causes of a person’s behavior when evaluating a person (Fletcher, Danilovics, Fernandez, Peterson, & Reeder, 1986). Conservatives, on the other hand, are more disposed to evaluate a person’s behavior in terms of its 13

Wänke and Wyer (1994) found that participants had better recall of behaviors that were ideologically inconsistent with the liberal or conservative orientation of the actor. However, this occurred only when the actor’s ideology was different from participants’ own (i.e., when the actor was an “outgroup” member). When the actor’s ideology was the same as participants’ (i.e., the target was an “ingroup” member), participants’ recall of the target’s behavior was similar regardless of whether it was inconsistent or consistent with his ideology (.42 vs. .48, respectively).

13

164

CHAPTER 7

consequences. Note that this difference corresponds to different levels of moral development, as defined by Kohlberg (1976). (For other evidence that liberals and conservatives differ in this regard, see Emler, Renwick, & Malone, 1983; Fishkin, Keniston, & MacKinnon, 1973; Nassi, Abramowitz, & Youmans, 1983.) Based on this evidence, we expected that liberals would be more inclined than conservatives to consider the attitudes that underlie a person’s behavior when forming an impression of the person. On the other hand, conservatives should be more disposed to evaluate behaviors in terms of their desirability without considering the attitudes and opinions they reflect. Furthermore, these individual differences should be particularly evident under conditions that emphasize the use of these different judgmental criteria. To evaluate this possibility, participants who had been identified as having a liberal or conservative political orientation were exposed to information about a target person who was initially described as either liberal or conservative and who had ostensibly engaged in a number of behaviors that were normatively either liberal (e.g., “organized a sit-in at a nuclear power plant,” “raised money for a public housing project,” etc.) or conservative (“tried to prevent a theatre from showing an antireligious movie,” “donated $1,000 to the National Rifle Association,” etc.). In behavior-focus conditions, participants were told that we were specifically interested in how people form impressions of others on the basis of their behavior. In opinion-focus conditions, they were told that we were interested in how people form impressions of persons on the basis of their opinions as reflected in their behavior. Participants after receiving the information and evaluating the target person were asked to recall the behaviors they had read and finally, were asked to evaluate the desirability of the behaviors that were described in the information they had received. Table 7.2 shows the proportion of behaviors that liberals and conservatives recalled as a function of their perceptions of the behaviors’ desirability and the criterion they were asked to use in forming their impressions. When liberals were told to base their impressions on the target person’s opinions, they recalled a greater proportion of behaviors they considered to be undesirable than behaviors they considered to be desirable (.64 vs. .37, respectively). This was also true of conservatives who were told to base their judgments on behaviors (.66 vs. .35, respectively). However, this difference was not apparent when liberals were told to focus on the target’s behaviors per se, or when conservatives were told to consider the opinions underlying these behaviors. In short, participants only thought more extensively about behaviors whose implications were considered undesirable when they were told to employ the criterion that they were likely to use spontaneously to evaluate these behaviors. When they were

165

PRAGMATIC INFLUENCES TABLE 7.2 Mean Proportion of Behaviors Recalled as a Function of Perceived Desirability, Participants’ Ideology, and Instructional Conditions Opinion-Focus Conditions

Desirable behaviors Undesirable behaviors Difference

Behavior-Focus Conditions

Liberal Participants

Conservative Participants

Liberal Participants

Conservative Participants

.37 .64 –.27

.39 .33 .06

.46 .54 –.08

.35 .66 –.31

Note. Adapted from Wänke and Wyer (1996).

told to use a criterion that differed from the one they typically employed, this spontaneous cognitive activity was disruptive, and so recall differences were not evident.

PRAGMATIC INFLUENCES ON IMPRESSION FORMATION IN INFORMAL CONVERSATIONS The pragmatic implications of communications are particularly likely to come into play when the communications are exchanged in informal conversations. People may spontaneously identify the pragmatic implications of a person’s comments when they perceive the comments to be either uninformative or untrue. However, their identification of these implications can also be motivated by more specific goals they have at the time they engage in the conversation. People who observe or take part in a social interaction are often disposed to form general impressions of the interaction participants. In some cases, they are likely to form these impressions spontaneously, without any more specific goal in mind other than to get to know the individuals better. In other cases, they might do so intentionally in order to interpret the individuals’ comments about the topic at hand. For example, a person’s disparaging remarks about another could convey valid information about the individual being described. On the other hand, they could also indicate that the speaker is an unpleasant individual who criticizes others indiscriminately regardless of their merits. Similarly, a person who expresses a favorable opinion of a new television program either could have discriminating tastes or could like almost anything, including mindless situation comedies that most people would find dull and offensive. In these circumstances, listeners are likely to assess a speaker’s general personality characteristics in order to construe the implications of his or her remarks. These possibilities have largely been ignored in much of the research on the cognitive underpinnings of person impressions (for reviews, see Ander-

166

CHAPTER 7

son, 1981; Carlston & Smith, 1996; Wyer & Carlston, 1994). In this research, participants are usually asked to form an impression of someone on the basis of trait and behavior descriptions that are conveyed out of their social context. In doing so, they are expected to take the information at face value, independently of its source. To this extent, the implications of this research for person impression formation outside the laboratory may be limited. The research described in the remainder of this chapter calls attention to several of these limitations. To provide a perspective on this research, we first review briefly the conclusions that have been drawn in the typical research paradigm in which person impression formation has been investigated. We then describe the results of a number of studies in which similar information is conveyed in the context of informal conversations under conditions in which the pragmatic implications of the information is taken into account. Finally, we describe the role of nonverbal and paralinguistic features of the communications exchanged in a conversation that influence recipients’ perceptions of their implications and, therefore, the impressions they form of both the topic being discussed and the persons who are discussing it. Impression Formation in Nonsocial Contexts Extensive research has been conducted on the impressions that people form of someone on the basis of information that describes the person’s traits and behaviors, and rigorous theories have been proposed to describe the processes that underlie these impressions (Carlston & Smith, 1996; Wyer & Srull, 1989). In a typical study (cf. Srull, 1981; Wyer, Bodenhausen, & Srull, 1984; for a review, see Srull & Wyer, 1989), participants are asked to form an impression of a target person on the basis of a favorable or unfavorable trait description followed by a number of behaviors. Some behaviors are descriptively and evaluatively consistent with the initial trait description of the person and others are inconsistent with this description. The target person is typically fictitious, and the trait description is conveyed in a way that suggests that its validity is not open to question. (For example, the person might ostensibly be a character in a novel, and the trait adjectives might have been used by the author to describe him; see Wyer & Martin, 1986.) The mental representation that is formed under these conditions, which presumably exemplifies a generalized person representation of the sort described in chapter 1, has been conceptualized using an associative network metaphor (Hastie & Kumar, 1979; Srull & Wyer, 1989). That is, the concepts formed from information are represented by nodes, and associations between them (formed by thinking about the concepts in relation to one another) are represented by pathways. Stronger associa-

PRAGMATIC INFLUENCES

167

tions, resulting from more extensive thought, are designated by wider paths. The accumulated evidence (Srull & Wyer, 1989) suggests the following process.14 1. People who receive an initial trait description of a target person form an evaluative concept of him on the basis of this description. If they later learn about specific behaviors the target has performed, they encode the behaviors as either favorable or unfavorable and think about their implications for the concept they formed of him earlier. This establishes associations between the behaviors and this concept. Thus, suppose a person is described by a set of favorable traits, followed by the following sequence of favorable and unfavorable behaviors: b+, b+, b-, b+, b-, b-, b+. Then, a favorable concept of the person, P+, is formed on the basis of the traits, and the behaviors become associated with it as shown in Fig. 7.2. 2. When a behavior is encountered that is evaluatively inconsistent with the central person concept, people attempt to understand why the behavior occurred. In doing so, they think about the behavior with reference to other behaviors the person has performed, establishing associations between the inconsistent behavior and others. Assuming that each inconsistent behavior (b-) in our example is thought about with reference to the two behaviors that precede it in the sequence, the interbehavior associations would be those shown in Fig. 7.2. 3. In addition, people who encounter behaviors that are inconsistent with their general concept of the target attempt to bolster their confidence in the validity of this concept. They do this by reviewing behaviors that are consistent with it, strengthening the association of these behaviors with the concept, as also shown in the figure. Numerous studies provide support for this conceptualization and its implications. For example, note that as a result of people’s attempts to reconcile the occurrence of a behavior that is inconsistent with their initial impression of the person, the behavior becomes more interconnected to other behaviors than the consistent behaviors are. To this extent, the behaviors may be more easily recalled. Bolstering, however, strengthens the 14 Srull and Wyer (1989; see also Gordon & Wyer, 1987) postulated that in addition to the processes described here, people interpret the behaviors they read about in terms of the traits they exemplify, leading trait-behavior clusters to be formed and stored separately from the more general person representation. The role of these clusters is of considerable importance in conceptualizing a number of phenomena, including the mental representation information about commercial products (Park & Wyer, 1993) and the impact of information one is told to disregard (Wyer & Budesheim, 1987; Wyer & Unverzagt, 1985). However, these representations are not central to the issues of concern in this chapter, so they are not described in detail.

14

168

CHAPTER 7

FIG. 7.2. Metaphorical representation formed of a target person described by a set of favorable traits and a series of behaviors that are evaluatively consistent (b+) or inconsistent (b-) with the person concept formed from these traits. Thicker lines denote stronger associations between the elements involved. Each inconsistent behavior is thought about in relation to the two behaviors that precede it in the sequence.

association of evaluatively consistent behaviors with the central concept, making these behaviors more accessible in memory. Thus, the influence of inconsistency resolution and bolstering should be reflected by the ease of recalling of inconsistent and consistent behaviors, respectively. Many studies (e.g., Hastie & Kumar, 1979; Srull, 1981; Srull, Lichtenstein, & Rothbart, 1985; Wyer, Bodenhausen, & Srull, 1984) show that inconsistent behaviors are recalled better than consistent ones, suggesting that inconsistency resolution takes priority over bolstering. (See Abelson, 1959, for a similar assumption in conceptualizing the processes of resolving inconsistencies among beliefs and attitudes.) However, when people are given more time to think about the information they receive, either as it is presented or subsequently, bolstering may predominate, leading consistent behavior to gain a recall advantage (Wyer, Budesheim, Lambert, & Martin, 1989; Wyer & Srull, 1989; see also Wyer & Martin, 1986). Effects of Social Context on Impression Formation Processes The person memory model developed by Srull and Wyer (1989) has been applied successfully in conceptualizing numerous social phenomena, including the impact of information that one is told to disregard (Wyer &

PRAGMATIC INFLUENCES

169

Budesheim, 1987; Wyer & Unverzagt, 1985), the combined influence of stereotypes and idiosyncratic trait descriptions on person impressions (Wyer & Martin, 1986), and the different impressions that are formed of persons and groups (Srull, 1981; Wyer & Gordon, 1982). Nevertheless, the processes implied by the theory are restricted to conditions in which information is presented out of its social context. Quite different considerations arise when a person’s traits and behaviors are described in the context of a conversation. For one thing, an acquaintance’s trait descriptions of someone in the course of a conversation are matters of opinion and, as such, may not be valid. Consequently, listeners may attempt to establish the validity of these descriptions. To do this, they may construe the implications of the descriptions for characteristics of the speakers who provide them as well as for the individual being described. Wyer, Budesheim, and Lambert (1990) examined this possibility. In their studies, the trait and behavior descriptions that participants received were very similar to those presented in more traditional impression formation research (e.g., Wyer, Bodenhausen, & Srull, 1984). However, the information ostensibly pertained to an actual person and was provided by acquaintances of the person in the course of a conversation. Specifically, participants were told they would listen to a tape-recorded conversation between a male and a female student about a mutual acquaintance. The tape began with instructions to the two speakers to select the person they wanted to talk about and then, after doing so, to write down three general attributes they would use to describe this person. At this point, the tape stopped, and the listeners were given photocopies of the trait adjectives the speakers had had ostensibly written down. The favorableness of each speaker’s trait description of the target and the dimension to which this description pertained (intelligence or kindness) were varied independently. After participants had read the target descriptions, the taped conversation was restarted, and participants heard the experimenter tell the speakers to reminisce about things the target person had done that they had either seen or heard about. The speakers then proceeded to exchange anecdotes, interspersed with occasional pauses and promptings by the experimenter to make the conversation seem natural. Over the course of the conversation, each speaker mentioned 13 behaviors of which 6 were favorable (either kind or intelligent), 6 were unfavorable (unkind or unintelligent) and 1 (occurring near the beginning of the conversation) was neutral. These behaviors were very similar to those presented in studies by Wyer et al. (1984) and Wyer and Martin (1986), but were conveyed as they might be mentioned in a normal conversation. An excerpt of one conversation, about a target named Don, was as follows (for a complete transcript, see Wyer et al. (1990): Experimenter: Okay, then, why don’t we begin. Who wants to start off?

170

CHAPTER 7

M: Well, let’s see . . . well, he walks a mile to school each day, even when it’s freezing. I guess I thought of that because it’s so damn cold today. I asked him if he wanted a ride, but he said he didn’t. F: Yeah, he never seems to get sick. It must be because he works out 3 or 4 times a week—he’s in pretty good shape. M: [Laughs] Better than me, anyway. . . . I remember he won the university chess championship last fall. F: [Laughs] Yeah. . . . I remember we went out to dinner that night to celebrate and Don swore at the waiter at Eddies because he didn’t get served as quickly as he wanted. M: Yeah, I can just see him doing that . . . let’s see . . . well, he can’t seem to follow simple directions people give him. Remember that time last week at Jerry’s? F: Yeah, [Laughs] . . . Oh, by the way, do you know Jerry lost his job at the Union? M: No, really? Gee, that’s rough. F: Yeah. I thought of that ’cause he told me Don had loaned him 20 bucks for a date he had and stuff. M: [Pause] Jeez, this is a lot harder than I thought. [To experimenter] You want specific behaviors, right? Like I know he’s always smiling and saying hello to people when I walk down the street. But one particular instance doesn’t stand out. Experimenter: Well, try to be as specific as you can. M: [Pause] Hmm . . . well, let’s see . . . well, someone said he’d flunked a mechanics training course I took last summer—I don’t know why he took it, anyway. The other tape was identical except that the roles of the male and female speakers were reversed. We anticipated that people who listened to the conversation would spontaneously form impressions of the speakers as well as the target himself. To understand the nature of these different impressions, we employed two instructional conditions. In speaker-impression conditions, participants were explicitly told to form impressions of the speakers, and in target-impression conditions, they were told to form an impression of the person the speakers were discussing. In both cases, however, participants after listening to the tape reported their liking for the target person along a -5 to +5 scale. In addition, they reported their perception of how well each of the speakers liked the target and rated their own liking for the speakers. Finally, they recalled the behaviors they had heard the speakers mention.

PRAGMATIC INFLUENCES

1.

171

Speaker-Impression Conditions

Judgment Data. Participants evaluated each speaker more favorably when the speaker’s trait descriptions of the target were favorable than when they were unfavorable. Moreover, they inferred that the speakers liked the target to a greater extent in the former conditions. However, the favorableness of the speakers’ trait descriptions had contrast effects on participants’ own liking for the target. That is, partners liked the target less when both speakers’ trait descriptions were favorable than when both descriptions were unfavorable. In combination, these results suggest that participants based their impressions of the speakers on the speakers’ trait description of the target and then used these impressions as standards of comparison in judging their liking for the target himself, independently of the implications of the trait descriptions for his attractiveness. Recall Data. The recall data were particularly provocative. The proportion of behaviors recalled are shown in the top section of Table 7.3 as a function of the speaker who mentioned the behaviors and the consistency of the behaviors with each speaker’s trait description. When the two speakers’ descriptions of the target were evaluatively similar, participants had better recall of behaviors that were inconsistent with these descriptions (M = .444) than behaviors that were consistent with them (M = .391). When the two speakers’ descriptions differed in favorableness, however, participants had substantially better recall of the behaviors mentioned by a given speaker that were inconsistent with the trait description of the target by the other speaker than behaviors that were consistent with this description (.520 vs. .416). Thus, both sets of data suggest that participants had better recall of behaviors mentioned by one speaker if they were inconsistent with the trait description provided by the other speaker than if they were consistent with this description (.482 vs. .401). Although these data are confusing on first consideration, their interpretation is actually quite clear. That is, participants apparently used the speakers’ trait descriptions to form impressions of the speakers, and then used speakers’ descriptions of the target’s behaviors to confirm the validity of these impressions. Note that if a behavior mentioned by one speaker is evaluatively inconsistent with the trait description of the target provided by the other speaker, this suggests that the description is not a valid characterization of the target but rather, reflects a general disposition of the speaker to describe people favorable or unfavorably. Thus, these behaviors confirm the impression that participants formed of the speaker on the basis of his or her description of the target. Participants who sought to confirm this impression thought more about these behaviors and, therefore, recalled the behaviors better.

172

CHAPTER 7 TABLE 7.3 Proportion of Behaviors Recalled as a Function of the Speaker Who Mentioned Them and the Behaviors’ Evaluative Consistency With Each Speaker’s Trait Description of the Target

A. Speaker-impression conditions Behaviors mentioned by male Consistent with male’s trait description of target Inconsistent with male’s trait description of target M Behaviors mentioned by female Consistent with male’s trait description of target Inconsistent with male’s trait description of target M B. Target-impression conditions Behaviors mentioned by male Consistent with male’s trait description of target Inconsistent with male’s trait description of target M Behaviors mentioned by female Consistent with male’s trait description of target Inconsistent with male’s trait description of target M

Consistent With Female’s Trait Description of Target

Inconsistent With Female’s Trait Description of Target

M

.430

.514

.412

.403 .416

.438 .476

.421

.351

.430

.390

.527 .439

.451 .441

.489

.396

.437

.416

.422 .409

.424 .430

.423

.437

.528

.482

.409 .423

.472 .500

.441

Note. In each quadrant of the table, diagonal cells are those in which the two speakers’ trait descriptions of the target are evaluatively similar, and off-diagonal cells are those in which the descriptions are evaluatively dissimilar. Adapted from Wyer, Budesheim, and Lambert (1992).

Thus, suppose the female speaker described the target in terms of favorable traits and the male described him in terms of unfavorable traits. Then, the mental representation that presumably results from the cognitive activity described in the preceding paragraph can be captured in an associative network of the form shown in Fig. 7.3a, where F+ and M- refer to concepts of the male and female speakers, respectively, and the favorable and unfavorable behaviors mentioned by the speakers are denoted b+ and b-, respectively.

PRAGMATIC INFLUENCES

173

2. Target-Impression Conditions In these conditions, participants were told to form an impression of the target rather than of the speakers. Despite these instructions, the results were very similar to those observed in speaker-impression conditions. For example, the favorableness of the speaker’s initial trait descriptions of the target influenced both participants’ perceptions that the speakers liked the target (1.58 vs. -0.93, when both speakers’ trait descriptions were favorable vs. unfavorable, respectively) and also participants’ liking for the speakers (1.04 vs. -0.42, respectively). However, in contrast to earlier studies in which trait descriptions of the target provided the primary basis for evaluations of him (Srull & Wyer, 1989), participants liked the target less when both speakers described him favorably (M = -0.33) than when they did not (M = 0.75). Thus, these data, like those obtained in speaker-impression conditions, indicate that participants used the speakers’ trait descriptions of the target to form impressions of the speakers themselves, and then, having done so, used their impressions of the speakers as comparative standards in evaluating the target without considering the favorableness of the speakers’ trait descriptions for characteristics of the target himself. The recall data in this study were also similar to those observed in speaker-impression conditions except that in this case, participants were primarily disposed to confirm their impression of the female speaker rather than the male. These data are summarized in the bottom half of Table 7.3. When both speakers’ trait descriptions of the target were similar in favorableness, participants had somewhat better recall of behaviors that were evaluatively inconsistent with these descriptions (M = .448) than behaviors that were consistent with them (M = .417). When the two speakers’ descriptions differed in favorableness, participants were more inclined to recall behaviors that were inconsistent with the female speaker’s description of the target (M = .482) than behaviors that were consistent with these descriptions (M = .416), and this was true regardless of whether the behaviors were mentioned by the male or the female. Thus, both sets of data could suggest that participants organized the information around an evaluative concept of the female speaker. Thus, they thought more extensively about behaviors she had personally mentioned that were inconsistent with this concept in an attempt to reconcile their occurrence. On the other hand, they thought more about behaviors the other speaker mentioned that were inconsistent with this concept and, therefore, bolstered their perception that the female’s trait description of the target reflected a general disposition to judge people favorably or unfavorably independently of the target’s actual attributes. The representation that participants may have formed under this condition, therefore, would resemble that shown in Fig. 7.3b.

174

CHAPTER 7

FIG. 7.3. Representations formed under conditions in which participants (a) were told to form impressions of speakers and (b) were told to form impressions of the target. F+ and M- denote favorable and unfavorable concepts of the female and male speakers, respectively, and b+ and b- denote favorable and unfavorable descriptions of the target’s behaviors, respectively.

3. Summary Considered in combination, the results obtained under speaker-impression and target-impression conditions suggest that participants who overhear a conversation about a person are inclined to form impressions of the speakers, and that this is true regardless of whether they are explicitly told to do so or are asked to form an impression of the person the speakers are discussing. Moreover, they tend to use the speakers’ descriptions of the target’s behavior to confirm these impressions. Then, once these impressions are formed, participants use them as comparative standards in evaluating the person the speakers are discussing rather than basing their impressions on the literal implications of the speakers’ descriptions of him.

PRAGMATIC INFLUENCES

175

The Influence of Modesty and Politeness Norms on Impression-Related Information Processing

1. General Considerations Participants in informal conversations are expected not to offend the people to whom they are communicating. They are also expected to be modest and not to toot their own horns unnecessarily. Consequently, people who hear someone make a statement that appears to violate one of these norms may attempt to reinterpret it in a way that is more consistent with their normative expectations. These processes may be similar to those that result from violations of informativeness and truthfulness norms. For example, the interpretation of a statement as hostile or a tease may depend in part on whether the statement violates a norm to be polite. This, in turn, may depend on whether the target of the statement is present at the time the statement is made. Thus, for example, a person’s comment that a colleague has trouble balancing his checkbook is likely to be seen as a tease if the comment is directed to the colleague himself. In the colleague’s absence, however, a politeness norm might not be applicable, and so the statement is more likely to be interpreted as antagonistic. However, the effects of politeness norms on comprehension are not independent of its effects on truthfulness. Shortly after the publication of my 1974 book on cognitive organization and change, a friend remarked, “Congratulations on your book. I’m looking forward to the publication of the English language version.”15 Because I interpreted the statement as a tease rather than an expression of hostility, I found it amusing. However, this reaction was probably diluted by my lack of confidence in my writing skills and, therefore, my belief that the statement’s implications might be valid. If I had had more confidence in my writing ability, I might have construed the implications of the statement to be definitely false. I might then have appreciated the irony much more. A related implication of this analysis is that statements are more likely to be interpreted as teases if their literal implications are extremely disparaging than if they are less so. This is because they are more likely to be seen as invalid. I may be more disturbed by a mildly negative comment about my writing ability than by a comment that my writing is slightly superior to that of a chimpanzee. This is because I am inclined to believe that the first remark is valid and thus to take it literally, whereas I perceive that the second statement is clearly invalid and therefore, interpret it as a tease. 15

15

Appreciation is extended to John McCarty for this example.

176

CHAPTER 7

2. Effects of Politeness and Modesty Norms on Reactions to Information The conceptualization we propose implies that people who perceive that a communication violates a norm-based expectation may think about the communication more extensively in an attempt to understand the reason it occurred. The fact that people have better recall of opinion statements they regard as invalid (Wyer et al., 1992) provides indirect evidence of this more extensive processing when information violates a norm of accuracy. However, similar effects should be evident when other norms are violated. A series of studies by Wyer, Budesheim, Lambert, and Swan (1994) confirmed this possibility. Participants listened to a tape-recorded conversation that was ostensibly between two male student volunteers who knew each other well. In contrast to the study by Wyer et al. (1990), however, this conversation was about one of the participants rather than a third party. Over the course of the conversation, one person (O) described both favorable and unfavorable things that the other, target person (T) had performed, whereas the target conveyed both favorable and unfavorable things that he personally had done. Some participants were asked to form an impression of O on the basis of O’s description of the target (other-impression, other-focus conditions); others were told to form an impression of the target based on O’s description of him (target-impression, other-focus conditions), and still others were asked to form an impression of the target on the basis of the target’s descriptions of himself (target-impression, target-focus conditions). The tape was constructed in much the same way as the one employed by Wyer et al. (1990). That is, it began with the experimenter giving instructions to the two conversation partners (referred to in the tape as Don and Bob). After deciding which of the persons would be the topic of discussion, O was asked to write down three trait adjectives describing the target, and T was asked to write down three adjectives describing himself. The tape was then temporarily interrupted, and participants were given either a favorable or unfavorable trait description of the target that was ostensibly prepared by the person on whom they were told to focus their attention. (Thus, they were given the target’s trait description of himself under target-impression, target-focus condition, but were given O’s description of the target in the other two conditions.) The tape was then restarted and the conversation was played in which the two participants exchanged anecdotes about T’s behaviors. These behaviors were identical to those described by the two speakers in Wyer et al. (1990) with minor variations. Thus, an excerpt analogous to the one employed in the earlier study was as follows, where T and O refer to the target and the other speaker, respectively:

PRAGMATIC INFLUENCES

177

Experimenter: Okay, then, why don’t we begin. Why don’t you start, Don? O: Well, let’s see . . . well, you walk a mile to school each day, even when it’s freezing. I guess I thought of that because it’s so damn cold today. Every time I offer you a ride, you always say no. T: Yeah, I never seem to get sick. It must be because I work out 3 or 4 times a week—I’m in pretty good shape. O: [Laughs] Better than me, anyway. . . . I remember you won the university chess championship last fall. T: [Laughs] Yeah . . . I remember we went out to dinner that night to celebrate and I swore at the waiter at Eddies because I didn’t get served as quickly as I wanted. O: Yeah . . . I remember after we left, and we were supposed to go to that party at Jerry’s, you were trying to follow the directions he’d given you to get to his house and you got us totally lost! . . . Oh, by the way, do you know Jerry lost his job at the Union? T: No, really? Gee, that’s rough. O: Yeah. I thought of that ’cause he told me you had loaned him 20 bucks for a date he had and stuff. T: [Pause] Jeez, this is a lot harder than I thought. [To experimenter] You want specific behaviors, right? Like I know I’m always smiling and saying hello to people when I walk down the street. But one particular instance doesn’t stand out. Experimenter: Well, try to be as specific as you can. T: [Pause] Hmm . . . well, let’s see . . . well, I flunked a mechanics training course I took last summer—I don’t know why I took it, anyway. Participants after hearing the conversation were asked to indicate how well they liked the two speakers, and then were asked to recall as many of the behaviors as they could. We expected that participants would have better recall of statements that O made that violated a politeness norm, but would have better recall of statements the target mentioned that violated a modesty norm. This was in fact the case. Table 7.4 shows the mean proportion of favorable and unfavorable behaviors that participants recalled in each instructional condition as a function of the person who mentioned them. Participants had better recall of behaviors that O mentioned if they were unfavorable than if they were favorable (.49 vs. .39, respectively), but had better recall of behaviors the target himself mentioned if they were favorable than if they were unfavorable (.45 vs. .38, respectively). The first difference was more pronounced when participants were explicitly told to focus their attention on the behaviors that O mentioned, and the second was more pronounced when partici-

178

CHAPTER 7 TABLE 7.4 Proportions of Favorable and Unfavorable Behaviors Recalled as a Function of Task Objectives and the Person Who Mentioned Them Task Objectives

Behaviors mentioned by other (O) Favorable Unfavorable Behaviors mentioned by target (T) Favorable Unfavorable

Target Impression, Target Focus

Target Impression, Other Focus

Other Impression, Other Focus

.42 .45

.40 .54

.35 .49

.47 .45

.50 .54

.37 .34

Note. Adapted from Wyer, Budesheim, Lambert, and Swan (1994).

pants were told to focus their attention on behaviors that T mentioned. However, the differences were evident in all three instructional conditions and did not depend on the favorableness of the initial trait descriptions that the speakers provided. In other words, participants spontaneously thought more extensively about comments that violated norms to be polite and modest. Furthermore, this was true even when the comments were not directly relevant to the impression formation objectives that participants were pursuing, and regardless of whether the statements were consistent or inconsistent with more general trait descriptions of the target.

THE EFFECT OF NONVERBAL BEHAVIOR AND CONVERSATIONAL STYLE ON IMPRESSION FORMATION, JUDGMENT, AND INFORMATION SEEKING The studies by Wyer et al. (1990, 1994) provide evidence that people who hear a conversation pay attention to the pragmatic implications of things that are said about a person rather than to their literal implications. Moreover, they pay particular attention to statements that violate normative expectations for the sorts of statements that are typically made in the type of situation at hand. However, these expectations can pertain not only to the content of the information that is exchanged in the situation but also to how it is exchanged. People often manifest nonverbal behaviors that lead recipients to infer that the information they convey has implications that differ from those implied by its content alone. Moreover, the style in which people conduct a conversation (e.g., whether they ask other participants ques-

PRAGMATIC INFLUENCES

179

tions or elaborate answers to others’ questions) could have implications for both their interest in the conversation and their liking for the persons with whom they are conversing. These implications, in turn, could influence people’s impressions of both the topic under discussion and the discussants. Two quite different sets of studies bear on these possibilities. Nonverbal Influences on Perceptions of a Message’s Implications The nonverbal behaviors that accompany a person’s comments in a social interaction can obviously have an impact on recipients’ interpretation of these comments. An individual’s disparaging remark to another is more likely to be interpreted as a tease if the communicator is smiling than if he or she is not. The frequency and duration of eye contact can often be used as an indication of feelings of intimacy and of the intensity of feelings that are conveyed by a communicator’s statements and, therefore, can influence the interpretation of these statements (Ellsworth & Carlsmith, 1968). Other, less controlled behaviors (e.g., eye contact, voice pitch, etc.) can influence recipients’ perceptions of the communicator’s truthfulness and, therefore, the validity of the information being communicated (for summaries of the impact of nonverbal behaviors on these and other perceptions, see DePaulo & Friedman, 1998). A recipient’s perception of the validity of a message can also be inferred from manifestations of uncertainty. Speech style characteristics that reflect hesitancy (stammering, a high frequency of “uhhs” and “ers,” etc.) could result from lack of confidence in how the information should be conveyed. On the other hand, recipients might misattribute these characteristics to lack of confidence in the validity of the information itself. For example, a person who has read a story that advocates a particular point of view might try to convey its contents to someone who opposes this view in a way that the recipient will not find too offensive. Similarly, a student’s dissertation advisor might wish to convey the essence of a colleague’s devastating criticisms of a thesis proposal to the candidate in a way that is substantively correct but is more dispassionate in tone. The communicator’s difficulty in attaining these objectives could be reflected in characteristics of his or her speech style. However, these characteristics could be misinterpreted by the recipient as uncertainty about the validity of the information being conveyed. Gruenfeld and Wyer (1993; reported in Wyer & Gruenfeld, 1995) demonstrated these effects under conditions in which the difference between communicators’ objectives and the ostensible purpose of the original information was unobtrusively manipulated. Participants read a passage describing an unfamiliar disease. They were told that the passage had been taken from either a newspaper (whose objective is presumably to convey new and in-

180

CHAPTER 7

teresting information) or an encyclopedia (whose goal is to preserve archival knowledge). Regardless of the passage’s source, however, participants after reading it were told to convey its content to a naïve recipient in a way that was either as interesting as possible or as accurate as possible. Thus, in some cases, the assumed objectives of the message’s original source and those of the communicator were ostensibly similar, and other cases, these objectives differed. After delivering the message (which was tape recorded), both communicators and naïve recipients evaluated the seriousness of the disease and reported their specific beliefs about its prognosis, cause, and treatment. These latter beliefs were coded so that more positive values indicated stronger beliefs in the content of the original passage. Communicators’ own beliefs and opinions about the disease were not affected by either the source of the passage or their own objectives in communicating it. Moreover, a content analysis of communicators’ speech revealed no differences in the actual content of the information they conveyed in different conditions. Nevertheless, communicators exhibited more characteristics of poor speaking style (e.g., a greater frequency of “uhhs” and “errs”), and took longer to deliver the speech (an indication of rambling), when their communication objectives differed from those of the original message source than when they were similar. Correspondingly, recipients reported less strong beliefs and opinions about the disease and its seriousness in the former condition than the latter. These results are perhaps not too surprising. However, their importance in the present context lies in their implications that communicators take into account not only their own objectives in conveying a message but also the pragmatic implications of the source of the information they are transmitting, and that these factors combine to influence recipients’ perceptions of the validity of the information being communicated. The Impact of Conversational Style on Liking for the Communicator The speech style characteristics identified by Gruenfeld and Wyer seem likely to influence reactions to communications exchanged in an actual conversation. In this context, they might influence participants’ impressions of one another as well as the topic under discussion. Additional factors are likely to come into play, however. People who engage in an informal conversation are typically expected to converse in a way that will make it easy for others to respond. One obvious way of accomplishing this is to ask questions of one another. A related technique is to elaborate answers to one another’s questions, thereby increasing the likelihood that the response will stimulate a relevant idea in the listener. These techniques may be acquired through social learning and be applied spontaneously, without

PRAGMATIC INFLUENCES

181

much deliberation. When the techniques are not applied, however, recipients are likely to find the conversation awkward and difficult to conduct, and these reactions could affect their liking both for the conversation itself and for the other participants. These effects are particularly likely to occur in get-acquainted conversations, where participants are motivated to learn about one another and to identify areas of mutual interest. For example, suppose a participant in such a conversation asks, “Where are you from?” If the other responds, “Chicago. How about yourself?” the first person finds it quite easy to continue the dialogue. Alternatively, if the other responds, “Well, I live in Chicago now, but I just moved here from Florida,” the questioner might be stimulated to ask about Florida and why the person moved, or might describe his or her own recent visits to the state. However, suppose the other simply responds “Chicago,” without elaborating or asking a question in return. Then, the first individual is burdened with the task of finding a different topic to discuss or a different question to ask. This burden increases the difficulty of continuing the conversation and this difficulty, in turn could decrease the person’s liking for both the conversation and the interaction partner. But other factors can enter into the picture as well. An individual’s failure to elaborate answers to another’s questions, or the failure to ask questions in return, could also be interpreted as an indication that the individual is not really interested in interacting with the questioner. This inference could also decrease the questioner’s liking for the individual independently of the effects of these conversational characteristics on the ease of conducting the conversation per se. Wyer, Swan, and Gruenfeld (1995) investigated these effects in a getacquainted conversation. College students who were previously unacquainted took part in the study in same-sex pairs. However, only one member of each pair was assigned the role of the actual subject. The subject’s partner was recruited as an accomplice. Subjects and their partners arrived at the experiment in different rooms so they would not see one another prior to the experiment. Subjects were told we were interested in how people engage in get-acquainted conversations, and that they would be asked to take part in a 5-minute conversation with another student much as they might if they met the person for the first time at lunch or on a break between classes. However, we indicated that to ensure that some of the information exchanged in each conversation was similar, we would like them to ask their partner five questions at some point during the conversation: 1. Where are you from? 2. Do you like school? 3. Where do you live on campus?

182

CHAPTER 7

4. What is your major? 5. How are you doing in your classes? Subjects were given a list of the questions as a reminder, but were told they could ask them in any order and to work them into the conversation in a way that seemed natural. They were further told that their partner did not receive any questions but was being told simply to conduct the conversation as naturally as possible. Partners were given similar instructions about the study’s purpose. However, we went on to indicate that people vary in the amount of information they give when they are asked about themselves and in whether they ask questions in return, and that we were interested in how these factors influenced impressions. We then gave the partner the same list of five questions we had given to the subject and indicated that when the subject asked these questions, he or she should respond in one of four ways. In elaboration, question-reciprocation conditions, partners were told: We would like you to do two things when your partner asks you these questions. First, rather than simply answering the questions with a yes-or-no answer, elaborate your answer in one or two sentences. Then, after doing so, ask the other person the same question in return. For example, when your partner asks you if you like school, you might say, “Yes, I like the social life but the tests are hard. How about you?” Or, when you are asked where you are from, you might say, “I’m from the North side of Chicago, but before that I lived in Detroit. Where are you from?” Answer the questions truthfully, but don’t go into too much detail. Keep your answer to one or two sentences, and then ask your partner the same question that you were just asked. Except for your responses to the five questions that I have passed out to you, however, carry on the conversation as naturally as possible.

Under no-elaboration, question-reciprocation conditions, partners were told to restrict their answers to one or two words but then to ask the other the same question in return. Instructions in elaboration, no-question-reciprocation and no-elaboration, no-question-reciprocation conditions, were similar to those in the first two conditions except that partners were told not to ask a question in return. The subject was then ushered into the partner’s room and seated on the other side of a partition so they could not see one another. The participants were then introduced and left alone for 5 minutes to engage in their conversation. Then, after completing the interaction, participants were asked to infer how much they enjoyed the conversation, how easy it was to conduct it, and how much they thought they would like their partner if they got to know him or her better. They also predicted how interested their partner was in them as a person and how much they thought the partner liked

PRAGMATIC INFLUENCES

183

them. Finally, they judged the partner with respect to several personality attributes. Of these, the most relevant to his or her behavior in the conversation was aloofness. We expected that both partners’ elaboration and their reciprocation of questions would influence participants’ ease of conducting the conversation, and that this, in turn, would infer their liking for the conversation and, as a result, their liking for the partner. However, we also speculated that if subjects found that their partner did not elaborate and reciprocate their questions, they would interpret this behavior as an indication that their partner had little interest in the interaction and might not like them, and that these perceptions might also influence their liking for their partner. In fact, both possibilities were evident. The effects of conversational style can be seen most easily from path analyses. Figure 7.4 shows the significant paths connecting the two communication style characteristics to the participants’ estimates of the ease of conducting the conversation (Ease), their enjoyment of the conversation (Enjoy), their perceptions that their partner (O) was interested in them (O int S) and liked them (O like S) and their liking for the partner (like O). Females’ perception of the ease of conducting the conversation influenced both males’ and females’ liking for their partner through its mediating influence on both their enjoyment of the conversation and their inference that the partner liked them. However, whereas males’ perception of the ease of conducting the conversation was primarily a function of their partner’s reciprocation of their questions, females’ perceptions were influenced primarily by the degree to which their partner elaborated responses to their questions. In addition, partners’ elaborations also affected subjects’ perceptions that the partner was interested in them, and this factor also increased their liking for their partner. On the other hand, partner’s conversational style had no impact on participants’ perceptions of their aloofness, nor did these perceptions influence liking for the partners. The gender differences obtained in the study are provocative. As Wyer et al. (1995) point out, women are typically more motivated than men to establish a sense of “connectedness” to the persons with whom they interact (Chodorow, 1979; Gilligan, 1982; Tannen, 1990).Their partners’ elaboration of answers to their questions may have increased their feelings of connectedness, as evidenced by its impact on these subjects’ perceptions that their partner was interested in them and liked them. In contrast, males placed less emphasis on these factors. Perhaps because they are less socially skilled than females, however, they found that the ease of conducting the conversation was easier when their partners asked them questions. In summary, partners’ counternormative communication style influenced both men’s and women’s liking for their partner through its mediating impact on their perception of ease of conducting the conversation. In

184

CHAPTER 7

FIG. 7.4. Path diagrams of the effects of the partner’s conversational style on subjects’ liking for their partner (O) and the variables that mediate these effects. Elab = O’s elaboration of responses to the subject’ question. Quest = O’s reciprocation of the subject’s question. Ease = subjects’ perception of the ease of conducting the conversation. OintS = subjects’ perception that O was interested in them. O like S = subjects’ perception that O liked them, and like O = subjects’ liking for the partner.

addition, females attributed their partner’s conversational style to a lack of interest in them and this attribution was also a major determinant of their liking for her. Influence of Counternormative Conversational Style on Information Seeking We have assumed that when people encounter an expectancy-violating communication, they are likely to search for other information that might explain its occurrence. This cognitive activity may be reflected in the recall

PRAGMATIC INFLUENCES

185

of the counternormative messages that give rise to it, as suggested by Wyer et al. (1994). However, the assumption has other implications as well. That is, if people seek additional information in an attempt to explain counternormative messages, the information they identify in the course of this search may have greater influence on their judgments than it otherwise would. Evidence of this possibility was obtained in a second study by Wyer et al. (1995) under conditions similar to those described in the experiment just described. That is, female participants first engaged in a get-acquainted conversation with a female partner who responded to their questions by either (a) both elaborating their answers and asking them questions in return (normative-behavior conditions) or (b) neither elaborating nor reciprocating the participant’ questions (counternormative-behavior conditions). Before engaging in the conversation, however, some participants (ostensibly by mistake) were exposed to a description of their partner that another participant had provided in a previous experimental session. In some instances, this description indicated that the partner was sociable—a trait to which the partner’s conversational style was relevant and, therefore, could provide an explanation for the partner’s behavior. In other cases, the description indicated that the participant was high in integrity—an attribute that was irrelevant to the partner’s interaction behavior. Participants in a third, control condition received no trait information at all. We expected that when the partner’s responses to subjects’ questions were normatively consistent with the behavior that was likely to occur in get-acquainted conversations, subjects would base their liking for her on the interaction and that the other’s description of her would have little impact on these judgments. When the partner’s behavior was counternormative, however, participants should seek an explanation for it and, therefore, should pay more attention to the trait descriptions. As a consequence, these descriptions should have more impact on their liking for their partners than would otherwise be the case. Results confirmed these hypotheses. Subjects’ evaluations of the partner are shown in Table 7.5 as a function of the partner’s conversational style and the relevance of the target descriptions provided. When partners’ conversational style was normative, subjects evaluated her favorably regardless of whether trait descriptions of the partner were available. (In fact, their evaluations were nonsignificantly less favorable when trait descriptions were provided, suggesting that, if anything, the traits detracted from the impact of the interaction.) When the partner’s behavior was counternormative, however, participants evaluated her relatively unfavorably when no trait descriptions were provided, but trait descriptions substantially increased their evaluations. Note that this was true even when the traits were irrelevant to the target’s interaction behavior. In other words,

186

CHAPTER 7 TABLE 7.5 Subjects’ Liking for Their Partner as a Function of the Partner’s Conversational Style and Initial Trait Description Initial Trait Description of Partner ConversationIrrelevant

ConversationRelevant

None

M

8.56 6.63 7.50

8.78 6.78 7.78

8.95 3.88 6.56

8.76 5.89

Normative conversational style Counternormative conversational style M Note. Adapted from Wyer et al. (1995).

trait descriptions of the partner only had an influence on subjects’ evaluations of her when her behavior in the conversation deviated from expectations and, therefore, stimulated subjects to seek other information that might account for it. Moreover, this search led the information to have an impact regardless of whether it was relevant or irrelevant to the explanation that subjects were seeking.

PRAGMATIC COMMUNICATION IN CLOSE RELATIONSHIPS The discussion in this chapter has focused largely on the pragmatic influences of communications that are exchanged among casual acquaintances. When people know one another very well, the pragmatic implications of their shared messages may be difficult for others to discern. As noted earlier, married couples can often convey feelings to one another through a statement that seems quite innocuous to observers who are unaware of the pool of shared knowledge that the partners bring to bear on its interpretation (Watzlawick et al., 1967). Thus, for example, a wife who asks her spouse at a party what time it is might be interpreted as making a serious request for information. However, it could be interpreted by her husband as an indirect expression of anger because of his failure to leave the party early enough for her to see David Letterman. On the other hand, miscommunication often arises between partners in close relationships as well as casual acquaintances. This is particularly true in the case of emotions. One reason for this could be that the expression of some emotions is inherently ambiguous. However, other factors play a role as well. A study by Gaelick, Bodenhausen, and Wyer (1985) provided insight into these possibilities. In an initial session of their experiment, married couples engaged in a 10-minute tape-recorded discussion of a problem they were having in their relationship. Then, in a second session, each partner

PRAGMATIC INFLUENCES

187

viewed the tape of the conversation and identified three statements that were made in the conversation that had had an important effect on their feelings toward one another. Finally, in a third session, partners reviewed the segments of the tape containing the statements that both they and their spouse had identified. Partners rated statements they had personally made in terms of the feelings they intended to convey, their expectations for how the partner would interpret the statements, and how they thought the partner would respond. They rated statements their partner had made in terms of the feelings their partner intended to convey, how their partner thought they would respond, and their actual response. Factor analyses of these ratings revealed they fell along two independent dimensions pertaining to love and hostility. Several interesting results emerged. In general, partners attempted to convey the emotion they perceived their spouse had conveyed to them. However, they were only accurate in perceiving their spouse’s hostility. Consequently, feelings of hostility were actually reciprocated, but feelings of love were not. One implication of this is that hostility was more likely to escalate over the course of the conversation than feelings of love. Second, when wives communicated in a way they intended to be affectively neutral, their husbands interpreted their statements as expressions of hostility. In contrast, when husbands communicated in a way they intended to be neutral, their wives interpreted their statements as expressions of love. Gaelick et al. (1985) interpreted these miscommunications in terms of the stereotyped social role expectations that men and women hold for one another. Specifically, women are expected to be loving and affectionate. Therefore, when women responded in a way they intended to be neutral, their husbands interpreted this counternormative comment as hostile and, as noted earlier, were likely to respond hostilely in return. Men, however, are expected to be hostile and aggressive. Consequently, when they intended to convey neutral affect, their wives interpreted this expectancy-deviant statement as an indication of love, and presumably reciprocated this emotion. Unfortunately, however, because expressions of love were typically misperceived, these attempts to deescalate the conflict were unlikely to do much good.

CONCLUDING REMARKS Although this chapter has covered a lot of ground, it clearly does not provide a complete account of the factors that theoretically influence people’s perceptions of the pragmatic implications of the information they receive in social situations. For example, the extensive research on social attribution, stimulated by theories of Jones and Davis (1965) and Kelley (1967, 1987),

188

CHAPTER 7

bears indirectly on the situational factors that influence the implications that people draw from a person’s statements and behavior. Research on characteristics of the source of information that affect the impact of this information (for a review, see Eagly & Chaiken, 1993) also makes salient the need to take into account the influence of pragmatic factors (e.g., the intentions of the communicator) in conceptualizing the nature of this influence. A discussion of this research is beyond the scope of this volume. Nevertheless, the work we have summarized suffices to point out the importance of understanding these processes in the context of actual social situations. At the same time, the pragmatic influences of information cannot be understood independently of the more fundamental processes that underlie the spontaneous comprehension of information and an identification of its validity or its redundancy with prior knowledge. The extent to which pragmatic implications of a message are taken into account often depends on the recognition that the message’s literal implications violate expectations for the content and form of communications that are likely to occur. In the research we described in this chapter, these expectations typically derived from people’s prior knowledge about either the topic of the communication (and, therefore, whether the message was redundant with prior knowledge or false) or, alternatively, the source of the communication and the type of situation in which it is conveyed. Expectations for the meaning of a communication can also be based on prior communications that have been conveyed in the same situational context. Moreover, initial portions of a message can establish expectations that are violated by later portions. The communications we considered in this chapter were typically single statements (news headlines, witticisms, responses to questions, etc.). As we emphasized in earlier chapters, however, many communications that we experience in a social context consist of sequences of thematically related events, or narratives. In these instances, information that occurs early in a narrative provides a context for interpreting aspects of the narrative that occurs subsequently. However, when these latter aspects violate violates based on the context that was established by earlier features of the information, special circumstances can arise that have not yet been considered. The next chapter addresses these circumstances in some detail.

C H A P T E R

8 The Dynamics of Humor Elicitation: The Effects of Informational Context on the Interpretation and Elaboration of Narratives

The communications we considered in the previous chapter consisted largely of single statements. The interpretation of these statements was based largely on previously acquired knowledge about not only the topic to which the statements pertained but also the type of situation in which the statements were made. Similar factors come into play when information describes a temporally related sequence of events and is conveyed in the form of a narrative or story. There are some additional considerations, however. People theoretically comprehend descriptions of a series of situationally constrained events or states of affairs by constructing an episode model of the sequence of occurrences as a whole. In doing so, the interpretation they give to each description in the sequence is likely to be influenced in part by the concepts and knowledge that they have used to interpret the earlier ones. To give an example, consider the following pairs of statements: A. B.

The pilot was killed when his landing gear malfunctioned. Flying airplanes can be dangerous. Mary’s husband and children were killed when a plane overshot the runway and crashed into their home near the airport. Flying airplanes can be dangerous.

The second statement in B is likely to be interpreted quite differently than the second statement in A. This is presumably because readers draw upon different concepts and knowledge in constructing an event model of the first statement in each pair and, having done so, apply these concepts and knowledge to the second statement as well. 189

190

CHAPTER 8

In some instances, however, the concepts and knowledge that are activated and used to interpret the events that occur at the beginning of a sequence cannot be applied to the events that come later. In such cases, recipients must reinterpret the initial events in terms of different concepts in order to construct a complete understanding of the sequence of events as a whole. In the preceding example, suppose the statement, “Flying airplanes can be dangerous” is presented first. An individual who happens to interpret this statement as equivalent in meaning to “It can be dangerous to fly an airplane” might then encounter the statement “Mary’s husband and children were killed. . . .” To comprehend the information as a whole, the recipient would presumably reinterpret the first statement as referring to the hazards of living near an airport. The comprehension processes that are involved in this example are similar to those described in chapter 7 when a statement violates normative expectations. The only difference is that the expectations are not only activated by the social context in which it occurs. In addition, they result from the interpretation that has been given to other information in the communication in which the statement is embedded. Reinterpretations can also occur in the course of comprehending direct experiences. A waiter whose outward appearance gives the impression of elegance and sophistication but who is observed to spill soup in a customer’s lap may be reconceptualized as a pretentious oaf. However, reinterpretations of written or oral communications are more common. Particularly frequent examples occur in jokes and stories whose punch line stimulates a reinterpretation of the events leading up to it. For example, consider the following story: A young Catholic priest is walking through town when a prostitute accosts him. “How about a quickie for twenty dollars?” she asks. The priest, puzzled, shakes her off and continues on his way, only to be stopped by another prostitute. “Twenty dollars for a quickie,” she offers. Again, he breaks free and goes on up the street. Later, as he is nearing his home in the country, he meets a nun outside her convent. “Pardon me, sister,” he asks, “but what’s a quickie?” “Twenty dollars,” she says, “The same as it is in town.”

The events described in the punch line of this story stimulate the reinterpretation of two features of the previous information. First, one’s characterization of the nun is revised to include her service as a prostitute. Second, the question, “What’s a quickie?”, which was first assumed to be equivalent to “What does a ‘quickie’ mean?”, is reinterpreted as equivalent to “What’s a quickie cost?” Not all reinterpretations of information elicit amusement, as our example of flying airplanes testifies. In fact, the cognitive and motivational un-

DYNAMICS OF HUMOR ELICITATION

191

derpinnings of humor elicitation are not widely understood. This is surprising in light of the central role that humor plays in social communication. It is a rare conversation in which at least one participant does not respond with amusement to another person’s comments or behavior. Jokes, witticisms, and other humorous events are commonplace in social interaction situations and can have an impact on the quality of the interactions. Although jokes are often told for the purpose of entertaining, they can also be used strategically to decrease tension in a heated discussion (Kane, Suls, & Tedeschi, 1977) or to enliven a boring one. In short, the transmission and comprehension of humor are central features of social experience. Theoretical and empirical analyses of social interaction processes have largely ignored these processes. Several theories of humor elicitation have been proposed, beginning with the work of Freud (1928, 1960). However, most research on humor (for summaries, see Chapman & Foot, 1976, 1977; Goldstein & McGhee, 1972; McGhee & Goldstein, 1983) has focused on its motivational bases (cf. LaFave, Haddad, & Maeson, 1976; Zillman & Cantor, 1976) and its personality correlates (Lefcourt & Martin, 1986; Mindess et al., 1985). Yet, only a few attempts have been made to explicate the cognitive processes that underlie perceptions that something is funny (Apter, 1982; Long & Graesser, 1988; Suls, 1972, 1977). And of these, only Apter’s conceptualization is applicable to situations in which humor is elicited spontaneously in informal social interaction. In this chapter, we provide a theoretical analysis of humor elicitation in the context of the comprehension processes we have described in earlier chapters. Our theory avoids many of the deficiencies of existing formulations while at the same time providing an example of the potential applicability of our more general conceptualization of comprehension phenomena to information processing outside the laboratory. According to this conceptualization, humor elicitation is a by-product of the comprehension processes postulated in chapter 4 and in some cases, of higher order comprehension processes similar to those discussed in chapter 7. As we have noted, however, merely the reinterpretation of previously acquired information in the light of new information is not a sufficient condition for humor elicitation. Thus, several additional factors not discussed in earlier chapters must be taken into account. We first circumscribe the conditions in which humor is spontaneously elicited by information that people receive and consider the effects of more deliberative, elaborative processes that occur once the humor-eliciting interpretation of information has been identified. We then apply the conceptualization to a specific area in which an understanding of humor elicitation is particularly important, namely, the amusement elicited by jokes and stories that perpetuate a social stereotype or are likely to be considered offensive for other reasons.

192

CHAPTER 8

HISTORICAL BACKGROUND To provide a perspective on the conceptualization to be proposed in this chapter, a brief review of earlier theories of humor elicitation may be helpful. These theories fall into two categories, one of which focuses on the motivational antecedents of humor elicitation and the other of which concerns the cognitive processes that give rise to it. Considered in isolation, each theory is insufficient to account for the numerous circumstances in which amusement may occur. (A possible exception to this rule is Apter’s [1982] theory to be discussed presently.) Nevertheless, the factors that the various theories postulate to influence humor elicitation are generally compatible. That is, the factors postulated by several theories could contribute to humor elicitation, although not necessarily for the reasons the theories assume. Motivational Theories of Humor 1. Arousal-Reduction Theories Responses to humor have sometimes been conceptualized in terms of a release of tension or reduction in arousal. According to Freud (1905/1960, 1928), for example, people’s responses to humor-eliciting stimulus events is motivated by the need to reduce tension or arousal they are inhibited from expressing directly. This arousal, which was often assumed to be aggression- or sex-related, might either be induced by the stimulus itself or exist before exposure to the stimulus. For example, the humor a joke elicits could vary with the intensity of suppressed emotions that have previously become associated with the type of stimulus to which the joke is relevant. Although Freud’s conceptualization can account for some humoreliciting experiences, it is clearly not sufficient to explain the humor that is elicited by all types of communications. For example, the conceptualization appears to apply only under conditions in which features of the humoreliciting stimuli are similar to those to which the suppressed emotion is relevant. Given the wide diversity of stimuli that a given individual finds humorous, one would have to postulate a very large number of suppressed emotions in order to argue that this is a necessary antecedent of humor elicitation. A conceptualization of humor elicitation by Berlyne (1969, 1971) is broader in scope. He assumed an inverted-U relation between physiological arousal and the experience of pleasure. That is, pleasure increases with arousal up to a point and then decreases, ultimately reaching a level at which it becomes aversive. Berlyne viewed a joke as a scenario that induces arousal beyond its optimal value, followed by a punch line that de-

DYNAMICS OF HUMOR ELICITATION

193

creases this arousal to a relatively more pleasant level. This increase in pleasantness is experienced as amusement. Although broader than Freud’s, Berlyne’s formulation is also limited in generality. For one thing, arousal cannot be reduced unless it already exists. Thus, Berlyne’s theory seems applicable only when a situation-specific build-up of arousal has occurred before the humor-eliciting event is encountered. This might be true in the case of stories that set the recipient up for a punch line and, therefore, might induce a degree of cognitive tension. However, witticisms and other unexpected events that elicit humor spontaneously in informal conversations occur in the absence of any prior increase in arousal. 2. Superiority and Disparagement Theories A second group of theories (Bergson, 1911; LaFave, Haddad, & Maeson, 1976; Zillman & Cantor, 1976) assume that people derive pleasure from feelings of mastery, and that laughter at another’s deformities or misfortunes reflects an attempt to establish or maintain these feelings. These reactions are most likely when there are no social repercussions of expressing them. Thus, they may occur when the target of the disparagement is generally regarded as socially undesirable or is someone to whom people are indifferent. Americans are more inclined to be amused by a joke if it disparages Texas Aggies (a group that is of no particular interest to anyone but affiliates of Texas A&M) than if it disparages Jews, African Americans, or women. Disparagement theories are particularly useful in accounting for sick or ethnic humor. On the other hand, they have trouble accounting for self-referent humor that is elicited by one’s own ineptitude or by other events that reflect unfavorably on oneself. Although some people undoubtedly do find it difficult to laugh at themselves, this is by no means universal. In summary, motivation-based theories can potentially account for the humor that is elicited in certain circumscribed situations. However, they are insufficient to account for the spontaneous elicitation of humor that occurs under conditions in which the specific motives postulated by these theories do not exist. Moreover, the theories do not address the cognitive mechanisms that give rise to humor and that distinguish humor-eliciting communications from those that are simply disparaging, aggressive, or arousal-reducing but do not elicit amusement. Incongruity Resolution Theories Many theories (cf. Apter, 1982; Koestler, 1964; Suls, 1972) assume that amusement is stimulated by the awareness that a stimulus event has two alternative interpretations, each of which could potentially apply but are in some sense incongruous. Identification of the incongruity, however, may of-

194

CHAPTER 8

ten require access to a large body of knowledge about the type of persons and events described. For example, consider a joke from the Antioch Humor Test (Mindess et al., 1985): A blind man enters a department store, picks up his dog by its tail and begins swinging it over his head. A clerk hurries over and says, “Can I help you, sir?” “No, thanks,” the man replies, “I’m just looking around.”

In this joke, two planes of thought are brought together by the punch line, one of which pertains to the idiomatic meaning of “looking around” and the other of which concerns its literal meaning. However, an appreciation of the joke also requires knowledge that blind individuals often rely on seeingeye dogs to do their “looking.” 1. Suls’ Formulation of Humor Elicitation A more formal statement of incongruity-resolution processes, provided by Suls (1972, 1983), is consistent with the conceptualization of comprehension processes we proposed in chapters 4 and 5 as well as more general theories of prose comprehension (e.g., Graesser, 1981). Suls assumes that when people begin to read a story, the persons and events described activate a body of conceptual knowledge (e.g., event or episode models as conceptualized in this volume) that can be used to interpret it. This knowledge, once activated, is brought to bear on the interpretation of information that is conveyed subsequently. However, recipients may later encounter information that cannot be understood in terms of this knowledge. When this occurs, they must identify alternative concepts and knowledge structures that can be used to comprehend the new information in the context of the old. This process often requires a reinterpretation of the original information, and the generation of this reinterpretation elicits humor. Suls likens joke comprehension to a problem-solving task in which comprehension is analogous to solving the problem and amusement is analogous to the pleasure that people get from arriving at the solution. This analogy has additional implications. Most obviously, people must recognize that a “problem” exists (i.e., that the information cannot be interpreted in terms of previously acquired concepts and knowledge). More important, the problem must be neither so easy that the pleasure derived from solving it is minimal nor so difficult that its solution requires excessive effort. This suggests that humor is more likely to be elicited by jokes that are moderately difficult to comprehend than by jokes that are either too easy or too hard. We consider this possibility more fully in the conceptualization we outline later.

DYNAMICS OF HUMOR ELICITATION

195

2. Apter’s Reversal Theory Although inconsistency resolution can be an important antecedent of humor elicitation, it is not a sufficient condition for its occurrence. Not all reinterpretations of previously acquired information in light of new information are amusing. Scientific discoveries, or creative problem solving more generally, often involve the application of a new set of concepts to previously acquired knowledge in a way that reconciles this knowledge with new information that becomes available. Although these events can be very pleasureful, they are not amusing. A conceptualization proposed by Michael Apter (1982) is compatible with Suls’ (1983) but specifies both the necessary and the sufficient conditions for humor to be elicited. Apter explicitly recognizes that the social experiences that elicit humor include not only the persons, objects, and events to which a joke or story refers but also the communicator and aspects of the social situation in which the information is encountered. By taking these contextual factors into account, Apter is able to explain the humor that is elicited not only by jokes and cartoons but also by witticisms and fortuitous life experiences that are not intended to be funny at all. Apter (1982) assumed that people who encounter a social experience attempt to interpret it in terms of concepts and knowledge about both its referents and the situational conditions that surround it. In doing so, they arrive at a tentative understanding of the experience as it actually exists from the perspective of the individuals involved in it. To acquire this understanding, they may often infer unmentioned attributes of these individuals or of the situation more generally. (Thus, if a protagonist is described as a lawyer, he or she might be inferred to have characteristics that are typical of lawyers, etc.) Moreover, if a communication is conveyed verbally, people’s understanding of it may take into account not only the situation described in the communication but also the purpose for which the communication is transmitted (e.g., the goals of the communicator). Observed behavior is likewise understood in terms of the presumed goals of the actor and the situational constraints that are placed on the behavior’s occurrence. When an experience is composed of a sequence of events, however, the concepts and knowledge that are activated and used to interpret initial aspects of the experience may turn out to be inapplicable. (The sentence about flying airplanes described at the beginning of this chapter provides an example.) Then, a reinterpretation of the initial features may be required in order to arrive at a coherent understanding of the experience as a whole. This interpretation may give rise to the realization that the situation, or the persons and objects involved in it, are not actually what they were assumed to be at first. This process is very similar to the inconsistency-resolution process proposed by Suls (1972). However, Apter postulates that for the reinterpreta-

196

CHAPTER 8

tion of a situation to elicit humor, two things must be true. First, the reinterpretation must not replace the original interpretation that was made. That is, the perception of the experience that results from the reinterpretation must not change one’s perception of the experience that purported to exist at the outset. Rather, both interpretations must be taken into account simultaneously. (For a similar assumption, see Koestler, 1964.) Second, the perception of the experience that is established by the new information must in some sense be diminished in importance or value relative to the conditions that were first assumed. To see the applicability of the diminishment and nonreplacement principles, reconsider the “quickie” joke described earlier. To reiterate, this joke actually exemplifies two types of shifts in interpretation. One, purely semantic shift occurs in the interpretation of “What’s a quickie?” The second shift occurs from the perception of the nun as a chaste and devout woman to the perception of her as a prostitute. Note, however, that the reinterpretation of “What’s a quickie?” that is implied by the nun’s response does not negate the interpretation that was apparently intended by the priest. Moreover, the nun, although turning out to be a prostitute, remains a nun. In other words, the new perception of the reality of the situation does not invalidate the appearance of the situation that existed before the reality was revealed. Finally, note that the nun’s holiness is diminished as a consequence of being a prostitute on the side. Thus, both of the conditions that Apter (1982) postulated to be necessary for human elicitation are met. Diminishment should not be confused with disparagement. Although disparaging reinterpretations may often be diminishing, not all diminishing reinterpretations are disparaging. In the Wizard of Oz, for example, people are amused by the lion who purports to be ferocious but who turns out instead to be meek. The lion’s true attributes are less unfavorable than his purported ones. They are nevertheless more mundane and, therefore, elicit amusement. 3. Summary Apter’s conceptualization can account for a number of humor-elicitation phenomena that are not easily explained by other formulations. To give some examples: Puns. Puns purport to convey interesting information but are revealed to be silly and, therefore, trivialize the communication itself. For example: Two Eskimos sitting in a kayak were chilly. However, when they lit a fire in their craft, it sank, proving once again that you can’t have your kayak and heat it too.

DYNAMICS OF HUMOR ELICITATION

197

A group of chess enthusiasts were standing in a hotel lobby discussing their recent tournament victories. After about an hour, the manager went up to them and asked them to disperse. “But why?” they asked. “Because,” he said, “I can’t stand chess nuts boasting in an open foyer.” Mahatma Gandhi, as you know, walked barefoot most of the time, which produced an impressive set of calluses on his feet. He also ate very little, which made him rather frail and with his odd diet, he suffered from bad breath. This made him a super callused fragile mystic hexed by halitosis.

In each case, the humor is elicited by the realization that the communication is not intended to be informative but rather, sets up the reader for a play on words. Shaggy Dog Stories. These stories are characterized by a lengthy description of events that appear to lead up to something interesting or exciting but turn out to be totally mundane. In this case, the humor does not arise from an interpretation of the stories’ semantic content. Rather, it results from a reinterpretation of the situational context in which the information is presented. That is, the story’s length and content gives the appearance of leading to something of interest and importance, whereas the “punch line” reveals it to have no interest whatsoever. Slapstick. There are many forms of slapstick humor. A potential challenge to the diminishment assumption could be the Laurel and Hardy variety, in which amusement is elicited by protagonists’ repeatedly bludgeoning one another with two-by-fours. Certainly hitting someone with a stick is not inherently funny. The diminishment in this situation presumably arises from the realization that the protagonists are not, in fact, killed or even maimed by one another’s actions. That is, they only appear to hurt one another. Thus, the actual situation turns out to be more mundane than the purported one. Expectancy Deviations. In a study by Nehrhardt (1976), blindfolded participants were asked to judge a series of weights. After several exposures to weights that were fairly similar to one another, they encountered a weight that was either much lighter or much heavier than the others. On receiving this weight, subjects typically smiled or laughed. Assuming that participants were amused rather than simply registering surprise, the question is why. It seems reasonable to assume that participants who encountered the deviant weight inferred that they were being tricked and that the experiment was not a serious study of weight judgment after all. In other words, participants interpreted the situation as a whole as less im-

198

CHAPTER 8

portant than they had originally interpreted, and this reinterpretation elicited amusement. In summary, the diminishment assumption not only helps to account for a number of humor-eliciting phenomena that are hard for other theories to explain, but also accounts for conditions in which experiences are reinterpreted but do not elicit amusement. For example, scientific discoveries, the attachment of meaning to a person’s dreams, and the reinterpretation of events in mystery novels all involve a reinterpretation of a situation in light of new information. In these cases, however, the reality that is implied by the new information is of greater importance or value than the original, and so amusement is not experienced.

A COMPREHENSION–ELABORATION THEORY OF HUMOR ELICITATION The conceptualization of humor elicitation that Wyer and Collins (1992) proposed has much in common with incongruity-resolution theories and borrows particularly heavily from Apter’s (1982) formulation. At the same time, it is derived largely from the assumptions surrounding the comprehension of information outlined in previous chapters. Specifically, we assume that amusement is the by-product of comprehension processes similar to those described in chapters 4 and 5. However, the amount of humor that a joke elicits can depend on the difficulty of comprehending it. Moreover, it can be increased or decreased as a result of the amount and type of cognitive elaboration that is performed after its humor-eliciting interpretation has been identified. We first outline the basic assumptions of the formulation and provide empirical evidence of its implications. We then turn more specifically to the role of cognitive elaboration in humor elicitation and its implications for reactions to stories that people consider offensive. In our discussion, we focus largely on the humor elicited by jokes and stories to which event and episode models pertain. As Wyer and Collins (1992) indicated, however, the conceptualization is potentially applicable to the humor that is elicited by social experiences in general. Comprehension Processes Although the comprehension processes we assume to underlie humor elicitation have been elaborated in previous chapters, it may be worthwhile to reiterate them in the present context. This will again be done in terms of postulates. The first postulate is simply a recapitulation of the processes described in chapter 4.

DYNAMICS OF HUMOR ELICITATION

199

Postulate 8.1. New experiences are spontaneously interpreted in terms of previously formed event representations (e.g. situation models) with which they have common features. When more than one alternative knowledge representation is applicable for interpreting an experience, the representation that is most easily accessible in memory is typically applied. The result of this cognitive activity is a newly formed situation model of the event being comprehended.

When people have no goal in mind other than to comprehend the information they receive, the concepts and knowledge they apply are typically those they have used most frequently and recently in the past (cf. Higgins, Bargh, & Lombardi, 1985; Srull & Wyer, 1979; see Postulates 2.4 and 2.5). If recipients have a more specific goal in mind, however, this goal may activate concepts that are relevant to its attainment, and these concepts may then be used instead. The second postulate reiterates the processes that underlie the construction of episode models, and formalizes the effect of previously encountered stimulus events on the processing of information about subsequent ones. Postulate 8.2. Once a mental representation has been formed of an experience, this representation, along with the concepts and knowledge that were activated in the course of constructing it, is used to comprehend thematically related experiences that occur subsequently.

Thus, for example, information that a man enters a restaurant activates knowledge about things that are usually found in restaurants and the events that occur, and this knowledge provides the basis for comprehending events that occur subsequently. In some cases, the information may be quite redundant with previously acquired knowledge and therefore, as noted in chapter 6, it may not be retained (Trafimow & Wyer, 1993). Other events may not be redundant with one’s knowledge in the domain of concern but can nevertheless be interpreted in terms of concepts that are drawn from this knowledge. Thus, people do not often spill their drinks in a restaurant. Nevertheless, information that the man in our example spilled a glass of wine on his suit is likely to stimulate the construction of an event model in terms of concepts that compose one’s knowledge of restaurants and the things that are found there. On the other hand, suppose the man entered the restaurant, took off his clothes and started playing the guitar. These behaviors cannot be easily understood in terms of concepts that compose one’s knowledge of restaurants. Thus, they may stimulate a reassessment of the situation being described. (For example, recipients might conjecture that the restaurant was located in a nudist colony, and the man was an entertainer and not a guest.)

200

CHAPTER 8

Similarly, the initial reference to the nun in the “quickie” joke mentioned earlier activates expectations for her attributes that make her later comment to the priest difficult to interpret. Consequently, recipients identify a different domain of knowledge (e.g., knowledge about prostitutes) in order to comprehend it, with the result that the nun is reconceptualized as being a prostitute as well. These processes are captured by a third postulate, which essentially recapitulates assumptions made by Suls (1972, 1983) and Apter (1982): Postulate 8.3. When information about an event cannot be interpreted in terms of concepts drawn from the same domain of knowledge that was applied to previous events, recipients attempt to identify concepts and knowledge in a different domain that is applicable to both the new event and the preceding ones. If these concepts can be found, the previously learned events are reinterpreted in terms of them.16

This reinterpretation would not be made spontaneously, but would require higher order, goal-directed processing. Postulates 8.1 to 8.3 are generally applicable to the comprehension of information about a thematically related sequence of events. For humor to be elicited as a result of reinterpretations of the sort described in Postulate 6.3, additional factors must be considered. We describe these factors in turn along with empirical evidence bearing on their influence. The Role of Diminishment in Humor Elicitation 1. General Considerations The first factor we assume to underlie humor elicitation was suggested by Apter (1982) and has already been discussed. Postulate 8.4. Humor is elicited by a reinterpretation of a stimulus event only if the features of one or more features of the event are diminished in value or importance relative to that implied by the original interpretation.

Like Apter (1982), we further assume that diminishment can occur along many dimensions and at several levels of generality. For example, the reinterpretation of a stimulus event might paint a more mundane (or, in some cases, less desirable) picture of a person or event than the original interpretation implied. Alternatively, the reinterpretation could render the event it16 Note that Postulate 8.3 could be applied not only to stories but also to observed experiences or single statements that people make in a social context. To this extent, this postulate governs processes similar to those described in chapter 7.

16

DYNAMICS OF HUMOR ELICITATION

201

self, or the circumstances surrounding its occurrence, more mundane or trivial than the original interpretation implied. Thus, the reinterpretation of a statement describing person A’s behavior toward person B might diminish the importance of either an attribute of A, an attribute of B, A’s behavior toward B, the statement that describes the behavior, or the situational context in which the statement is made. Postulate 6.4 applies at any or all of these levels. 2. Empirical Evidence To the extent that a diminishing reinterpretation of the characteristics of a person or event is unfavorable, Postulate 6.4 would be consistent with disparagement or superiority theories of humor elicitation described earlier (cf. Zillman & Cantor, 1976). As we have noted, however, not all diminishing reinterpretations of an event are disparaging. Events that are initially interpreted as extremely aversive can elicit humor if they are later revealed to be less so. Shurcliff (1968) provided a good example. Some participants were told they would be asked to pick up and hold a white rat. Others were led to believe they would extract blood from the rat with a syringe. The latter condition was elaborately staged, with participants being required to wear a lab coat and being warned that the rat might bite. In both conditions, however, participants upon picking up the rat found that it was made out of rubber. After having this experience, participants were asked to rate the humor it elicited. Participants who had expected to extract blood from the rat rated the situation as more amusing than those who had only expected to handle it. This finding is consistent with Postulate 8.4. That is, the reinterpretation of the rat as a rubber toy led to a greater diminishment of the situation as a whole in the first condition, and the greater amusement it elicited could be attributed to this fact. An ambiguity in interpreting Shurcliff’s findings arises from the fact that the situations he constructed elicited anxiety. To this extent, his findings would be consistent with an arousal-reduction theory of humor elicitation. However, several studies in our own laboratory do not have this ambiguity. A study by Collins and Wyer (reported by Wyer & Collins, 1992) evaluated the effects of diminishment employing variations of the quickie joke described earlier. To reiterate, diminishment could come into play in two ways in comprehension of the story. First, the nun’s response to the priest’s question (“What’s a quickie?”), which reveals she is a prostitute, diminishes her status as a devout and chaste individual. Second, her response stimulates a reinterpretation of the priest’s question that diminishes its importance. Both factors could contribute to the humor that was elicited. To evaluate this possibility, Collins and Wyer constructed four versions of the story. One version was identical to the joke described earlier. In a second, the nun’s response to the priest’s question was “I’ll show you, but it

202

CHAPTER 8

will cost you $20, just like in town.” Thus, this response preserves the original meaning of the question rather than stimulating a reinterpretation of it. The third and fourth versions were similar to the first two except that the person the priest asked was a third prostitute rather than a nun; these versions, then, eliminated the shift in perception of the respondent. Results shown in Table 8.1 are very clear. That is, the original story, in which both diminishing shifts in meaning were present, was judged as amusing. However, eliminating either shift in interpretation decreased amusement, and eliminating both reinterpretations decreased it still further. In a second study by Collins and Wyer, participants read a story that could be interpreted in two different ways, one of which was less likely to be identified than the other. One story, for example, was likely to be interpreted spontaneously as a conversation about the best way to administer harsh physical punishment to children, but could also be interpreted as a discussion of the best way to open a jar of pickles. A second story was most likely to be interpreted as a man’s comments to a woman in the course of making love in the shower, but could also convey his comments in the course of washing a dog. No indication was given at the beginning of the story about the nature of its subordinate theme. In some versions, however, a statement was inserted near the end of the story that was anomalous when considered in terms of concepts activated by the dominant theme but made sense in terms of the subordinate one (specifically, “But honey, you know there’s nothing tougher than getting into a jar of pickles,” and “Honey, bring me the flea powder,” in the two stories, respectively). In other versions, this concluding statement was omitted. Participants were told to read the story for understanding as they would if they encountered it in a magazine or novel. After doing so, however, they reported how amused they were by it. Participants judged the stories more amusing when the statement that activated the subordinate, trivializing theme was present than when it was not. TABLE 8.1 Mean Humor Elicited by “Quickie” Joke as a Function of the Effects of the Punch Line on Interpretation of Story Features Effect of Punch Line on Perceptions of Nun

Effect of punch line on interpretation of statement Change in meaning No change in meaning

Change

No Change

5.60 3.33

3.67 2.18

Note. Judgments reported along a scale from 0 (not at all humorous) to 10 (extremely humorous). Adapted from Wyer and Collins (1992).

DYNAMICS OF HUMOR ELICITATION

203

Considered in isolation, this finding could have several alternative interpretations. For example, humor might be elicited simply by any reinterpretation of the story regardless of whether or not it was diminishing. This problem was eliminated in a study by Linda Isbell (reported in Isbell, Wyer, & Collins, 2002). Participants read a version of the “pickle” story that was similar to that employed by Collins and Wyer but varied in terms of the relative salience of the two alternative interpretations. The story was given a title that identified either the serious theme (“Disciplining the Children”) or the mundane one (“Getting a Pickle”). Participants read this story along with three filler stories with instructions that (a) the stories could each be about two quite different situations, (b) each story was given a title that identified one of the situations, and (c) they should generate an alternative title that reflected a different interpretation. (The three filler stories described situations that were similar in importance; e.g., a game of cards vs. a woodwind ensemble rehearsal.) After generating an alternative title for each story, they gave their reactions to the story along several 0–10 scales, one of which pertained to the amusement they experienced. As expected, participants were more amused when their reinterpretation of the story was more mundane than the original title conveyed (M = 6.12) than when it was less so (M = 4.29), and this was true regardless of which version of the story they read. Thus, the results of these studies converge on the conclusion that stories elicit more amusement if they are reinterpreted in a way that diminishes the importance of either their referents or the stories themselves. At the same time, the results do not indicate that diminishment is a necessary condition for humor elicitation. In the preceding experiment, for example, both reinterpretations that participants generated could have elicited some humor, differing only in magnitude. The study of people’s responses to witticisms, described in chapter 7, is worth reconsidering in this context. To reiterate, participants read some scenarios in which a statement was either likely to be taken literally or likely to be viewed as ironic. Moreover, the intended meaning of the ironic statement was either less favorable than its literal meaning (thereby diminishing the value of its referent) or more so. Diminishing reinterpretations elicited more amusement than enhancing ones (6.3 vs. 4.6 along a 0–10 scale), consistent with Postulate 8.4. On the other hand, both types of ironic statements were interpreted as substantially more amusing (M = 5.5) than their true counterparts (M = 1.7).17

17 A second aspect of these data is also worth noting. That is, statements with unfavorable implications were judged more amusing than statements with favorable implications, and this was true regardless of whether the statements were taken literally (3.5 vs. 1.9) or ironic (6.3 vs. 4.6). This finding, which would be consistent with disparagement theories (Zillman & Cantor, 1976), suggests that inconsistency resolution is also not a necessary condition for humor elicitation.

17

204

CHAPTER 8

Thus, these findings could suggest that diminishment is not a necessary condition for humor elicitation. On the other hand, the realization that a statement does not mean what it says could diminish the value or importance of the statement independently of its implications for the referent and, therefore, could elicit amusement for this reason. Therefore, in light of the ability for Postulate 8.4 to account for a wide range of humor elicitation phenomena that would otherwise be difficult to explain (slapstick humor, puns, etc.), it seems desirable to retain it pending more compelling evidence against its validity. Effects of Comprehension Difficulty 1. Theoretical Considerations The diminishing reinterpretation of information is often quite easy to identify. In some cases, however, the reinterpretation may be more difficult to generate, and may depend on recipients’ special knowledge of the persons and events involved. For example: Descartes and two of his buddies go into a bar. The two friends both order a scotch and soda. The bartender turns to Descartes and asks, “You, too?” Descartes replies, “I think not,” and immediately disappears.

The humor elicited by this joke presumably derives from the reinterpretation of Descartes’ reply in the context of his philosophical conclusion, “I think, therefore, I am.” However, someone who is not very familiar with Descartes’ philosophy might find the joke very bewildering. Even when a joke’s humor-eliciting interpretation can be identified, it might not be perceived as funny if an excessive amount of cognitive activity is required to understand it. On the other hand, jokes that are too easy to understand are also unlikely to elicit much amusement. Thus, as suggested by Suls’ (1972, 1983) problem-solving analogy, jokes that are either too easy or too difficult to comprehend may be less amusing than those that are moderately difficult. To formalize: Postulate 8.5. The amount of amusement that is potentially elicited as a result of reinterpreting a stimulus event is a nonmonotonic (inverted-U) function of the time and effort that is required to make this interpretation.

In a sense, this postulate is not new. Many years ago, McClelland, Atkinson, Clark, and Lowell (1953) also hypothesized a nonmonotonic relation novelty of a stimulus (and, therefore, the difficulty of understanding it in terms of previously formed concepts and knowledge) and judgments of its pleasantness. The question is why this relationship exists. Freud (1905/

DYNAMICS OF HUMOR ELICITATION

205

1960) suggested that the sheer pleasure of engaging in cognitive activity in relation to humor contributes to its gratification. Perhaps as comprehension difficulty increases up to a point, recipients feel more challenged, and their success in comprehending the information is rewarding (White, 1959). Beyond this optimal level, however, recipients might begin to feel stupid or incompetent, and so their enjoyment decreases. Therefore, humor elicitation might decrease correspondingly. 2. Empirical Evidence The results of several studies are consistent with Postulate 8.5. Of particular relevance is a study by Zigler, Levine, and Gould (1967). Participants read cartoons that systematically varied over four levels of comprehension difficulty. Both subjects’ facial expressions as they read the cartoons and their subsequent preference rating of the cartoons were recorded. An index of participants’ difficulty of comprehending the jokes was also obtained. This latter measure increased as expected over the four difficulty levels. However, preference rankings of the cartoons were higher when they were moderately difficult to comprehend than when they were either very easy to comprehend or very difficult (but not impossible) to comprehend. Codings of participants’ facial expressions revealed a similar pattern. Collins and Wyer (reported in Wyer & Collins, 1992) also supported the comprehension difficulty postulate using stimulus materials similar to those described earlier. Participants read a story that was likely to be interpreted spontaneously as concerning a serious situation (either making love in the shower or abusing children) but could alternatively be interpreted as concerning a mundane one (giving a dog a bath or opening a pickle jar, respectively). However, the salience of the mundane interpretation was varied in two ways. First, a statement with an interpretation that required concepts associated with the subordinate theme either was or was not inserted near the end of the story. Second, the story was preceded by a title that disposed subjects to think of either the serious theme or the mundane one. (Thus, the story that could be interpreted as either love making in the shower or giving a dog a bath was titled either “Mary’s bath” or “Spot’s bath.” Correspondingly, the story that could concern either child abuse or how to open a pickle jar was titled “Getting Out of a Pickle” or “Getting a Pickle”). According to Postulate 6.5, participants should be more amused when it was moderately difficult to identify the subordinate, diminishing interpretation than when it was either very easy or very difficult. This was in fact the case, as shown in Table 8.2. Pooled over the two stories, more humor was elicited when the humor-eliciting theme was suggested in the text than when it was not (6.36 vs. 4.21), suggesting that this theme was easier to identify in the former case than the latter. However, when the theme was suggested in the text, mentioning the theme in the title

206

CHAPTER 8 TABLE 8.2 Humor Elicited by Stories as a Function of Allusion to Subordinate Theme in the Title and Text Theme Suggested in Title

Lovemaking vs. dog’s bath story Subordinate theme suggested in text Subordinate theme not suggested in text Child abuse vs. pickle story Subordinate theme suggested in text Subordinate theme not suggested in text

Subordinate

Dominant

6.17 576

6.85 4.32

5.76 3.87

6.67 2.94

Note. Judgments are reported along a scale from 0 (not at all humorous) to 10 (extremely humorous). Adapted from Wyer and Collins (1992).

as well (thus making its identification even easier) decreased amusement (5.96 vs. 6.76, when the theme was versus was not conveyed in the text, respectively). In contrast, when the theme was not suggested in the text and so identifying it was more difficult, mentioning the theme in the title increased amusement (4.82 vs. 3.61, respectively). As an additional test of Postulate 8.5, we constructed different versions of the following joke: Q. A.

What did Adam say to Eve in the Garden of Eden? Stand back! I don’t know how big this thing gets.

The first interpretation of Adam’s response, which is induced by the warning, “Stand back!”, connotes danger. The implications of this interpretation are diminished by the subsequent realization that the “thing” is only an erection rather than a real danger and that the warning results from Adam’s sexual naivete. To understand the joke, however, one must realize that the “thing” is, in fact, an erection, and must know that Adam and Eve have no prior knowledge of sex. Identifying these concepts and knowledge requires cognitive effort, the magnitude of which should depend on whether the relevant information is explicitly provided. To evaluate this possibility, we constructed eight versions of the joke that varied in the explicitness of information bearing on three types of knowledge that were necessary to understand it: (a) the “thing” referred to an erection, (b) it was Adam’s first such experience, and (c) Adam did not know its size. Thus, the punch line when all three pieces of information were provided was “Stand back! This is my first erection and I don’t know how big it gets!” In contrast, the punch line when none of the three pieces of information was provided was simply “Stand back!”

207

DYNAMICS OF HUMOR ELICITATION TABLE 8.3 Humor Elicited by Adam and Eve Joke as a Function of the Information Mentioned in the Punch Line Ignorance of Size Mentioned

Erection mentioned Erection not mentioned

Ignorance of Size Not Mentioned

“First Time” Mentioned

“First Time” Not Mentioned

“First Time” Mentioned

“First Time” Not Mentioned

5.57 6.00

6.43 8.00

2.63 2.12

3.25 1.00

Note. Judgments are reported along a scale from 0 (not at all humorous) to 10 (extremely humorous). Adapted from Wyer and Collins (1992).

Humor ratings are shown in Table 8.3 as a function of the amount and type of information conveyed in the punch line. The joke was generally more amusing when the punch line explicitly stated that Adam did not know the size of the erection. More important, however, is the fact that when Adam’s lack of knowledge was not explicit, providing one or both of the other pieces of information increased the ease of understanding the joke and, therefore, increased the humor it elicited. When Adam’s lack of knowledge was explicitly mentioned, however, providing the other two pieces of information presumably reduced cognitive effort below the optimal level required to appreciate the joke. Therefore, it decreased humor ratings. 3. Situational and Individual Differences in Comprehension Difficulty The ease of identifying a humor-eliciting interpretation is not only a function of the explicitness of the information itself. Situational and individual difference factors can play a role as well. For example, situational variables that increase the accessibility of comprehension-relevant concepts in memory should increase the humor elicited by difficult-to-comprehend stimulus experiences but should decrease the humor elicited by easy-to-comprehend experiences. On the other hand, factors that distract recipients from comprehending the information in terms of these concepts should have the opposite effects. Two studies provide indirect support for this possibility. In a study by Schick, McGlynn, and Woolam (1972), participants were exposed to two sets of cartoon strips. All strips were originally from the “Peanuts” series and so the characters and their personalities were well known. However, half of the strips were redrawn to make the characters unfamiliar. Participants were exposed to several strips in succession and asked to indicate how

208

CHAPTER 8

amused they were by each. People are likely to have some difficulty comprehending strips containing unfamiliar characters as they do not have a previously formed body of knowledge to draw upon for use in construing the implications of the strips and their situational context. However, comprehension difficulty should decrease as the characters become more familiar, and so the humor the strips elicit should increase. On the other hand, suppose the characters in the strips are already familiar. Then, increases in the number of strips should not have this effect. In fact, humor judgments of the redrawn cartoon strips increased as a function of the number of exposures to the strips. In contrast, the humor elicited by the original “Peanuts” strips was high at the outset and did not change as a function of the number of exposures. A perhaps more interesting finding surrounds the combined effects of familiarity and participants’ chronic anxiety (as inferred from the Taylor Manifest Anxiety Scale). Chronic anxiety might function as internal noise that increases participants’ difficulty in comprehending the stimuli. If so, it should increase the humor elicited by familiar-character cartoon strips, which in the absence of this noise are below the optimal level of comprehension difficulty. However, anxiety might decrease the humor elicited by unfamiliarcharacter strips, which may be above the optimal level of difficulty in the absence of distraction. This was in fact the case. It can be difficult to interpret the effects of situational manipulations of comprehension difficulty without an a priori understanding of the difficulty level of the stimuli to be comprehended independently of these manipulations. A study by Goldstein, Suls, and Anthony (1972) provides an example. Subjects were initially shown photographs of either aggressive stimuli or automobiles as part of an aesthetic preference task, thereby activating concepts that were associated with the domain to which the pictures pertained. Later, in an ostensibly unrelated experiment, participants judged the funniness of cartoons in either the same domain to which the pictures they saw were relevant or in the other, unrelated domain. Cartoons elicited more amusement in the former condition than the latter. This could suggest that the concepts activated by the photographs facilitated the comprehension of the cartoons that participants encountered later and, therefore, increased the ease of identifying their humor-eliciting features. According to Postulate 8.4, however, this increase should only occur if the cartoons are fairly difficult to comprehend in the absence of this facilitating influence. If the cartoons were very easy to comprehend in the absence of priming, experiences that further increase the ease of comprehending them should have precisely the opposite effect. Thus, Goldstein et al.’s finding would only be consistent with the conceptualization we propose if the stimuli were of the first variety.

DYNAMICS OF HUMOR ELICITATION

209

COGNITIVE ELABORATION PROCESSES The humor that people experience in the course of identifying a communication’s diminishing reinterpretation theoretically occurs automatically. However, their spontaneous amusement may be increased or decreased as a result of deliberate cognitive elaboration that the individuals perform subsequently. If the individuals have constructed an event or episode model in the course of comprehending the situation described, for example, they may elaborate the mental image they have formed of the situation and the unmentioned events that occur subsequently. Their cognitive elaboration of the humor-eliciting aspects of the situation is likely to increase the amusement that the story elicits. On the other hand, some stories are considered offensive by the persons who encounter them. This could occur because the jokes are tasteless, are socially inappropriate in the situation in which they are told, or cast aspersions on a person or group that the recipients hold in high regard. In such cases, recipients’ postcomprehension cognitive activities may not pertain to the humor-eliciting reinterpretation per se. Rather, recipients may think about the motives of the person who conveyed the story and, if the story is disparaging, the storyteller’s attitude toward the protagonists. This humorirrelevant cognitive elaboration is likely to decrease the amusement the story elicits. These possibilities are formalized in the following postulate: Postulate 8.6. The amount of humor that is elicited as a result of reinterpreting a stimulus event is a monotonic function of the amount of cognitive elaboration of the event and its implications that occurs subsequent to its reinterpretation. 1. If recipients’ processing objective at the time is simply to comprehend and enjoy experiencing the event, cognitive elaboration of the event will typically increase the humor it elicits. 2. If recipients’ goal is more restricted, the event is elaborated in terms of its implications for this more specific objective. In this case, cognitive elaboration could either increase, decrease, or have no effect on the amount of humor elicited, depending on whether the humor-eliciting reinterpretation of the event is relevant to the attainment of this objective.

There is a qualification on the applicability of this postulate, however. The humor-eliciting aspects of some jokes are easy to elaborate. For example: A Texas Aggie and two friends are marooned on a desert island without food or water. Suddenly, the sky opens and a voice says, “Each of you may have one wish.” One friend says, “I wish I were in the arms of my loved one.” Immediately, he is gone.

210

CHAPTER 8

The second friend says, “I wish I were in the bosom of my family.” Immediately, he is also gone. The Texas Aggie thinks for a moment and says, “Gee, I wish my friends were with me now.”

The humor elicited by this story presumably results from the realization that the Aggie’s wish essentially negates the effects of the friends’ wishes, returning them to the same abysmal situation they had been in before. Once readers of the story identify these implications, they are likely to form a mental image of the friends’ reactions to finding themselves back on the island again. To this extent, the opportunity to think about the joke is likely to increase the amusement it elicits. In contrast, consider the following: Q. A.

Why did the Texas Aggie want people to save their burned-out light bulbs? He needed them for the darkroom he was building.

Although this joke might be mildly amusing, it has little elaboration potential and so thinking about the joke after comprehending it should have little impact on amusement. (Moreover, note that both jokes might be seen as disparaging the intelligence of Texas A&M University students, to whom the name “Texas Aggie” is typically applied. Students from this university might be inclined to elaborate the nonhumorous implications of both jokes, and this could decrease their amusement.) Effects of Repetition Some jokes continue to elicit amusement even when they are repeated one or more times. This can be true despite the fact that the punch line is well remembered and, therefore, no longer elicits a new interpretation of the stimulus events being described. Other jokes, however, are less likely to bear repeating, even though they might have been considered funny at the time they were first heard. Suls (1972) conjectured that repetition effects reflect a general tendency for novel stimuli to become better liked as they become more familiar (Zajonc, 1968; but see McClelland et al., 1953). If this were so, however, repetition should have similar effects on all jokes. Thus, this conceptualization cannot easily explain why some repeated jokes continue to elicit humor but others do not. It seems more reasonable to interpret repetition effect in terms of Postulate 8.6. If a joke has high elaboration potential, all possible implications of it are unlikely to be considered at the time the joke is first encountered. To this extent, repeating the joke might stimulate a different subset of implications than it did the first time, and these new implications could elicit hu-

DYNAMICS OF HUMOR ELICITATION

211

mor. Thus, the “desert island” joke described earlier is likely to elicit somewhat different mental images and elaborations each time it is told or thought about. Eventually, with still further repetitions, no new elaborations are likely to come to mind and so humor is no longer elicited. (All jokes become stale eventually.) However, jokes that have low elaboration potential to begin with (e. g., the “burned out light bulb” joke) may elicit little humor when they are repeated even once. Cognitive elaborations may be aided by external stimulation. In fact, much of the humor that is generated by professional comedians results from their ability to stimulate their audience to elaborate the implications of a humor-eliciting event with which they are already familiar. Moreover, many movies and stories concern protagonists whose humor-eliciting behavior, although initially unexpected, is repeated in one form or another throughout. Don Quixote, for example, repeatedly behaves in foolish ways while appearing to be distinguished and chivalrous. Inspector Clouseau, the French detective created by Peter Sellers in the Pink Panther movies appears sophisticated but is actually a buffoon. In such cases, the audience soon becomes aware of the protagonist’s actual characteristics, and so his or her later behavior is not unexpected. Yet, the behavior continues to provide amusement. For example, the first Pink Panther movie one sees leads the ostensibly sophisticated Inspector Clouseau to be reinterpreted as an inept detective who only appears sophisticated. Once this is established, new elaborations of the reinterpretation are encountered as the film presents further instances of Clouseau appearing sophisticated but being in reality a buffoon. The repeated instances of Clouseau’s humor-eliciting behavior constitute externally generated elaborations of one’s initial reinterpretation of the character and situation and function in much the same way as self-generated elaborations. These elaborations elicit humor for much the same reasons that self-generated elaborations do. These considerations help to conceptualize the humor elicited in a large number of situations in which the comprehension processes we postulate might otherwise seem irrelevant. That is, many everyday situations appear to elicit amusement although they do not require a reinterpretation. A colleague who is already known to be incompetent but who (perhaps like Inspector Clouseau) behaves in a pompous fashion elicits humor whenever he or she inadvertently says or does something that betrays this incompetence. Neither the new event nor the colleague’s attributes are reinterpreted. However, the colleague’s behavior essentially constitutes an externally generated elaboration of the implications of a past event (an earlier instance of the colleague’s buffoonery that occurred in the context of his or her pomposity) that did stimulate a diminishing reinterpretation at the time it occurred.

212

CHAPTER 8

Situational Influences on Cognitive Elaboration and Humor Elicitation Individual differences undoubtedly exist in the disposition to be offended by jokes that are considered socially inappropriate in the circumstances in which they are told, or that perpetuate a negative stereotype. These differences may be reflected in the humor the jokes elicit. However, these differences, which presumably result from differences in the type of cognitive elaboration that people perform after the implications of the jokes have been comprehended, should only be apparent when recipients are motivated or able to think about the jokes after their humor-eliciting interpretation has been identified. In the absence of this postcomprehension cognitive activity, people should be amused regardless of their disposition to be offended by them. A study with Linda Isbell (reported in Isbell, Wyer, & Collins, 2002) suggested this possibility. In this experiment, we obtained participants’ reactions to jokes that we expected to be more offensive to women than to men and, therefore, to produce different types of cognitive elaboration. Two jokes, which described explicit sexual activity (specifically, fellatio and masturbation), were ones that men typically find amusing but women find embarrassing or offensive. A third joke was scatological: Q. A.

Why do farts smell? For the deaf.

Sex differences in reactions to the latter joke were less clear a priori. Females might find the joke more embarrassing than men. On the other hand, both men and women might consider the joke to disparage people with disabilities. Moreover, cognitive elaboration of the joke could stimulate unpleasant olfactory images and might decrease humor for this reason as well. Be that as it may, we reasoned that if people are encouraged to think about the implications of these jokes, they should judge a joke to be more amusing if they elaborate its humor-eliciting aspects than if they think about its humor-irrelevant aspects. However, suppose participants give their spontaneous reactions to the joke without engaging in this postcomprehension elaboration. Then, they should be equally amused by the joke regardless of the sort of elaboration they might otherwise be disposed to perform. 1. Method We told participants we were interested in reactions to stories of the sort they encounter in daily life, and that to study this, we would like them to react to a number of stories of the sort they might hear in situations out-

213

DYNAMICS OF HUMOR ELICITATION

side the laboratory. With this preamble, participants in spontaneous-reaction conditions were told we were particularly interested in people’s spontaneous responses to stories, and that after reading each story, they should report their immediate reactions to it without thinking about it carefully. In contrast, participants in thought conditions were told we were interested in how people respond to a story after they have had time to think about it, and were urged to think carefully about the stories before reporting their reactions. Participants in each condition read a series of eight stories that included the three jokes described earlier, two jokes that were void of content that participants were likely to consider offensive, and two stories describing situations that were not expected to be amusing. Participants read each story/joke and then rated it along four 0–10 scales pertaining to their emotional reactions: sad, angry, amused, and offended. 2. Results The amusement elicited by inoffensive jokes was similar regardless of participants’ sex or instructional conditions (M = 5.10). Moreover, encouraging participants to think about these jokes did not appreciably increase their amusement. This could indicate that the innocuous jokes we selected were not sufficiently high in elaboration potential for an increase to be detected. Reactions to the sexual and scatological jokes are of greater interest. The amusement elicited by these jokes is shown in the left half of Table 8.4 as a function of participant sex and presentation conditions. As we expected, men were more amused by sexual jokes when they thought about them than when they did not (7.08 vs. 4.94, respectively), whereas women were less so (2.62 vs. 4.64, respectively). Therefore, although men were more amused by the jokes than women under thought conditions (7.02 vs. 2.62, respectively), this difference was not at all apparent under spontaneTABLE 8.4 Mean Ratings of Inoffensive, Scatological, and Sexual Jokes as a Function of Instructional Conditions and Distraction Amusement Ratings

Sexual jokes Males Females Scatological joke Males Females

Offensiveness Ratings

Spontaneous Reaction Conditions

Thought Conditions

Spontaneous Reaction Conditions

Thought Conditions

4.94 4.64

7.08 2.62

1.87 1.12

2.07 5.44

5.16 5.63

2.87 3.00

2.62 1.62

1.29 4.78

214

CHAPTER 8

ous-reaction conditions (4.94 vs. 4.64). The interaction of participant sex and instructional conditions was significant, F(1,42) = 4.55, p < .05. Gender differences in reactions to the scatological joke were not evident. In fact, both men and women reported being less amused by the joke when they were told to think about it (M = 2.94) than when they were told to give their spontaneous reactions to it (M = 5.40), F(1,42) = 3.97, p < .05, and this difference did not depend on participant sex (F < 1). Thus, both groups of participants appeared to elaborate humor-irrelevant implications of the joke under thought conditions. The humor-irrelevant cognitive elaboration that occurred when participants had a chance to think about the jokes was presumably stimulated in part by their perception that the jokes violated social standards of taste or portrayed activities they considered undesirable. However, the jokes’ offensiveness was apparently not the only determinant of the elaboration they performed. Participants’ offensiveness ratings of the offensiveness of the three types of jokes are summarized in the right half of Table 8.4. Participants judged both sexual and scatological jokes to be inoffensive when they were asked to report their spontaneous reactions. However, whereas men continued to consider these jokes to be inoffensive when they were told to think about them, women increased their judgments of the jokes’ offensiveness in these conditions. These conclusions are confirmed by marginally significant interactions of instructional conditions and participant sex in analyses of reactions to both sexual jokes and the scatological joke (in each case, p < .07). However, the fact that men were less amused by the scatological joke under thought conditions despite not being offended by it suggests that the humor-irrelevant elaboration they performed in this condition was not a result of the joke’s offensiveness per se. Correlational analyses confirmed these conclusions. That is, the correlation between amusement ratings and offensiveness ratings, computed separately for each experimental condition, was generally low (between -.16 and -.45) and was significant in only one case. 3. Effects of Distraction If the different effects of instructions on reported amusement are mediated by differences in the cognitive elaboration that participants performed when they were told to think about them, distracting participants from engaging in this cognitive activity should decrease the magnitude of these effects. To examine this possibility, we constructed an additional condition in which participants received instructions similar to those administered under thought conditions of the main experiment. In addition, however, participants were told that we wished to simulate conditions outside the laboratory in which people hear stories when they are preoccupied with other

DYNAMICS OF HUMOR ELICITATION

215

things, and so we would like them to listen to a tape-recorded conversation while reading the stories. (The conversation was an intrinsically interchange between a man and a woman that was taken from the opening scenes of Edward Albee’s Who’s Afraid of Virginia Woolf?). Results were consistent with expectations. Specifically, distracting participants from thinking about sexual jokes reduced the sex difference in amusement that occurred under thought conditions (7.08 vs. 2.62 for men and women, respectively; see Table 8.4) to a level (5.12 vs. 3.94, respectively) that did not significantly differ from the level we observed under spontaneous reaction conditions (4.94 vs. 4.64, respectively). Correspondingly, distracting both men and women from thinking about the scatological joke, which they reported as unamusing under thought conditions (M = 2.93) increased their amusement to a level (M = 4.25) similar to that observed when participants gave their spontaneous reactions (M = 5.38). It therefore seems reasonable to conclude that distraction disrupted the cognitive elaboration that participants performed when they were told to think about the jokes and, as a result, it decreased the effect of this elaboration on their amusement. Sexual Humor in Social Interaction The preceding experiment provides evidence that both situational and individual difference factors combine to influence the type of cognitive elaboration that people perform in response to a joke and the amusement they experience as a result of this elaboration. The conclusions drawn from the study are nonetheless limited by the failure to take into account the social context in which the jokes are conveyed. People may respond quite differently to sexual jokes that are told by members of their own sex than to those that are told by members of the opposite sex. This could be particularly true of females, who may interpret a man’s spontaneous relating of such a joke to be rather boorish and insensitive. An experiment by James Collins (reported in Isbell, Wyer, & Collins, 2002) provided evidence concerning this contingency and, moreover, gave insight into the cognitive processes that underlie reactions to the jokes. The study is particularly provocative as the jokes were ostensibly conveyed spontaneously under conditions that were not part of the experiment itself. Specifically, participants were male and female undergraduates who took part in the experiment in mixed sex groups. The experiment was conducted by either a male or female undergraduate who was ostensibly helping the faculty member obtain reactions to some humor materials he was preparing for a future experiment. At this point, however, the experimenter spontaneously remarked, “Say, that reminds me. I just heard a new joke. Maybe he (Dr. Collins) would like this one. Let me try it out on you . . .” The

216

CHAPTER 8

experimenter then proceeded to convey one of two jokes. One, sexual joke, was the “Adam and Eve” joke described earlier in this chapter (see Table 8.3). The other was irrelevant to sexuality. After this aside, the experimenter indicated that before going on with Dr. Collins’ experiment (s)he would like the participant to perform two unrelated task. One was a filler task that was intended to disguise the relatedness of the studies. The second was a word association task in which participants were asked to select one of three words that was most similar to a fourth. One alternative in each set had sexual overtones. For example: Feeling: touch tingle stimulate Oral: eating sex communication Excited: winning aroused horny

Finally, ostensibly as part of the main study, participants were asked to report the amusement elicited by approximately 45 humorous passages, 12 of which had sexual content. Collins assumed that the sexual joke told by the experimenter would stimulate male participants to engage in humor-relevant cognitive elaborations regardless of whether the joke was told by a man or a woman. However, females were expected to engage in these elaborations when the joke was told by another woman, but to engage in humor-irrelevant elaboration when the joke was told by a man. The nature of these elaborations is likely to be reflected in the type of concepts that were accessible in memory at the time participants performed the word association task and, therefore, the frequency of sexuality-related words they would select. Data bearing on this possibility are shown in the top half of Table 8.5. Both male and female participants made a greater number of sexuality-related word associations TABLE 8.5 Mean Sexuality-Related Word Associations and Amusement Elicited by Sexual Jokes as a Function Participant Sex, the Experimenter’s Sex, and the Type of Joke the Experimenter Told Sexual Joke Told by Experimenter Male Experimenter Sexuality-related word associations Male participants 3.75 Female participants 1.00 Amusement elicited by sexual jokes Male participants 5.78 Female participants 4.40

Nonsexual Joke Told by Experimenter

Female Experimenter

Male Experimenter

Female Experimenter

5.40 4.86

3.00 2.00

1.89 1.92

6.28 6.96

5.13 5.52

4.42 3.96

DYNAMICS OF HUMOR ELICITATION

217

when they had heard a female tell a sexual joke than when they had heard her tell a nonsexual joke, indicating that the female’s joke activated sexuality-relevant concepts. Correspondingly, as shown in the bottom half of Table 8.5, they were more amused by the sexual jokes they encountered in these conditions. In contrast, male participants’ sexuality-related word associations increased only slightly when a male experimenter told a sexual joke, and female participants’ sexuality-related associations actually decreased in these conditions. As a result, the male experimenter’s telling of a sexual joke only slightly increased males’ amusement in response to the sexual jokes they encountered later, and decreased women’s amusement.18 Combined Effects of Elaboration and Comprehension Difficulty Although we have considered the effects of comprehension difficulty and cognitive elaboration separately, these effects are actually interdependent. For one thing, people cannot elaborate the implications of information they are unable to interpret. Moreover, the experience of difficulty in comprehending the information could itself stimulate cognitive elaborations. For example, if recipients believe they have had more difficulty understanding a joke than they should, they might be stimulated to think about themselves and their competence rather than about the humor-eliciting aspect of the information. Such humor-irrelevant elaborations could offset any amusement that the joke would otherwise elicit. A third consideration arises from the fact that people are typically neither willing nor able to devote an unlimited amount of time to the processing of any given stimulus event (Sperber & Wilson, 1986; Taylor & Fiske, 1978). This is particularly true in a social interaction in which a continuous stream of events commands attention. Suppose people are unwilling to spend more than a given amount of time thinking about a stimulus event. To this extent, the more time and effort they devote to comprehending a stimulus event, the less time they will devote to a subsequent elaboration of its implications. If this is so, the combined effects of comprehension difficulty and cognitive elaboration on humor elicitation can be described more precisely. Suppose the total time that people are willing to devote to (a) the comprehension of an experience and (b) the cognitive elaboration of its humor18

Our interpretation of these results of course assumes that participants’ cognitive elaborations of the experimenter’s joke activated concepts that influenced their interpretation of the sexual jokes they encountered later, thus increasing the humor the jokes elicited. This assumption is supported by a significant correlation between the amusement they reported in response to these jokes and the number of sexuality-related word associations that participants made (r = .39, p < .05).

18

218

CHAPTER 8

eliciting implications is a constant. Wyer and Collins (1992) showed that under these conditions, the elaboration potential of a stimulus not only has a direct effect on the humor elicited but also influences the relation between humor elicitation and comprehension difficulty. Specifically, less comprehension time is required to maximize the humor an experience elicits when the elaboration potential of the experience is high than when it is low.

REACTIONS TO DISPARAGING HUMOR Jokes and stories often perpetuate a negative stereotype or for other reasons disparage the persons or groups to which they pertain. Some people are amused by such jokes and others are not. Substantial research has been devoted to the identification of personality and individual difference factors that account for these differences. Much of this research has been stimulated by the assumption that laughing at others’ deficiencies is a manifestation of displaced hostility or aggression (Freud, 1905/1960). Alternatively, it may reflect a desire to maintain a feeling of superiority by derogating others or by otherwise calling attention to their flaws (Wills, 1981; Zillman, 1983; Zillman & Cantor, 1976). One obvious implication of this assumption is that people are more amused by jokes that disparage individuals who are particularly threatening to their self-esteem. Thus, for example, jokes that disparage minority group members are more likely to elicit amusement in persons who feel their economic or social status is being threatened by these groups than in individuals who feel more secure. The possibility that humor is elicited by disparagement per se cannot be entirely discounted (see Footnote 18). According to the conceptualization we propose, however, the humor elicited by a joke or story is a joint function of the identification of its humor-eliciting implications in the course of comprehending it, and the amount and type of cognitive elaboration that occurs subsequently. To this extent, individual differences in responses to disparaging humor are potentially traceable to (a) the type of previously formed concepts and knowledge that people bring to bear on their comprehension of a joke and (b) their motivation to engage in humor-eliciting or humor-irrelevant cognitive elaboration, assuming they have an opportunity to do so. Effects of Comprehension Difficulty People are usually less amused by a joke that disparages a stereotyped group to which they personally belong than a group to which they are indifferent. This difference could be due in part to the type of cognitive elabora-

DYNAMICS OF HUMOR ELICITATION

219

tion that people perform in the two conditions. However, it could be localized at the comprehension stage as well. For example, jokes often mention the ethnic group to which they pertain at the outset. People who do not expect to be offended by a disparaging joke about this group may spontaneously activate a body of knowledge about attributes that are stereotypically associated with the group and use this knowledge to comprehend the rest of the joke or story. In such cases, activating the stereotype may facilitate comprehension of the joke and, perhaps, increase the humor that it elicits. On the other hand, people who identify with the group may anticipate that the joke is going to offend them, and may think about the motives of its source. This humor-irrelevant activity could interfere with the identification of the joke’s humor-eliciting interpretation. Whether or not this interference increases or decreases amusement, however, depends on the difficulty of comprehending the joke more generally. The extent to which mentioning a stereotyped group activates humorirrelevant cognitive activity can also depend on characteristics of the storyteller. A joke that pertains to Jews is more likely to stimulate this activity in a Jewish listener than in a Ku Klux Klan member. On the other hand, it is less likely to do so if the joke-teller is Jewish than if (s)he belongs to the KKK. However, not all jokes that refer to members of a stereotyped group are disparaging. If people who anticipate that a joke will perpetuate a negative stereotype engage in cognitive activity that interferes with comprehension, it could affect the amusement the joke elicits even if the joke turns out not to perpetuate the stereotype at all. Consider the following joke, for example: A bus passenger notices that the woman sitting beside him is staring at him. “I’m sorry to bother you,” she says, “but I wonder if you would mind my asking you a personal question. Are you by any chance Jewish?” “I certainly don’t mind your asking,” replies the man good-naturedly, “but no, I don’t happen to be Jewish.” The woman continues to stare at him, however, and asks again, “Are you sure you’re not Jewish?” “Yes, I’m sure,” says the man, now somewhat irritated. “My mother is Roman Catholic, and my father is Japanese. There’s really no chance that I am Jewish.” This still doesn’t satisfy her, and as the trip continues, she becomes even more insistent. “Now, I know that you’re Jewish. Tell me you’re Jewish. Admit it!” Finally, after repeated denials, the man, only to keep from being pestered further, says, “Okay. Have it your way, I’m Jewish.” “Well,” the woman responds, “you certainly don’t look Jewish.”

Thus, the joke does not disparage Jews, but rather, disparages the stereotype. However, this does not become clear until the punch line. The allusion

220

CHAPTER 8

to the ethnic group in the material preceding the punch line could stimulate Jewish recipients to engage in cognitive activity that interferes with their later comprehension. To this extent, they might consider the joke to be less funny than non-Jewish recipients would, even though the joke’s implications are consistent with their a priori values. (This conjecture assumes that the joke is already moderately difficult to understand. If it were very easy to understand, the increases in comprehension difficulty resulting from this activity could increase Jewish individuals’ appreciation of the joke rather than decreasing it.) Effects of Cognitive Elaboration Although individual differences in attitudes and values undoubtedly influence the comprehension of a joke or story, their primary effects are likely to result from their impact on the cognitive elaboration that occurs subsequently. Suppose individuals encounter the following joke: Q. A.

What is the difference between a Jew and a canoe? A canoe tips.

This joke is extremely easy to understand. Its humor derives from the different meanings of “tips,” coupled with knowledge of the stereotype of Jews as being tight with money. Thus, differences in the humor the joke elicits are likely to be due largely to differences in the type of cognitive elaboration that recipients perform subsequent to its comprehension. That is, people who are generally offended by the perpetuation of the stereotype may think about why the joke is told and may question the attitudes and motives of the storyteller. However, people who personally believe the stereotype, or are not concerned with perpetuating it, might not engage in this humor-irrelevant elaboration. There are two qualifications on this prediction, however. First, differences in reactions to the joke should only be evident when people have an opportunity to elaborate its implications after comprehending it. When this opportunity does not exist, recipients may be equally amused by the joke regardless of their feelings about the group being disparaged. (This could explain why people often laugh at jokes they hear at a party that they would find unamusing in a less cognitively demanding situation in which they had more opportunity to think about the jokes’ implications.) A second contingency surrounds the elaboration potential of a joke. When a joke’s humor-eliciting reinterpretation evokes visual images, or if the situation it implies can be elaborated, individuals who are not offended by the joke may elaborate its humor-eliciting implications and, therefore, may be more amused by the joke than they might otherwise be. However,

DYNAMICS OF HUMOR ELICITATION

221

jokes of the sort provided in our example, which are simply plays on words, have low elaboration potential. The amusement elicited by these jokes is unlikely to increase with the opportunity to engage in elaboration regardless of recipients’ attitudes toward the stereotyped group. In fact, if individuals have nothing to distract them, any additional postcomprehension thoughts they might have about the joke could pertain to the source of the joke, and perhaps the validity of the stereotype, and these thoughts could occur even if these individuals do not identify with the group being disparaged. As a result, the amusement experienced by these individuals, like persons who are disposed a priori to elaborate the humor-irrelevant implications of the joke, might decrease when they have an opportunity to think about the joke more extensively. A second study with Linda Isbell (see Isbell et al., 2002) examined this possibility. Participants under thought and spontaneous reaction conditions similar to those in Isbell’s earlier study were asked to read and give their reactions to six stories. Four of the stories were nonhumorous, but two others were jokes that had negative implications for the protagonist’s intelligence. One of these jokes (e.g., the “desert island” joke described earlier) had high elaboration potential, whereas the other (e.g., the “burnedout light bulb” joke) had low elaboration potential. Moreover, the protagonist in one joke was described as Polish and in the other was described as a Texas Aggie. (The type of protagonist associated with each joke was of course counterbalanced.) The midwestern college students who participated in this study were undoubtedly aware of the stereotype of Polish as unintelligent. However, they were typically not of Polish extraction themselves, and were unlikely to be offended by the jokes. (Participants’ reactions to the jokes in the experiment confirmed this assumption; 65% of the participants reported being “not at all” offended by these jokes, or 0 along the 0–10 scale used to report their judgments.) Although Texas Aggies are sometimes stereotyped as unintelligent as well, this stereotype was generally unknown in the student population from which participants were drawn. Based on the assumptions outlined earlier, we expected that encouraging participants to think about when Polish jokes with high elaboration potential would stimulate them to elaborate the humor-eliciting aspects of the jokes, leading them to be more amused by the jokes than they would otherwise be. In contrast, encouraging participants to think about Polish jokes with low elaboration potential should stimulate them to elaborate the jokes’ negative implications for the stereotyped group to which the jokes refer, and thus should lead them to be less amused by the jokes than they would otherwise be. Results were consistent with these assumptions. Participants reported being equally amused by Texas Aggie jokes regardless of whether they were told to think about them (M = 4.16) or to give their spontaneous reac-

222

CHAPTER 8

tions (M = 4.52). In contrast, thinking about Polish jokes with high elaboration potential increased the amusement the jokes elicited from 4.04 (under spontaneous reaction conditions) to 5.13 (under thought conditions), whereas thinking about Polish jokes with low elaboration potential decreased amusement from 4.96 to 3.86. These differences were reflected in an interaction of elaboration potential and instructional conditions in an analysis of reactions to Polish jokes alone, F(1,40) = 4.27, p < .05. Thus, the results provide insight into people’s reactions to jokes that perpetuate a negative stereotype. Thinking about a joke that perpetuates a negative stereotype can obviously decrease the amusement experienced by persons who are offended by the joke. However, thinking about such a joke can also decrease the amusement experienced by persons who are not offended by the joke, provided the joke’s humor-eliciting implications cannot easily be elaborated. The Effects of Social Context on Responses to Sexist Humor Humor-irrelevant cognitive elaborations are particularly likely when participants themselves belong to the stereotyped social category and, therefore, are apt to be offended by it. However, this is not the only factor that can influence humor-irrelevant elaboration. In some cases, recipients might perceive that a joke-teller is motivated by a desire to ingratiate his or her audience, or alternatively, to convey hostility toward the person or group that the joke disparages. They might think about the implications of these motives regardless of whether or not they are personally offended by the joke. Consequently, they might be relatively less amused by the joke than they would be if the joke-teller were simply trying to be funny. In other cases, people might be aware of the motives that lead a joke to be told and still elaborate its humor-eliciting implications. (This is particularly true if the joke’s elaboration potential is high, as suggested by the results of Isbell’s experiment described earlier.) The effects of these contextual factors may be particularly evident (at least, among college students) when jokes exploit a negative stereotype of men or women. Consider the following joke: A man and a woman are in the elevator together. The woman suddenly starts taking off her clothes and says, “Make me feel like a woman.” The man takes off his clothes, throws them on the floor, and says, “Here, fold these.”

Women’s reactions to this joke, which perpetuates a negative stereotype of women, are likely to differ from men’s reactions to it. However, the na-

DYNAMICS OF HUMOR ELICITATION

223

ture of these reactions may depend on both the sex of the communicator and the social context in which the joke was originally told. The effects of these factors may be mediated by their influence on recipients’ assumptions about the speaker’s motives and, therefore, on the amount and type of cognitive elaboration that they perform. 1. Perception of the Motives for Telling Sexist Jokes The motives that people perceive to underlie the communication of sexist jokes were investigated by Isbell et al. (2002). Male and female college students were introduced to the study with instructions that we were particularly interested in the motives that people attribute to individuals who tell jokes that convey negative stereotypes of men and women. With this preamble, we gave participants a short questionnaire containing four (female-bashing) jokes that perpetuated a negative stereotype of women and four other (male-bashing) jokes that conveyed a negative stereotype of men. Each joke was preceded by an indication of the sex of both the person who told the joke and the individual to whom it was told. (Each joke was associated an equal proportion of times with each of the four speaker–audience configurations.) Participants were asked to imagine that each joke had been told in the situational context described and to estimate the likelihood that the joketeller was (a) just trying to be funny, (b) trying to gain social approval, or (c) being intentionally hostile or aggressive. They also estimated the likelihood that the joke-teller personally believed in the stereotype being conveyed. These estimates, reported along scales from 0 (not at all) to 10 (extremely), are shown in Table 8.6. as a function of the joke-teller’s sex, the sex of the audience, and the type of joke that was told. Between-cell comparisons, shown in the table, suggest the following conclusions: 1. Participants in all cases inferred that the joke-teller’s primary motive was to be funny (M = 7.30). However, they perceived this motive to be particularly strong (M = 8.08) when someone told a joke that disparaged the opposite sex to a member of his or her own sex (i.e., when a man told a femalebashing joke to another man, or a woman told a male-bashing joke to another woman). 2. Joke-tellers were perceived to have less strong personal beliefs in the stereotype implied by a joke when the joke disparaged members of their own sex (M = 3.16) than when it disparaged members of the opposite sex (M = 5.10). However, this difference was evident both when men told a malebashing joke (2.28 vs. 4.97) and when women told a female-bashing joke (4.09 vs. 5.23).

224

CHAPTER 8

3. Joke-tellers were judged more likely to be seeking approval when they either told a male-bashing joke to a female or told a female-bashing joke to a male (M = 5.81) than when they told these jokes to members of the sex that was being disparaged (M = 3.19). This was true regardless of whether the joke disparaged males (5.75 vs. 3.37) or females (5.87 vs. 3.00). 4. Persons who told male-bashing jokes were attributed less hostility if they were male (M = 1.37) than if they were female (M = 3.97). In contrast, persons who told female-bashing jokes were seen as hostile regardless of their sex (3.27 vs. 3.05 when the jokes were told by men and women, respectively). In addition, joke-tellers were attributed greater hostility if they told a joke to members of the sex being demeaned (M = 3.32) than if they told it to members of the opposite sex (M = 2.51). This difference was evident regardless of whether the joke disparaged men (3.07 vs. 2.26) or women (3.56 vs. 2.75). 2. Context Effects on Amusement and Offensiveness If the motivational attributions described in Table 8.6 stimulate humorirrelevant elaboration, they should decrease the humor the sexist jokes elicit. As our earlier experiments indicate, however, this elaboration may not always occur. To understand the effects of contextual factors on humor elicitation, it is necessary to distinguish between the audience to whom a speaker intends to communicate and the actual recipient. To this end, Isbell et al. (2002) constructed stimulus materials using the same eight jokes employed in the previous study (Table 8.6). One stimulus replication comprised four filler TABLE 8.6 Motives Attributed to Tellers of Sexist Jokes as a Function of the Joke-Teller’s Sex, the Sex of the Audience, and Joke Type Male Joke-Teller Male Audience Motivation for telling male-bashing jokes To be funny Belief in stereotype Social approval Hostility Motivation for telling female-bashing joke To be funny Belief in stereotype Social approval Hostility

Female Joke-Teller

Female Audience

Male Audience

Female Audience

M

7.67ab 2.72b 3.69b 1.66b

7.47ab 1.84b 6.75a 1.09b

6.59b 5.53a 3.06b 4.47a

8.12a 4.41a 4.75b 3.44a

7.46 3.62 4.56 2.66

8.04a 4.58ab 6.46a 2.79

7.00ab 5.88a 2.62b 3.75

7.08ab 4.17b 5.29a 2.71

6.46b 3.92b 3.38b 3.38

7.14 4.63 4.43 3.16

Note. Means in each row with unlike superscripts differ at p < .05.

DYNAMICS OF HUMOR ELICITATION

225

jokes had nothing to do with sex-role stereotypes, two male-bashing jokes and two female-bashing jokes. The second replication was composed of the same filler jokes and the remaining four stereotype-related jokes. Participants were told we were interested in how people respond to jokes that were conveyed in social contexts outside the laboratory, and that to study this, we had collected jokes in a number of informal conversations that had taken place on campus. On this pretense, we asked participants to give their reactions to eight jokes, preceded in each case by a description of the context in which the joke had ostensibly been told. (For example, one joke had ostensibly been told by a male graduate student to two female students at a coffee house, another had been told by a female undergraduate to three undergraduate men at a fraternity party, etc.) Participants were told to imagine they had actually overheard the joke being told conveyed in this context, and to indicate both how amused they would be by the joke and how offended they would be. These ratings were made along scales from 0 (not at all) to 10 (extremely). Each of the four filler jokes was ostensibly told to a mixed audience. The joke-teller and audience assigned to the four target jokes were varied in a partial latin square design so that pooled over participants and stimulus replications, each target joke was told about the same number of times in situations that represented each possible combination of speaker sex and audience sex. The top half of Table 8.7 shows the amusement elicited by each type of joke as a function of the sex of the joke-teller, the sex of the audience, and the sex of the participants. The bottom half of the table conveys the offensiveness of the jokes. The implications of these data are conveyed most easily by considering separately participants’ reactions to (a) jokes that disparaged their own sex and (b) jokes that disparaged the opposite sex. Reactions to Jokes That Disparaged Participants’ Own Sex. First consider jokes that perpetuate a negative stereotype of males. Men were less amused by these jokes when they were told by a woman to a group of men (M = 2.43) than they were in any of the other three joke-teller/audience conditions (M = 4.19). Moreover, they were nonsignificantly more offended by the jokes in this condition (M = 3.00) than in the other three conditions (M = 1.86). The data summarized in Table 8.6 suggest why this might be so. That is, male-bashing jokes are perceived to convey greater hostility when women tell them to a man (M = 4.47) than they are under other conditions (M = 2.06). In addition, joke-tellers are attributed stronger beliefs in the stereotype in the former condition than in others (5.53 vs. 2.99). It seems reasonable to suppose that men who encounter a woman telling a male-bashing joke to men are offended by the joke as a result of these attributions and

226

CHAPTER 8

cognitively elaborate its humor-irrelevant implications. Consequently, they are less amused by the joke than they would otherwise be. Note, however, that male-bashing jokes are also perceived to reflect hostility, and a strong belief in a negative male stereotype, when a woman tells these jokes to other women (see Table 8.6). Despite this fact, men were not offended by the jokes in this condition (M = 1.63). Moreover, they were more amused by the jokes in this condition than they were in other speaker–audience conditions. Thus, men may have attributed hostility to a woman who told a male-bashing joke, and may have thought that she personally believed in the stereotype she was perpetuating. Nevertheless, they were only offended when she told the joke to men and, therefore, appeared to be intentionally insulting. Women’s reactions to jokes that disparaged females showed a different pattern. As shown in Table 8.7, women were not very amused by femalebashing jokes in any condition (M = 3.64). Moreover, although women were somewhat less amused when joke-tellers told these jokes to someone of their own sex than they told them to someone of the opposite sex, this difference was not reliable (p > .10). Perhaps women are generally threatened by the perpetuation of a female stereotype and, therefore, are not amused by female-bashing jokes regardless of the context in which the jokes are TABLE 8.7 Males’ and Females’ Ratings of Jokes as Amusing and Offensive as a Function of Joke Type, the Sex of the Joke Teller, and the Sex of the Intended Audience Male Joke-Teller Male Audience

Female Audience

Female Joke-Teller Male Audience

Female Audience

Amusement Ratings Male participants Male-bashing joke Female-bashing joke Female participants Male-bashing jokes Female-bashing jokes

4.21 4.18

4.03 4.72

2.43 3.51

4.33 5.07

6.17 3.37

6.80 3.77

5.96 4.04

6.89 3.39

Offensiveness Ratings Male participants Male-bashing jokes Female-bashing jokes Female participants Male-bashing jokes Female-bashing jokes

1.46 1.00

2.50 1.98

3.00 1.93

1.63 1.43

0.90 2.43

0.87 3.70

1.57 3.54

0.39 3.23

DYNAMICS OF HUMOR ELICITATION

227

told. In contrast, men are not particularly threatened by the perpetuation of a male stereotype. Therefore, they only fail to be amused by male-bashing jokes when the joke-teller is intentionally insulting. Jokes That Disparaged the Opposite Sex. Men (but not women) were less amused by female-bashing jokes when the jokes were told to women (M = 3.84) than when they were told to men (M = 4.90), and this was true regardless of the joke-teller’s sex. Correspondingly, women (but not men) were less amused by male-bashing jokes when these jokes were told to men than when they were told to women (6.06 vs. 6.85). Thus, participants appeared to be more amused when they perceived the joke-teller to be seeking social approval (see Table 8.6). Conceivably both men and women recognized this motivation in all conditions. However, they only elaborated the humor-irrelevant implications of the jokes when the jokes disparaged members of their own sex. 3. Conclusions The results of Isbell et al.’s studies are complex in detail. They nevertheless confirm the general conclusion that the social context in which a stereotype-perpetuating joke is told can influence the amount of humorirrelevant cognitive elaboration that occurs in response to it and, therefore, the amount of humor that the joke elicits. Moreover, although this elaboration is likely to concern the reason why the joke was conveyed, it does not necessarily result from perceptions that the joke is particularly offensive. In some instances, jokes were reported as funny despite being offensive. In other cases, jokes were not considered very funny despite being perceived as relatively inoffensive. Perhaps the offensiveness of the jokes was not a determinant of the humor-unrelated cognitive elaboration that occurred, but instead, was a consequence of this elaboration. That is, the social context in which a joke is told might create expectations for the joke’s content, and a violation of these expectations might stimulate attempts to construe the speaker’s motives. This cognitive activity could decrease the humor the joke elicits independently of the motive that the speaker is ultimately attributed.

CONCLUDING REMARKS Humor is a fundamental ingredient of social communication in everyday life. A specification of the conditions in which it is elicited is therefore a necessary component of any theory of real-world information processing. We have argued that the experience of amusement is partly a result of the ba-

228

CHAPTER 8

sic processes of comprehension outlined in previous chapters of this book. That is, it occurs when the episode models that have been formed spontaneously in the course of comprehending a sequence of events must be modified in order to understand the sequence as a whole, thereby stimulating higher order, goal-directed information processing in order to perform this modification. However, although the reinterpretation of information that results from this processing is a necessary antecedent of humor elicitation, it is not sufficient. In addition, the reinterpretation that occurs must generally diminish the value or importance of either the events described in the information being processed, the participants in these events, the situational context in which the information is conveyed or, in some cases, the information itself. Moreover, the humor that is elicited by a social experience is also a function of the cognitive elaboration that occurs after its humor-eliciting reinterpretation has been identified. The amount and type of this elaboration, which depends on both the stimulus information itself and the prior knowledge and values of the recipient, can either add to or offset the amusement that might otherwise be experienced. The formulation we have proposed in this chapter does not intend to be the final word. However, it provides a framework within which future research and theorizing on humor elicitation can be conceptualized and investigated. At the same time, the comprehension and elaboration processes that underlie humor elicitation can also come into play in responding to social information more generally and in making judgments and decisions on the basis of this information. The last section of this volume is devoted to a discussion of these processes. Although our discussion in this chapter has focused on the antecedents of humor, we have ignored a consideration of its consequences. This concern, which is obviously important, is beyond the scope of this book. One obvious by-product of amusement, however, is the experience of positive affect. To this extent, it could have an impact on information processing similar to that of positive feelings more generally. These effects are also discussed in the final section of this volume.

P A R T

III INFERENCE PROCESSES

C H A P T E R

9 Crimes, Vacations, and Political Candidates: The Construction and Use of Narratives in Social Judgments

The discussion of comprehension processes in the preceding chapters was focused largely on the construction of narrative representations of social experience. To the extent that these representations occupy the bulk of our social knowledge (Schank & Abelson, 1995), they are correspondingly the primary basis for judgments and decision to which this knowledge is relevant. In this section, we consider the conditions in which narrative representations are brought to bear on judgments and decisions and the processes that underlie their use. Narrative representations can enter into inference processes in several ways. As suggested in chapter 3, some inferences occur spontaneously when comprehending a sequence of events in terms of a previously formed narrative representation (Spiro, 1977; see also Graesser, Singer, & Trabasso, 1994). However, narrative-based inferences can also be the result of conscious goal-directed activity. In some cases, a narrative account of real and imagined events might be particularly relevant to the objective at hand (e.g., planning how to behave in order to please a date, construing the events that lead up to a murder in a detective novel, or predicting the consequences of taking a new job). In other cases, a narrative representation is not essential. Nevertheless, information that is conveyed in a form that facilitates the construction and use of a narrative may be easy to comprehend in terms of previously acquired knowledge, and so inferences may be based on this type of representation rather than other types. The next two chapters discuss narrative-based inference processes in some detail. In the present chapter, we consider the conditions in which narrative representations of new information are constructed and used to 231

232

CHAPTER 9

make inferences about their referents. However, the construction of these representations often requires the activation and use of previously formed knowledge structures, many of which constitute implicit theories about the causes and consequences of social events. The use of these implicit theories, and of previously formed narrative representations more generally, are discussed in chapter 10.

THE IMPACT OF NARRATIVES ON JURY DECISION MAKING One of the first systematic explorations of narrative-based inference processes was conducted by Pennington and Hastie (1986, 1988, 1992) in research on jury decision making. Jurors are presumably motivated to determine whether a person is guilty or innocent of a crime. (This situation is not one that normally occurs in daily life. However, similar motives are likely to operate when reading a news story, watching a movie, or reading an Agatha Christie mystery novel.) To make this determination, people may attempt to reconstruct the sequence of events that led up to the crime as well as the crime itself and its aftermath. This reconstruction can sometimes be difficult. For one thing, the events are not always described in the order they occurred. In movies and mystery novels, for example, events that occurred early in the sequence are often not revealed until near the end. Courtroom testimony is typically organized according the witnesses who provide it, and is not conveyed in the order in which the events described are likely to occur. In such cases, people are likely to find it cognitively taxing to fit the various pieces of evidence into a scenario that permits their temporal and causal relatedness to be evaluated. As a result, they may be inclined to use other strategies to evaluate the evidence. In a jury trial, for example, they might attach more importance to the credibility of witnesses who provide individual pieces of evidence rather than considering how well the evidence fits into a coherent account of the crime. Pennington and Hastie (1986, 1988, 1992) provided compelling evidence of narrative-based decision processes and the factors that influence them. Participants in one study (Pennington & Hastie, 1988) read the transcript of a murder trial and were asked to think aloud as they arrived at a decision about the defendant’s guilt. Content analyses revealed that participants’ thought protocols were more likely to have a “story” structure than to have other possible (e.g., categorical) structures. A further study confirmed this conclusion. That is, participants who had read the testimony and reported

CONSTRUCTION AND USE OF NARRATIVES

233

their verdict were given a recognition memory test containing some statements that had actually been made in the course of the testimony and others that had not. Participants were more likely to report a statement as having been mentioned in the story if its implications were consistent with the narrative-based causal scenario on which they based their verdict than if its implications were inconsistent with this scenario. A second series of studies (Pennington & Hastie, 1988) examined these inference processes more directly. Participants listened to the transcript of an actual murder trial containing testimony for both the prosecution and the defense. In witness-order conditions, the evidence favoring a given verdict was organized according to the witness who provided it, as in the original trial. In story-order conditions, the evidence for a given side was presented in the order it became relevant in the sequence of events that had allegedly occurred. (That is, testimony about events that preceded the incident was presented first, followed by testimony about the incident itself, the arrest, the autopsy, etc.). This was done in a 2 ´ 2 design so that the order of presenting the prosecution testimony and the order of presenting the defense testimony were varied independently. Participants after receiving the testimony reported both their verdict and their confidence that their decision was correct. When the defense and prosecution testimonies were presented in different orders, 73% of the participants recommended the verdict implied by the evidence that was conveyed in story order. When the evidence favoring both verdicts was presented in the same way, the percentage of participants who favored a guilty verdict did not depend on how the evidence was ordered. However, participants were much more confident of their judgments when the evidence was in story order. In short, evidence that was conveyed in an order that facilitated the construction of a narrative account of the crime was apparently easier to comprehend, and so recipients were more confident of its implications. Consequently, they were more likely to base their judgments on this evidence than on evidence conveyed in other ways. This conclusion has further implications. For example, if people have formed a narrative of the sequence of events that occur, they may be more receptive to additional information that is consistent with the implications of the narrative as a whole than to information that conflicts with these implications. Moreover, this could be true regardless of the credibility of the information’s source. In contrast, if people are unable to construct a coherent narrative representation of the information, they may construe the implications of each piece separately. In this case, the credibility of source of this additional information should be given more weight.

234

CHAPTER 9

A COMPARISON OF NARRATIVE-BASED VERSUS NONNARRATIVE-BASED INFORMATION PROCESSING General Considerations A narrative account of a crime is of particular relevance in assessing the likelihood that the defendant was motivated or able to commit it. Participants in Pennington and Hastie’s studies may conceivably have tried to construct a narrative account of the events that occurred regardless of how the information was presented, but simply found it easier to attain this objective when the information was conveyed in temporal order. As these authors point out, however (Hastie & Pennington, 1995), many of the decisions we are called upon to make in daily life are not typically based on narratives. Rather, they are based on spatial, functional, hierarchical, and argumentative criteria. The mental representations that are formed in the course of applying these criteria may be quite different. Thus, for example, computer commands might be cognitively organized on the basis of functional considerations, and the arguments bearing on a particular issue might be organized according to the alternative points of view they support. Goal-related differences in the representations that people form of information, which were discussed in chapter 6, are considered more generally by Pennington and Hastie (1993). Even within a knowledge domain, a number of alternative strategies might be used to compute a judgment or decision. For example, the evaluation of a consumer product could be based on either an imagined sequence of events that surround the use of the product and its consequences or a piecemeal analysis of the product’s individual features (cf. Fiske & Pavelchak, 1986). In such cases, the strategy that people use, and the mental representation they construct in the course of applying it, may depend on situational or informational factors that exist at the time the information is received. Two factors—the format in which the information is presented and the presence or absence of pictures—are of particular interest. 1. Effects of Presentation Format on Processing Strategies When information describes a temporally ordered sequence of events that are situationally constrained, people may spontaneously construct an episode model of this sequence in the course of comprehending it (Postulate 4.1). Under these conditions, people are likely to base their judgments on the implications of the model as a whole without analyzing its individual features. In contrast, suppose individuals receive the same information in an ostensibly unordered list. In this case, they may not attempt to assess

CONSTRUCTION AND USE OF NARRATIVES

235

the temporal or thematic relatedness of the events unless this relatedness is particularly relevant to the goal they are pursuing. Instead, they may adopt a piecemeal processing strategy. That is, they may assess the evaluative implications of each piece of information independently and integrate these implications using a mechanistic procedure similar to that described by algebraic models of information integration (cf. Anderson, 1971, 1981; Fishbein & Hunter, 1964). Both holistic and piecemeal processing strategies may become very well learned as a result of numerous past experiences in which they have been applied. Personal experiences are inherently in the form of a narrative, and people have undoubtedly acquired procedures for construing their implications that they activate and apply automatically to new experiences that are similar in form. Verbal descriptions of temporally related events of the sort that compose a personal experience could likewise stimulate a holistic inference process. However, people also learn procedures for aggregating a number of unrelated items. This learning could occur in both formal situations (e.g., solving arithmetic problems in school) and informal ones (e.g., making a decision on the basis of a consideration of its possible consequences, or estimating whether one has enough money to pay for the groceries one wishes to purchase). Once learned, these procedures could be used to construe the implications of information that is conveyed in a manner similar to that in which the procedures have typically been applied in the past. To this extent, the procedures could function as “if [X], then [Y]” productions of the sort proposed by Anderson (1983; see also Smith, 1984, 1990), which are activated and applied spontaneously whenever the eliciting conditions for their application [X] are met. The format in which information is received could be one of these eliciting conditions. 2. The Role of Visual Imagery The mental representations that people form in the course of using these procedures are likely to differ in an important respect. Event and episode models consist in part of visual images (Postulate 4.2). Pictures of the events to which these models refer are likely to facilitate the construction of the models and to increase the detail with which their referents are encoded into memory. To this extent, the pictures should increase the extremity of the evaluations that are based on these models. In contrast, the information that is conveyed in a list, which is not temporally related, may be represented only semantically, as a network of features. Moreover, the piecemeal integration procedure that people activate and use to integrate the implications of these semantic features may not normally involve a consideration of mental images. To this extent, pictures might not facilitate a computation of evaluations on the basis of this strategy. In fact, they could even interfere.

236

CHAPTER 9

The role of pictures and visual images in the computation of evaluations is surprisingly unclear. Lampel and Anderson (1968) applied an information integration model in assessing the way in which pictures and personality trait descriptions combined to influence female participants’ judgments of a man’s datability. Their data were consistent with an averaging model in which the weight attached to the man’s physical appearance was inversely related to its favorableness. (More generally, women appeared to perceive the man’s physical attractiveness to be a necessary but not sufficient condition for dating him.) In Lampel and Anderson’s study, however, pictures were an independent source of information about the person being rated and, therefore, were used in addition to the verbal information provided. When pictures simply elaborate the same attributes that are described verbally, their impact is not as clear. Studies of consumer behavior suggest that when pictures accompany nonnarrative text (e.g., product attribute descriptions), they often have little impact on judgments over and above the effect of the text information alone (Edell & Staelin, 1983; see also Costley & Brucks, 1992). In fact, this impact may only be appreciable when recipients have little personal interest in the verbal information they receive about the products they evaluate (Miniard et al., 1991). (In these latter conditions, recipients might tend to base their judgments on pictures alone without bothering to assess the implications of the verbal information at all.) Two series of experiments by Adaval and her colleagues examined the interactive effects of pictures and information presentation format in some detail. In one set of studies (Adaval & Wyer, 1998), participants evaluated vacations of the sort they might personally experience. In a second set (Adaval, Isbell, & Wyer, 2003), participants evaluated politicians on the basis of information about events that occurred over the course of their lives. Thus, the two sets of studies differed in the extent to which situationspecific episode models were likely to be formed in the course of comprehending the event descriptions and, therefore, in the likelihood that the events would elicit visual images spontaneously. The Role of Narratives in Judgments of Imagined Personal Experiences In the studies by Adaval and Wyer (1998), participants read two travel brochures. One brochure described a vacation in India and the other, a vacation in Thailand. Each brochure began with a general overview of the vacation, followed by descriptions of 12 places that the vacationers would visit. These events were always conveyed in the same order. In one brochure, however, the events were described in a narrative, and in the other, they

CONSTRUCTION AND USE OF NARRATIVES

237

were conveyed in an ostensibly unordered list. In narrative-format conditions, for example, the India vacation brochure began: On your vacation, you will start out from the capital of India, Delhi, and move on to see the Taj Mahal. Later, you will go west and see the palaces and temples in the colorful deserts of Rajasthan . . . before heading south. Further south, you will visit the beaches of Goa, the tropical forests and backwaters of Kerala, and . . . complete your trip at the southernmost tip of India.

This was followed by written narrative descriptions of the places to be visited. For example: Only a short trip from Delhi is Agra, the home of the Taj Mahal. A mausoleum built by Shah Jahan for his empress, the Taj is widely regarded as the most beautiful man-made structure in the world. Remarkable at all times of the day, you will visit as the sun rises above the early morning mists and return on a moon-lit night when the Taj seems to float unattached above the blue-green pools in front of it.

In contrast, the India brochure under list-format conditions described aspects of the vacation without indicating the order in which the events described would occur. This brochure began with a simple listing of highlights: Some features of your vacation experience are: · a visit to the capital, Delhi · the cool mystery of the forests · the forest tribes · the Taj Mahal at Agra · palaces and temples in Rajasthan, etc.

Moreover, the individual places and situations were described in a similar manner: · Agra, home to one of the most beautiful man-made structures in the world. · the Taj Mahal—a mausoleum built by Shah Jahan for his empress. · a beautiful spectacle both when the sun rises above the early morning mists and on moonlit nights when the Taj seems to float above the blue-green pools in front of it.

238

CHAPTER 9

1. Experiment 1 In the first study, an attractive picture accompanied the verbal description of each event. However, the relative dominance of the pictures and verbal descriptions varied. In some cases, the pictures were large and the verbal descriptions were in small (10-point) font. In other cases, the pictures were relatively small and the verbal descriptions were in large (18-point) font. To ensure that format and picture-word dominance were not confounded with information content, the particular vacation representing each combination of these variables was counterbalanced. Participants after reading both brochures evaluated the extent to which the brochure pertaining to each vacation made them want to go there. Then, they reported the difficulty they had had in imagining each of the two vacations and finally, recalled the events contained in each brochure. Processing Efficiency. The ease of processing the information under each presentation format condition was inferred from both participants’ self-reported ease of imagining each vacation and their memory for events that would occur in it. Data pertaining to these indices are shown in the first two sections of Table 9.1. When the text information was dominant, participants found it fairly easy to imagine the vacation regardless of format, and had relatively better recall of the places they would visit when the information was conveyed in a list. When the information was conveyed in a narrative, however, increasing the dominance of the picture decreased the difficulty of imagining the vacation and increased recall of the places to be visited. When the same information was presented in a list, on the other hand, increasing the dominance of the pictures had the opposite effects. In other words, pictures appeared to facilitate the processing of information that was presented in a narrative. However, they interfered with the semantic, piecemeal processing of this information when it was conveyed in a list. Effects on Judgments. The combined effects of format and pictures in evaluations of the vacations are consistent with the processing differences just described. The bottom section of Table 9.1 shows evaluations of the first vacation that participants considered.19 When the text was dominant, participants rated the vacation somewhat less favorably when the information about it was conveyed in a narrative than when it was described in a list. However, increasing the dominance of pictures increased evaluations 19 Participants appeared to use the same criteria to judge the second vacation they considered that they used to judge the first one, even though the information was conveyed in a different format. Thus, the effects of format on judgments of the second vacation were the mirror image of its effects on judgments of the first. (This was true in the second experiment to be reported as well.)

19

239

CONSTRUCTION AND USE OF NARRATIVES TABLE 9.1 Effects of Information Presentation Format and Picture-Text Dominance on Recall and Judgments

Difficulty of imagining vacation Pictures dominant Text dominant Proportion of places recalled Pictures dominant Text dominant Evaluations of first vacation presented Pictures dominant Text dominant

Narrative Format

List Format

3.2 4.5

6.3 4.7

.52 .48 7.3 6.5

.37 .55 4.1 7.9

Note. Adapted from Adaval and Wyer (1998).

of the vacation in the first case and decreased it in the second. As a result, evaluations of a vacation when pictures were dominant were much more favorable when the vacation was described in a narrative than when it was described in a list. 2. Experiment 2 A second experiment confirmed these conclusions. In this study, the text-dominant condition was replaced by one in which participants received only text information and no pictures were presented at all. In addition, some participants were explicitly told to imagine themselves actually having the experience described on each page of the brochure. We reasoned that if the effects of picture results from the mental images they stimulate explicitly telling participants to form their own mental images of the events might have similar effects. Results confirmed this reasoning. Participants’ evaluations of the first vacation they considered are shown in Table 9.2 as a function of presentation format, the presence of pictures and imagination instructions (see Footnote 20). When the vacation was described in a narrative, participants evaluated it more favorably when both pictures were presented and they were told to imagine the events described (M = 8.1) than they did under other conditions (M = 6.4). When the vacation was described in a list, however, participants evaluated it somewhat less favorably in the first condition than in other conditions (5.8 vs. 6.1). Thus, both studies converge on the conclusion that people form mental images of events that are conveyed in a narrative, and that situational conditions that encourage the construction of these images increase the extremity of their judgments. On the other hand, these same conditions interfere with the piecemeal processing that under-

240

CHAPTER 9 TABLE 9.2 Evaluations of Vacations as a Function of Presentation Format, Presence of Pictures and Imagination Instructions Narrative Format

List Format

8.1 6.3

5.8 6.8

6.4 6.5

6.1 5.5

Imagination instructions Pictures presented No pictures No imagination instructions Pictures presented No pictures Note. Adapted from Adaval and Wyer (1998).

lies people’s computation of evaluations on the basis of listed information and, therefore, decrease the extremity of these evaluations. 3. Effects of Information Inconsistency Research in other domains suggests that people who are asked to judge a stimulus give greater weight to unfavorable information about the stimulus than to favorable information about it (Birnbaum, 1974; Skowronski & Carlston, 1989; Wyer, 1973, 1974). However, this may only be true when each piece of information is considered in isolation. If the information is conveyed in a narrative, people presumably base their evaluations on the implications of the sequence of events as a whole, and so unfavorable information that is embedded in otherwise favorable information is likely to have less impact. Supplementary data obtained in Adaval and Wyer’s first experiment confirmed this speculation. In some conditions, descriptions of two relatively unpleasant aspects of the vacation (e.g., poor living accommodations, an arduous bus ride, etc.) were included among the favorable event descriptions. Although adding these undesirable features decreased participants’ evaluations of the vacation, this decrease was much greater when the information was listed (from 6.0 to. 4.7) than when it was conveyed in a narrative (from 6.9 to 6.4).

THE INFLUENCE OF NARRATIVES ON POLITICAL IMPRESSION FORMATION We speculated that the format in which information is presented could activate a cognitive “production” that is applied automatically, with little awareness. On the other hand, it could also induce an implicit expectation for the type of processing strategy that should be applied, and participants might consciously attempt to comply with this expectation. In Adaval and

CONSTRUCTION AND USE OF NARRATIVES

241

Wyer’s studies, it is unclear which was the case. It was also unclear whether the facilitating effects of pictures are restricted to conditions in which episode models are likely to be formed from the information presented or whether their effects generalize to conditions in which the narrative information is unlikely to activate these models (e.g., abstract descriptions of events that are not temporally or situationally constrained). Adaval, Isbell, and Wyer (2003) provided evidence of this generalizability. Some of the more important judgments that people are called upon to make outside the laboratory are their evaluations of candidates for public office. These judgments are presumably based in part on the candidates’ stands on social and political issues. Ever since the 1960 Kennedy–Nixon debates, however, it has been widely recognized that voters’ evaluations of politicians are influenced at least as much by the global images of the candidates that are largely unrelated to the candidates’ political ideology or specific issue positions (Englis, 1994). These images are undoubtedly stimulated in part by a candidate’s physical appearance, as conveyed in a magazine or on television. However, they can also be influenced by descriptions of the individual’s activities over the course of his or her lifetime that have implications for intelligence, morality, or steadfastness. Thus, information that a candidate had smoked marijuana in college, or performed a heroic deed in World War II, can have an impact on inferences about the candidate’s character and, therefore, judgments of his or her suitability for office at a much later point in time. Adaval et al. (2003) examined the processes that underlie the construction of the images that people form of politicians using procedures similar to those employed in the vacation studies. The first two studies confirmed Adaval and Wyer’s earlier findings and, in doing so, provided evidence that the processing strategies that participants employed under different presentation format conditions were activated and applied automatically, with little awareness. Two other studies used recognition response times to understand more specifically the cognitive representations that were formed of information under conditions in which the strategies were employed. A final study explored the role of both pictures and mental imagery in political information processing and clarified further the processes that contributed to judgments in the earlier experiments. Participants in an initial study were told we were interested in the impressions people form of famous personalities. On this pretense, they read brochures about two political figures who had been well known during the period they held office. Each brochure began with an overview of the major events in the politician’s life, followed by 12 additional pages each describing an event or situation in more detail. As in the vacation studies, the information in one brochure was conveyed in a narrative and the information in the other was conveyed in the form of a list. In narra-

242

CHAPTER 9

tive-format conditions, for example, the brochure describing one politician (Thomas Winters) began: Thomas Winters was a well-known political figure between 1950 and 1975. He was a veteran of World War II and served as an executive of General Motors before becoming Governor of Michigan. He then served two years as a U. S. Senator, and ended his career as a special envoy to China.

This paragraph was followed by a series of paragraphs, each describing a different event that occurred during the politician’s career and the point at which it occurred. For example: He left General Motors to become Governor of Michigan. There, he showed sensitivity to public interests. Upon assuming office, for example, he went on television to oppose the construction of a nuclear waste processing plant near Detroit that would contaminate the city’s water supply.

Other activities included urging the government to halt the bombing in Vietnam, donating his summer home for use by a charitable organization, hosting the Pope during his visit to America, and helping to revise the state budget to provide support for crime prevention. In contrast, the brochure under list-format conditions described the events in the politician’s life in bullet form and did not indicate their temporal relatedness. Thus, the brochure pertaining to Winters began: Thomas Winters was a well-known political figure between 1950 and 1970. He was: · a member of the U. S. Senate · a World War II veteran · a General Motors executive · Governor of Michigan · special envoy to China

Although the individual events were conveyed in the same order they were presented in narrative-format conditions, they were also conveyed in bullets that had no temporal implications: · Was sensitive to the interests of the public while Governor of Michigan. · Went on television to oppose the construction of a nuclear waste processing plant that would contaminate the city’s water supply.

Information about each life event was presented on a different page of the brochure. In picture conditions, the information was accompanied by a black-and-white photograph. In some cases, the picture showed the politi-

243

CONSTRUCTION AND USE OF NARRATIVES

cian engaged in an activity related to the event (giving a speech, talking to someone, etc.) whereas in other cases, it portrayed an event-related situation in which the individual was not involved. (Thus, for example, a statement that the politician had headed a committee to investigate how to decrease violent crime was accompanied by a policeman at the scene of a murder.) The pictures were taken from books and magazines. (Pictures of Henry Kissinger were used for one of the two politicians, and pictures of Robert McNamara were used for the other. Pretesting indicated that neither politician’s face was familiar to the college student population from which participants were drawn.) Participants after reading the brochures reported the favorableness of their impressions of each politician along a scale from -5 (very unfavorable) to +5 (very favorable) and estimated the difficulty of imagining the sequence of events that occurred in the person’s life. Finally, they were asked to recall the events they had read about. As the aforementioned examples indicate, many of the events described in the information were not situationally and temporally constrained. Moreover, they were widely separated in time and occurred in unrelated situations. Therefore, the events were unlikely to be integrated into a single episode model. Nevertheless, our results were similar to those observed by Adaval and Wyer (1998). As shown in the top half of Table 9.3, participants in the absence of pictures evaluated politicians less favorably when the information about them was conveyed in a narrative than when it was listed. However, introducing pictures increased evaluations in the former condition and decreased it in the latter. As a result, evaluations of the politicians when pictures were presented were more favorable when the information was conveyed in a narrative than when it was listed. Although these differences were small in magnitude, the interaction of format and pictures was reliable (p < .05). Recall data, shown in the bottom half of the table, had a similar pattern. That is, pictures increased the recall of events that were described in a narrative but not when they were listed, suggesting that they TABLE 9.3 Impressions of Politicians and Number of Events Recalled as a Function of Format, the Presence of Pictures, and Presentation Order

Impressions of politician Pictures No pictures Number of items recalled Pictures No pictures

Narrative Format

List Format

3.98 3.58

3.63 3.95

5.79 5.04

4.92 4.98

Note. Adapted from Adaval et al. (2002, Experiment 1).

244

CHAPTER 9

facilitated memory for the events in the former case but not the latter. In contrast to Adaval and Wyer’s studies (see Footnote 20), these effects were evident for both the first politician that participants considered and the second. Deliberative Versus Spontaneous Influences of Format on Information Processing Strategies If participants intentionally use a holistic or piecemeal strategy when they encounter information in a narrative or a list, respectively, explicit instructions to employ the strategies should override the effects of format. On the other hand, suppose the strategies have the status of cognitive productions that are activated and applied automatically when information to which they are applicable is encountered. Then, they may have an influence regardless of participants’ conscious attempts to employ other criteria. To evaluate these possibilities, participants in a second experiment were exposed to stimulus materials identical to those in the first study. In this case, however, we told participants explicitly which processing strategy we would like them to use. Under piecemeal-instruction conditions, we indicated: Sometimes people form impressions of a person by focusing on specific events in a person’s life (e.g., winning a lottery). At other times, however, they try to think about the person’s life as a whole and whether it was admirable, successful, happy, etc. In this study, we would like you to use the first strategy. That is, when reading about the political figures described in the brochures, try to imagine the specific events that occurred in each politician’s life and their implications for your impression of the politician. Use these individual events as a basis for your impression.

Under holistic-instruction conditions, participants were told: Sometimes people. . . . happy, etc. In this study, we would like you to use the second strategy. That is, when reading about the political figures described in the brochure, try to imagine each politician’s life as a whole and use this as a basis for your impression.

1. Judgments and Recall In fact, telling participants what processing strategy they should use had little impact on the judgments they made. Table 9.4 shows participants’ evaluations as a function of instructions, presentation format, and the presence of pictures. Results replicated those of the first experiment. That is,

245

CONSTRUCTION AND USE OF NARRATIVES TABLE 9.4 Impressions of Politician as a Function of Format, Task Demands, the Presence of Pictures, and Presentation Order

Schematic instructions Pictures No pictures Piecemeal instructions Pictures No pictures Mean Pictures No pictures

Narrative Format

List Format

3.94 3.46

3.61 4.03

3.76 4.00

3.12 4.02

3.85 3.73

3.37 4.03

Note. Adapted from Adaval et al. (2002, Experiment 2).

adding pictures slightly increased evaluations of politicians who were described in a narrative and decreased evaluations of politicians based on information conveyed in a list. As a result, the advantage of a narrative over a list format was significantly greater when pictures were presented (3.85 vs. 3.37) than when they were not (3.73 vs. 4.02), F(1,56) = 7.49, p < .01. These effects did not significantly depend on the instructions that participants were given concerning the strategy they should use (p > .10). The effects of task demands cannot be ruled out entirely, however. Note that pictures facilitated impressions to the greatest extent when the information was conveyed in a narrative and participants were told to apply a schematic criterion (3.94 vs. 3.46 under picture and no-picture conditions, respectively). Correspondingly, they interfered with impressions to the greatest extent when participants were told to base their judgments on specific events and the information as conveyed in a list (3.12 vs. 4.02, respectively). These data therefore suggest that participants did attempt to apply the criteria they were told to use, and these conscious efforts added slightly to the spontaneous effects that were induced by the format of the information. However, neither the effect of task demands nor its interaction with pictures was statistically significant. Thus, the impact of these demands was very small in relation to the effects of presentation format per se. Recall data also confirmed the results of the first experiment. That is, participants recalled a significantly greater number of events when the information was conveyed in a narrative than when it was conveyed in a list (4.91 vs. 4.22), and this effect was greater when pictures were presented (4.99 vs. 3.85) than when they were not (4.82 vs. 4.58). However, these effects also did not depend on instructional set.

246

CHAPTER 9

2. Self-Report Data The failure for participants’ judgments to be appreciably influenced by explicit instructions concerning the criteria they should use suggests that the strategies that are elicited by different presentation formats were applied unintentionally. In fact, the use of these strategies, and the impact of pictures on their effectiveness, may occur without awareness. Several aspects of the supplementary data we collected suggest this conclusion. For one thing, participants’ self-reports indicated they attempted to apply the same strategy in evaluating the second politician they saw that they had applied in judging the first one, even though the information about the second politician was conveyed in a different format. In fact, however, their actual judgments of the two politicians were quite dependent on format in much the same way. Second, the effect of format on participants’ actual evaluations of the politicians depended on whether pictures were presented but did not depend significantly on task demands (see Table 9.4). In contrast, its effect on the criteria that participants reported using as a basis for these evaluations depended on task demands but not on the presence of pictures. Moreover, analyses of participants’ estimates of the extent to which pictures facilitated the impressions they formed did not significantly depend on either task demands or presentation format. These results therefore argue against the possibility that the effects of presentation format reflect a conscious attempt to comply with implicit format-based expectations concerning the criteria to be used in making judgments. Rather, different presentation formats appear to activate different processing strategies that participants apply automatically, and that have effects independently of the strategies they actively try to use. The presence of pictures may nevertheless facilitate or interfere, depending on the particular strategy that participants actually employ. The Role of Pictures in Narrative-Based Processing: A Reconceptualization Although the interactive effects of pictures and presentation format in the preceding studies were very similar to those observed by Adaval and Wyer (1998), they nevertheless raise questions about the content and structure of the mental representations that underlie these effects. The assumption that the event descriptions conveyed under narrative-format conditions were stored in memory as a sequence of temporally related events was confirmed by Adaval et al. (2003) in a study of recognition memory. Participants were exposed to stimuli under conditions comparable to those constructed in the first experiment by Adaval et al. (2003). Rather than judging the politicians, however, they were given a timed recognition memory test

CONSTRUCTION AND USE OF NARRATIVES

247

for the verbal descriptions of the events presented. The order in which the recognition items were presented was systematically varied. We assumed that if two presentation items were associated in memory, exposure to one of these items would spontaneously activate the second, and so the speed with which the second item could be identified would be increased. When the original information had been conveyed in a narrative, participants were quicker to identify an event if it was preceded by a description of the event that had come before it in the presentation sequence (M = 3.68 s) than if it was preceded by a different item (M = 4.16 s). When the items had been conveyed in a list, however, this difference was negligible (4.21 s vs. 4.18 s). These data suggest that the events were organized in memory in a temporal sequence when they were conveyed in a narrative, but that this was not the case when the events were presented in a list. The influence of pictures on the representations that participants formed is less clear. We assumed that when information was conveyed in a narrative, participants would form mental images of the events and that these images would become part of the representations of these events that they stored in memory. Unlike Adaval and Wyer’s (1998) studies, however, many of the pictures used in the present experiment were only indirectly connected to the event descriptions they accompanied. (E.g., the verbal description of Harrison as displaying courage under enemy gunfire in World War II was accompanied by a picture that showed him in an army uniform but did not portray him as actually engaged in combat. Similarly, the verbal description of him as showing sensitivity to the needs of the poor and disabled, and as initiating legislation to provide assistance to homeless children, was accompanied by a picture of him seated a desk rather than as actively involved in this behavior.) Thus, although the pictures may have facilitated the construction of mental images of the events presented, their specific content may not itself have been part of the representation that was formed and stored in memory. A second recognition memory study confirmed this speculation. In this study, both pictures and verbal event descriptions were used as test stimuli. Participants did not recognize an event description any more quickly if it had been preceded in the recognition series by the picture that had accompanied it in the original stimulus materials than if it had been preceded by a novel picture, and this was true regardless of whether the events had been conveyed in a narrative (1.84 s vs. 1.81 s) or in a list (1.66 s vs. 1.60 s).20 In other words, the specific picture that accompanied the event description

20

The generally shorter recognition times in this study than the preceding one is attributable to the fact that American students participated in this study and Hong Kong Chinese students in the earlier one. The latter students generally took longer to read the English phrases than Americans did.

20

248

CHAPTER 9

was not associated with the description in memory under either format condition. These data therefore suggest that pictures played a somewhat different role in the representations formed under these conditions than they played in Adaval and Wyer’s (1998) research. The nature of this role is suggested by a conceptualization of visual information processing by Barsalou (1993). That is, a photograph of a person might stimulate the formation of a “perceptual symbol” that people mentally manipulate along with other symbols to construct a visual image of the person’s activities in not only the situation to which the picture pertains but other situations as well. (This could occur in much the same way that a previously formed mental picture of a colleague’s physical appearance stimulates an image of his or her actions in a story that someone tells.) Moreover, this perceptual symbol may be applied across situations, establishing a stronger connection between the events that occur in a narrative than would otherwise exist. Pictures of a person could create such a symbol and, therefore, could facilitate the construction of a visual image of events in which the individual is involved even if they are not themselves directly related to these events. If this is the case, however, pictures of the person should facilitate the construction of a coherent image-based narrative representation of the person’s life even when the pictures do not become associated with the specific events that accompany them. For that matter, pictures of the politician should have similar effects even if they do not accompany the written event descriptions at all, but are presented at the outset, before any of these descriptions are provided. These speculations were confirmed in an additional study described in the next section. In this study, pictures did not accompany the verbal descriptions of the politicians’ life events. In some cases, however, four photographs of the politician were provided at the outset, before the event descriptions were presented. Each picture provided a clear image of the politician’s physical appearance but contained few other indications of the context in which the pictures were taken. Relative to conditions in which no pictures at all were presented, introducing pictures at the outset increased the favorableness of participants’ evaluations of the politicians, and this was true both when the information was conveyed in a narrative (3.56 vs. 2.88) and when it was listed (3.35 vs. 2.56). Thus, pictures that were presented at the outset had the same influence under narrative-format conditions that they had when they were conveyed in the context of the event descriptions. Furthermore, presenting pictures before the event descriptions in list-format conditions eliminated the interference effects that occurred when the pictures accompanied these descriptions, leading the pictures to have a positive influence on judgments under these conditions as well.

CONSTRUCTION AND USE OF NARRATIVES

249

These results indicate that the influence of pictures in Adaval et al.’s (2003) other studies did not derive from the information they provided about the events themselves. Nor did it result from the evaluative implications of the pictures per se; considered out of context, the pictures were evaluatively neutral and conveyed very little information about the politicians they depicted. Rather, the pictures stimulated the formation of “perceptual symbols” (Barsalou, 1993) that people used to construct mental images of the politicians engaging in the events they later read about, in much the same way that the picture of a character in a novel helps the reader to imagine verbal descriptions of the character’s actions in the text. This facilitation can occur both when the events are conveyed in a narrative and when they are not. On the other hand, accompanying listed events with pictures appears to interfere with the piecemeal integration processes that are activated by this presentation format and occur on-line, as the information is presented and a judgment is computed. The adverse effects of this interference override the beneficial effects that the pictures might otherwise have. THE ROLE OF VISUAL IMAGERY ON VERBAL INFORMATION PROCESSING: ADDITIONAL CONSIDERATIONS The conclusions drawn from the results just described call attention to the need to examine more closely the impact of visual images on the processing of verbal information. Two quite different studies in the political domain bear on this impact. The first identifies the influence of individual and situational differences in the tendency to form visual images on the impact of verbal event descriptions. The second shows the way in which visual images can influence the implications that people extract from other types of verbal information. Individual and Situational Differences in Visual Imagery and Its Effects on Information Processing To reiterate, the influence of pictures on judgments appears to occur because the pictures facilitate the construction of mental images of events that are described verbally. If this is so, other factors that influence the construction of images should have similar effects. Some evidence of this was obtained in the vacation study we described earlier (Adaval & Wyer, 1998, Experiment 2). That is, the effects of instructions to imagine the events were similar to the effects of presenting pictures. This possibility was examined in additional conditions of the experiment described at the end of the previous section. In these conditions, we took into account by both instructional demands to form mental images of the events and chronic individual differences in the disposition to construct them spontaneously.

250

CHAPTER 9

To reiterate, participants made judgments of politicians on the basis of verbal materials identical to those we employed in the earlier studies. In this case, however, pictures of the politicians were either presented at the outset, before participants read verbal descriptions of the politician’s life events, or not at all. In addition, participants in imagination conditions were explicitly told to “imagine what the person must have looked like and what he must have done at each stage of his life. . . In other words, try to imagine that you are watching a movie . . . and form a picture in your mind of the events described.” In contrast, participants under no-imagination conditions received instructions similar to those in the earlier experiment, without further elaboration. Finally, participants after making judgments of the politicians were administered a questionnaire developed by Sheehan (1967) to assess individual differences in information-processing style several items of which assessed the disposition to use visual images in responding to information (e.g., “I like to picture future events or situations in my mind,” “When I think of someone I know, I often picture in my mind what they look like,” “My thinking often consists of mental pictures or images,” etc.). Participants whose scores were above and below the mean response to these items were classified as visualizers and nonvisualizers, respectively. Participants’ impressions of each politician were analyzed as a function of picture conditions, imagination instructions, presentation format, and the disposition to form visual images. No effects of format were significant (p > .10). However, the effects of pictures, imagination instructions, and chronic dispositions to form images were highly interactive, F(1,47) = 5.11, p < .03. These effects are summarized in Table 9.5. Visualizers evaluated a politician more favorably when they were told to imagine the events that ocTABLE 9.5 Evaluations of Politicians as a Function of Imagination Instructions, the Presence of Pictures, and Disposition to Form Visual Images Visualizers

Pictures Narrative format List format Mean No pictures Narrative format List format Mean

Nonvisualizers

Imagination Instructions

No Imagination Instructions

Imagination Instructions

No Imagination Instructions

3.33 3.50 3.42

3.08 2.83 2.96

2.73 3.07 2.90

4.04 3.87 3.96

3.08 3.71 3.40

3.10 2.67 2.88

3.83 3.54 3.68

2.67 2.46 2.56

Note. Adapted from Adaval et al. (2002, Experiment 4).

CONSTRUCTION AND USE OF NARRATIVES

251

curred in his life than when they were not, and this was true regardless of the format in which the information was conveyed. Moreover, pictures had no effect whatsoever on these participants’ evaluations. That is, visualizers apparently formed clear images of the events even in the absence of pictures, and so making the pictures available had little additional effect. In contrast, pictures had a positive influence on nonvisualizers’ evaluations when these participants were not explicitly told to imagine the politician’s life events. Furthermore, imagination instructions had a positive effect provided no pictures were presented. When nonvisualizers were asked to imagine the politician’s life events and pictures were presented, however, the two sources of mental images appeared to interfere with one another, thus eliminating the positive effect that each factor had in isolation. As noted earlier, the verbal information we presented typically described abstract events that were unlikely to elicit visual images spontaneously. Nevertheless, chronic visualizers found it easy to construct images when they were asked to do so, and the presence of pictures neither helped nor hindered them. Therefore, the images that these participants formed increased the extremity of their judgments regardless of whether or not pictures were presented. Nonvisualizers, who do not normally form visual images on the basis of verbal information, were also benefited by doing so provided no pictures were presented. When pictures were presented, however, these individuals apparently found it difficult to integrate their implications into the images they formed, and this led them to form less clear impressions of the politicians than they might otherwise have done. Effects of Image-Based Representations on the Processing of New Information The preceding series of studies focused on how visual and verbal information about a politician combined to affect his “image.” Once this image is formed, however, it can have an impact on the way other information is processed and consequently on the implications drawn from this information. A study by Wyer, Budesheim, Shavitt, Riggle, Melton, and Kuklinski (1991) is of interest in this regard. Nonacademic employees were recruited for a study of the way people make judgments of political candidates on the basis of information of the sort they receive during an election campaign. On this pretense, they received two types of information about a member of the U.S. House of Representatives who had recently run for the Senate in a neighboring state. First, participants were shown a videotaped nonpolitical speech of the candidate’s remarks at a bicentennial celebration at which he was asked to present an award to a local dignitary. The speech, delivered by an accomplished character actor, was identical in content in all conditions but was delivered in either a forceful, articulate manner that conveyed a favorable im-

252

CHAPTER 9

age or in a bumbling manner, with inappropriate pauses, fidgeting, and other mannerisms that conveyed an unfavorable image. The second type of information was conveyed in an audiotaped portion of a radio program that had ostensibly been sponsored by the League of Women Voters. In this program, commentators reported the candidate’s votes on ten bills that had recently come before the House. Votes on six of these bills (e.g., a proposal to increase military spending by 15%, a proposal to allow prayer in public schools) conveyed either a consistently conservative or consistently liberal ideology. (Votes on the remaining four bills had no ideological significance.) This information was conveyed in three conditions. In no-delay conditions, participants listened to the radio program describing the candidate’s issue stands immediately after they watched the videotape of his speech. Then, after doing so, they reported their impression of the candidate along a 100-point “feeling thermometer.” The procedure in two other conditions was similar. In delayed-information conditions, however, a 24-hour delay was introduced between the videotaped speech and exposure to the candidate’s issue positions, with judgments being made immediately after exposure to the issue positions. In delayed-judgment conditions, participants listened to the candidate’s issue stands immediately after seeing the speech, as they did under no-delay conditions. However, a 24-hour delay was induced between this information and judgments. In all conditions, participants after evaluating the candidate reported their personal positions on each of the issues to which the candidate’s votes pertained. Finally, they indicated their party affiliation and ideological orientation. These latter data, in combination with the candidate’s issue stands, were used to define two independent variables. First, participants’ reported ideology was coded as either similar or dissimilar to the candidate’s, as implied by the liberal or conservative orientation of his issue stands. Second, participants were classified as either generally in agreement or generally in disagreement with the candidate’s specific issue positions, based on the proportion of ideologically relevant issue stands with which they agreed. (Each participant agreed with at least one liberal and one conservative issue position, regardless of his or her general ideology. Therefore, ideological similarity and agreement level were relatively independent.) Evaluations of the candidate are shown in Table 9.6 as a function of each informational variable and delay conditions. These evaluations were obviously affected by the favorableness of the candidate’s image as conveyed in the videotape. Moreover, this effect was greater when the image was salient at the time of judgment (i.e., under no-delay conditions) than when it was not. The effects of agreement and ideological similarity are of greater interest, however. When the candidate’s image was not salient at the time information

253

CONSTRUCTION AND USE OF NARRATIVES TABLE 9.6 Candidate Evaluations as a Function of Delay Conditions, Image, Agreement With Candidate’s Issue Stands, and Ideological Similarity No-Delay Conditions Candidate’s image Favorable 51.0 Unfavorable 35.6 Participants’ agreement with issue stands Agree 42.3 Disagree 33.2 Ideological similarity to candidate Similar 49.6 Dissimilar 26.0

Delayed-Information Conditions

Delayed-Judgment Conditions

53.8 44.6

47.7 42.9

64.5 30.9

48.1 40.1

48.1 47.3

57.3 31.0

Note. Adapted from Wyer et al. (1991).

about his issue stands were presented (i.e., under delayed-information conditions), participants based their evaluations of the candidate on their agreement with his issue positions, and the candidate’s similarity to them in ideology had virtually no effect. However, when participants learned of the candidate’s issue stands immediately after they viewed his image-inducing speech (i.e., under no-delay and delayed-judgment conditions), they based their evaluations on the candidate’s general ideology, and their agreement with him on specific issues had little influence. This was true under both nodelay and delayed-judgment conditions. Thus, unlike the direct effects of the candidate’s image on judgments, the indirect effects of image on the processing of issue information were not a function of its salience at the time of judgment. Rather, they depended on the salience of the candidate’s image at the time the issue stand information was conveyed. In short, the salience of the candidate’s image at the time his issue stands were learned altered the way in which the implications of these issue stands were construed. When a global image of the candidate was not salient to them, participants assessed their agreement with his stands on specific issues and based their judgments on this criterion independently of the ideological implications of the candidate’s positions. When the candidate’s image was salient to participants at the time they heard about his issue positions, however, they applied a global criterion in assessing the implications of his issue positions as well. Consequently, their agreement with the candidate on specific issues had relatively little effect.21 21

An alternative interpretation of these results might be that participants experienced overload when the candidate’s videotaped speech and his issue stands were conveyed in temporal proximity and, therefore, they devoted less cognitive effort to an assessment of the candidate’s

21

254

CHAPTER 9

Summary and Conclusions The research described in the preceding sections provides several insights into the impact of narrative descriptions of a person or object and the influence of visual images on the processes that underlie this impact. At the same time, it places constraints on the conditions in which these processes operate. To summarize: 1. When people’s processing objectives require knowledge of the temporal sequence of events that occurred in a particular situation (e.g., the events surrounding a crime), presenting information in a narrative facilitates the construction of this sequence. Consequently, it increases recipients’ confidence in the conclusions they draw from the information and, therefore, increases the impact of this information on judgments. 2. When an evaluation does not require the construction of a temporal sequence of events, this sequence may only be constructed if the events are described in a narrative that conveys their temporal relatedness. Otherwise, recipients may employ other strategies for combining the implications of the information that can often be equally effective (e.g., they may make a piecemeal assessment and integration of the individual pieces of information). 3. Pictures increase the impact of verbally described events when the events are conveyed in a narrative. This is true regardless of whether the events are situationally and temporally constrained (i.e., if they stimulate the construction of an episode model) or refer to more general experiences that take place over a period of time. However, pictures interfere with the piecemeal integration of the implications of events that are conveyed in a list, and consequently decrease the impact of these events on judgments. 4. Pictures can increase the impact of verbal event descriptions when they are conveyed at the outset as well as when they accompany these descriptions. Furthermore, this increase can occur regardless of the format in which the descriptions are presented. 5. The information-processing strategies that are stimulated by different presentation formats may be activated and applied spontaneously, with little conscious awareness. Therefore, their effects are independent of explicit instructions or situational demands to employ alternative strategies. The influence of pictures on the use of these strategies also appears to occur without awareness.

issue positions. If this were so, however, they would presumably have been inclined to use the candidate’s image as a heuristic, leading it to have greater effect on judgments than it otherwise would. In fact, however, the candidate’s image had no greater effect under delayed-judgment conditions (when the two types of information were presented together) than under delayedinformation conditions. Therefore, this alternative interpretation does not seem viable.

CONSTRUCTION AND USE OF NARRATIVES

255

6. Instructions to imagine the events that are described verbally can influence the impact of these events in much the same way as pictures. However, when pictures and imagination instructions are both provided, people who are not normally disposed to form visual images may have difficulty integrating pictures into the images they form when they are explicitly asked to do so. In this case, therefore, pictures can decrease the impact of the information relative to conditions in which the pictures are not presented. 7. Narrative representations that are formed on the basis of observed events can induce a holistic processing strategy that, once activated, influences the processing of verbal information that is conveyed subsequently. In particular, it stimulates the use of more global criteria for judgments, and correspondingly decreases the use of criteria that require more analytic, piecemeal processing. These conclusions may have more general implications. For example, information is likely to have greater impact on judgments and decisions if it stimulates individuals to form a narrative-based account of an experience, and to form mental images of the events that occurred, than if it does not. This possibility, which was suggested by results of the study by Reyes et al. (1980), was also recognized by Nisbett and Ross (1980). To borrow their example, a person who is considering the purchase of a particular brand of dishwasher may be less influenced by statistical evidence of its superior maintenance and repair record than by a neighbor’s vivid description of her unique experience with the machine, which broke down 2 weeks after she purchased it, flooded her kitchen and dining room, and ruined a valuable rug. As noted earlier, narrative-based information processing may not always generate more extreme evaluations than other well-learned processing strategies. Nevertheless, as this anecdotal example suggests, when narrativebased representations are easy to construct and are likely to elicit visual images, they are generally more likely to have an impact than abstract information whose implications are more difficult to assimilate and interpret.

GENERAL EFFECTS OF NARRATIVE-BASED PROCESSING ON EXPLANATION AND PREDICTION As Pennington and Hastie’s research exemplifies, narratives are often constructed in the service of both prediction and explanation. Many examples of this possibility can be drawn from research in traditional areas of psychology, including (a) the impact of constructing explanatory narratives on predictions of future events, (b) counterfactual reasoning, and (c) attribution phenomena. Much of this research was not conducted for the purpose

256

CHAPTER 9

of delineating the role of narratives in judgment. Nevertheless, a conceptualization of the research from the perspective proposed in this chapter has implications for the situational and informational factors that determine which of several alternative narrative representations are likely to be constructed and used. The Impact of Constructing Explanatory Narratives on the Prediction of Future Events People are likely to predict that an event is likely to occur if they can construct a plausible narrative of the events that might lead up to it. This implies that increasing the accessibility in memory of a plausible narrative account of the event should increase beliefs that the event will occur. Several studies bear on this possibility. Ross, Lepper, Strack, and Steinmetz (1977) asked some participants to read a clinical case study with instructions to explain why the individual might have committed suicide after leaving therapy. Others read the same passage with instructions to explain why the protagonist might have donated a sizable sum of money to the Peace Corps. Participants after generating their explanations were asked to predict the likelihood that the protagonist had engaged in a number of behaviors after leaving therapy, one of which was the event they had explained. Participants were told that the case study had been taken from a textbook and that neither the experimenters nor anyone else had any knowledge of what had actually happened to the individual. Nevertheless, participants typically predicted that the event they had explained was more likely to have occurred than the events they had not explained. Participants who were asked to explain an occurrence presumably constructed a plausible narrative of the sequence of events that led up to it and then, having done so, used this narrative as a basis for their later predictions. Results of other studies are consistent with this interpretation. Sherman, Skov, Hervitz, and Stock (1981), for example, arbitrarily asked participants to explain why they might either succeed or fail on an anagrams task and then, after generating their explanation, asked them to predict their actual performance on the task. Participants predicted the occurrence of the outcome they had arbitrarily been asked to explain. Moreover, their actual task performance confirmed their prophecy. (That is, participants who had been asked to explain why they might succeed, and predicted success, performed better than those who had been asked to explain why they might fail.) Participants apparently used the implications of their narrative-based prediction as a standard at the time they actually performed the task, motivating them to attain the performance level it implied. Narrative-based explanations are likely to be generated spontaneously when individuals receive information that is inconsistent with expectations.

CONSTRUCTION AND USE OF NARRATIVES

257

In a study by Ross, Lepper, and Hubbard (1975), participants were asked to distinguish between true and false suicide notes. They were given bogus feedback on each trial concerning whether they were correct or incorrect. Some participants were led to believe they had been correct on 12 of 15 trials, whereas others were told they had been correct only three times. Later, however, participants were debriefed, being told that the feedback they received bore no resemblance to their actual performance. The debriefing was quite compelling, with participants being shown the preprepared schedule the experimenter had used to deliver the trial-by-trial feedback. It was therefore impossible for participants to believe that the feedback they had received was contingent on their actual performance. Nevertheless, participants were more likely to predict they would do well on a similar task in the future if they had been led to believe that they had performed well on the first task than if they had been led to believe they had done poorly. Direct evidence of the cognitive processes that produced these effects was not obtained in this study. It nevertheless seems reasonable to suppose that participants who received feedback that deviated from their expectations spontaneously attempted to explain it and, in doing so, selectively retrieved a body of self-knowledge about their past experiences that provided a plausible narrative-based causal account of it. Later, they used this representation as a basis for their predictions without considering the validity of the feedback that stimulated its construction. The Influence of the Ease of Constructing a Narrative on Its Impact The use of a narrative as a basis for judgment may depend on not only its accessibility in memory but also the ease with which it can be applied. Examples of this contingency are suggested by research on the impact of counterfactual reasoning. People who have a negative experience are particularly likely to ponder the reasons for its occurrence. In doing so, they may construct a scenario of how the event might have been avoided. If the story they construct permits them to conclude that the event occurred for circumstances beyond their control, they may be inclined to dismiss the event without further ado. However, if they can easily construct a scenario whereby they could have avoided the unpleasant experience, they may ruminate about their failure to have done so, and may experience upset and regret. A well-known experiment by Kahneman and Tversky (1982) provided an example. Suppose Al and Bob are scheduled for different flights that leave at the same time. They travel from town in the same taxi, are caught in a traffic jam and arrive 30 minutes after the scheduled departure. Upon

258

CHAPTER 9

reaching the terminal, Al learns that his fight left on time. However, Bob finds that his flight was delayed by 25 minutes and left only 5 minutes before he got there. Who is more upset? Ninety-six percent of the participants who are asked this question agree that Bob would be more upset than Al. Presumably, the reason is that Bob can easily construct a scenario whereby he could have saved 5 minutes and caught the plane, whereas Al is less easily able to do so. For similar reasons, people are more likely to be bothered by a mishap if it results from an atypical behavior than if it results from a typical one. For example, people imagine being more upset by having a traffic accident while driving home on a route they take infrequently than by having an accident while driving home on their normal route. Moreover, people may be more upset by getting robbed by a hitchhiker if they only rarely pick up hitchhikers than if they regularly do so (Kahneman & Tversky, 1982). In each case, the difference results from the fact that it is easier to imagine how the event could have been avoided in the first instance than it is in the second. The ease of constructing a narrative can also come into play in other judgments and decision phenomena. Research by Miller (1977) provided an example. That is, it is easier to imagine how one’s donation of money would help a particular family whose home has been destroyed by fire than to imagine how it would help to alleviate the suffering of thousands of people whose homes have been destroyed in Bosnia. Therefore, people are more likely to contribute money in the first case than the latter. Miller’s (1977) research provided evidence that this is the case. Evidence that this contingency is recognized outside the laboratory is suggested by the appeals of charitable organizations to “adopt” a child, a whale, or even a highway. This strategy presumably stimulates the construction of a narrative in which one’s own behavior helps an individual person, animal, or highway maintenance project. Because this narrative is easy to construct, it may increase the willingness to provide assistance. When two or more alternative narratives are equally easy to construct, the one that is generated may depend on a number of factors that affect the relative accessibility of the knowledge that enters into their construction. Several of these factors, which could be informational, situational, or motivational, are discussed in the sections that follow. Informational Determinants of Narrative Construction 1. Framing Effects The narrative that is constructed from information for use in making an inference can depend to a large extent on features of the information itself that make different sets of concepts accessible in memory. Examples of this

CONSTRUCTION AND USE OF NARRATIVES

259

contingency abound. To give a well-known example from research on decision making, consider the following two choice situations: 1. One thousand people are in danger of being infected by a deadly virus. You have the choice of administering two serums. One serum, A, is effective but in short supply. If you administer A, 350 persons will be saved. If you choose B, there is a probability of .65 that everyone will be saved but a probability of .35 that no one will be saved. 2. One thousand people are in danger of being infected . . . If you administer A, 650 persons will die. If you choose B, there is a probability of .35 that everyone will die but a probability of .65 that no one will die. The choice alternatives are actually identical in each case. In the first case, however, the description of the alternatives activates concepts associated with saving lives, and so people are likely to construct scenarios in terms of these concepts. As a result, they are more likely to prefer A, which guarantees that lives will be saved, than B. In the second case, on the other hand, the descriptions emphasize the loss of life and activate concepts associated with the avoidance of this negative outcome. In this case, therefore, people typically prefer B, which describes a situation in which no one will die, than A. The implications of the narratives that people use to make inferences on the basis of verbal information can also depend on the perspective from which the narrative is formed. (For evidence of perspective differences in the mental representations of verbal information, see Black, Turner, & Bower, 1979, described in chap. 4.) A study by Read (1985) provides an example. Consider the following vignette: Helen was driving to work along a three-lane road, where the middle lane is used for passing by traffic from both directions. She changed lanes to pass a slow-moving truck, and quickly realized that she was headed directly for another car coming in the opposite direction. For a moment, it looked as if a collision was inevitable. However, this did not occur. Please indicate in one sentence how you think the accident was avoided.

The situation is objectively similar for both cars. However, the story as written focuses the reader’s attention on Helen, leading a narrative to be constructed in which she is the central figure. Therefore, people who are called upon to explain how the accident was avoided typically attribute the responsibility of avoiding the accident to Helen rather than the other driver. If the story had been written from a different perspective, this tendency would presumably have been less likely.

260

CHAPTER 9

2. Effects of Conversational Norms on Narrative Construction A quite different informational influence arises from considerations elaborated in chapter 7. That is, the representation that participants construct of a social interaction is likely to be guided in part by normative expectations concerning the type of information they expect to be exchanged (Grice, 1975; Higgins, 1981; Sperber & Wilson, 1986). These expectations can influence the concepts and knowledge that listeners use to interpret not only the statements a person asks during a conversation but also the responses to these statements. This interpretation can often require the construction of a narrative. Kahneman and Miller (1986) recognized this possibility in the context of norm theory. For example, consider the following example: Q. A.

Why did Joan pass the history examination? She went to bed early the night before.

People who overhear this conversation are likely to assume that the questioner is seeking new information. That is, they may infer that Joan’s passing the exam is somewhat unexpected and, therefore, requires an explanation. They may further assume that the explanation is informative; that is, it describes behaviors or situational factors that do not normally exist. In interpreting the answer, therefore, they may construct a narrative of Joan as someone who typically stays out late before exams and is usually too tired to perform adequately, but who in the present instance went to bed at a reasonable hour and was sufficiently rested the next morning to do well. In contrast, consider the conversation: Q. A.

Why did Joan pass the history examination? She stayed up late the night before.

This conversation might stimulate the construction of a scenario of Joan as someone who normally doesn’t study much but, in this particular instance, decided to burn the midnight oil and consequently performed well. These narrative-based explanations, which are generated in an attempt to comprehend the situation, are likely to influence judgments of the protagonist as well as predictions of her behavior in other situations. Situational Determinants of Narrative Construction 1. Effects of Perspective on Attributions of Responsibility The differences in narrative construction suggested by the preceding examples result from the way in which events are described verbally. However, situational variables may have analogous differences when narrative

CONSTRUCTION AND USE OF NARRATIVES

261

representations are formed on the basis of direct observation. Early studies of social attribution phenomena (Regan & Totten, 1975; Storms, 1973) provide examples. In Storms’ (1973) study, some participants watched a videotape of a get-acquainted conversation between themselves (the target) and another person. The tape was also watched by persons who were not themselves involved in the conversation. The videotape was made from a vantage point that focused attention on either the target or the other. Participants after viewing the tape were asked whether the target’s behavior was attributable to characteristics of the target (personality, mood, etc.) or the situation (e.g., the other participant). Participants were more likely to attribute the target’s behavior to dispositional factors when the target was visually prominent in the tape than when (s)he was not. This was true regardless of whether the participants were themselves in the role of the target. Analogous conclusions can be drawn from a study by Regan and Totten (1975). In this case, participants observed a videotape of a conversation involving two other persons with instructions either to observe the target person or to empathize with this person. Thus, the same information was available to all participants. Nonetheless, participants who were told to observe the target presumably constructed a representation of the interaction in which the target was the focus of attention. In contrast, observers who were told to empathize with the target were likely to imagine interaction from the target’s perspective and, therefore, to construct a narrative that was focused on the conversation partner. The content of these narratives then provided the basis for their attributions. Thus, the former participants attributed the target’s behavior to dispositional characteristics, whereas the latter participants attributed his behavior to situational factors. Motivational factors can also influence the perspective from which a narrative is constructed. In a study by Wolfson and Salancik (1977), observers watched an actor perform an achievement task. Some observers expected to engage in the task themselves at a later time, whereas others did not. The first group of participants presumably formed a representation of the situation that focused on the task, whereas the second group formed a representation that focused on the actor. Consistent with this possibility, the first group of participants was more inclined than the second group to attribute the actor’s behavior to the task, and were less inclined than the second group to attribute it to the actor himself. 2. Contextual Influences on Narrative-Based Explanations The type and implications of narratives that individuals use as a basis for explanation and prediction can be influenced in part by situational factors that activate different sets of concepts and knowledge for use in con-

262

CHAPTER 9

structing these narratives. Norenzayan and Schwarz (1999) provided provocative evidence of these differences in an analysis of the expectations that participants generate in psychology experiments. Participants read a New York Times article about either (a) a disgruntled postal employee who went on a mass murder rampage or (b) a man who was accused of murder for bombing a federal office building. The letterhead of the questionnaire being administered suggested that the study was being conducted by either the “Institute of Personality Research” or the “Institute of Social Research.” In each case, participants after reading the story listed five reasons why the protagonist might have committed the crime. It seems reasonable to suppose that people who believed that the study was sponsored by a personality research institute would activate concepts associated with characteristics of the actor and, therefore, would construct a narrative account of the individual in terms of these characteristics. In contrast, persons who were told that the study was being conducted by a social research institute are more likely to activate concepts about social (i.e., situational) factors and to construct a narrative in terms of these features. These narratives should then influence the explanations that participants generate. This was in fact the case. Participants gave more dispositional explanations when they believed that the study was conducted by a personality research institute (M = 4.4) than when they believed it was conducted by a social research institute (M = 2.6), and gave fewer situational explanations in the former case than the latter (5.8 vs. 9.4). It is important to mention these differences occurred not only when the nature of the research institute was identified before participants read the story but also when it was identified afterwards. As the authors point out, this could indicate that the differences in attributions reflected a tendency for participants to tailor their responses to the apparent interests of the individuals to whom they were communicating, and might not reflect a difference in the underlying representation they formed of the interaction. This possibility cannot be discounted. However, it seems equally likely that participants mentally reviewed and reconstructed the crime after they had read the story, and that they based their explanations on this reconstruction rather than the representation they had formed earlier. Be that as it may, these considerations have implications for the interpretation of numerous psychology experiments in which experimenters implicitly or explicitly induce expectations for the criteria that should be applied in providing information and, therefore, the narrative-based representations they are likely to construct for use as a basis for their responses. As Schwarz, Strack, Hilton, and Naderer (1991) argued, studies that are ostensibly conducted by psychologists can dispose participants to construct explanations in terms of personality characteristics of the protagonists rather than employing other criteria that are less relevant to psychology. If

CONSTRUCTION AND USE OF NARRATIVES

263

the studies were conducted in a different situational context, this might not be the case. To demonstrate this, Schwarz et al. introduced some participants to a judgment task with instructions similar to those employed in an earlier experiment by Kahneman and Tversky (1973): A panel of psychologists have interviewed and administered personality tests to 30 engineers and 70 lawyers . . . On the basis of these interviews, thumbnail descriptions of the [100 individuals] have been written. In your form, you will find 5 descriptions, chosen at random. . . . For each description, please indicate the probability that the person described is an engineer. . . . the same task has been performed by a panel of experts who were highly accurate in assigning probabilities . . .

A second condition was similar except that the first line of the instructions referred to “researchers” rather than to “psychologists,” and the last line referred to “statisticians” rather than “experts.” Following these instructions, participants read a sketch of the following type; Hans K. is 45 years old. He is married and has four children. He is generally conservative, careful and ambitious. He shows no interest in social and political issues. He spends most of his time on his many hobbies, which include . . . solving mathematical puzzles.

The authors speculated that participants who believed that the story was associated with psychologists would construct an account of the individual in terms of personality and life style characteristics and would use this account as a basis for predicting his vocation. In contrast, associating the task with statistics was expected to lead participants to use a statistical criterion (e.g., base rates) as a basis for prediction. This was in fact the case. That is, .76 of the participants predicted that the individual was an engineer when the study referred to psychologists, whereas only .55 of the participants made this prediction when the instructions referred to statisticians. These data obviously do not provide evidence that a narrative representation of the target person was constructed in one case but not the other. The results nevertheless provide a general indication that contextual factors can play an important role in the criteria that underlie judgments and, to this extent, the likelihood that narrative-based explanations are used. Motivational Influences on Explanation-Based Narrative Constructions The scenarios that people construct in the course of explaining events may not be entirely dispassionate. This is particularly true when the events are ones in which people are personally involved in the events and the explana-

264

CHAPTER 9

tions they give could reflect positively or negatively on themselves. Individuals in Western cultures typically make defensive attributions. That is, they take responsibility for their own success but attribute their failures to other factors that have relatively few negative implications for their self-esteem (Harvey & Weary, 1984; Zuckerman, 1975). These factors may determine the narratives that people construct. The self-serving motives that underlie the construction of scenarios about oneself may be attributed to others as well. To give yet another example from Tversky and Kahneman (1982):

Tom and Jim were both eliminated from a tennis tournament. Both were eliminated on a tiebreaker. Tom lost when his opponent served an ace. Jim lost on his own unforced error. Who will spend more time thinking about the match that night?

Tom can presumably explain his loss in terms of consequences beyond his control. Jim, on the other hand, may continue to rehearse his error and to generate alternative ways in which it could have been avoided. Therefore, as Tversky and Kahneman’s findings imply, participants expect Jim to be more likely to think about the match than Tom. The assumption that people are more motivated to generate scenarios that reflect positively on themselves, and avoid constructing narratives that reflect negatively, is indirectly confirmed by Arkin, Gleason, and Johnston (1976). Participants in this study were induced to administer therapy to another person to eliminate a mild phobic reaction. The therapy was actually the same in all conditions. However, some participants were led to believe that the treatment they would use was the only one available whereas others were ostensibly given several alternatives from which they could choose. In all cases, participants were told that the treatment they would use had been typically used successfully or unsuccessfully by others. However, participants after administering the treatment themselves were told they had either succeeded or failed. After receiving this feedback, they estimated the extent to which they were personally responsible for the outcome. In principle, participants should believe they are more responsible for the outcome if they personally chose the therapy they use than if they are required to use it, and more responsible if their outcome differs from others’ (e.g., if they succeed when others fail, or fail when others succeed) than when it does not. As shown in Table 9.7, however, this was only true when participants were told they had failed. Participants took responsibility for success regardless of whether they had freedom of choice over the therapy they used and regardless of the outcomes that others had attained.

265

CONSTRUCTION AND USE OF NARRATIVES TABLE 9.7 Mean Attributions of Responsibility for Success and Failure in Administering Therapy as a Function of Freedom of Choice and Others’ Outcomes Personally Succeeded

Choice over therapy To be administered No choice

Personally Failed

Others Succeeded

Others Failed

Others Succeeded

Others Failed

6.63 6.38

6.13 6.75

7.13 4.13

4.25 4.63

Note. High scores indicate greater attributions of responsibility to oneself. Adapted from Arkin, Gleason, and Johnston (1976).

This study obviously does not provide direct evidence that participants’ attributions were mediated by a narrative representation that they constructed of the situation. It is nevertheless plausible to assume that participants who did well spontaneously activated a scenario of themselves as competent individuals who typically succeed in achievement activity and considered this scenario to be a sufficient explanation for the outcome. Therefore, because the scenario reflected positively on themselves, they used it as a basis for judgments without searching for alternative explanations. In contrast, participants were reluctant to construct a scenario that suggested they were responsible for failure if a plausible alternative could be generated. Therefore, they only accepted responsibility for their failure when they had personally chosen the therapy and had failed when others had done well. Arkin et al.’s study provides only one example of a more pervasive tendency for individuals to construct representations of themselves and others that confirm a perception of themselves as admirable individuals and of the world as a place in which they can be happy and successful. This tendency, which has been documented in detail by Kunda (1990), is discussed more fully in the next chapter.

CONCLUDING REMARKS The studies reviewed in this chapter confirm the assumption that people often attempt to construct a narrative representation of events on the basis of information they receive about them, and that their judgments are based on implications of this representation. The events may be ones they imagine themselves or another person experiencing, events they either read about, or ones they directly observe. It would obviously be wrong to assume that all social inferences are based on these representations. At the

266

CHAPTER 9

same time, the influence of narrative constructions of social experiences on judgments is incontrovertible. However, narrative accounts of a new experience, and the inferences that are made on the basis of these narratives, do not occur in a cognitive vacuum. These accounts are constructed, and implications are drawn from them, on the basis of previously formed representations that specify the temporal and causal relations among events of the sort under consideration. The use of these representations, which may often constitute implicit theories about the world in which one lives, is discussed in the next chapter.

C H A P T E R

10 The Impact of Implicational Molecules and Implicit Theories on Inferences About Oneself and Others

The construction of narrative representations from new information obviously requires prior knowledge about the type of persons, objects, and events to which the information refers. This knowledge, which often has implications for features that are not specified in the information presented, can sometimes include episode models of specific past experiences that have features in common with the new one. However, it can also consist of generalized narrative representations that are composed of object and event concepts that the new information exemplifies. As noted in chapter 1, these latter representations often function as implicit theories that individuals apply in comprehending new experiences and construing their implications. This chapter reviews theory and research in a number of areas in which implicit theories are likely to play a role, including reconstructions of the past, the use of judgmental heuristics, perceptions of social support, the correlates of marital satisfaction, and the effects of stereotypes on overt behavior. Before embarking on a discussion of specific inference phenomena, however, it may be useful to provide a simplified conceptualization of the processes we assume to underlie the use of previous acquired knowledge representations in making inferences.

GENERAL CONSIDERATIONS The generalized narrative representations that function as implicit theories may be abstractions of a number of specific instances of the sequence of events they characterize (although, as we indicated in chap. 6, these repre267

268

CHAPTER 10

sentations may be formed less often than one might intuitively expect). Others may be formed in the course of goal-directed cognitive activity that leads the events that compose them to be associated. Many theory-related narrative representations may be similar to “story skeletons” of the sort proposed by Schank and Abelson (1995). Other, simpler theories may be conceptualized as implicational molecules of the sort described earlier in this volume. These molecules theoretically consist of two or more verbally coded segments that combine to describe a temporally and causally related sequence of events or states of affairs. Although the nature of implicational molecules has been discussed in earlier chapters, their centrality to the issues of concern in the present chapter arranges a brief recapitulation of their structure and use in making judgments and decisions. In combination, the segments of a molecule often exemplify a generalization about the world in which we live. To reiterate an earlier example, the generalization that smoking causes lung cancer might be represented in the two-segment molecule: [P smokes; P has (will get) lung cancer].

Similarly, the generalization that people get what they deserve might be conveyed in a “just-deserts” molecule; i.e.: [P does a bad thing; P encounters misfortune]

or, alternatively, [P does a good thing; P has good fortune].

The use of such molecules to make inferences is theoretically governed by a completion principle. This principle, which is consistent with our previous discussion of narrative-based comprehension processes, can be conveyed as a postulate: Postulate 10.1. If the information available about a specific situation instantiates all but one of the propositions that compose a molecule, an instantiation of the remaining proposition will be spontaneously inferred.

The completion principle theoretically holds regardless of which propositions are instantiated by the information available and which must be inferred. Thus, in the previous example, people who learn that a person has committed a nefarious act might use a just-deserts molecule to infer that the person will be punished or will otherwise experience misfortune. However, they might also use this molecule to infer that a person who has expe-

THE IMPACT OF IMPLICATIONAL MOLECULES

269

rienced misfortune has done a bad thing or is reprehensible for other reasons. This possibility is consistent with the just world hypothesis (Lerner et al., 1976; Lerner & Simmons, 1966). In support of this hypothesis, Walster (1966) found that people evaluate the innocent victim of an automobile accident more unfavorably if the individual was seriously injured than if he or she was only slightly harmed. Research with similar implications was reported by Lerner and Simmons (1966) and also by Wyer, Bodenhausen, and Gorman (1985) in a study described in chapter 3. A study by Spiro (1977) also suggested an application of the completion principle. Protagonists read a story about an engaged couple with instructions that the experiment concerned “reactions to situations involving interpersonal relations.” In some versions of the story, the man informed the woman that he didn’t want children, whereupon the woman expressed considerable upset and a bitter argument ensued. After reading the story, participants engaged in routine activities unrelated to the experimental task. During this period, the experimenter, who ostensibly knew the couple incidentally noted that they eventually married and were still happily together. Participants who are motivated to comprehend the couple’s relationship are likely to construct a scenario of how this unexpected outcome might have occurred. In doing so, they might invoke a previously formed “kissand-make-up” molecule of the form: [People love one another; people fight; people resolve conflict; people make up].

To this extent, they may infer an instantiation of the proposition “people resolve conflict,” which is not implied by the information available. For example, they might speculate that one of the partners had a change of heart, that the woman found she couldn’t have children, or that some other unforeseen event occurred that stimulated the couple to resolve their conflict. The inferred event might then be added to the mental representation that participants construct of the couple’s relationship and store in memory. Participants after completing the activities they were asked to perform were dismissed. However, they returned for a second experimental session a few days or several weeks later, at which time they were asked to recall the story they had read. They were explicitly told to report only things that were mentioned in the story and not to include any personal reactions or inferences they might have made. Nonetheless, participants made frequent errors, the number of which increased over time. These errors were typically of the sort one might expect as a result of attempts to reconcile the experimenter’s “incidental” comment with the original information. For example, one person recalled that “the problem was resolved when they found that [the woman] could not have children anyway.” Another reported that

270

CHAPTER 10

although one person thought the matter was important, the other did not (see Spiro, 1977, for additional examples). The completion principle has potentially broad applicability. However, additional considerations arise in applying it to the inference phenomena of concern in this chapter. For one thing, more than one implicit theory can potentially be brought to bear on a particular observation. Or, in terms of our present discussion, a proposition may be contained in more than one molecule. This is particularly likely when an event has more than one possible cause. For example, people’s theories about poor academic performance might be reflected in the molecules: [P doesn’t study; P gets poor grades]

and [P is stupid; P gets poor grades].

In such instances, the information that a particular person received a poor grade could activate either of these molecules. Assuming that each molecule is equally applicable, the molecule that comes to mind first is most likely to be applied. A second consideration is that the inferences made on the basis of an implicit theory may often be probabilistic. In some cases, the information available may not be sufficient to determine a theory’s applicability. (In the case of an implicational molecule, for example, more than one segment might be uninstantiated.) In other cases, the information to which a theory is applied might not necessarily be valid. People who hear that an acquaintance did not study for an exam could use the molecule described earlier to predict that the person will get a bad grade. At the same time, they are undoubtedly aware that a failure to study does not always lead to bad grades. Moreover, they might not be completely convinced that the information they have received about the individual’s study behavior is valid. Therefore, although they might use an implicational molecule to generate a prediction, they may not be completely certain their prediction is correct. When an implicational molecule consists of only two propositions, a formal model of probabilistic inference proposed by Wyer and Hartwick (1980; see also Wyer, 1974; Wyer & Goldberg, 1970) takes this possibility into account. The conceptualization, which is essentially an extension of McGuire’s (1960) probabilogical model, provides a quantitative description of the relations among beliefs in causally related propositions. Suppose an implicational molecule has been formed as a result of associating two propositions, A (e.g., “war is good for the economy”) and B (“America is likely to go to war within the next 5 years”). If this molecule is accessible in memory

THE IMPACT OF IMPLICATIONAL MOLECULES

271

at the time an individual is asked to predict the likelihood of B, the individual is likely to identify and use A as a basis for this inference. Suppose, however, that the individual is uncertain that A is true. Then, the person is likely to consider not only the likelihood that B would occur not only if A is true but also if A is not true. Thus, in our example, people who are asked the likelihood that American will go to war might base their belief on not only the likelihood that this will occur if war is good for the economy, but might also take into account the possibility that America would go to war even if war were not good for the economy. The relative weight attached to these two conditional beliefs may be a function of the relative likelihood that A is or is not true (i.e., that war is or is not good for the economy, respectively). The quantification of this process proposed by Wyer (1974; Wyer & Goldberg, 1970) assumes that if beliefs are defined in units of probability, then the belief in the event described by one proposition, B, is related to beliefs in a possible cause of this event, A, according to the equation: P(B) = P(A)*P(B/A) + P(~A)*P(B/~A),

[10.1]

where P(B) is the belief that A is true, P(A) and P(~A) [= 1 - P(A)] are beliefs that A is and is not true, respectively, and P(B/A) and P(B/~A) are beliefs that B is true if A is and is not true, respectively. This equation provides a remarkably accurate quantitative description of the relations among these beliefs and, therefore, the influence of beliefs in one proposition on beliefs in a second (for summaries of this evidence, see Wyer & Carlston, 1979; Wyer & Hartwick, 1980). In one study, for example (Wyer, 1970), participants read scenarios composed of two parts. The first part led participants to believe that a particular event (A) was unlikely to occur and, in addition, implicitly conveyed the likelihood that a second event, B, would occur if A did and did not occur. The second part of the story provided additional information that increased beliefs that A would occur without changing its implications for B. In neither part of the story, however, was the target event itself explicitly mentioned. One story, for example, concerned the likelihood that a student riot will occur at a particular university (A), and the likelihood that the university president would be fired (B). The first part of the story described the president as coercive, establishing a low initial belief that a riot would occur. However, the second part provided information that substantially increased the belief that a riot would occur without affecting perceptions of whether the president would remain in office. After reading each part of the story, participants reported their belief that B would occur, followed by their belief in A and the two conditional beliefs defined in Equation 10.1. These estimates, along a scale from 0 (not at all likely) to 10 (extremely likely), were divided by 10 to convert them to

272

CHAPTER 10

units of probability. The predicted belief in B, P(B), was then computed for each participant separately by combining his or her beliefs in the manner described by the right side of Equation 10.1. Mean obtained values of P(B) are plotted as a function of mean predicted values in Fig. 10.1. Changes in beliefs in B after reading the second part of the story are also shown in this figure. In each case, the standard error of the difference between mean predicted and mean obtained values was less than .05 (half a scale unit). This accuracy was obtained without the use of any ad hoc curve-fitting parameters. Further evidence bearing on the validity of Equation 10.1 as a description of the relations among beliefs in causally related propositions is provided by research on the “Socratic effect” (McGuire, 1960; Rosen & Wyer, 1972). That is, people’s a priori beliefs in a set of propositions may not conform to the relations specified in Equation 10.1 because the propositions have not recently been thought about in relation to one another. However, if people’s beliefs in the propositions are called to their attention in close temporal contiguity, they may recognize their inconsistency and, consequently, may modify one or more of the beliefs to eliminate this inconsistency. This should be indicated by an increase in the quantitative accuracy of Equation 10.1. Rosen and Wyer (1972) found evidence that this is the case. That is, participants reported their beliefs in causally related propositions in two experimental sessions 1 week apart. The accuracy of Equation 10.1 in describing the relations among these beliefs was significantly greater in the second session than it was at first, suggesting that the consistency of the beliefs increased over time. A later study (Henninger & Wyer, 1976) is potentially more relevant to the concerns of this chapter. Participants in some conditions were asked to report their belief in a proposition occupying the position of A in Equation 10.1 (i.e., an antecedent) before reporting their belief in B (the consequent). These participants appeared to use A as a basis for their belief in B when they encountered it later in the questionnaire (as inferred from the accuracy of Equation 10.1 in describing their relationship). Other participants, however, reported their belief in B at the outset, when A was not salient to them. These participants typically used a different criterion for evaluating the validity of B, and so their belief in this proposition was inconsistent with the belief in A that they reported later. Henninger and Wyer’s findings suggest that Equation 10.1 might be used to determine empirically the criteria that people bring to bear on their beliefs in a proposition they are asked to evaluate. Suppose two or more implicational molecules could potentially provide the basis for an inference. A comparison of the quantitative accuracy of Equation 10.1 in describing the relations among the beliefs in the propositions composing each molecule could provide an indication of which molecule was actually activated

273

FIG. 10.1. Beliefs in a Proposition B, and changes in these beliefs, as a function of predicted values based on Equation 10.1. Based on data reported by Wyer (1970).

274

CHAPTER 10

and used. The utility of the equation is obviously restricted to theories that concern only two causally related events or states. Many implicit theories are more complex. However, if these theories can be broken down into subsets of causally related propositions, the conceptualization could be useful in evaluating the assumptions underlying the application of these theories as well.

EFFECTS OF IMPLICIT THEORIES ON RECONSTRUCTIVE MEMORY Implicit theories may sometimes be more accessible in memory than the actual events that led them to be formed. To this extent, they may often be retrieved and used as a basis for reconstructing the events that occurred in a particular instance without consulting the original memory trace of the events at all. That is, suppose people are asked to recall a specific event to which an implicit theory is relevant. They may find it easier to employ the theory to reconstruct the event than to retrieve the event itself, which may be relatively less accessible in memory. (This speculation is consistent with the effect of generalized representations of an observed experience on memory for the experience; see Adaval & Wyer, 2003, discussed in chap. 6.) Reconstructing the Past Several intriguing demonstrations of the role of implicit theories in memory have been reported by Michael Ross (1989) in the context of discussing the impact of these theories on the recall of personal experiences. Two studies are particularly provocative. In one study, female participants who had previously reported their typical emotional reactions during the period of their menstrual cycle were asked to keep a daily diary of their moods over the course of a month. At the end of the month, they were asked to recall the moods they had experienced during this period. Participants’ recall was better predicted by their implicit theories about their emotional reactions during the time of their menstrual cycle than by the actual feelings they had reported experiencing at this time. Students in a second study (Conway & Ross, 1984) participated in a program that they believed would increase their study skills. After participating, they were asked to recall their preprogram estimates of their skills. Their recall was governed primarily by their implicit theories that the program would be effective. Thus, participants whose skills after participating did not actually change over the course of the program recalled their preprogram skills as lower than they actually were, consistent with their theory that they had improved.

THE IMPACT OF IMPLICATIONAL MOLECULES

275

In a study by Goethals and Reckman (1973), students participated in a group discussion of bussing. The discussion was dominated by a confederate whose opinion contrasted with the attitudes that participants had reported in an earlier session. The confederate’s view had a substantial influence on not only participants’ postdiscussion attitudes but also their recall of the attitudes they had reported earlier. Thus, participants apparently employed an implicit theory that their attitudes were stable over time, and therefore used their postdiscussion attitudes to infer what their attitude must have been before the discussion took place. This interpretation was confirmed by Ross (1989). Specifically, participants after reporting their attitudes toward the position advocated in a persuasive message were asked to list the thoughts they had had while recalling the attitude they had reported 1 month earlier. Responses of more than 50% of the participants suggested the use of an implicit “temporal consistency” theory (e.g., “I answered the question now and assumed that my opinion probably hadn’t changed in a month or so.”). Self-Perception Theory Ross (1989) was among the first to elaborate the role of implicit theories in reconstructive memory and judgment. However, Bem’s (1972) well-known theory of self-perception is based on very similar assumptions. Bem argued that people who are asked to report their attributes, or alternatively, their attitude toward a social issue, do not perform an exhaustive review of the large amount of self-knowledge they have stored in memory that bears on this characteristic. Rather, they retrieve the most recent judgment-relevant information that comes to mind and base their judgments on the implications of this information alone. In many instances, the most accessible judgment-relevant information is a recent behavior they have performed. Under these circumstances, people construe the implications of this behavior for the judgment they are asked to make, and resort to additional information only if they consider its implications to be unclear or unreliable. Bem further assumes that people use the same principles to infer their own attributes that they use to infer others’. Or, in terms of the conceptualization we are proposing, both inferences about oneself and inferences about others may be based on the same implicit theory about the reasons people engage in attribute-relevant activity. For example, suppose one person agrees to write an essay in favor of capital punishment even though he is given the opportunity to write about something else. In contrast, a second person is told to write the essay without being given the chance to refuse. An observer of the events is likely to activate an implicit theory that people voluntarily advocate positions with which they agree, as reflected in the molecule:

276

CHAPTER 10

[P believes X; P chooses to advocate X].

In the second case, the observer may activate an implicit theory that people do things because they have to; e.g., [A controls P; A desires X; P advocates X].

The first theory has implications for the actor’s belief, but the second theory does not. Consequently, the observer may infer that the first person is more likely to believe in capital punishment than the second. However, now suppose the two individuals in our example are asked to report their own beliefs in capital punishment. If people use similar implicit theories to infer their own attributes from their behavior, the first person should report a stronger belief in capital punishment than the second. Although there are obviously alternative interpretations of this difference (Festinger, 1957), Bem’s conception is certainly plausible. Bem and McConnell (1970) constructed an intriguing empirical demonstration of self-perception processes that also has implications for the role of implicational molecules in reconstructive memory. Some participants wrote a counterattitudinal essay under conditions in which they were given the right to refuse, whereas others were told to write the essay without being given a choice. Then, some participants in each condition were asked to report their attitude toward the position they had advocated. These participants reported more favorable attitudes toward the position under freechoice conditions than under no-choice conditions, consistent with numerous other studies. Other participants, however, were asked to recall the attitude they had reported in an earlier experimental session. These participants’ recall of their prebehavior attitudes was affected in the same way that postbehavior attitudes of the first group of participants were influenced. That is, participants under free-choice conditions recalled their prebehavior attitudes as similar to the position they advocated in the essay they had written subsequently. This was not the case under forced-choice conditions. Participants in this study, like those in Ross’s studies and the research by Goethals and Reckman (1973), may have invoked an implicit theory about the attitudinal antecedents of their behavior and used this theory to infer not only their postbehavior attitudes but also what their attitudes must have been before engaging in this behavior. Further Considerations The interpretation of research on the effects of behavior on attitudes is often equivocal. That is, it is often difficult to know whether these effects occur after the behavior has been performed or in the course of deciding to

THE IMPACT OF IMPLICATIONAL MOLECULES

277

engage in it. In the latter case, the results would not indicate an effect of the behavior per se, but rather, would simply reflect the results of cognitive deliberations that were involved in deciding whether or not to engage in it. A study by Albarracin and Wyer (2000) circumvented this ambiguity. Participants were told that the experiment was designed to test a new computerized procedure for identifying “unconscious behavioral tendencies.” The procedure ostensibly consisted of presenting statements about social issues subliminally and having participants respond to them without being consciously aware of their content. Participants were told that although the stimulus statements would appear to be only flashes of light, they would elicit unconscious feelings that would give rise to a more conscious “intuition.” They were informed that to provide a measure of their behavior, they should “follow their intuition” and generate a “yes” or “no” response to each statement, which would then be interpreted by the computer as a vote either in favor of or against the issue to which it pertained. With this preamble, participants were ostensibly exposed to 14 statements that ostensibly concerned seven different university policies. They were told that the statements might express either support for or opposition to a policy and that the computer program would take this into account in determining the implications of their votes. After the 14 “statements” had been judged, participants received computerized feedback about both the policies to which they were subliminally exposed and their responses to these policies. Participants’ a priori positions on six of the policies (e.g., “maintaining civil liberties on campus,” “receiving free tickets to sports events,” “raising tuition,” etc.) were self-evident. Participants were told that they had voted in favor of or in opposition to each of these policies, depending on which vote was normative. (This established the credibility of the assessment procedure.) The seventh policy, “instituting comprehensive examinations on campus,” was one on which participants held varying opinions. Independently of these opinions, however, some participants were told they had voted in favor of instituting the exams and others were told they had voted against it. Participants after receiving the feedback were told that to understand their unconscious decisions, we needed to know their personal feelings about the policies and, on this pretext, asked them to report their attitudes toward instituting the exams. In addition, they estimated the likelihood and desirability of a number of possible consequences of instituting the exams. Finally, we indicated that because participants had an opportunity to think about comprehensive examinations during the experiment, we wanted to see how students might actually vote in a referendum that was likely to be held later in the semester. Participants were then left alone with instructions to select a slip of paper that represented their choice and to place it in a secured ballot box. (The box appeared to be partially full. In fact, how-

278

CHAPTER 10

ever, it contained only blank slips of paper, and so we could determine after the experiment how the participants had voted.) Thus, the procedure ensured that participants had not thought about comprehensive examinations before they received feedback about their ostensible voting behavior. Nevertheless, participants reported more favorable attitudes toward the institution of the exams if they had ostensibly voted in favor of instituting them than if they had voted against. Furthermore, this difference was evident regardless of the attitudes that participants had reported before to the experiment. Finally, their postbehavior attitudes, once formed, were used as bases for their actual voting decisions that they made at the end of the experiment. (Path analyses indicated that behavior feedback did not have an effect on participants’ judgments of specific behavioral consequences, nor did it have direct influence on participants’ voting behavior. Rather, this latter influence was mediated by the impact of the feedback on participants’ attitudes.) This study provided no direct evidence of the cognitive mediators of the effects we observed. It nevertheless seems reasonable to assume that participants invoked an implicit theory that people publicly advocate positions they favor, as embodied in the implicational molecule: [P favors X; P votes in favor of X].

Participants may have used this molecule not only to infer their attitude toward comprehensive examinations from their “unconscious” behavior, but also, once their attitude was inferred, as a basis for the overt behavioral decision they made at the end of the experiment.

MOTIVATIONAL INFLUENCES ON IMPLICIT THEORY CONSTRUCTION AND USE As we have suggested, the associated cognitions that compose the theories that people form and use can often result from the experience of these cognitions, or their referents, in temporal proximity. However, this is certainly not the whole story. As Arkin et al.’s (1976) study indicated, people are often motivated to interpret their experiences in a way that reflects favorably on themselves and their ability to be successful in the world in which they live. This may be a manifestation of a more general tendency to construct and apply implicit theories about oneself and the world that confirm the desire to have a happy and successful life. Kunda (1990) summarized substantial evidence in support of this possibility. For example, although college students know that 50% of all first mar-

THE IMPACT OF IMPLICATIONAL MOLECULES

279

riages end in divorce, they nevertheless are convinced that they personally will remain married to their first spouse for life (Kunda, 1987). This suggests that individuals who are motivated to believe that they will be successful in marriage, construct implicit self-theories that support this belief. To demonstrate this, Kunda (1987) gave participants information about a target person who was either happily married or divorced and whose demographic and personality characteristics either matched or did not match those of the participants themselves. After receiving this information, participants indicated which of the target’s attributes were most likely to contribute to his or her marital status. Participants were more inclined to attribute the success of happily married targets to characteristics that matched their own than to characteristics that did not. Correspondingly, they were less likely to attribute the failure of divorced targets to the former characteristics. Thus, they constructed theories about themselves that were consistent with the outcome they wished to attain. People who are motivated to construct and maintain a self-serving theory may selectively search memory for information that supports it. In a study by Santioso, Kunda, and Fong (1990), some participants were told that extroversion was conducive to success after leaving college, whereas others were told that introversion was more conducive to success. Then, in an ostensibly unrelated study, they were asked to list all of the behaviors they had performed in the past along a related trait dimension (shy vs. outgoing). Participants listed more behaviors that were congruent with the trait they were told was conducive to success than behaviors that were incongruent with this trait. In a second study, participants after exposure to the first task were shown a series of behaviors and were asked in each case to press a button as soon as they thought of a personal experience that exemplified it. Participants responded more quickly to behaviors that exemplified the success-related trait than to behaviors that exemplified the opposite. Thus, people who are told that a particular trait is associated with success selectively searched memory for personal experiences that confirmed their possession of this trait. Consequently, these experiences came to mind more easily when they were called upon to report instances of their behavior at a later point in time. Research in other paradigms has similar implications. For example, people are more likely to report engaging in a particular activity (drinking coffee, brushing one’s teeth, etc.) if they are told that the activity is healthy than if they are told it is not (Ross, McFarland, & Fletcher, 1986; B. Sherman & Kunda, 1989). These studies do not provide direct evidence that these responses were mediated by the construction of an implicit theory. However, it seems reasonable to suppose that the construction of self-serving theories is a result of similar processes.

280

CHAPTER 10

THE ROLE OF IMPLICIT THEORIES IN HEURISTIC-BASED JUDGMENTS The use of heuristics in making inferences is well established (for reviews, see Kahneman, Slovic, & Tversky, 1982; Nisbett & Ross, 1980; Sherman & Corty, 1984). As Wyer and Srull (1989) noted, many of these heuristics may reflect a general tendency to treat conditional relationships as biconditionals. That is, people assume that if A implies B, then B implies A (for evidence, see Wyer, 1977). Thus, for example, if people believe that members of a certain category have a certain cluster of attributes, they infer that individuals with this cluster of attributes belong to the category. Alternatively, if they believe that events that are likely to occur are easy to imagine, they infer that events that are easy to imagine are likely to occur. The latter, “availability” heuristic (Tversky & Kahneman, 1973) is of particular interest in the present context. This heuristic can be conceptualized as an implicit theory in the form of an ease-of-retrieval molecule; specifically: [X has occurred frequently (infrequently); instances of X are easy (difficult) to recall].

Thus, people may use this molecule not only to infer the ease of recalling instances of an event from knowledge of how often it has occurred, but also to infer the frequency of an event’s occurrence from the difficulty of recalling instances of it. Two quite different sets of phenomena can be conceptualized as resulting from an application of this molecule. One concerns the effects of television on perceptions of social reality. The other concerns the effects of ease of retrieving knowledge about oneself and others on judgments of the attributes to which this knowledge is relevant. Research on each set of phenomena is discussed in turn. Perceptions of Social Reality Americans watch an average of more than 4 hours of television each day (Nielsen, 1995). As a result, television is a major source of the episode models that people form. On the other hand, people often fail to distinguish clearly between the sources of the knowledge they acquire (Hasher, Goldstein, & Toppin, 1977; Jacoby, Kelly, Brown, & Jasechko, 1989; Johnson, Hastroudi, & Lindsay, 1993). As a result, they are likely to use the episode models they form while watching television as a basis for judgments without considering the context in which these models were constructed. L. J. Shrum and his colleagues (O’Guinn & Shrum, 1997; Shrum, O’Guinn, Semenik, & Faber, 1998; Shrum, Wyer, & O’Guinn, 1998) have examined im-

THE IMPACT OF IMPLICATIONAL MOLECULES

281

plications of this possibility in the context of their research on the “cultivation effect” (Gerbner, Gross, Morgan, & Signorielli, 1994). That is, people tend to overestimate the occurrence of events in the real world that are disproportionately represented on television. Moreover, the magnitude of their overestimation increases with the amount of television they watch. Instances of events that occur frequently on television are easily accessible in memory, and this is particularly true for heavy viewers. Therefore, if individuals are asked to infer the frequency of occurrence of these events, and if they apply the ease-of-retrieval molecule, heavy users are likely to make higher estimates than light viewers do. Thus, for example, heavy viewers are relatively more likely than light viewers to overestimate the proportion of Americans who belong to a country club or have a swimming pool in their back yard (O’Guinn & Shrum, 1997). On the other hand, they are also more likely to overestimate the incidence of violent crime and the size of the police force. Several studies by Shrum and his colleagues support this interpretation. In one study (O’Guinn & Shrum, 1997), frequent and infrequent viewers of soap operas were asked to estimate the incidence of characteristics that are associated with an affluent life style (belonging to a country club, owning a luxury car, having a swimming pool, etc.). Frequent viewers not only estimated the incidence of these characteristics to be higher than infrequent viewers did, but also made their estimates more quickly. These findings are consistent with the hypothesis that frequent television viewers find instances of these characteristics easier to retrieve from memory and estimate their incidence to be greater as a result. Other interpretations of these effects should of course be considered. For example, a third variable, such as education level or socioeconomic class, could have independent effects on both television watching and frequency estimates, producing an artifactual relation between these variables that would not otherwise exist. However, O’Guinn and Shrum (1997; see also Shrum, Wyer, & O’Guinn, 1998) found that the relation between frequency estimates and television viewing was evident even when the potentially confounding effects of these variables were controlled. Moreover, Shrum et al. (1998) found that when participants were asked to report their television viewing habits before making frequency estimates (thus calling their television watching to their attention), the impact of television viewing on frequency estimates was significantly reduced. The effects of watching television on frequency estimates can also be reduced or eliminated by increasing participants’ motivation to make correct judgments (Shrum, 1999). Thus, people appear able to distinguish between events they see on television and those they learn about from other sources, and thus can avoid the bias produced by television when they are motivated to do so. Normally, however, this motivation does not exist.

282

CHAPTER 10

Although the effects of television on perceptions of social reality are well established, a qualification on these effects might be noted. Hamilton and Gifford (1976) found that the incidence of infrequently occurring events is overestimated to a greater extent than the incidence of more common ones. This is presumably due in part to the fact that rare events are thought about more extensively (for evidence that the recall of information increases with amount of processing, see Craik & Lockhart, 1972). In some cases, this effect could override the effects identified by Shrum and his colleagues. On the other hand, people are typically passive viewers of television, and are likely to encode and store the information they receive with minimal cognitive elaboration. To this extent, the effects identified by Hamilton and Gifford (1976) may not be pronounced. The effects of television on perceptions of social reality have potentially important implications. For example, individuals who are frequently exposed to television violence may come to view violence and aggression as normal aspects of life and, therefore, may have more tolerance of aggression than they otherwise would. Zillman and Bryant (1982) found that exposing naïve college students to heavy doses of sadomasochistic pornography over a period of several weeks increased their tolerance of rape. Moreover, these effects were still evident several months after the main study had been completed. Other effects of the media on attitudes and values are worth examining from this perspective.

Effects of Ease of Retrieval on Perceptions of Self and Others Shrum et al.’s research suggests that the ease of recalling instances of an event increases estimates of its frequency. To the extent people apply the ease-of-retrieval molecule, difficulty in retrieving instances should have the opposite effect. An intriguing demonstration of this effect was constructed by Norbert Schwarz and his colleagues (for a review, see Schwarz, 1998). In a typical study (Schwarz et al., 1991), participants were asked to recall either 6 or 12 instances of assertive behavior. After doing so, they rated their assertiveness. Participants obviously recalled more behaviors when they were asked to recall 12 than when they were asked to recall 6. Therefore, one might expect participants to judge themselves to be more assertive in the former condition than the latter. On the other hand, it is much more difficult to generate 12 instances of the attribute than 6 instances. Therefore, to the extent individuals base their judgments on the ease-of-retrieval molecule, they should infer that they engage in assertive behavior less frequently in the former condition than the latter. Thus, they should infer that they are less

THE IMPACT OF IMPLICATIONAL MOLECULES

283

assertive when they have been asked to recall 12 instances than when they were asked to recall only 6. This was in fact the case. The use of an ease-of-retrieval heuristic as a basis for judgment is quite pervasive (for a review, see Schwarz, 1998). Further examples are described later in this chapter. Its implications can be quite ironic. For example, people may be less likely to believe that a proposition is true if they have attempted to generate a large number of reasons for its validity than if they have thought about only a few. Research by Wänke, Bless, and Biller (1996) supported this speculation. Some participants were asked to generate either 3 or 7 arguments that either favored or opposed a specific issue, after which they were asked to report their own position on the issue. Other, yoked participants read the arguments that individuals in the first group had written. The latter participants reported themselves to be more in favor of the position advocated when they had read 7-argument responses than when they had read 3-argument responses, confirming the assumption that the substantive implications of the 7-argument sets were relatively more persuasive. Nevertheless, the participants who had actually generated the arguments judged themselves to be less in favor of the position when they had generated 7 arguments than when they had generated only 3. Thus, the effects of ease of retrieval overrode the effects of actual knowledge. People do not always ignore the implications of their past knowledge, of course. However, their computation of a belief on the basis of these implications is cognitively effortful. Consequently, they may only do so when ease of retrieval is likely to be an unreliable criterion. In other conditions of Schwarz et al.’s (1991) research, for example, participants generated instances of assertiveness in the presence of distracting background music. In this case, participants apparently attributed their difficulty of generating instances to the distraction and to their lack of knowledge. In these conditions, therefore, they judged themselves more assertive when they had generated 12 instances rather than 6. 1. Effects of Information About Others’ Ease of Retrieving Information As suggested in our discussion of self-perception processes, people can use implicit theories to construe the implications of others’ behavior as well as their own. Support for this assumption in the present paradigm was obtained by Menon and Raghubir (1998) in a study of consumer judgment. Participants read an advertisement that described the attractive features of a computer. Later, they were asked to recall either 2 or 8 of these features. Either before or after the recall task, however, they were given information about the difficulty that others had in recalling the product’s attributes. If

284

CHAPTER 10

participants use the ease-of-retrieval molecule to infer the implications of this information, they should infer that the computer has more attractive features when others find it easy to recall these features than when they did not, and should judge the product more favorably in the former condition. Support for this prediction is conveyed in Table 10.1, which summarizes participants’ product evaluations as a function of the number of features they were personally asked to recall, the difficulty that others ostensibly had in recalling attributes, and the point at which the latter information was provided. Participants generally evaluated the product more favorably when others ostensibly found it easy to recall features than when they found it difficult. However, participants’ use of the ease-of-retrieval molecule to construe the implications of their own behavior was evident only when they recalled the product’s features before learning about the difficulty that others had. In this condition, they evaluated the product more favorably when they had recalled 2 attributes than when they recalled 8. When they were told about others’ difficulty at the outset, however, they appeared to ignore the difficulty they personally experienced and based their judgments on the number of features they could recall, thus reporting more favorable evaluations when they were asked to recall 8 features rather than 2. In summary, people appear to use the ease-of-retrieval molecule to infer the incidence of information in memory when (a) alternative bases for judgment are not easily accessible in memory and (b) the difficulty they have in recalling the information is unlikely to be due to extraneous situational factors. However, when alternative reasons for their subjective feelings are called to their attention, or when other bases for judgment (e.g., the difficulty that others have had recalling attributes) are easily available, people apply other criteria instead. TABLE 10.1 Effects of Subjective Difficulty of Recalling Product Attributes, and Perceptions of Others’ Difficulty, on Product Evaluations Two Attributes Recalled

Eight Attributes Recalled

5.9 5.4 5.6

4.9 4.4 4.9

5.4 4.9

5.3 4.3 4.8

5.9 4.9 5.4

5.6 4.6

Others’ difficulty mentioned after recall Others find recall easy Others find recall difficult M Others’ difficulty mentioned before recall Others find recall easy Others find recall difficult M Note. Adapted from Menon and Raghubir (1998).

M

THE IMPACT OF IMPLICATIONAL MOLECULES

285

2. The Role of Ease of Retrieval in Hindsight Bias The role of ease of retrieval in the construction and use of implicit theories is demonstrated in a series of studies by Sanna and Schwarz (2003; Sanna, Schwarz, & Small, 2003; Sanna, Schwarz, & Stocker, 2002). People who know that an event has occurred often overestimate the likelihood that they would have predicted it (Fischoff, 1975, 1982; for a review, see Hawkins & Hastie, 1990). This “hindsight” bias presumably occurs for reasons similar to that we assumed to underlie the effects reported by L. Ross et al. (1975). That is, individuals who are told that an event has occurred attempt to generate a plausible explanation for it and, if this can be done easily, may conclude that the occurrence of the event was foreordained. In doing so, they ignore the possibility that if the event had not occurred, they might have generated an explanation of its nonoccurrence just as easily. If this is the case, however, stimulating individuals to generate additional explanations for the event, which could be difficult, might decrease or reverse the effects observed by Fischoff (1975). Correspondingly, stimulating people to generate explanations of why the event might not have occurred could increase the magnitude of the hindsight bias. Sanna and Schwarz (2003) found this to be the case. Participants in one series of studies read a story about a military conflict and were arbitrarily told that one of the two antagonists had been victorious. Participants were then told to generate either 2 or 10 thoughts about how this outcome might have been averted. Participants who generated thoughts about the alternative outcome of the conflict decreased their belief that this outcome could have occurred. However, this effect was much greater when they had generated 10 thoughts (which was difficult to do) than when they had generated only 2. Thus, generating explanations for how the alternative outcome could have occurred increased beliefs that the actual outcome was inevitable, thus magnifying the hindsight bias that participants manifested. In a second series of studies (Sanna, Schwarz, & Small, 2003), however, participants thought about why the outcome they were given should have occurred. In this case, participants who generated many thoughts decreased their belief that the outcome was inevitable, thus eliminating the hindsight bias that was otherwise evident. In summary, people may often spontaneously generate an implicit theory for why an outcome will occur, and this theory may underlie their belief that the outcome was foreordained (Fischoff, 1975, 1982). However, generating additional explanations for the outcome, which is difficult, can decrease their belief that the outcome was inevitable, whereas generating reasons why the outcome might not have occurred is likely to strengthen this belief. Thus, as also implied by Wänke et al.’s (1996) findings noted earlier, the greater the number of thoughts that individuals generate in support of a position actually weakens their belief that the position is valid.

286

CHAPTER 10

INDIVIDUAL DIFFERENCES IN IMPLICIT THEORIES To reiterate, more than one implicit theory can often be used to account for a given social experience. In such instances, individual and situational differences may arise in the theory that people happen to apply and, therefore, in the conclusions they draw from the experience. These differences are likely to be reflected in their judgments and behavioral decisions. Implicit Theories of Personal Achievement Individual differences in implicit theories of social judgment and behavior were suggested by Carol Dweck and her colleagues (Dweck, 1991; Dweck & Leggett, 1988; Erdley & Dweck, 1993) in an analysis of individuals’ responses to success and failure. Based on Dweck’s earlier work on learned helplessness (for a summary, see Dweck & Leggett, 1988), they postulated that some individuals believe that performance on an achievement task can typically be increased by exerting effort, whereas others are disposed to belief that performance is largely a reflection of abilities that are fixed and unable to change. The first belief might be captured by an implicit theory of the form: [P performs task; P fails; P tries harder; P succeeds].

The second, however, could reflect the theory: [P has low ability; P performs task; P fails; P tries harder; P fails again].

Thus, suppose people perform a task and fail. This event (“P fails”) instantiates a segment of each of these theories. Individuals who activate and use the first theory to interpret the event may infer that they did not work hard enough and that if they exert more effort, they might ultimately succeed. As a result, they may be stimulated to try the task again. However, suppose individuals activate the second theory instead. Then, they are likely to infer that their failure reflects low ability and that repeating the task will have the same outcome. Therefore, they may be disinclined to try a second time. Individual differences in the chronic accessibility of these theories, and their effects on responses to failure, have been elaborated by Dweck and Leggett (1988). Incremental Versus Entity Theories of Personality and Ability Dweck (1991; Dweck & Leggett, 1988) suggest that the aforementioned performance-related theories exemplify two more general theories of personality and behavior. Incremental theories imply that people’s attributes and

THE IMPACT OF IMPLICATIONAL MOLECULES

287

abilities are malleable and can be modified, whereas entity theories imply that attributes are fixed and resistant to change. These different theories have numerous implications for judgment and behavior, many of which have been examined by Dweck, Chiu, and Hong (1995a, 1995b; Dweck, Hong, & Chiu, 1993; Chiu, Dweck, Tong, & Fu, 1997; Chiu, Hong, & Dweck, 1997; Hong, Chiu, Dweck, & Sacks, 1997). Hong et al. (1997), for example, found evidence that people who were classified as entity theorists on the basis of an independent measure were more influenced by a target person’s personality and ability test scores than incremental theorists were, suggesting that they attached relatively more significance to measures that ostensibly assess stable attributes and abilities. People may also have incremental vs. entity theories of morality (implying that individual’s moral character can or cannot be easily modified) and of the world in which they live. A series of studies by Chiu et al. (1997) provided evidence that individuals with entity theories of morality and the world are more inclined than individuals with incremental theories to evaluate others’ behavior in terms of its fulfillment of obligations rather than in terms of individual rights. Thus, for example, participants in one study were asked to give open-ended descriptions of how they would handle hypothetical classroom situations. Some scenarios described socially desirable behavior (e.g., “The teacher asked Jerry to remove old papers and notices on the students’ notice boards, and Jerry does what she requested as soon as possible.”). Others described undesirable actions (e.g., “The teacher asked Peter to remove old papers and notices. . . . A week has passed and Peter does not do it.”). Entity theorists, who believe that the fulfillment of responsibilities is normative, were less inclined than incremental theorists to indicate that Jerry should be rewarded for fulfilling his obligations. Correspondingly, they were more inclined than incremental theorists to recommend that Peter be punished for not meeting his responsibilities. As Dweck et al. (1995a, 1995b) pointed out, the implicit theories that individuals apply may often be domain specific. Moreover, several alternative theories may coexist, the implications of which can conflict. The individual differences in implicit theories that Dweck and her colleagues identified in the aforementioned studies could reflect differences in the chronic accessibility of these theories. To this extent, situational factors that make one or another theory more accessible in memory might sometimes override these chronic differences. Although this possibility was not examined in the studies cited in this section, evidence that the effects of transitory situational factors can often dominate the effects of chronic accessibility (at least in the short run) is provided by other research described presently.

288

CHAPTER 10

IMPLICIT THEORIES OF SOCIAL SUPPORT: THE EFFECT OF PERSPECTIVE ON THEORY ACTIVATION The implicit theories that people acquire are often the result of accumulated past experiences in a particular type of situation that have been sufficiently involving to stimulate the construction of a generalized event representation of the sequence of events that take place. Individuals are therefore likely to differ in the nature of these theories, depending on the past experiences they have typically had. However, the implications of their theories can also depend on the perspective from which the experiences in question were viewed. An interesting example of these contingencies is provided by Eric Mankowski’s research on perceptions of social support (Mankowski & Wyer, 1996). Imagine the following situation: Russell has learned he got a D on the psychology midterm. He is near tears when he gets back to his room where his roommate, Chris, is studying. He tells Chris the bad news. “If I don’t get good grades, I’ll have to drop out of school. . . . I can’t seem to do well here.” “Wow,” Chris replies, “What do you say we go out for a pizza and beer? I haven’t had a good day, either.” They go out for pizza. . . . Russell tells Chris, “Remember that girl I met at your party? She’s right over there. . . . We were supposed to go to a movie together, but she never showed up. . . . I really thought she was interested in me. . . . I don’t know if I can eat with her sitting there.” “Oh, she won’t even notice us here,” Chris says. “You should get fixed up with someone. . . . Mark might have a suggestion. He sure seems to have no problem getting dates.”

This story is ambiguous in terms of the social support it describes. Chris’s suggestion to get a pizza, for example, could be viewed as either a desire to distract Russell from thinking about his bad grade or, alternatively, as insensitivity to Russell’s need for sympathy and reassurance. Readers’ perception of the story’s implications could therefore depend in part on the semantic concepts that they bring to bear on its interpretation. However, their construal of the story’s implications could also depend on the similarity of the situation it describes to their own past experiences. For example, people could use a theory about their own experiences as a standard of comparison in evaluating others. Thus, they might perceive Chris’s behavior to be relatively more supportive if they have often experienced rejection themselves than if they have not. This tendency, on the other hand, is likely to depend on the whether the new experience is viewed from the perspective of the potential support provider or the point of view of the recipient.

THE IMPACT OF IMPLICATIONAL MOLECULES

289

These contingencies were identified by Mankowski and Wyer (1996). Participants were told that the study was concerned with person impression formation. On this pretense, they were asked to read a story similar to the one just described, in which one person was in need of support and the other was in a position to provide it. In third-person perspective conditions, the story was written from the perspective of a disinterested observer. In provider-perspective conditions, however, the provider was referred to as “you” and the recipient was identified by a name that was ambiguous with respect to gender (Chris or Terry). In recipient-perspective conditions, the recipient was referred to as “you” and the potential provider was identified by name. After reading the story, participants in each perspective condition wrote down three adjectives that they would use to describe the provider and then rated the provider with respect to several traits that were relevant to social support (supportive, warm, rejecting, accepting, sensitive, etc.). Finally, participants were completed a general index of the support they perceived to be personally available (Cutrona & Russell, 1987). Participants’ perceptions of support availability could reflect two things. First, it could indicate the frequency with which participants had personally encountered support-relevant situations in the past and, therefore, the likelihood that semantic concepts associated with social support were chronically accessible in memory. To this extent, participants’ perceptions of support availability should have a positive impact on their use of support-related concepts to interpret the provider’s ambiguous behavior, and this should be true in all three perspective conditions. On the other hand, participants’ perceptions of support availability could also indicate the likelihood that they have constructed an implicit theory of themselves as recipients of support. If participants who read the story from the recipient’s perspective activate and use this theory as a standard of comparison in construing the story’s implications, they should perceive the situation o be less supportive when their self-referent theory implies that they typically receive support than when it does not. This contrast effect of the story’s implications on judgments could offset the positive effect of semantic concepts on its interpretation that occurs at an earlier stage of processing. Results were consistent with this analysis. Table 10.2 shows the supportiveness implied by both participants’ open-ended trait descriptions of the support provider and their ratings of the provider’s supportiveness. When participants took the perspective of either the provider or third person, their perceptions of the provider’s supportiveness increased with their perception of their own support availability. When they took the role of the recipient, however, this was not the case. In these latter conditions, the contrast effect that resulted from using an implicit theory of their personal experiences as a standard of comparison appeared to offset the positive ef-

290

CHAPTER 10 TABLE 10.2 Perceptions of the Provider’s Supportiveness as a Function of Perspective and Social Support Availability Perspective

Supportiveness of trait descriptions High support availability Low support availability Difference Supportiveness of trait ratings High support availability Low support availability Difference

Provider

Third Person

Recipient

0.39a 0.18 0.21

0.14 –0.26 0.40

–0.44 –0.26 –0.18

0.81b 0.35 0.46

0.26 –0.44 0.70

–1.07 –1.16 0.09

Note. Adapted from Mankowski and Wyer (1996). aScores reflect the mean difference between the number of trait descriptions that implied support and the number that implied lack of support. bScores reflect the mean difference in ratings of traits implying support (reported along a 0–10 scale) and ratings of traits implying rejection.

fects of semantic concepts on the interpretation of the scenario that was evident in other perspective conditions.

CULTURAL DIFFERENCES IN IMPLICIT THEORIES Individual differences in the chronic accessibility of implicit theories are likely to result in part from differences in the socialization practices to which individuals have been exposed (cf. Miller, 1994; Nelson, 1993; see chap. 6) and, therefore, the frequency with which the theories have been applied in the past. To this extent, differences may exist in the theories that are typically applied by members of different subgroups within a society (e.g., men vs. women, Blacks vs. Whites, etc.). More general cultural differences may exist as well. The precise nature of this difference is somewhat controversial (cf. Briley & Wyer, 2001; Heine, Lehman, Markus, & Kitayama, 1999; Markus & Kitayama, 1991; Triandis, 1995). However, certain general differences appear to exist. For example, members of Western and Asian cultures are typically characterized as differing in terms of individualism and collectivism (Hofstede, 1991; Triandis, 1995). Individualism is characterized by a focus on individual freedom and independence, a concern with personal goals with little consideration of others, independently of others, and competitiveness. In contrast, collectivism is typified by a focus on oneself in relation to the groups to which one belongs, and a concern with group well-being rather than per-

THE IMPACT OF IMPLICATIONAL MOLECULES

291

sonal goals. As Briley and Wyer (2001) pointed out (see also Ho & Chiu, 1994; Rhee, Uleman, & Lee, 1996; Triandis & Gelfand, 1998), the individualism–collectivism construct is multidimensional, and the dimensions that are often assumed to distinguish individualists and collectivists (i.e., individuality, cooperation, competitiveness, self-sacrifice, etc.) are not strongly related (Briley & Wyer, 2001). Nevertheless, a general cultural difference in the tendency to think of oneself as an independent individual vs. a member of a group is fairly pervasive (Markus & Kitayama, 1994). However, the implicit theories that reflect these different cultural orientations are not the only ones that people acquire. As Hong et al. (2000) suggested, representatives of a given culture are often exposed to norms and values that pervade other cultures as well. Moreover, many individuals are bicultural, having lived for many years in more than one country, or being residing in a country (e.g., Hong Kong) in which both Western and Asian norms and values are pervasive. Hong et al. (2000) conceptualize culture as a dynamic construct, the effects of which can depend on situational factors that influence the extent to which cultural norms and values are accessible in memory. Several studies provide evidence of this dependence. Attributions of Causality Hong et al. (2000) showed that the implicit theories that Hong Kong participants bring to bear on the interpretation of information and the inferences they draw from this information can be induced by exposing participants to cultural symbols of Western and Asian cultures prior to the judgment task. For example, Westerners (individualists) typically acquire an implicit theory of causality that places responsibility on the individual, whereas Asians (collectivists) acquire theories that place responsibility on the society or social context in which the behavior occurs. Evidence supporting this difference was reported by Morris and Peng (1994; see also, Choi, Nisbett, & Norenzayan, 1999). Nevertheless, both implicit theories are likely to exist in memory, and the theory that is activated and applied may be influenced by situational factors that affect its accessibility. To demonstrate this, Hong et al. constructed stimulus materials consisting of a school of fish in which one fish was swimming ahead of the others. Thus, the situation could be interpreted as an indication that either the first fish was leading the others (implying a dispositional cause of the fish’s behavior) or, alternatively, that the first fish was being chased by the others (a situational cause of its behavior). Hong Kong Chinese undergraduates were exposed to these pictures with instructions to explain why the one fish was swimming ahead of the others. Before doing so, however, they were exposed to either a series of American cultural icons or a series of Chinese icons as part of an ostensibly different experiment. Exposure to

292

CHAPTER 10

American icons increased the tendency to assign a disposition cause to the fish’s behavior relative to control conditions, and correspondingly decreased the tendency to assign a situational cause to it. However, exposure to Chinese cultural icons had the opposite effects. Theories of Personal Causality A series of studies by Oishi, Wyer, and Colcombe (2000) identified cultural differences in the situational factors that activate implicit theories about the determinants of one’s own behavior, and demonstrated that these differences are reflected in people’s use of their current life satisfaction to predict the future. People are likely to use this criterion if they activate a person-focused theory of causality (i.e., a theory that life experiences are determined by stable personal characteristics that persist over time). However, they are less likely to do so if they apply a situation-focused theory (i.e., the theory that life experiences are influenced by unforeseen situational factors). Like Hong et al., we reasoned that people are likely to have both implicit theories in memory regardless of their cultural background. However, we expected that the situational factors that activate these theories would differ. Specifically, Heine and Lehman (1997; Heine et al., 1999) noted that representatives of individualistic cultures exhibit strong self-enhancement motives, as evidenced by a desire to take responsibility for positive experiences but to attribute negative experiences to transitory situational factors (Harvey & Weary, 1984; Zuckerman, 1975; see also the study by Arkin et al., 1976, described in chap. 9). If this is so, calling European Americans’ attention to positive experiences they have had in the past (for which they presumably take responsibility) should activate a theory that people are personally responsible for the events that befall them. Therefore, it should increase the likelihood that these individuals see themselves as responsible for both their current and future life situation, and should increase their tendency to use their current life satisfaction to predict the future. However, calling these individuals’ attention to negative life events (which they attribute to situational factors) should activate a theory that life events are the result of unpredictable factors over which they have little control, and so it should decrease their tendency to use current life satisfaction to predict the future. These predictions should not hold for Asians, however. Members of Asian societies tend to be self-effacing, and to take responsibility for personal failure but attribute success to external circumstances (Fry & Ghosh, 1980; Yamauchi, 1988). If this is so, calling Asians’ attention to positive and negative life experiences should have the opposite effects that it has on Americans. That is, thinking about positive life events should activate a sit-

293

THE IMPACT OF IMPLICATIONAL MOLECULES

uation-based theory, whereas thinking about negative life events should activate a person-based theory. Consequently, Asians should be less inclined to use their current life satisfaction to predict their future in the former condition than the latter. Oishi et al. (2000) confirmed these predictions. Participants were European Americans and Asian Americans whose life satisfaction was assessed in an ostensibly unrelated experiment. Participants first completed a lifeevent inventory in which they were asked to write about either a positive life experience they had had or a negative one. Then, after doing so, they were given a list of 10 positive and 10 negative events (getting the best grade on an exam, finding that your friends were talking about you behind your back, etc.), and estimated the likelihood that each event would occur to them in the near future. Multiple regression procedures were used to analyze the favorableness of participants’ predictions of the future as a function of a number of orthogonal contrasts corresponding to the main and interactive effects of life satisfaction (a continuous variable), cultural background and the valence of the life experience that participants recalled. This analysis yielded a significant interaction of all three variables. The implications of this interaction can be seen in the left half of Table 10.3, which shows estimated values of the favorableness of participants’ future outlook as a function of culture, the valence of the stories that participants wrote about themselves, and current life satisfaction. European Americans’ life satisfaction had a greater effect on their future outlook when they had written about a positive life experience than when they had written about a negative one. In contrast, Asian Americans’ life satisfaction had more influence on their future outTABLE 10.3 Regression-Based Estimates of Future Outlook as a Function of Current Life Satisfaction, Cultural Background, and the Valence of Stories About Self and Other Valence of Stories About Self

European Americans High current life satisfaction Low current life satisfaction Difference Asian Americans High current life satisfaction Low current life satisfaction Difference

Valence of Stories About Others

Favorable

Unfavorable

Favorable

Unfavorable

0.36 –2.04 2.40

1.26 –0.26 1.52

1.44 1.18 0.26

1.84 –0.38 2.22

0.80 –0.32 1.12

1.54 –1.10 2.64

1.74 0.56 1.18

0.98 0.24 0.74

Note. Adapted from Oishi et al. (2000).

294

CHAPTER 10

look when they had written about a negative experience than when they had written about a positive one. A second study was analogous to the first but focused on the effects of activating concepts about others. If European Americans are motivated by a desire to be superior to others, they may tend to attribute others’ successes to situational factors but others’ negative outcomes to dispositional factors (Snyder, Stephan, & Rosenfield, 1976). In contrast, if Asians are typically self-effacing and tend to promote others’ interests at their own expense, they may be inclined to attribute others’ successes to dispositional causes (Yamauchi, 1988). To this extent, a person-based theory of causality should be activated in European Americans by stimulating them to think about others’ failures, but should be activated in Asian Americans by inducing them to think about others’ successes. To explore this possibility, participants participated in a study that was very similar to the first. Rather than writing about a personal experience, however, they read a newspaper story describing either positive or negative job prospects for the current year’s college graduates. (Participants were freshmen, and so the story was not personally relevant.) Regression analyses of these participants’ future outlook as a function of experimental variables yielded a three-way interaction of cultural background, life satisfaction, and the valence of others’ experiences, the implications of which are shown in the right half of Table 10.3. The pattern of these data is exactly the opposite of the first set of data. That is, European Americans’ future outlook was most strongly influenced by their current life satisfaction when they had read about others’ negative experiences, but Asians’ future outlook was most strongly influenced by life satisfaction when they had read about others’ positive experiences. Cultural Differences in Inference Strategies The preceding discussion suggests cultural differences in the content of the implicit theories that are applied by members of different cultures. However, more general differences may exist in the conditions in which these theories are used. Choi et al. (1998) suggested that general cultural differences between Western and Asian cultures exist in the tendency to use analytic reasoning in generating explanations for social events. To this extent, the inference processes that are implicitly assumed in much of our discussion of the use of implicational molecules might not generalize over cultures. Some support for this possibility was obtained in a study by Norenzayan and Kim (2000) in research on the Socratic effect (McGuire, 1960). In an initial experimental session, European Americans and Asians completed a belief questionnaire containing causally related propositions of the sort to which Equation 10.1 is applicable. They then returned 1 week later and com-

THE IMPACT OF IMPLICATIONAL MOLECULES

295

pleted the questionnaire a second time. Based on research and theorizing by McGuire (1960; Rosen & Wyer, 1972), Norenzayan and Kim assumed that participants might show inconsistencies in their beliefs initially, but that making the beliefs salient in temporal proximity would lead them to revise these beliefs to make them more consistent. Thus, their beliefs should be more consistent in the second session than they were at first, based on Equation 10.1. This was true of European Americans, thus replicating findings by Rosen and Wyer (1972). However, Asians’ beliefs were just as inconsistent in the second session as they were at the outset. This suggests that Asians, unlike Westerners, either did not perceive their beliefs in the propositions to be inconsistent or, alternatively, employed inference processes that differed from those described by Equation 10.1.

THE ROLE OF IMPLICIT THEORIES IN MARITAL SATISFACTION The theories we construct and apply in evaluating our life experiences can often have social consequences. For example, people often have narrativebased theories about close relationships. However, substantial individuals exist in the nature of these theories. Some people, for example, may believe that marriage partners fall in love quickly but that these feelings inevitably deteriorate over time, ultimately leading to separation or divorce. Others may believe that love takes a long time to develop but that once established, it is maintained. Still others may believe that people’s feelings change nonmonotonically. For example, partners’ affection for one another decreases over the early years of a relationship as romantic love dissipates but then increases again in later years as the relationship matures. Several theorists have conceptualized the role that narrative representations can play in the dynamics of close relationships (Forgas, 1991; Holmes & Murray, 1995; Miller & Read, 1991; Schank & Abelson, 1995). Murray and Holmes (1996), for example, found that marriage partners often reconstruct stories about their personal relationships in order to make them consistent with the implications of their narrative-based theories about relationships in general (see also Holmberg & Holmes, 1994). However, it seems likely that if partners have different theory-based expectations for the typical progress of close relationships, they may differ in how they evaluate their own relationship. These differences could create marital conflict and dissatisfaction. Research by Gohm and Wyer (1998) bears indirectly on this possibility. We assumed that individuals’ theories about the prognosis of relationships over time could be inferred from their perceptions of change in the feelings that characterize these relationships, and that differences in marital satis-

296

CHAPTER 10

faction might be related to these theories. To this end, we identified 32 married couples who had been in their relationship for an average of 9 years, and whose marital satisfaction (based on the Locke–Wallace, 1959, scale of marital adjustment) had been determined in an earlier study. In 16 of these couples, both partners expressed satisfaction with their relationship, and in the other 16, both partners reported being dissatisfied. Participants were told that the study’s purpose was to understand people’s perceptions of how interpersonal relationships were likely to change over time. On this pretense, participants plotted six graphs showing the changes that typically occurred in marriage partners’ feelings toward one another during the first 10 years of their relationship. Three graphs described the typical man’s feelings of romantic love, commitment to the marriage, and affection for his partner. Three others described the typical woman’s feelings along the same dimensions. To aid them in constructing their graphs, participants were given axes in which the x-coordinate pertained to the year of the relationship and the y-coordinate referred to the attribute being rated along a scale from 0 (none) to 10 (very much). Partners were told to place an “x” above the number denoting each year of the relationship to indicate their perception of the partner’s feelings during that year. Satisfied partners’ theories about the prognosis of marital relationships were expected to be more similar to one another than dissatisfied partners’ theories. Data pertaining to this hypothesis are conveyed in Figs. 10.2 and 10.3, which show the composite graphs drawn by satisfied and dissatisfied partners along each dimension. Analyses of these data revealed an interaction of relationship satisfaction, the partner being rated, and attribute dimension, F(2, 60) = 4.29, p < .05 and an higher order interaction involving these variables, participant sex, and time, F(18, 540) = 1.88, p < .01. These interactions are most clearly interpretable by considering data for satisfied and dissatisfied couples separately. As expected, satisfied husbands and wives had generally similar theories about how a typical marriage partner’s feelings were likely to change. For example, they agreed that the typical partner’s feelings of romantic love decreased to a much greater extent over time than his or her feelings of affection. Moreover, they agreed that the typical man’s feelings of love would tend to increase during the later years of the relationship without a corresponding increase in affection, whereas the typical woman’s feelings of affection were likely to increase in later years without a corresponding increase in feelings of romantic love. In contrast, dissatisfied partners’ perceptions of the typical marriage differed much more dramatically. These differences were reflected in an interaction of respondent sex, the sex of the partner being rated, attribute dimension, and time, F(18, 270) = 2.80, p < .01, in an analysis of judgments by dissatisfied partners alone. Three aspects of these data are noteworthy.

THE IMPACT OF IMPLICATIONAL MOLECULES

297

FIG. 10.2. Satisfied partners’ perceptions of the typical husband’s and typical wife’s feelings over the first 10 years of a relationship.

1. Dissatisfied husbands and wives both perceived the typical marriage partner’s feelings of romantic love to decrease substantially over the first 8 years of the relationship. However, whereas dissatisfied wives expected these feelings to increase in later years, dissatisfied husbands did not. 2. Dissatisfied husbands perceived the typical man’s commitment during early years of the relationship to correspond more closely to his feelings of romantic love than to his feelings of affection. In contrast, dissatisfied wives perceived the typical man’s feelings of commitment to parallel his feelings of affection, and to be maintained despite a decrease in his feelings of romantic love.

298

CHAPTER 10

FIG. 10.3. Dissatisfied partners’ perceptions of the typical husband’s and typical wife’s feelings over the first 10 years of a relationship.

3. Dissatisfied husbands believed that the typical woman’s commitment to her relationship corresponded more closely to her feelings of romantic love than to her feelings of affection. In contrast, dissatisfied wives perceived the typical woman’s commitment to parallel her feelings of affection and to differ markedly from her feelings of romantic love. The implications of these findings for an understanding of marital relationships are necessarily equivocal. It is unclear, for example, whether the differences in dissatisfied partners’ perceptions are a cause or an effect of their feelings about their own relationship. Be that as it may, however, the

THE IMPACT OF IMPLICATIONAL MOLECULES

299

results demonstrate the utility of comparing the implicit theories that are likely to characterize happy and unhappy couples. Differences in these theories may be worth more careful attention in future research.

EFFECTS OF IMPLICIT THEORIES ON BEHAVIORAL DECISIONS The preceding sections discuss the impact of people’s implicit theories on their judgments of self and others. Perhaps the most important influences of these theories, however, are on behavioral decisions. This influence is evident in Dweck’s work on learned helplessness, noted earlier. That is, people who apply an incremental theory of ability are likely to conclude that they can improve their performance on a task by exerting more effort and, therefore, they may persist. In contrast, people who invoke an entity theory may attribute their failure to an inherent lack of ability and conclude that they are likely to fail again regardless of their effort. Therefore, they may not bother trying (Dweck, 1975; Dweck & Leggett, 1988). We have already noted that when two or more implicit theories can potentially be brought to bear on judgments, the theory that is applied may be determined not only by its chronic accessibility but also by transitory situational factors that activate concepts and knowledge with which they are associated. This is true of theories that have implications for behavior as well. Many situational influences on behavior can be conceptualized in terms of the mediating effects of situational variables on the activation of different implicit theories. In this section, we restrict consideration to three quite different areas in which implicit theories potentially come into play in behavioral decisions. Although the research reported in each case does not unequivocally demonstrate the mediating influence of these theories, a consideration of the research from this perspective often raises additional issues that might not otherwise be identified. Helping Behavior Several factors that underlie decisions to provide help have already been noted in our discussion of the just world hypothesis. That is, people are more likely to provide help if they can construct an image-based episode model of how their behavior might benefit the recipient than if they cannot (Miller, 1977). However, other factors also enter into the picture. As noted by Hong et al. (2000), people may have implicit theories about not only the dispositional antecedents of their behavior but also situational influences, and both type of theories could affect their own behavioral decisions as well as their perceptions of others. These theories can have different impli-

300

CHAPTER 10

cations. For example, if people are asked to help another person, two types of theories could influence their decision to help. One, person-focused theory might be captured by the molecule: [P is a kind person; O asks P for assistance; P helps].

A second, recipient-focused theory might have the form: [O requests assistance; O’s request is (is not) legitimate; P helps (does not help)].

Note that if people who are asked to help activate the first theory and consider themselves to be kind, they are likely to provide help independently of other considerations. If they activate the second theory, however, their decision to help may depend on his or her perception of the request’s legitimacy. An imaginative study by Langer and Abelson (1972) can be conceptualized in terms of these considerations. Shoppers were approached outside a supermarket by a woman who had ostensibly hurt her leg. In some cases, the woman asked the shopper to call her husband to come and pick her up. This request was likely to be considered legitimate under the circumstances. In other cases, she asked the shopper to call her boss and tell him she would be late—something that was not sufficiently important to justify asking a stranger. The second variable manipulated in this study was the order in which the woman (a) described her adversity and (b) made her request. Specifically, under request-first conditions, the woman said: “Would you do something for me? Please do me a favor and call my husband to ask him to pick me up. My knee is killing me. I think I sprained it.” Under needfirst conditions, however, the woman’s words were: “My knee is killing me. I think I sprained it. Would you do something for me? Please call my husband and ask him to pick me up.” Thus, the descriptive content of the woman’s statements was identical in the two cases. Under request-first conditions, however, the victim’s initial statement (“Would you do something or me . . .”) focuses attention on the recipient and, therefore, is likely to activate a person-focused theory of the sort described earlier. To this extent, shoppers who typically consider themselves to be kind are likely to grant the request regardless of other considerations. In contrast, the woman’s initial statement under need-first conditions (“My leg is killing me . . .”) focuses attention on the victim and, therefore, is likely to activate the second theory, in which the legitimacy of the request comes into play. In this case, therefore, shoppers should comply with the request only when the request appears legitimate. Langer and Abelson’s findings are consistent with this analysis. That is, the likelihood

THE IMPACT OF IMPLICATIONAL MOLECULES

301

of helping the victim increased with the request’s legitimacy under needfirst conditions but not under request-first conditions. Effects of Anticipated Outcomes on Decisions People who are confronted with a decision are likely to base it on an implicit theory about the consequences of the alternative courses of action available. These theories are often domain specific. People’s theories about the consequences of betting on a horse race are likely to differ from their theories about the consequences of studying for a final examination. However, more general theories can exist as well. As we discussed in chapter 3, Higgins (1997) postulates two general motivational orientations that can underlie goal-directed information processing. One, promotion focus, is characterized by an emphasis on the rewards that can potentially result from a behavioral decision irrespective of the costs that might be incurred. The other, prevention focus, is characterized by a concern with avoiding negative consequences of one’s behavior independently of its potential benefits. These different orientations could be embodied in two implicit theories, each of which has implications for a different behavioral objective: [P performs A; A has positive consequences; P attains benefits],

and [P performs B; B avoids negative outcomes; P avoids misfortune].

In many instances, the behavior that maximizes the likelihood of positive outcomes also increases the risk of negative ones. (E.g., Investing in stocks can potentially reap substantial benefits but runs the risk of a substantial loss as well.) In these instances, the two theories can have different implications. For example, suppose an individual is confronted with a choice between (a) a vacation in which one is likely to have new and exciting cultural experiences but will require living in sleazy hotels and eating poor food, and (b) a vacation that is much less exotic but where the accommodations are not elegant but are adequate. A person who uses the promotion-focus molecule to predict the individual’s behavior should infer that the vacationer will choose the first option. However, a person who activates the prevention molecule should predict the vacationer to choose the second alternative. This possibility becomes of particular interest when people apply the theories to their own behavioral decisions. That is, the decision of an individual who is personally confronted with the two vacations just described might depend on which of the two alternative theories happens to be applied. Several situational and individual difference factors could influence

302

CHAPTER 10

the relative accessibility of the theories and, therefore, the likelihood they are applied. One factor, discussed in chapter 3, can be the salience of one’s group membership. That is, consciousness of belonging to a group appears to activate feelings of responsibility to others and, therefore, to increase attention to negative consequences of behavior. Thus, it may increase the relative accessibility of a prevention-focus molecule that, once activated, influences decisions in both interpersonal and individual choice situations (Briley & Wyer, 2002). Situational factors that emphasize the importance of success as opposed to the importance of avoiding failure could have a similar influence. Finally, chronic individual differences in the accessibility of the two theories can also exist (Aaker & Lee, 2001; Briley et al., 2000). (This research is also discussed in chapter 3.) However, the implicit theories on which decisions are based can pertain to not only the objective consequences that are likely to occur but also the emotional reactions that occur in response to these consequences. People may anticipate feeling happy if their decision has positive consequences but disappointed if it has negative consequences. In addition, they might anticipate experiencing regret if they decide against an option that would have benefited them and might expect to feel relief if they decide to avoid an option that would have turned out badly if they had taken it. The decision they make might depend to a greater extent on their implicit theories about the occurrence of these reactions than by the outcomes that elicit them (cf. Bell, 1982; Loomes & Sugden, 1986; Mellers, Schwartz, & Ritov, 1999). However, both situational and individual differences undoubtedly occur in the conditions in which these theories are invoked. Fong and Wyer (2003) examined these differences in simulated choice situations similar to those that participants were likely to encounter in daily life. Both North American and Hong Kong Chinese students participated. In one study, participants imagined a financial situation in which they could decide either (a) to make an investment that could potentially yield a considerable profit but also ran the risk of taking a substantial loss, or (b) to make a bank deposit that would yield a low rate of interest but would avoid the risk of loss. In a second study, they imagined an academic situation in which they could decide whether (a) to study a particular topic that would ensure a high grade on an exam if the topic actually appeared but a low grade if it did not, or (b) to study other topics, guaranteeing an average grade regardless of whether or not the special topic was on the exam. In each experiment, the magnitude of the gain or loss that might occur as a result of taking the risk, and information about the decisions that others had typically made, were varied over conditions. Participants after reading the scenario reported the likelihood that they would choose the risky option and estimated the risk of doing so. Then,

THE IMPACT OF IMPLICATIONAL MOLECULES

303

they estimated the emotional reactions they would experience if they (a) took the risk and benefited, (b) took the risk and incurred misfortune, (c) did not take the risk and avoided misfortune, or (d) did not take the risk and missed out on an opportunity to benefit. These emotional reactions were analogous to happiness, disappointment, relief and regret, respectively. Participants’ decisions in each study were typically influenced by their perception of the risk involved in doing so, the importance of the consequences, and the decisions that others ostensibly made. However, the effects of these variables were largely mediated by their influence on participants anticipated emotional reactions. Regression analyses indicated that participants’ anticipated reactions accounted for a sizable proportion of variance in decisions over and above the effects of situational variables, whereas the latter variables contributed only a small and often nonsignificant proportion of variance in decisions over and above the effects of anticipated emotional reactions. In other words, participants’ implicit theories about the emotional reactions they would experience in response to alternative decision outcomes were the primary determinant of their choices. However, the relative impact of the four anticipated reactions varied over both scenarios and cultural groups. For example, decisions in the financial situation were based primarily on the consequences of taking the risk (i.e., happiness and disappointment), whereas decisions in the academic situation were influenced by anticipated relief and regret as well. Moreover, Americans’ decisions in the latter situation were influenced to the greatest extent by the happiness they anticipated if they benefited as a result of taking the risk, and to the least extent by the relief they anticipated if they avoided a negative outcome by not taking the risk. In contrast, Chinese participants’ decisions were influenced least by the happiness they anticipated as a result of receiving a positive outcome and most by the relief they anticipated as a result of avoiding a negative one. These results suggest that chronic cultural differences exist in the tendency to invoke theories that focus on positive versus negative decision outcomes. To this extent, they confirm conclusions drawn by Briley et al. (2000; see also Aaker & Lee, 2001). Effects of Stereotypes on Social Behavior One of the most heavily researched areas in social psychological research surrounds the determinants and effects of social stereotypes (for reviews of theory theoretical analyses, see Bodenhausen & Macrae, 1998; Fiske, 1998; Hamilton & Sherman, 1994). In most of this research, a stereotype has been conceptualized as a cluster of traits and behavioral dispositions that are associated with a social or ethnic category (women, African Americans,

304

CHAPTER 10

fraternity members, etc.). Generalized representations of this sort presumably influence persons’ attitudes toward individual members of the category and the attributes they assign to these members (Hamilton & Sherman, 1994). Moreover, a trait-based stereotype that is activated in one situation can later influence the interpretation of information about other individuals to whom the stereotype is objectively inapplicable (Devine, 1989; Lepore & Brown, 1997). Finally, active attempts to suppress the use of the stereotype can have boomerang effects, actually increasing the use of the stereotype as a basis for judgment once these conscious suppression mechanisms are deactivated (Bodenhausen & Macrae, 1998). An understanding of the impact of trait-based stereotypes on social judgments, and the conditions in which they are applied, is of obvious importance. As Reid and Wyer (1998) noted, however, there is surprisingly little evidence that these stereotypes directly influence individuals’ actual behavior toward members of the stereotyped group. The classic study by LaPiere (1934) provides an example of a situation in which individuals’ behavioral decisions in a social context do not correspond to the attitudes and behavioral dispositions they express outside of this context. In fact, it seems somewhat implausible that people who encounter a member of a stereotype group would expend the cognitive energy required to activate a trait-based stereotype and construe its implications in anticipation of deciding how to behave toward members of the group. Instead, they are likely to activate a situation stereotype of the events that are likely to occur in the situation at hand, and to base their decision on the implications of this event representation. A situation stereotype is likely to be activated by a configuration of features that exist in the situation rather than by individual attributes. Thus, its activation can depend on features of persons involved in a situation and features of the situation in combination, rather than on either set of features in isolation. For example, suppose a woman sees an African-American man approaching her late at night on an empty Chicago street. She is unlikely to activate a trait-based stereotype of African Americans as aggressive or hostile before deciding to cross the street. Rather, she is more likely to activate a prototypic event representation of the events that occur in this particular situation (i.e., “People who see a woman walking alone late at night are likely to mug or rape them, and so it is best to avoid contact with them if possible.”) This stereotypic event representation may function as an implicit theory that is activated and applied in the particular situation at hand. If the woman encountered this same individual in a university library she would be far less likely to engage in avoidance behavior. This is because a different situation stereotype is activated that has quite different implications for behavior.

THE IMPACT OF IMPLICATIONAL MOLECULES

305

FINAL COMMENT The research reviewed in this chapter has cut across a very wide variety of concerns, ranging from counterfactual reasoning and reconstructive memory to perceptions of social reality, cultural differences in future outlook, perceptions of social support, marital satisfaction, and stereotyping. The role of implicit theories in these phenomena is admittedly somewhat conjectural. That is, few studies provide direct evidence of the cognitive basis of the judgments and decisions that were made. To the extent these theories exist, however, they are a useful construct in integrating a diversity of empirical findings within a common conceptual framework. Much of our discussion was guided by the assumption that many different theories can often be brought to bear on a judgment or decision, and that the theory that is activated and applied is determined by its relative accessibility. The determinants of knowledge accessibility summarized in chapter 3 obviously come into play in predicting the conditions in which these theories are used. Although the phenomena discussed in this and the preceding chapter do not pretend to be exhaustive, they are representative of those in which narrative representations are likely to be constructed and used as a basis for judgments and decisions in daily life. However, an obvious and important factor that influences judgments and behavior in social situations has been totally ignored, both in these chapters and elsewhere in this volume. We now turn to a consideration of this factor.

C H A P T E R

11 The Role of Affect in Information Processing

Our focus throughout this volume has been on the way we comprehend information about people, objects, and events of the sort we encounter in daily life. In our discussion, we have emphasized the role of narrative representations of knowledge in conceptualizing new information and in making judgments of the people and events to which the representations refer. We have also considered the role of pictures and visual images in the comprehension and use of information. In doing so, however, we have virtually ignored an obvious but important factor that pervades information processing outside the laboratory. Specifically, many situations we encounter in daily life elicit affective reactions. These reactions can be stimulated by direct experiences with a person, object, or event with which they are associated. However, imagining these stimuli, recalling past experiences in which they occurred, or writing stories about them can also elicit these reactions. When these affective reactions are experienced, they can influence judgments and overt behavior toward the stimuli that gave rise to them. Thus, we behave antagonistically to a person who makes us angry, and attempt to increase our contact with someone who makes us feel happy. The cognitive determinants and consequences of affect and emotion have been a major concern of social psychological research and theorizing for more than 25 years. The issues investigated have ranged from the influence of affect on the encoding and organization of new information in memory (Bower, 1981, 1991) to its impact on the retrieval of previously acquired concepts and knowledge (Blaney, 1986; Niedenthal & Setterlund, 1994), its use as a basis for judgments and decisions (Clore, 1992; Schwarz & Clore, 1983, 1988), and its influence on the strategies that individuals use when perform306

THE ROLE OF AFFECT

307

ing cognitive tasks (Isen, 1987; Mackie & Worth, 1989; Nisbett, Schwarz, & Bless, 1991). These issues have been investigated in research on not only judgment and decision making but also creativity (Isen, Daubman, & Nowicki, 1987; Isen, Johnson, Mertz, & Robinson, 1985), communication and persuasion (Albarracin & Wyer, 2001; Bless, Bohner, Schwarz, & Strack, 1990; Mackie & Worth, 1989l; Wegener, Petty, & Smith, 1995); stereotyping (Bodenhausen, 1993; Bodenhausen, Sheppard, & Kramer, 1994), self-evaluations (Levine, Wyer, & Schwarz, 1994); political judgment (Abelson, Kinder, Peters, & Fiske, 1982; Ottati & Isbell, 1996; Ottati & Wyer, 1993), and consumer behavior (Adaval, 2001; Pham, 1998; Pham, Cohen, Pracejus, & Hughes, 2001). Despite the diverse conditions in which affect has been shown to influence information processing, the cognitive mechanisms through which it exerts this influence may be more limited than is often assumed. These mechanisms have been difficult to isolate. However, the research reported to date converges on two general conclusions: 1. Positive and negative affect can be preconditions for cognitive operations for cognitive productions of the sort that compose procedural knowledge (J. Anderson, 1983; E. R. Smith, 1990) and that govern behavior in specific types of situations to which they are relevant. Once activated, these productions can influence responses to new information without conscious awareness. 2. The affective reactions that one experiences at a given moment can be used as information about one’s attitude toward oneself, other people, situations with which one is confronted, the outcomes of behavior, or the appropriateness of certain strategies for attaining behavioral outcomes. It can also influence perceptions of the validity of other affect-eliciting information and, therefore, can influence the weight that is attached to this information in making judgments. The validity of these conclusions is discussed with reference to research and theory in several different areas, with particular emphasis on the impact of affect on behavioral decisions. In this context, a performance feedback model is proposed that accounts for the influence of affect on behavior and judgment in a number of different situations that occur in both the laboratory and daily life. GENERAL CONSIDERATIONS Definition Wyer, Clore, and Isbell (1999) conceptualized affect as a configuration of positively or negatively valenced subjective reactions that a person experiences at a given point in time and perceives as either pleasant or unpleas-

308

CHAPTER 11

ant feelings. These feelings can be elicited by proprioceptive cues (e.g., facial expressions; see Strack, Martin, & Stepper, 1988; Zajonc, Murray, & Inglehart, 1989), physical stimulation, or drugs. More commonly, however, they are internal responses to a set of new or previously formed cognitions. Several general classes of affective reactions can be distinguished (for a detailed discussion of these distinctions, see Clore, Schwarz, & Conway, 1994; Ortony, Clore, & Collins, 1988; Wyer et al., 1999). 1. Moods and Emotions Wyer et al. (1999) assumed that emotions are affective states that occur in response to cognitions about persons, objects, and events, and that are consciously attributed to a specified source. In contrast, moods are typically experienced in the absence of specific cognitions about their source and are not attributed to any particular cause unless one is called upon to do so. Thus, the clusters of subjective reactions that exemplify emotions and moods are similar except that one is consciously attributed to a given source and the other is not. As Wyer et al. (1999) suggested, feelings that are elicited by a stimulus event take time to dissipate. Therefore, feelings that are initially experienced as an emotion can become a mood once their source is no longer considered. The preceding remarks suggest an important distinction between emotions and affect per se. That is, emotions, unlike moods or affective reactions more generally, have a strong cognitive component. Forgas (2002), for example, assumed that emotions are defined in terms of a cognitive appraisal of a situation, and that “physiological and autonomic reactions characteristic of that emotion” are responses to this appraisal (p. 104). Thus, emotions, unlike other affective states, are necessarily accompanied by cognitions about the situational factors that give rise to them. To this extent, there may be concepts in memory whose features include a representation of one’s affective reactions as well as alternative names of both the emotion and its behavioral manifestations. The concept angry, for example, might be metaphorically represented as shown in Fig. 11.1. The concept itself is denoted by a central node to which the features that define and characterize it (a cognitive appraisal of the conditions that elicit the emotion, a mental representation of the affective reactions that are experienced, names, behaviors, etc.) are associatively linked. The concept of a mood might be similar to that shown in the figure except that a characterization of its eliciting conditions would not be included among its features. In each case, thinking about one or more of the features associated with it could presumably activate the mood or emotion concept. Thus, consciousness of the cluster of physiological reactions that define an emotion concept might activate it, in much the same way as the features of

THE ROLE OF AFFECT

309

FIG. 11.1. Metaphorical mental representation of the emotion concept “angry.”

any concept can activate this concept. Note, however, that these physiological reactions per se do not activate the concept. This activation occurs only if the physiological reactions are thought about or otherwise cognitively represented. 2. Affect Versus Evaluation It is important to distinguish between affective reactions to a stimulus and evaluations of the stimulus along a good–bad dimension (e.g., attitudes; see Eagly & Chaiken, 1993, 1998). Although one’s feelings about a person or object can often be the basis of such an evaluation, not all evaluative judgments are based on this criterion (cf. Adaval, 2001; Wyer et al., 1999; Zanna & Rempel, 1988). The distinction between affect and evaluation is not always made in theories of affect-based information processing (cf. Abelson et al., 1982; Bargh, Chaiken, Govender, & Pratto, 1992; Fiske & Pavelchak, 1986). In the present context, however, this distinction is important to keep in mind. For one thing, evaluations are cognitions and, as such, can be features of the mental representations of the entities to which they refer. Affective reactions, however, may not be. 3. Affect Versus Cognition As implied by our discussion of moods and emotions, people have concepts of different types of affect, and cognitions about different configurations of subjective reactions can be features of these concepts (see Fig. 11.1). These latter cognitions, however, may not always have verbal labels. Moreover, it is important to distinguish between cognitions about one’s affective reactions and the affective reactions per se. Although affective reactions may sometimes be learned responses to cognitions, they are not part

310

CHAPTER 11

of the cognitive system. This distinction is made by LeDoux (1994) in conceptualizing the relation of emotion and memory: It is important to distinguish emotional memory [the learned or conditioned emotional significance of an object that automatically triggers an emotion] and memory for emotion. The latter is a declarative, conscious memory of an emotional experience. It is stored as a fact about an emotional episode. Emotional memory (mediated by the amygdala) and memory for emotion (mediated by the hippocampus) can be reactivated in parallel on later occasions, thereby giving new declarative memories an emotional flavor. . . . Because emotional memory and memory for emotion are stored through different brain systems, it is possible that the information stored will not always coincide. (p. 312)

Thus, experiencing affect is not the same thing as conceptualizing it. In fact, we often have subjective affective reactions that we do not consciously identify at all. And even if we are aware of the reactions, we might not attempt to explain their occurrence unless the reactions are intense or there is some particular reason to do so. Put another way, affective reactions may be responses to cognitions, but are not themselves cognitions. Because specific clusters of affective reactions are experienced as feelings, a cognitive representation of these feelings can often be a feature of a more general concept of an emotion or mood state (angry, happy, etc.), as shown in Fig. 11.1. Once such a concept is activated, it could theoretically be used by the Parser to interpret a new experience to which it is applicable, depending on whether features of the experience instantiate it. Thus, an emotion concept could be used to interpret the feelings one is personally experiencing. As suggested by Fig. 11.1, however, the concept could also be activated by other features, including those of the situational context in which these reactions occur (cf. Schachter & Singer, 1962). Nevertheless, these considerations do not imply that affective reactions themselves are part of an associative network of cognitions that compose semantic and episodic knowledge. Affective reactions can be responses to concepts and cognitions that are formed of a new stimulus event, to previously formed cognitions about a past event, or to thoughts about an imagined (e.g., future) event (LeDoux, 1994). Moreover, affective reactions can be the referent of concepts and cognitions about one’s own internal state, and these cognitions, once activated, can enter into subsequent information processing. However, the reactions themselves are no more part of the cognitive system than the referent of the concept “chair.” This means that if a mental representation is not formed of the affective reactions one experiences, the influence of these reactions on information processing may be quite limited. These distinctions might seem self-evident. However, they appear to be ignored in much of the theory and research on the impact of affect on infor-

THE ROLE OF AFFECT

311

mation processing. Several theories of affect and cognition (e.g., Bower, 1981; Forgas, 1995; Isen, Shalker, Clark, & Karp, 1978), as well as introductory textbooks (e.g., Baron & Byrne, 2000), appear to assume that affect functions as a concept in an associative network whose role in information processing is similar to that of other concepts. That is, affect can increase the accessibility of other concepts and knowledge that are directly or indirectly associated with it (i.e., concepts that are similar to it in valence). As a consequence, it can influence the interpretation of new information to which these concepts apply. For similar reasons, affective reactions can cue the retrieval of previously acquired knowledge representations whose features are similar in valence. Thus, the impact of affective reactions on information processing is far reaching. In contrast, if affective reactions are not themselves part of the cognitive system, their influence may be more limited. If these reactions are represented cognitively, they might increase the accessibility of the specific concepts they exemplify. However, they would not necessarily influence the activation and use of concepts and knowledge to which they are not directly related. Moreover, affective reactions should not even have this effect unless persons are conscious of the reactions and are stimulated to interpret them in cognitive terms (see Fig. 11.1). There are two ways, however, in which affective reactions per se are likely to play a role. First, affective reactions can be part of the precondition of an “If [X], then [Y]” production. That is, the precondition [X] could include not only cognitions but also proprioceptive stimuli that, in combination, could activate a behavioral sequence that is performed with minimal conscious awareness. Second, many judgments or decisions are based on feelings. For example, if we experience positive or negative affective reactions when we encounter or think about a particular person, we might attribute these reactions to our feelings about this person and must use these feelings as a basis for estimating how much we like this individual. More generally, subjective reactions can often be used as a source of information about the person, object, or event that ostensibly elicits them. This assumption, which is implicit in early research and theory on attitude assessment (Thurstone, 1959) and interpersonal attraction (Byrne, 1971; Clore & Byrne, 1974), has been stated more formally in a theory of affect as information proposed by Schwarz and Clore (1983, 1988, 1996; for an early version of the theory, see Wyer & Carlston, 1979). There are several contingencies on the use of affect as a basis for judgment that are noted later in this chapter. For example, affective reactions are only likely influence judgments for which they are considered on a priori grounds to be relevant. The informational influence of affective reactions is nonetheless pervasive, having been demonstrated to occur in judging life satisfaction and self-esteem, jokes and car-

312

CHAPTER 11

toons, task performance, consumer products, and political candidates. These effects are discussed in some detail later in this chapter. Methodological Considerations Before discussing the implications of research and theory on the influence of affect in information processing, a methodological problem is worth noting. In much of this research, the affective reactions that people experience have been manipulated experimentally. This is done in several ways. That is, people might be asked to write about a past experience that had elicited positive or negative feelings at the time it occurred and that re-elicits these feelings when thinking about it (Schwarz & Clore, 1983), to watch a movie that elicits positive or negative emotions (Martin, Achee, Ward, & Wyer, 1993), or to read aloud statements that reflect strong emotions (e.g., “This is the greatest day of my life!”, “I wish I could go to bed and never get up,” etc.; see Velten, 1968). Alternatively, participants might be given positive or negative feedback about their performance on a test (Forgas & Bower, 1987). These procedures are typically successful in inducing the affective reactions they are intended to stimulate. As the aforementioned examples indicate, however, the procedures activate semantic concepts and knowledge that could also have an influence on judgments and behavior. Consequently, the interpretation of research that employs these procedures is often equivocal. This contention is elaborated in the following discussion.

AFFECT AND CONCEPT ACCESSIBILITY The influence of affective reactions (or, more accurately, a mental representation of these reactions) on the accessibility of concepts and knowledge in memory is presumably governed by the processes described in chapters 2 and 3. In particular, the representation of a particular configuration of reactions should only activate concepts and thoughts that the configuration exemplifies. There is no reason to suppose that the reactions will influence the accessibility of concepts and knowledge to which they are not directly related. Two quite different bodies of research bear on this matter, each of which is discussed in turn. Effects of Priming Affect on Semantic Concept Accessibility Research by Niedenthal and her colleagues (Niedenthal, Halberstadt, & Setterlund, 1997; Niedenthal & Setterlund, 1994) is consistent with this conclusion. In these studies, participants were induced to feel either happy or

THE ROLE OF AFFECT

313

sad by having them listen to upbeat or gloomy music for 12 to 15 minutes. Pretesting established that the music elicited specific feelings of happiness and sadness rather than positive or negative affect more generally. In the main experiments, however, participants were led to believe that the music was being played to study the relation between auditory and visual perception, and were given no indication of the type of affect that the music was intended to induce. Thus, the procedure avoided the interpretative ambiguities described in the preceding section. If participants who hear the music are aware of the feelings it elicits, they may form a mental representation of these feelings, and this representation could increase the accessibility of the specific mood or emotion concept the feelings exemplify. However, it should not influence the accessibility of concepts pertaining to other types of feelings, or to positively or negatively valenced concepts more generally. Results confirmed these predictions. After listening to the music, participants performed a word recognition task in which a number of positively and negatively valenced words were presented. However, some of these words (e.g., “delight,” “joy,” “weep,” “despair,” etc.) pertained to happiness and sadness whereas others (e.g., wisdom, luck, crime, injury, etc.) did not. Participants quickly identified words that described the particular type of affective reactions they were experiencing. However, the time they took to identify words that described other types of positive or negative affect (Niedenthal et al., 1997), or to identify positively or negatively valenced words that were not directly related to affective reactions (Niedenthal & Setterlund, 1994), was unaffected by their reactions. In short, Niedenthal’s research suggested that if people experience affective reactions that exemplify a particular mood or emotion concept of the sort described in Fig. 11.1, these reactions can increase the accessibility of other features of this concept (e.g., the name of the mood or emotion, or of behaviors that exemplify it). However, these reactions do not increase the accessibility of other concepts and knowledge to which the emotion concept is descriptively unrelated. In fact, even the effects that Niedenthal et al. observed could be of limited generality. In these studies, both participants’ affective reactions and the words they were given to identify in the lexical decision task may have combined to activate the mood or emotion concept they exemplified. In Fig. 11.1, for example, both the representation of the physiological reactions that define the concept “angry” and the name of this concept may resonate with the concept, and the resonance from these two sources in combination may activate it. In the absence of the activation produced by the concept’s name, however, it is conceivable that participants’ representation of their affective reactions might not produce sufficient resonance to activate the concept. As we noted earlier, people may often experience feelings

314

CHAPTER 11

without necessarily labeling or interpreting them. This labeling may only occur spontaneously when the feelings are sufficiently intense that people are motivated to understand and explain them. Automatic Evaluation Effects Niedenthal et al.’s findings are consistent with the conceptualization we propose. At the same time, they conflict with conclusions drawn by Forgas (1995, 2002). Forgas proposed that priming a particular affect-related concept should increase the accessibility of other concepts with which similarly valenced affective reactions are associated. Research on the “automatic evaluation effect” (cf. Bargh, Chaiken, Govender, & Pratto, 1992; Bargh, Chaiken, Raymond, & Hymes, 1996; Fazio, Sanbonmatsu, Powell, & Kardes, 1986) might appear consistent with this possibility. In this research, participants are typically exposed to a positively or negatively valenced word (e. g., lovely, putrid, etc.) at below recognition threshold. Immediately afterwards, participants in some studies are asked to categorize a second word as either favorable or unfavorable (Bargh et al., 1992). Participants in other studies are simply asked to pronounce the second word, without being given any indication that its valence was of concern (Bargh et al., 1996). In each case, participants typically respond more quickly to the second word when its evaluative implications are similar to those of the word that preceded it than when its evaluative implications are dissimilar (Bargh et al., 1992, 1996). These findings could be interpreted as evidence that the first word elicits an affective reaction and that this reaction cues the retrieval of other words that are associated with similar reactions. Thus, participants identify the second word more quickly or more slowly, depending on whether the affective reactions the word elicits are similar to or different from the reactions elicited by the word that preceded it. As noted earlier, however, a distinction should be made between the affective reactions that a stimulus elicits and its evaluative implications. If affective reactions were involved in the processes identified by Bargh and his colleagues, observed effects should increase in magnitude with the extremity of the evaluative implications of the words involved and, therefore, the intensity of the affect they elicit. In fact, this is true under very limited conditions (Bargh et al., 1992; Chaiken & Bargh, 1993). Moreover, responses to valenced semantic concepts are influenced by exposure to normatively positive and negative stimuli that participants have never before seen (Duckworth, Bargh, Garcia, & Chaiken, 2002). This finding indicates that the effects do not reflect a previously learned association between the concepts and the affect they elicit. Wyer, Clore, and Isbell (1999) argued that the effects obtained by Bargh et al. are likely to reflect a categorization process that people spontaneously

THE ROLE OF AFFECT

315

perform at the time they encounter stimuli that is not affect eliciting in and of itself. That is, people may spontaneously categorize stimuli as either good or bad. (This automatic categorization process could reflect an unlearned, evolution-based disposition to identify stimuli as potentially harmful or not harmful in order to survive.) However, affective reactions might not provide the basis for this categorization. In fact, these reactions might only be elicited by a cognitive appraisal of the stimulus and its implications that occurs after the stimulus has been categorized. If this is so, the effects observed by Bargh et al. (1992; see also Fazio et al., 1986) would not conflict with the implications of Niedenthal et al.’s (1997) findings. This conclusion can be summarized in the following postulate: Postulate 11.1. Affective reactions may activate a specific affect-relevant concept that can be used to interpret them. However, affective reactions do not spontaneously influence the accessibility in memory of other similarly valenced semantic concepts and declarative knowledge.

This postulate, however, runs counter to claims that affective reactions do have an influence on the recall and interpretation of information that is evaluatively congruent with it (Bower, 1981; Forgas, 1995). In the following section, we review research that is often cited in support of these claims. Although the review is not exhaustive, it is sufficiently representative to convey the ambiguities that generally exist in the literature bearing on this matter.

THE INFLUENCE OF AFFECT ON THE RECALL AND INTERPRETATION OF INFORMATION The Recall of Previously Acquired Knowledge If affective reactions are features of a mental representation that people have formed and stored in memory, the experience of similar reactions at some later time should cue the retrieval of this representation. This effect should occur for the same reasons that linguistically and perceptually coded stimuli cue the retrieval of previously formed cognitions (see chap. 2). If affective reactions are not represented in memory, however, the experience of these reactions per se would not have this effect. Rather, these reactions should influence the recall of information only if the experience of these reactions is accompanied by perceptually or linguistically coded features similar to those of the knowledge representations to be retrieved. Thus, for example, a person who experiences a certain cluster of affective reactions might be stimulated to interpret this configuration in terms of an affect-related semantic concept such as “sad,” or “down in the dumps.” This

316

CHAPTER 11

concept, once activated, could cue the retrieval of a past experience to which the concept had been applied. In the absence of this mediating cognitive activity, however, the experience is unlikely to be recalled. This conclusion is not widely accepted. In fact, the assumption that affect does indeed cue the retrieval of valenced concepts and knowledge from memory is quite pervasive (cf. Bower, 1981; Forgas, 1992, 1995, 2002; Isen, 1984). However, empirical support for this assumption is very limited (Blaney, 1986). In a study by Bower, Gilligan, and Monteiro (1981), participants in some conditions read a story that described both positively valenced and negatively valenced events that occurred to the protagonists. Then, they were induced to feel either happy or sad by describing a life event that had elicited these emotions. If affective reactions per se cue the retrieval of previously acquired knowledge that is similar in valence, participants should recall aspects of the story that are evaluatively congruent with their feelings at the time of recall. In fact, however, the affect they were experiencing had no influence on the type of information they recalled. Mueller, Grove, and Thompson (1991) induced participants to feel either happy or sad after exposing them to a list of positively and negatively valenced words. This study also found no evidence that the affect that participants were experiencing influenced the valence of the information they recalled. In contrast, participants in a study by Bower (1981) were asked to recall childhood experiences after having been induced hypnotically to feel happy or sad. Participants recalled experiences that were evaluatively congruent with their mood. At the time of mood induction, however, participants were explicitly told what feelings they should try to experience. These instructions presumably activated semantic concepts of the feelings being induced, and these concepts, rather than the affective reactions to which they referred, could have cued the recall of experiences whose features had been encoded in terms of them. Our interpretation of these findings gains plausibility in light of a series of experiments by Parrott and Sabini (1990). In one study, participants listened to either upbeat or depressing music with explicit instructions to use the music to get into a happy or sad mood, respectively. They were then asked to recall an experience they had in high school. In these conditions, they were more likely to recall a pleasant high school experience when they were happy than when they were sad, consistent with Bower’s (1981) earlier findings. In a second study, however, participants were told that the music was being played to investigate aesthetic preferences, and no mention was made of the feelings it would elicit. In this case, however, the effect of participants’ feelings on recall was not only eliminated but reversed; that is, the first event that participants recalled was less pleasant when the music had made them feel happy than when it had not. Moreover, this effect

THE ROLE OF AFFECT

317

was replicated in nonlaboratory settings in which participants’ feelings were induced by naturally occurring events. For example, participants recalled less pleasant high school experiences on sunny days than on rainy ones, and less pleasant experiences after learning they had done well on an examination than after learning they had done poorly. Parrott and Sabini’s (1990) results should be evaluated in the context of Niedenthal et al.’s (1997) finding that listening to music increased the accessibility of mood-related words in a lexical-decision task even when explicit references to the mood or emotion being induced by the music were not made. As noted earlier, words that denote an emotion concept may activate this concept more quickly when the feelings that define the concept are also being experienced. In the absence of semantic labels, however, the feelings may not be sufficiently intense to activate the concept, as Parrot and Sabini’s (1990) findings indicate. This does not explain why the influence of induced affect on recall was actually reversed under conditions in which the affect was not labeled. The reason for this reversal can only be speculated. Unhappy individuals’ recall of positive events could result in part from a desire to distract themselves from their negative mood state by thinking about something that makes them happy (Parrott & Sabini, 1990; Rusting & DeHart, 2000). However, this would not explain the relatively better recall of negative events by persons who feel happy. Wyer and Srull (1989) suggested that when people are asked to recall an experience they have had at an earlier point in time (e.g., high school), they spontaneously activate a global concept of this period in their lives. This concept might spontaneously elicit affective reactions that differ from those they happen to be experiencing for other reasons, and this discrepancy may stimulate thoughts about the implications of this difference. In the conditions considered by Parrott and Sabini (1990), for example, the affect elicited by the “high school” might on average be less positive than the feelings that happy persons were typically experiencing at the time and, therefore, might spontaneously stimulate thoughts such as “I felt less happy in high school than I do now.” This thought, in turn, could cue the retrieval of a negative experience that confirms it. Similarly, “high school” might stimulate unhappy individuals to think, “I felt happier in high school than I do now,” and this cognition may cue the retrieval of a positive experience. Be that as it may, however, Parrott and Sabini’s (1990) results clearly indicated that affective reactions per se do not increase the recall of evaluatively congruent knowledge representations. This influence occurs only when the reactions are accompanied by affect-related semantic concepts that serve as retrieval cues for descriptive knowledge representations that contain them. In the laboratory, these concepts are often activated by the procedures that are used to induce different affective reactions. However,

318

CHAPTER 11

affect-related semantic concepts could also be activated spontaneously if the affect is intense and persons are stimulated to think about why they are experiencing it. In this regard, Bower and Mayer (1985) suggested that the affect induced in the laboratory is often not sufficiently intense to stimulate affect-congruent retrieval. Thus, the effects are often not detected (cf. Mueller et al., 1991). This speculation is consistent with our own contention that affective reactions are often not represented in memory, and that even if a representation is formed of these reactions, it might not be encoded in terms of a mood or emotion concept. This encoding might only be made if the feelings are intense or, alternatively, other features of the emotion concept are present as well. Selective Attention and Encoding If affect increases the accessibility in memory of semantic concepts and knowledge representations that are similar to it in valence (cf. Bower, 1981; Forgas, 1995, 2002), it should also increase the likelihood of comprehending and thinking about information that can be interpreted in terms of these concepts. This increased attention might be reflected in (a) longer time spent reading and thinking about the information, (b) better memory for the information, and (c) greater influence of the information on judgments to which it is relevant (Forgas, 2001). Evidence bearing on each of these factors is discussed in turn. 1. Effects on Processing Time In a study by Forgas and Bower (1987), participants were induced to feel either happy or unhappy by giving them positive or negative feedback about their performance on a measure of social adjustment. Then, they were told to form an impression of someone on the basis of both favorable and unfavorable behaviors pertaining to competence, intelligence, likeableness, and happiness. Participants spent more time reading and thinking about behaviors if the behaviors were evaluatively congruent with their feelings than if they were incongruent. Although data for behaviors pertaining to each attribute separately were not reported, the effects apparently generalized to behaviors that had nothing to do with affect per se but were only evaluatively consistent with participants’ affective state (e.g., behaviors concerning competence, intelligence, etc.). If Postulate 11.1 and the implications of Niedenthal et al.’s (1997) research are valid, Forgas and Bower’s findings must be due to factors other than the affect that participants were experiencing per se. In retrospect, these factors seem fairly obvious. That is, feedback about one’s social adjustment is likely to activate concepts about attributes that reflect this adjustment. As noted earlier, many of the behaviors that participants encoun-

THE ROLE OF AFFECT

319

tered in the later impression formation task were descriptively related to such attributes (likeableness, competence, etc.). It seems likely that these concepts increased participants’ attention to behaviors that exemplified them independently of the affect they were experiencing. To this extent, this increased attention could also facilitate the later recall of the behaviors. Forgas and Bower’s findings support this prediction as well. 2. Effects on Selective Encoding The best-known demonstration of the role of affect in the selective encoding of information was performed by Bower et al. (1981) in a study mentioned earlier. In other conditions of this study, participants were induced under hypnosis to feel happy or sad by recalling a past experience, and were given a posthypnotic suggestion to retain these feelings while they read a story. (Thus, unlike the conditions described earlier, participants were induced to feel happy or sad before reading the story rather than afterwards.) The story concerned one or two characters who experienced a series of both positive and negative life events. Participants had better recall of events that were evaluatively congruent with their feelings than events that were incongruent. The authors did not distinguish between events that were descriptively related to the affect participants were experiencing and events that were only similar in valence. Consequently, the consistency of their results with Postulate 11.1 cannot be directly evaluated. One aspect of their methodology is noteworthy, however. To induce affect, participants were explicitly told to experience the particular feeling state to which they were assigned (i.e., to be happy or sad) and to continue to do so once they were no longer hypnotized. Therefore, semantic concepts of these feeling states were clearly activated by the instructions. Semantic concepts that are activated during hypnosis remain accessible in memory despite instructions to “forget” the experiences that led to their activation (Kihlstrom, 1980). It is therefore likely that the semantic concepts associated with participants’ feelings, rather than the feelings themselves, produced the selective encoding effect that Bower et al. observed. Even when the induction of affect explicitly requires the activation and use of affect-related semantic concepts, it does not always influence the selective encoding of information. In a series of six experiments, Mueller et al.(1991) induced participants to feel happy or sad by having them utter statements that express these feelings (Velton, 1968). Then, they asked the participants to learn positively and negatively valenced words and, finally to recall the words. The statements that were used to induce different feeling states undoubtedly activated semantic concepts pertaining to these states. Nevertheless, the manipulation had weak and inconsistent effects on participants’ later recall of the words they were asked to learn.

320

CHAPTER 11

The Interpretation of Ambiguous Information As Postulate 2.4 implies and research summarized in chapter 3 indicates, concepts that have recently been activated in memory as a result of using them in one context can influence the interpretation of information presented in a later, ostensibly unrelated context. Therefore, if the affect that people experience spontaneously activates semantic concepts with similar evaluative implications, it should influence people’s interpretation of information through its mediating impact on the accessibility of these concepts. According to Postulate 11.1, however, affective reactions only influence the accessibility of the particular mood or emotion that the reactions exemplify. To this extent, the effects of inducing these reactions should be quite restricted if they occur at all. To examine these possibilities, Forgas, Bower, and Krantz (1984) induced participants under hypnosis to feel either happy or sad by having them recall a past experience that elicited these feelings. Then, participants observed a previously videotaped interaction between themselves and another person with instructions to identify behaviors that exemplified either social skill or the lack of it. Happy participants interpreted more behaviors as socially skilled, and fewer behaviors as unskilled, than unhappy participants did. However, the affect-induction task explicitly referred to the feelings that participants were to experience, creating interpretative ambiguities similar to those surrounding the study by Bower et al. (1981) noted earlier. Moreover, many of the experiences that participants recalled were likely to have concerned interpersonal relations, and consequently to activate concepts that had implications for social skills. These concepts could also have influenced participants’ interpretation of the behaviors they observed. A study by Innes-Ker and Niedenthal (2002) avoided this ambiguity. Suppose affective reactions influence the interpretation of information through their mediating effect on the accessibility of concepts to which the information is relevant. Then, similar effects should result from activating these concepts directly without inducing affective reactions. To examine this possibility, some participants were induced to feel either happy or sad as a result of listening to music. Others, however, were required to use emotionrelevant semantic concepts in the course of performing a sentence-construction task similar to that used to prime trait concepts in other studies (e.g., Srull & Wyer, 1979, 1980). Then, as part of an ostensibly unrelated experiment, participants read about a person whose behaviors could be interpreted as conveying either happiness or sadness; specifically: The sound of rain on the street was relentless. She thought that maybe it was a day to stay inside and read a novel, or chat on the phone with friends. . . .

THE ROLE OF AFFECT

321

The cat still had not come back from his night on the prowl, and M wondered where he was. Then the phone rang. . . . It was an old friend . . . who had recently left town. He was calling to say that things were going fine and that he liked his new job. . . . He would not be back in town over Thanksgiving but would probably be back over Christmas break. he said he had a great gift for her in mind.

As expected, participants who personally experienced an emotion judged the protagonist’s feelings as consistent with this emotion. However, when participants were exposed to emotion-related semantic concepts without actually experiencing the emotions themselves, these concepts had no influence on their judgments of the protagonist’s feelings. Taken at face value, these results are quite provocative. That is, they seem to suggest that the feelings people personally experience influence the interpretation of other persons’ emotional experiences, whereas the semantic concepts that are activated by these feelings do not. This conclusion would directly contradict Postulate 11.1. However, the interpretation of Innes-Ker and Niedenthal’s findings is not unequivocal. For one thing, the failure for primed semantic concepts to influence the interpretation of the information must be reconciled with the abundant evidence in other research that concepts activated through similar procedures do influence the interpretation of information (see Higgins, 1996, and chap. 3 of this volume). Note that primed semantic concepts normally influence the interpretation of information only if these concepts are descriptively applicable (Higgins, Rholes, & Jones, 1977; but see Stapel & Koomen, 2000, for an exception). In Innes-Ker and Niedenthal’s study, this condition might not have been met. For example, the emotion words employed in the priming stimuli were “ecstatic,” “pleasure,” “exuberance,” and “peaceful” on one hand, and “lonely,” “dejected,” “melancholy, “sorrow,” and “despondent” on the other. Although the concepts defined by these adjectives are not incompatible with the thoughts and behaviors of the protagonist in the story participants read, they are not so clearly applicable that readers are likely to use them spontaneously, even if these concepts are accessible in memory. This could account for the failure of the semantic task to have an impact. If this interpretation is correct, however, why did the feelings that participants personally experienced have an impact on their interpretations? It is conceivable that participants’ judgments were not mediated by their interpretation of the target information at all. Rather, their judgments may have been based directly on the feelings they were experiencing independently of the information in the story. That is, participants may have projected the feelings they were experiencing onto the target independently of the story content. This “informational” influence of affect is elaborated in the next section of this chapter.

322

CHAPTER 11

THE USE OF AFFECT AS INFORMATION Preliminary Considerations The research summarized in the previous sections calls into question the assumption that affective reactions per se exert an influence on judgments and behavior through its mediating influence on the accessibility of previously acquired concepts and knowledge in memory. These effects are restricted to concepts that are descriptively applicable to the specific emotion or mood that participants are experiencing, and do not extend to positively or negatively valenced concepts and knowledge in general. To this extent, the impact of affective reactions on judgments and decisions may occur for other reasons. Perhaps the most widely accepted influence of affective reactions is in one sense the most obvious. Many evaluations of a person, object, or event are based largely on our feelings about this referent. Thurstone’s (1959) attitude scaling procedures were an attempt to assess individuals’ affective reactions to stimuli independently of their specific beliefs about its attributes. The assumption that attitudes toward an object reflect one’s affective reactions toward it also underlies Fishbein’s (1967) theory of attitude formation. More generally, our liking for something is often an expression of the feelings we experience when encountering it. In this regard, Schwarz and Clore (1988) postulated that people’s evaluations of a stimulus are often based on a “how-do-I-feel-about-it?” heuristic. They suggest that affect is used as a basis for judgment in order to simplify an otherwise difficult task (see also Forgas, 1995). As Wyer et al. (1999) pointed out, however, affect can often be the primary basis for an evaluation rather than only being used when other relevant information is unavailable. People’s affective reactions can be used as information about a wide variety of things and, as such, can have a broad impact on information processing. For example: 1. Affect can be used as a basis for evaluating a particular person, object, or event. 2. Affect can provide the basis for an initial impression of an object that guides the attitude toward and processing of other available information about the object. 3. Affective reactions can often influence one’s perceptions of whether one is successful in attaining a goal one is seeking and, therefore, can influence the likelihood of persisting in goal-directed activity. The remainder of this chapter discusses these possibilities in detail.

323

THE ROLE OF AFFECT

An Initial Test An initial demonstration of the use of affect as information was performed by Schwarz and Clore (1983, Experiment 1). This study not only established a methodology for identifying the impact of affective reactions on judgments, but also identified conditions in which the reactions were likely to be employed. Schwarz and Clore (1983) recognized that the affect people experience at any given time can come from different sources. However, people often cannot distinguish easily between the affect elicited by one source and the affect elicited by another. Consequently, if they base their judgments of a stimulus on their affective reactions to it, they are likely to confuse these reactions with affect they are experiencing for other, irrelevant reasons, and so the latter affect can have an influence on these judgments. On the other hand, suppose people do not base their judgment on affect. Then, extraneous affect should have no impact in any condition. To examine these possibilities, Schwarz and Clore asked participants to write about a happy or sad past experience while they were seated in a strange-looking soundproof booth. Some participants were told that people often found that the booth made them feel mildly elated after they had been in it for a while. Others were told that the booth made people feel mildly depressed, and others were given no information about the booth’s effects. Participants after writing their essays were then asked to judge their satisfaction with their life as a whole. These judgments are summarized in Table 11.1 as a function of the type of experience participants wrote about and the ostensible effects of the booth. Participants who were given no information about the booth’s effects reported greater life satisfaction when they had written about a happy experience than when they had written about an unhappy experience. Considered in isolation, this difference might be attributed to the content of the story that participants wrote rather than the affect they were experiencing. If this were the case, however, the alleged effects of the booth on their feelings should have had no impact on their judgments. In fact, with participants who TABLE 11.1 Judgments of Life Satisfaction as a Function of the Affective Quality of a Recalled Personal Experience and Ostensible Characteristics of the Booth Ostensible Effect of Booth

Happy recalled experience Sad recalled experience

Depressing

No Effect

Elating

9.6 8.6

8.6 5.7

9.7 4.4

Note. Adapted from Schwarz and Clore (1983).

324

CHAPTER 11

had written about a sad experience, telling them that the booth would make them feel sad eliminated the negative impact of the story on their judgments of life satisfaction, whereas telling them that the booth would make them happy increased these judgments relative to no-information conditions. In other words, participants who were told that the booth made them unhappy attributed their feelings to the booth and did not use it as a basis for judgment. In contrast, participants who experienced unhappiness despite being told that the booth would make them happy believed that they must be particularly unhappy about their life as a whole and, therefore, judged themselves to be even less satisfied than they would otherwise. It is interesting to note that the effects of the booth were not apparent when participants had written about a positive life experience and, therefore, were feeling happy. These participants appeared to base their judgments of life satisfaction on their feelings regardless of the information they were given about the booth. The asymmetric effects observed in the two affectinduction condition could be due to factors similar to those underlying Arkin et al.’s (1976) research described in chapter 9. That is, participants whose affect had favorable implications for themselves were inclined to accept these implications and use them as a basis for judgment without seeking alternative explanations for their feelings. In contrast, people whose affect had undesirable implications resisted using it as a basis for judging themselves unless no alternative explanation for their feelings was available. The Role of Narratives in Affect Elicitation Affective reactions can be elicited in many ways, as noted earlier. Affect that is induced by proprioceptive feedback (cf. Strack, Martin, & Stepper, 1988; Zajonc, Murray, & Inglehart, 1989) can also influence judgments to which it is relevant (e.g., the amusement elicited by cartoons). The role of narratives in affect elicitation was confirmed in a series of studies by Strack, Schwarz, and Gschneidinger (1985). They found that when people listed past experiences in abstract terms that were unlikely to elicit affect, they used these experiences as standards of comparison in judging their current life satisfaction, reporting themselves to be less satisfied when the past experiences were favorable than when they were unfavorable. When participants wrote detailed narratives of their past experiences, however, this activity elicited affect, and this affect had a positive influence on satisfaction judgments. In a particularly provocative study (Strack et al., 1985, Experiment 3), participants were either asked to explain why a particular experience occurred or to describe how it occurred. Writing explanations for happy and sad events did not elicit affect, and the favorableness of the experiences that participants explained had contrast effects on the life satisfaction they

THE ROLE OF AFFECT

325

reported later. However, writing descriptions of how the experience occurred did elicit affect and consequently had effects on judgments similar to those reported by Schwarz and Clore (1983).

AFFECTIVE VERSUS DESCRIPTIVE BASES FOR JUDGMENT As noted, not all evaluations that people make are based on affective reactions (Zanna & Rempel, 1988). Many judgments are based on purely descriptive considerations. Thus, although we might like a pair of jeans because they feel good when wearing them, we might like a computer because it is fast and has a lot of storage. Similarly, we might evaluate a dinner partner favorably because her physical attractiveness and effervescent personality elicit positive affect, we might evaluate a research assistant favorably because she gets work done on time and records data accurately. Other judgments can be based on either affective or descriptive criteria, depending on which happen to be more accessible (Postulate 3.1). This latter contingency has been identified in research in quite different areas, concerning judgments of personal competence in one case and evaluations of commercial products in the other. Judgments of Personal Competence and Global Self-Esteem General characteristics such as life satisfaction or global self-esteem are typically not based on a single event, but rather, on more global feelings that reflected the accumulated influence of many experiences. Consequently, affective reactions may be a primary basis for judging these characteristics. In many cases, however, descriptive criteria are potentially applicable as well. In such instances, whether descriptive or affective criteria are used can depend on the relative accessibility of these criteria in memory. If the content of a narrative has direct implications for a judgment and if these implications are accessible in memory, the effects of these implications can override the influence of affect. Levine, Wyer, and Schwarz (1994) provided an example. Participants wrote about a happy or an unhappy past experience they had had in either an achievement situation or a social situation. After doing so, they estimated their competence in both types of situations and their social competence as well as their global self-esteem. We reasoned that judgments of competence in the domain to which their story directly pertained would be based on the descriptive implications of the story, as these implications were highly accessible in memory. However, their competence in the do-

326

CHAPTER 11

main to which their narrative was descriptively irrelevant should be based on the affect they were experiencing. This hypothesis is difficult to evaluate when descriptive and affective criteria for judgment have similar implications. However, a content analysis of participants’ narratives showed that although participants typically took responsibility for both positive and negative achievement experiences they wrote about, they attributed responsibility for both good and bad social experiences to situational factors (e.g., the other persons involved in the situation). Based on these differences, we predicted that participants’ judgments of their achievement competence would be influenced by the favorableness of the achievement experience they wrote about, but that their judgments of social competence would not be influenced by the type of social experience they described. On the other hand, participants’ judgments of both achievement competence and social competence should be influenced by the affect that was generated by writing about an experience in a domain to which the competence being judged was unrelated. Results shown in Table 11.2 confirm these predictions. That is, participants’ judgments of their achievement competence were influenced by the favorableness of the achievement experience they had written about, whereas their judgments of their social competence was not influenced by the favorableness of the interpersonal situation they had described. On the other hand, participants’ judgments of both achievement competence and social competence were influenced by writing about an affect-eliciting experience in the domain to which the competence they were judging was irrelevant. (Thus, somewhat ironically, judgments of social competence were more influenced by writing about an objectively irrelevant affect-eliciting experience than about a relevant one.) Finally, participants’ overall judgments of their self-esteem were influenced by the favorableness of the description they had written about regardless of the domain, suggesting that these global judgments were based on affect in all cases. TABLE 11.2 Judgments of Competence as a Function of the Domain and Affective Quality of Participants’ Recalled Experiences

Recalled experience in same domain Happy experience Unhappy experience Recalled experience in different domain Happy experience Unhappy experience

Judgment of Achievement Competence

Judgment of Social Competence

4.77 4.37

4.38 4.38

4.87 4.67

4.46 3.46

Note. Adapted from Levine, Wyer, and Schwarz (1994).

THE ROLE OF AFFECT

327

Consumer Judgment Judgments of competence can often be based on either descriptive or affective criteria, and so participants are likely to use whichever criterion comes to mind most easily at the time judgments are made. This may also be true in the consumer domain. Evaluations of commercial products, for example, are often based on utilitarian considerations as well as hedonic or affective criteria. A refrigerator, for example, is likely to be evaluated on the basis of its ability to perform the functions for which it is designed. On the other hand, a cashmere sweater, or strawberry shortcake, could be judged on the basis of either affective criteria (how it feels or tastes) or utilitarian considerations (durability, healthfulness, etc.). In the latter case, people are most likely to base their evaluations on affect if (a) the stimulus is one that is typically judged on the basis of affective (hedonic) criteria, (b) the information about the stimulus is conveyed in a way that spontaneously elicits positive or negative feelings, or (c) people’s objectives at the time of judgment require the use of affective criteria. The latter contingency was demonstrated by Pham (1998). Participants who had been induced to feel happy or unhappy by recalling a past experience were asked to read a description of a new movie and indicate if they would like to attend a preview of it. Before reading the description, some participants were given instructions that emphasized hedonic criteria for a decision (e.g., having a good time, enjoying life, etc.), whereas others were given a utilitarian reason for seeing it (its relevance to participating in a subsequent study for pay). Participants in the first condition expressed stronger intentions to see the movie if they were feeling happy than if they were not, suggesting that they used the affect they were experiencing as a basis for their judgment. In contrast, participants’ intentions in the second condition were unaffected by the affect they were experiencing. The use of hedonic or utilitarian criteria can also depend on their relative accessibility at the time of judgment. However, unless situational factors make descriptive criteria particularly relevant (as in Levine et al.’s 1994 study), these criteria may be relatively inaccessible. In contrast, the affective reactions can be activated very quickly, before a detailed cognitive analysis of the utilitarian features of the description is performed. (For a more general argument that affective responses to a stimulus as a whole occur more quickly than responses to its individual features, see Zajonc, 1980.) One implication of this is that if a stimulus can potentially be evaluated on the basis of both hedonic and utilitarian criteria, hedonic criteria may often take priority. That is, utilitarian considerations may come into play only when people have both the motivation to evaluate these criteria and the ability to do so. A study by Shiv and Fedorikhin (1999) supported this contention. Participants were given a choice between a chocolate cake and a fruit salad as a

328

CHAPTER 11

reward for participating in the study. The authors assumed that the sight of the chocolate cake would elicit positive affect, and that the choice of this option would reflect the use of these reactions as a basis for their decision. In contrast, the choice of a fruit salad was more likely to reflect the use of utilitarian (health-related) criteria. The authors further speculated that the latter criteria would come into play only if participants had the motivation and ability to think carefully about their decision. To evaluate this possibility, some participants were placed under high cognitive load at the time they made their choice by being asked to remember a multiple-digit number for use in a later part of the study. Others were put under low load. Participants in the latter condition tended to choose the fruit salad, suggesting that they based their decision on utilitarian (i.e., health-related) criteria. Participants who were distracted from thinking about their decision, however, were more inclined to choose the chocolate cake. The implications of these findings may extend beyond the domain of consumer judgment. For example, a faculty member who wishes to hire a research assistant may consider utilitarian criteria (background, experience, etc.) to be particularly relevant. In a personal interview, however, a candidate’s personality and physical attractiveness might elicit affective reactions that are used as a basis for judgment in addition to other considerations. Moreover, if the interviewer meets the candidate before receiving descriptive information about his or her qualifications, these affective criteria may predominate, and descriptive criteria may be taken into account only if the interviewer has the motivation and ability to think more carefully about the candidate’s credentials.

Effects of Initial Appraisals on the Use of Affective Versus Descriptive Criteria In most research on the use of affect as a basis for judgment (for reviews, see Clore, Schwarz, & Conway, 1994; Schwarz & Clore, 1996; Wyer, Clore, & Isbell, 1999), the influence of affect is assumed to occur at the time of judgment. If people make a spontaneous appraisal of a stimulus at the time they first encounter it (Lazarus, 1982), and if this appraisal elicits affective reactions, these reactions can give rise to an affect-based initial impression of the stimulus that influences participants’ later judgments over and above the effects of more specific information they receive later. This possibility was confirmed by Catherine Yeung (2003; Yeung & Wyer, in press). She noted that people’s first encounter with a product often occurs when they notice it in a store window or see a picture of it in a magazine, and that they spontaneously appraise the product on the basis of this encounter. If this appraisal elicits affect, it not only can have an impact on

329

THE ROLE OF AFFECT

evaluations of the product but also can bias the search for additional information about it. 1. Experiment 1 To examine these possibilities, participants in one experiment took part in three conditions. In no-picture conditions, participants who had been induced to feel either happy or unhappy as a result of writing about a life experience were given verbal descriptions of a pair of sports shoes consisting of three favorable and three unfavorable features. Some participants were told to base their judgment on hedonic criteria (how it would feel to wear the shoes), and others were told to use utilitarian criteria (durability, workmanship, etc.). Participants’ evaluations of the products in these conditions, shown in the top left panel of Table 11.3, indicate that the extraneous affect that participants were experiencing influenced their evaluations when they were told to use hedonic criteria but had little effect when they were told to use utilitarian criteria. These data, which are consistent with those reported by Pham (1998), confirmed the assumption that when participants are told to use hedonic criteria as a basis for evaluations, they imagine the feelings they would experience when using the product, and the affect they were experiencing for other reasons influence their assessment of these feelings. When participants were told to use utilitarian criteria, however, they did not consult their feelings and so the extraneous affect they were experiencing had little effect. In two other conditions of this study, however, participants were shown an attractive picture of the product before they received descripTABLE 11.3 Product Evaluations and Recall of Attribution Information as a Function of Induced Affect, Judgment Criterion, and Information Presentation Conditions Affect Induced Before Picture

No Picture Positive Affect

Negative Affect

Product evaluation (Experiment 1) Hedonic criterion 6.50 Utilitarian criterion 4.00 Product evaluation (Experiment 2) Hedonic product 7.11 Utilitarian product 3.41 Number of positive attributes recalled Hedonic product 2.55 Utilitarian product 1.14

Affect Induced After Picture

Positive Affect

Negative Affect

Positive Affect

Negative Affect

5.25 5.25

3.75 3.78

3.46 3.89

3.50 4.33

4.67 6.38 3.78 4.88 (Experiment 2) 1.32 2.13 1.80 1.62

4.38 3.63

5.78 4.11

5.30 4.48

1.32 1.29

1.68 1.32

1.59 1.26

4.10 4.12

Note. Adapted from Yeung and Wyer (in press).

330

CHAPTER 11

tions of its attributes. Moreover, the pictures in affect-picture conditions were shown after participants had been induced to feel happy or unhappily, whereas in picture-affect conditions, the pictures was conveyed at the outset. We assumed that the picture would stimulate participants to make a spontaneous appraisal of the product that elicits affective reactions, and that these reactions would then provide the basis for an initial impression of the product that they would use as a basis for their evaluation of the product they reported later. Moreover, this should be true regardless of the specific criteria they would apply in the absence of this affect-eliciting appraisal. Therefore, if participants are experiencing affect for other reasons at the time they made this appraisal, this affect should also influence the impression they form and the judgments they ultimately make on the basis of it. On the other hand, suppose extraneous affect is not induced until after participants have made an appraisal of the product and formed an impression on the basis of it. Then, this affect should have no impact on their initial impression and, therefore, should not influence the evaluation they report later. Data in the last two sections of Table 11.3 (top) confirm these predictions. That is, the extraneous affect that participants were experiencing had a positive impact on judgments when it was induced before pictures were presented, and this was true regardless of the criteria they were told to use. When extraneous affect was not induced until after participants saw the product, however, it had no influence on judgments whatsoever. 2. Experiment 2 A second experiment confirmed these findings and, in doing so, provided further evidence bearing on the assumptions underlying their interpretation. The design of this study was similar to the first with two exceptions. First, rather than explicitly instructing participants as to which criterion they should use, we asked them to evaluate products that they were likely to evaluate spontaneously on the basis of either hedonic criteria (salad dressing) or utilitarian criteria (backpacks). Second, participants after receiving the product information were told to recall all of the product’s features they could remember. (Because most participants could recall all six pieces, we considered only the first three features they recalled, and used the number of these features that were favorable as an index of the relative attention paid to these features in relation to unfavorable ones.) Product evaluations, shown in the middle third of Table 11.3, confirm conclusions drawn from the first study. That is, when a picture of the product was not presented, participants judged the hedonic product more favorably when they were feeling happy than when they were feeling unhappy, whereas they evaluated utilitarian products similarly regardless of the ex-

THE ROLE OF AFFECT

331

traneous affect they were experiencing. When participants had received a picture of the product, however, the extraneous affect they were experiencing at the time they saw the picture had an impact on their judgments of both types of products, whereas affect that was not induced until after they saw the picture had no impact. Note that this same pattern of effects was reflected in the favorableness of the attribute information that participants recalled, as shown in the bottom third of Table 11.3. The parallel effects of experimental variables on recall and judgments has two possible interpretations. First, participants who formed affectbased impressions selectively attended to attributes of the product that confirmed the evaluative implications of these impressions, and then used this attribute information as a basis for judgments. Second, participants who formed an initial impression of the product based their judgments directly on this impression independently of the specific attribute information presented. However, they thought more about information they received that was consistent with this impression in order to confirm its validity (for evidence of such confirmatory information seeking, see Bodenhausen & Wyer, 1985; Chernev, 2001; Snyder & Cantor, 1979). Although these alternative possibilities cannot be clearly distinguished on the basis of the data available, the correlation between product evaluations and the favorableness of the attribute information recalled was very low (r = .14), suggesting that in fact, the two measures are relatively independent. In any event, the two studies in combination clearly suggest that when participants read verbal descriptions of a product’s attributes without having seen a picture of it, they only based their evaluation of the product on their affective reactions if they believed that hedonic criteria were relevant to this evaluation. When they saw a picture of the product, however, the picture stimulated a spontaneous appraisal of the product that elicited affective reactions, and participants formed an initial impression of the product on the basis of these reactions. Then, they used this affect-based impression as a basis for their later evaluations of the product regardless of the criteria they would otherwise have applied. 3. A Contingency Yeung and Wyer’s (in press) results suggested that affect can exert its impact on judgments of a stimulus through its influence on the spontaneous appraisal that people make of the stimulus at the time they first encounter it. Not all initial appraisals elicit affect, however. Whether this is so could depend in part on previously learned associations of affect with features of the stimulus on which appraisals are based. When an appraisal does not elicit affect, subsequent judgments could still be determined by the initial impression that is formed on the basis of it. However, affective reactions may not have much influence.

332

CHAPTER 11

Yeung (2003) obtained evidence of this contingency in a study that paralleled the experiments described in the previous paragraphs. Rather than giving participants a picture of the product, however, she told them that the product they would judge was being produced by a company that was well known for either hedonic products (Pepsi) or utilitarian ones (Bonaqua, a brand of distilled water). Either before or after conveying this information, participants were induced to feel happy or unhappy by writing about a past experience. Finally, they were given information about a pair of sports shoes that were ostensibly about to be produced by the company whose name had been provided earlier. When the brand was associated with a product that was judged on the basis of affective criteria, results paralleled those obtained under comparable conditions of the earlier studies. That is, extraneous affect had a positive influence on participants’ evaluations of the brand extension when it was induced before participants learned of the brand, but not when it was induced afterward. When the brand was associated with a product that was judged on the basis of utilitarian criteria, however, the extraneous affect that participants were experiencing had no influence regardless of when it was induced. These results further strengthen the conclusion that judgments of a stimulus are often induced by an initial appraisal that is made at the time the stimulus is first encountered independently of more specific information about the stimulus that is conveyed afterwards. In these conditions, however, affective reactions only enter into this process when the appraisal elicits affect. Moreover, extraneous affect that participants happen to be experiencing only has an impact when it is experienced at the time the appraisal is made and, therefore, becomes confused with the affect that is automatically elicited by the appraisal-inducing stimulus.

AFFECT-CONFIRMATION PROCESSES In the research by Yeung (2003; Yeung & Wyer, in press) affect was presumably elicited by a spontaneous global appraisal of the product as a whole that was independent of any specific product features. However, affect can obviously be elicited by specific features of a stimulus as well. A politician’s position on a particular issue, for example, could elicit positive or negative affective reactions in someone who favors or opposes this position. In the product domain, a sweater’s material can convey how nice it would feel to wear it, and could elicit affect for this reason. Individuals who wish to evaluate such stimuli on the basis of affective considerations are likely to pay attention to features that elicit affect and use these features rather than others as a basis for judgment. If the individuals are also experiencing affect

THE ROLE OF AFFECT

333

for other reasons, however, this affect could have two effects. First, they might confuse their affective reactions to these features with the affect they are experiencing for other reasons, and consequently could interpret these features as more or less favorable. On the other hand, their extraneous feelings could influence the importance that they attach to the affect-eliciting features and, therefore, the weight they attach to these features when computing a judgment. This latter possibility is formalized in the affect-confirmation theory proposed by Adaval (2001). According to this theory, participants who base their judgments of a stimulus on hedonic considerations assess not only their affective reactions to the stimulus’s individual features but also the extent to which their perceptions of these reactions are valid. If extraneous affect that they happen to be experiencing is similar to the affect that is elicited by a stimulus feature, it may appear to confirm their reactions to this feature and, therefore, may give them more confidence in the implications of these reactions. On the other hand, if extraneous affect they are experiencing differs from that elicited by a stimulus feature, they might interpret this difference as ambivalence and thus might decrease their confidence in the validity of their assessment of the feature’s affective implications. Consequently, if individuals base their judgments of the stimulus on their affective reactions to it, they should attach greater weight to features that elicit affect similar to the feelings they are experiencing for other, unrelated reasons than to features that elicit different reactions. Empirical Evidence Two studies by Adaval (2001) evaluated these hypotheses. In one, participants who had been induced to feel either happy or unhappy received information about a pair of jeans consisting of either a favorable or an unfavorable brand name (Levis vs. Rustler) and either a favorable or an unfavorable attribute (preshrunk vs. not preshrunk). In some cases, participants were told to base their evaluations of the jeans on hedonic criteria (how it would feel to wear them) and the attribute was described in a way that emphasized its implications for comfort. In other cases, they were told to base their judgments on utilitarian criteria, and the attribute description emphasized its implications for the quality of construction. Participants after reading the product description evaluated the product and then reported their confidence that this evaluation was correct. Note that differences in the impact of the attribute information on product evaluations under different affect-induction conditions could indicate either that the attribute information was weighted more heavily in making their judgments or, alternatively, that the attribute information was interpreted differently (i.e., as more or less extreme) under these conditions. To

334

CHAPTER 11

avoid this ambiguity, Adaval inferred the relative weight attached to the attribute information from the impact of the brand information that accompanied it. If affect information receives relatively greater weight when its evaluative implications are consistent with the extraneous affect that participants were experiencing, the relative influence of the brand information that accompanies it should be relatively less.22 Data summarized in the top half of Table 11.4 confirm this hypothesis. That is, when participants were told to evaluate the product on the basis of hedonic criteria, brand information had less impact on their judgments when the attribute description was evaluatively consistent with the extraneous affect that participants were experiencing than when it was inconsistent, indicating that the attribute information received greater weight in the former condition. When participants were told to use utilitarian criteria, however, brand had greater influence in the former condition than the latter. This reversal suggests that participants in this condition may have intentionally avoided basing their judgments on affective criteria and, in fact, overcompensated for its influence, attaching less weight to the attribute information than they otherwise would. Interestingly, participants reported greater confidence in their judgments when the evaluative implications of the attribute information and those of the extraneous affect they were experiencing were consistent, and this was true regardless of the objective they were pursuing. Thus, participants were not only more confident about using their affective reactions to the attribute information as a basis for judgments when their evaluations were based on hedonic criteria but were also more confident about avoiding its use when the basis for their judgment was utilitarian. A second experiment confirmed these predictions under conditions in which the criterion that participants used was activated spontaneously by the type of product that participants were asked to consider. Normative data suggested that in the absence of explicit instructions, participants were likely to evaluate jeans on the basis of hedonic criteria and backpacks on the basis of utilitarian criteria. A second experiment was conducted using these two product categories but without instructing participants about the criteria they should employ. The bottom half of Table 11.4 shows that as expected, brand had less impact on evaluations of jeans when the evaluative implications of the attribute information were consistent with the extraneous affect that participants were experiencing, whereas it had greater impact on evaluations of backpacks in this condition. 22 The use of this criterion assumes that judgments are a weighted average of the implications of the two pieces of information considered separately (cf. N. Anderson, 1971, 1981). Although this assumption is often questionable (cf. Wyer & Carlston, 1979), the validity of the assumption in product evaluation situations is viable (cf. Bettman, Capon, & Lutz, 1975; Troutman & Shanteau, 1976).

22

335

THE ROLE OF AFFECT TABLE 11.4 Effects of Brand Favorableness on Product Evaluations as a Function of Processing Criterion and Attribute-Affect Consistency Attribute Consistent With Induced Affect Experimentally-induced judgment criterion (Experiment 1) Hedonic 2.21a Utilitarian 2.83 Spontaneously activated judgment criterion (Experiment 2) Jeans (hedonic) 3.0 Backpacks (utilitarian) 2.0

Attribute Inconsistent With Induced Affect

2.56 1.85 4.2 1.3

Note. Adapted from Adaval (2001). aEffect of brand favorableness is inferred from the difference in product evaluations when the brand was favorable and evaluations when the brand was unfavorable.

Implications for Political Judgment Adaval’s (2001) findings, and her affect-confirmation theory more generally, are noteworthy in the context of considerations raised earlier in this chapter. That is, the greater weight of affect-consistent information in judgments has often been attributed to the greater accessibility of affect-consistent semantic concepts and knowledge in memory and, therefore, the greater likelihood of encoding this information into memory (cf. Forgas & Bower, 1987). However, only two pieces of product information were presented in Adaval’s studies, and this information was available at the time of judgment. In other words, the differential accessibility of the information in memory cannot plausibly account for the effects. Moreover, the effects are unlikely to result from the encoding of the information in terms of evaluatively consistent semantic concepts at the time it was received (Niedenthal & Setterlund, 1994). To the extent affect-confirmation processes account for the results of Adaval’s studies, it seems likely that the influence of affect-consistent information observed in other research studies are influenced by similar processes. A study by Isbell and Wyer (1999) is worth considering in this regard. Happy and sad participants were asked to evaluate a political candidate on the basis of a news article that described his stands on a number of social issues. In addition, their intrinsic and extrinsic motivation to think carefully about the candidate’s issue positions was varied.23 Partici23

Intrinsic motivation was inferred from responses to a self-report index of the strength of participants’ partisanship (i.e., their party identification). To induce extrinsic motivation, participants were told that the individual they would evaluate was a current candidate for office in their area, and that we were consequently interested in whether they would vote for him. In contrast, low motivation was induced by telling participants we were primarily interested in whether the article was clearly written, independently of its specific content.

23

336

CHAPTER 11

pants after reading the news article reported their impressions of the candidate and then indicated their personal agreement with each of the issue stands he had espoused, along with an equal number of positions on which the candidate’s stands were unknown. When participants were neither intrinsically nor extrinsically motivated to evaluate the candidate, they judged him more favorably when they were feeling happy than when they were not. Moreover, they reported stronger agreement with the candidate’s issue stands in the former case than the latter. (In contrast, their agreement with issue positions that were not associated with the candidate were not influenced by the affect they were experiencing.) Thus, participants’ mood influenced their perceptions of the favorableness of the candidate’s issue stands and these perceptions, in turn, determined their liking for the candidate. This conclusion was confirmed by covariance analyses. Specifically, the effects of experimental variables on candidate evaluations were eliminated when their effects on issue agreement were covaried, whereas the effects of these variables on issue agreement remained significant when evaluations of the candidate were covaried. Isbell and Wyer (1999) concluded that participants interpreted their reactions to the candidate’s issue stands as generally more favorable when they were feeling happy than when they were not. However, affectconfirmation processes could also underlie these findings. For example, happy participants may have had more confidence in the validity of their reactions to issue stands that elicited positive affect (i.e., stands with which they agreed) than in the validity of their reactions to issue stands that elicited negative affect (stands with which they disagreed). Unhappy participants, on the other hand, may have had less confidence in their reactions to the former statements than their reactions to the latter. The differential weighting of the issue stands that resulted from these differences in confidence could have influenced candidate evaluations for much the same reasons suggested by Adaval (2001). Isbell and Wyer’s study does not distinguish between these interpretations.

ADJUSTMENTS FOR BIAS The aforementioned results by Isbell and Wyer (1999) were evident only among participants who were neither intrinsically nor extrinsically motivated to make accurate judgments. When people are more motivated to be correct, they may attempt to distinguish between the various sources of affect they are experiencing. Furthermore, if they believe their perception of their feelings toward the stimulus they are judging to be contaminated by

THE ROLE OF AFFECT

337

extraneous sources of affect, they might try to adjust for this biasing influence. As noted in chapter 10, however, people are often unclear how much adjustment is necessary. If people do not adjust enough, the extraneous affect they are experiencing would continue to have a positive influence on their judgments. If they adjust too much, on the other hand, it would appear to have a contrast effect. The latter appeared to be the case in Isbell and Wyer’s study. That is, although the extraneous affect that participants were experiencing had a positive impact on their judgments when they were unmotivated to make accurate evaluations, it had precisely the opposite effect on judgments by individuals who were either intrinsically or extrinsically motivated to be correct (see footnote 24). Effects of Ability Even if people are motivated to adjust for the biasing influence of extraneous affect, they may be unable to do so. Ottati and Isbell (1996), for example, found that extraneous affect had a positive influence on evaluations of a candidate by participants with relatively little knowledge about politics, but had a contrast effect on evaluations by participants with high political expertise. Perhaps both groups of participants adjusted for the biasing influence of the extraneous affect they were experiencing. However, participants with little expertise had difficulty evaluating the implications of the candidate’s issue positions and the cognitive resources they had to devote to this activity prevented them from adjusting sufficiently. On the other hand, experts found it easy to evaluate the implications of the candidate’s issue positions and so they could devote more effort to distinguishing their reactions to these stands from other sources of affect. However, these participants adjusted too much, resulting in a contrast effect of extraneous affect on their evaluations. Adjustment Versus Discounting Adjustments for bias are most likely to occur when people consider their affective reactions to the primary basis for their judgments of an object but cannot distinguish clearly between these reactions and affect from other sources. In many instances, however, nonaffective criteria for judgment may be as or more appropriate bases for judgment. In these instances, people who believe that the affective reactions are biased by extraneous factors might ignore their feelings entirely and default to other, nonaffective criteria. In this case, the affect they are experiencing would not have a contrast effect on judgments. Rather, it would have no influence at all.

338

CHAPTER 11

One example of discounting was obtained by Schwarz and Clore (1983, Experiment 2). In this study, residents of downstate Illinois were contacted by telephone on either a sunny or a rainy day and asked to participate in a short telephone interview in which they reported their general life satisfaction. Participants judged themselves to be more satisfied with their life on sunny days than on rainy ones, suggesting they based their judgments in part on the affective reactions activated by the weather. In other conditions, however, the interviewer, who was ostensibly calling from Chicago, incidentally asked participants, “By the way, how’s the weather down there?”, thereby calling participants’ attention to an extraneous source of their affect. In this condition, participants apparently ignored the affect they were experiencing and based their judgments on other criteria (e.g., their knowledge about life events that they retrieved from memory). Consequently, weather-induced affect had no impact on their judgments at all. Calling participants’ attention to an extraneous source of affect does not always lead participants to discount their feelings, however. When participants in Schwarz and Clore’s (1983, Experiment 1) study had written about an unhappy life event in a booth they believed would make them happy, they apparently inferred that they must be feeling even more negatively about themselves than did participants who were not given information about the booth. Consequently, they reported lower life satisfaction than other participants did. A study by Gasper and Clore (1998) found that chronically anxious individuals do not discount the influence of situational factors on the negative affect they are experiencing. Calling these individuals’ attention to extraneous sources of negative affect apparently made salient to them that their feelings were not due to transitory factors alone. Consequently, it increased the influence of the negative affect they were experiencing rather than decreasing it. In this regard, it is undoubtedly an oversimplification to assume that when both affective and nonaffective criteria for judgment are available, people may use one set of criteria or the other but not both. More generally, people use both types of criteria, with the relative weight attached to each depending on perceptions of its relative validity and also the time and effort that are required to construe its implications. As noted earlier, affective reactions to a stimulus can often occur spontaneously upon exposure to it, and so this criterion is likely to be weighted heavily when participants’ motivation or ability to think about other available criteria are low. However, if participants have the time and motivation to apply other relevant criteria, they may use these criteria as well. The impact of situational factors on the relative use of alternative criteria for judgment was identified by Shiv and Fedorikhin (1999) and also by Albarracin and Wyer (2001) in a study of communication and persuasion. The latter study is described in more detail presently.

THE ROLE OF AFFECT

339

AUTOMATIC INFLUENCES OF AFFECT ON INFORMATION PROCESSING People’s affective reactions can influence not only their judgments but also their overt behavior and decisions. These effects could be either deliberative or automatic. In the first regard, people are sometimes motivated to behave in a way that will maintain a positive mood or, alternatively, will rid themselves of negative feelings. As noted earlier, however, the affective reactions one experiences, like other internally generated stimuli, can activate a production of the sort described in chapter 1 (J. Anderson, 1983; E. R. Smith, 1984, 1990). That is, they can form part of the configuration of stimuli [X] in a production of the form “if [X], then [Y],” where [Y] is a sequence of responses that are elicited automatically once the preconditions are met. These effects may be captured in a postulate: Postulate 11.2. The affective reactions that one experiences, along with concepts and cognitions that are stimulated by an appraisal of a stimulus situation, can compose a precondition for a previously learned sequence of responses (i.e., a cognitive production). This response sequence may be activated spontaneously, with little conscious cognitive mediation, under conditions in which the precondition is met.

Despite the vagueness of Postulate 11.2, some general predictions seem tenable. One particularly interesting hypothesis concerns the use of holistic vs. piecemeal processing strategies to interpret information and make judgments. The impact of affect on the use of these strategies is described in the pages that follow. The Influence of Affect on the Use of Categorical Criteria Bless (2001; Bless et al., 1996) postulated that positive affect is associated with a disposition to use broad concepts and general knowledge structures to perform experimental tasks, whereas negative affect is associated with attention to situational details. These differences could result from social learning. That is, positive affect may become associated with conditions in which success has not required a careful scrutiny of the situation or task to be performed. In contrast, negative affect could become associated with failure and negative events that can only be avoided by attention to situational details (cf. Schwarz, 1990). As a result of these learned associations, the experience of positive affect can often stimulate a production that involves the use of general (e.g., categorical) criteria in responding to an achievement task, whereas negative affect may activate a production that involves a more detailed analysis of the task requirements and the information bearing on them.

340

CHAPTER 11

Furthermore, these alternative productions might be activated spontaneously when the affect-based preconditions for their occurrence are met, and might guide behavior with little if any conscious cognitive mediation. The different productions that are activated by positive and negative affect require different types of knowledge. The procedures activated by positive affect involve the use of general concepts and knowledge structures. In contrast, the procedures activated by negative affect may require more specific concepts that can be used to interpret informational details. This difference should be reflected in the type of concepts and knowledge that are used to interpret and organize new information. At least two studies support this possibility. Bless et al. (1996) reasoned that if people who experience positive affect spontaneously activate general knowledge structures to use in processing information, they should be able to process this information more efficiently. Moreover, their use of these knowledge structures should free cognitive resources for use in other, unrelated cognitive activity. To evaluate this possibility, participants who were induced experimentally to feel happy or unhappy listened to two tape-recorded stories about daily life events (e.g., making a telephone call from a public phone booth) while they performed a clerical task. Both recognition memory for the events mentioned in the audiotapes and accuracy in performing the clerical task were recorded. As expected, happy participants were more likely than unhappy ones to use a prototypic script (Schank & Abelson, 1977) to comprehend the taped information. That is, happy participants were more likely than unhappy ones to report prototype-consistent events as having been mentioned, but were less likely than unhappy participants to report atypical events. To the extent that participants apply a general prototype in comprehending the tape-recorded information, however, they should be able to devote more cognitive resources to the clerical task. This was also the case. Happy participants performed better on the secondary, clerical task than unhappy participants did. A quite different study by Dienes (1996) has similar implications. In this study, happy and unhappy participants were asked to learn a list of words that exemplified either easily identifiable concepts (animals, countries, etc.) or obscure ones (e.g., things made of wood). Learning the list could be facilitated by (a) attending to the details of the individual words, (b) mentally organizing the words in terms of the concepts they exemplify, or (c) both. (For evidence that the organization of information items into categories improves learning, see Hunt & Einstein, 1981.) Dienes reasoned that when the concepts exemplified by the stimulus words were obvious, both happy and unhappy participants would spontaneously identify and use these concepts to organize the stimuli. When the concepts were more obscure, however, only happy participants should identify them spontaneously, so only these participants’ learning of the stimulus words should be facilitated. Consis-

THE ROLE OF AFFECT

341

tent with this reasoning, happy participants had better memory than unhappy participants for the hard-to-categorize words but not for the easy-tocategorize words. The conclusions that persons who experience positive affect bring categorical and general knowledge structures to bear on tasks they perform does not imply that affective reactions per se cue the retrieval of these knowledge structures. Rather, this retrieval is mediated by cognitive productions that require different types of declarative knowledge. The preconditions for activating a production include situational features as well as affective reactions. In some cases, the cognitive productions that are activated by situation-specific features might not require the types of knowledge of concern in the studies by Dienes and Bless et al. In other cases, affective reactions might not be a feature of the preconditions that elicit goalrelevant productions. Consequently, there are many instances in which affect is likely to have little influence on the strategies that people use in responding to information. In this regard, Murray, Sujan, Hirt, and Sujan (1990) found that although happy participants used fewer categories to classify stimuli than control participants when they were told to focus on similarities, they used relatively more categories to classify the stimuli when they were told to focus on differences. These latter findings qualify the conclusion that happy persons typically use broader categories to classify stimuli than unhappy persons. Further experiments by Murray et al. (1990) indicated that differences in categorization observed in both task conditions were likely to result from a more general tendency for positive affect to increase the number of attribute dimensions that participants brought to bear on the particular task they were asked to perform. Thus, it increased their ability to make distinctions among the stimuli they were considering as well as to identify similarities. As Murray et al. suggest, these results could reflect mood-induced differences in the motivation to perform the task at hand rather than differences in the cognitive productions that participants spontaneously apply under the conditions being investigated. We consider this possibility in more detail shortly. Implications for Judgment If positive affect increases the disposition to rely on categorical criteria for comprehending and responding to stimuli, this disposition should be reflected in judgments to which the criteria are relevant. However, this influence could be localized at two stages of processing. First, if the information that is available for judging a stimulus consists of both categorical descriptions and more detailed features, positive affect might increase the relative weight that is attached to the categorical information at the time the implications of the two types of information are combined to form a judgment. An al-

342

CHAPTER 11

ternative possibility, however, is that positive affect influences the attention that is paid to categorical information at the initial, comprehension stage. That is, if individuals who experience positive affect think more extensively about the implications of categorical information at the time the information is received, they may interpret the implications of this information as more extreme, and this could also lead the information to have more effect on judgments that individuals report later. (For more general evidence that thought increases the extremity of judgments, see Tesser, 1978.) These two alternative possibilities are difficult to separate. That is, both effects imply that positive affect will increase the apparent influence of categorical information on judgments. However, a series of studies by Adaval (2003) suggested that, in fact, the effect is localized at the initial comprehension stage. In two studies, happy and unhappy participants evaluated a product on the basis of an unfavorable or unfavorable brand name (a categorical criterion) and a favorable or unfavorable specific attribute description. Judgments were also obtained of products described by each attribute feature separately. Brand information had more influence when participants were feeling happy than when they were not. To determine the reason for this difference, Adaval assumed a weighted averaging model of information integration of the sort proposed by Anderson (1971, 1981; see Footnote 23), and applied a parameter-estimation procedure to identify both the absolute weight attached to each piece of product information and its scale value (i.e., its evaluative implications). Analyses of these parameters, computed for each participant separately, showed that positive affect had no impact on the weight attached to brand information. However, it increased the extremity of the scale value assigned to it (That is, happy participants interpreted favorable brands as more favorable, and unfavorable brands as more unfavorable, than unhappy participants did.) Further studies provided confirming evidence of this conclusion using different criteria. These results are consistent with the results obtained by Bless et al. (1996), that people attend more to categorical bases for interpreting the stimuli they encounter. Moreover, the effects of this attention occur at the time the information is first received and interpreted and not at the time the information is used to make a judgment. MOTIVATIONAL INFLUENCES OF AFFECT ON INFORMATION PROCESSING One body of phenomena to which Postulate 11.2 is applicable has been identified in research on communication and persuasion. Much of this research suggests that people who feel happy are less influenced by the quality of the arguments contained in a persuasive message than unhappy people are (cf. Bless, Bohner, Schwarz, & Strack, 1990; Mackie & Worth, 1989).

THE ROLE OF AFFECT

343

Several factors could have been hypothesized to account for this difference. Mackie and Worth (1989), for example, proposed that positive affect is likely to activate a large number of concepts and knowledge representations in memory, taxing participants’ capacity to provide new information and making them less able to attend to details of the information available. This interpretation, however, is called into question by Niedenthal and Setterlund’s (1994) finding that affective reactions activate only those concepts that the specific type of reactions being experienced exemplify, and do not increase the accessibility of valenced concepts and knowledge in general. A somewhat more viable conceptualization is consistent with the use of affect as information. Specifically, Schwarz (1990; Schwarz, Bless, & Bohner, 1991) argued that because positive affect is typically associated with pleasant states of affairs, people who experience these feelings in a new situation may appraise the situation as benign and, therefore, may not pay careful attention to details. In contrast, negative affect is associated with aversive states of affairs and so the experience of it stimulates an interpretation of a new situation as potentially problematic. Although these associations could give rise to cognitive productions that are applied automatically, as noted in a previous section, they could potentially stimulate deliberative cognitive activity as well. Consequently, people may be less motivated to pay attention to the details of a message they receive when they feel happy than when they feel unhappy, and so the quality of the arguments contained in the message has less impact on their judgments. Although this interpretation seems plausible, it is contradicted by evidence that people engage in as much cognitive activity in responding to a persuasive message when they are in a good mood as when they are in a bad mood. In Bless et al.’s (1990) study, for example, thought listing data indicated that the overall amount of this activity did not depend on the mood that participants were in. That is, participants in a good mood tended to counterargue both weak and strong arguments (and thus were equally influenced by both),whereas participants in a bad mood elaborated strong arguments and counterargued weak ones, thus being differently persuaded by them. The total number of thoughts generated in the two mood conditions was the same in both cases, however. Research on impression formation by Isbell (2000) also suggested that happy and sad individuals do not differ in the amount of cognitive activity they perform in response to information, but differ only in the nature of this activity. A third conceptualization is based on the notion that people who are feeling happy are motivated to maintain their mood and, consequently, avoid activities that are likely to bring them down. Unhappy individuals, on the other hand, are motivated to divest themselves of the negative feelings they are experiencing, and engage in activities that distract themselves from thinking about these feelings and the events that gave rise to them. Because persua-

344

CHAPTER 11

sive messages are typically counterattitudinal, thinking about them is particularly likely to destroy one’s good mood. Consequently, people who feel happy tend to avoid this cognitive activity and, as a result, are insensitive to differences in the quality of arguments contained in the messages. If this interpretation is correct, processing differences should be eliminated or even reversed by leading happy persons to believe that thinking about the information they receive will be enjoyable and, therefore, will maintain or enhance their positive mood state. Wegener, Petty, and Smith (1995) showed this to be the case. Participants listened to one of two communications. One message advocated a position with which participants were inclined to disagree and was introduced with instructions that it was likely to make them unhappy. The quality of arguments contained in this message had less impact on participants’ attitudes when they were happy than when they were not. The other message advocated a position that participants typically favored and was introduced with a comment that reading it was likely to be enjoyable. In this case, the quality of the arguments had greater effect on participants’ attitudes when they were happy. Wegener et al.’s (1995) results therefore qualified the conclusion that people who experience positive affect are generally unmotivated to engage in extensive cognitive processing. There are also qualifications on the assumption that people who experience negative affect are motivated to process new information extensively. Cognitive activity may only help to eliminate negative affect if the affect is mild and nothing can be done about the events that produced it. Extremely unhappy individuals are likely to ruminate about the events that created their unhappiness instead of distracting themselves by engaging in irrelevant activity (Ellis & Ashbrook, 1988; Klinger, 1975; Martin & Tesser, 1996). To this extent, these individuals may be just as insensitive to details of affect-irrelevant information as happy people are. The information processing implied by this interpretation is deliberative. To this extent, it would not contradict the possibility that affective reactions can stimulate cognitive procedures of the sort noted in the previous section, which are performed automatically in situations in which they apply. However, the deliberative processes that are suggested by moodmaintenance and mood-repair motives have an alternative interpretation. That is, these processes could be the byproduct of a more general tendency to use affect as information. We now turn to this possibility. A PERFORMANCE-FEEDBACK MODEL OF AFFECT AS INFORMATION In most research that has been performed to evaluate the use of affect as information, people are assumed to make judgments of a particular person, object, or event, and the affect they experience at the time of judgment is

THE ROLE OF AFFECT

345

used as information about their reaction to the stimulus they are judging. However, people’s affective reactions can provide information about other things as well. In situations outside the laboratory, people may use their affect as a basis for behavioral decisions. For example, they may decide to engage in a behavior that gives them pleasure, while avoiding activities they consider to be aversive. In achievement situations, on the other hand, people’s decisions to continue on a task can also be influenced by their feelings about the quality of their performance. A woman who is writing a term paper, for example, might have little objective basis for deciding whether her output is sufficient to attain the grade she is seeking. Therefore, her decision to continue working on the paper may be based on her subjective feelings about the work she has done. The extraneous affect she happens to be experiencing at this time could influence her perceptions of these feelings and, as a consequence, her decision to stop working or continue. These possibilities were formalized by Wyer et al. (1999) in a performance-feedback model of affect. This conceptualization, which was stimulated in large part by Martin, Achee, Ward, and Wyer’s (1993) previous work, assumes that people spontaneously monitor their behavior and the outputs it generates to determine if they have satisfactorily attained the goal they are pursuing. If they conclude that this is the case, they terminate their goal-directed activity. Otherwise, they continue processing, using the same or a different strategy, until they generate an output they consider to be satisfactory. When objective criteria exist for deciding whether a goal has been attained, the decision to engage in goal-directed activity is obviously based on these criteria. More often than not, however, the criteria for goal attainment are ambiguous. When this is the case, people may base their decision on the affect they happen to be experiencing and attribute to their feelings about their attainment of the goal they are pursuing. The most interesting implications of this analysis derive from the fact that people’s perceptions of their affective reactions depend on the particular goal they happen to be pursuing. In the preceding example, the woman’s goal was to write a paper of high quality. In this case, she may ask herself if she feels the paper meets this standard and is more likely to stop work if her feelings are positive than if they are negative. On the other hand, suppose the woman focuses her attention on whether she is having fun rather than the quality of the paper she is writing. Then, her feelings are likely to have precisely the opposite effect. That is, she may be more inclined to stop writing if her feelings are negative than if they are positive. Two studies by Martin et al. (1993) provided direct support for these possibilities. In one, happy or sad participants were asked to form an impression of someone on the basis of a series of behaviors. Each behavior was printed on a separate card of a deck that participants were given. Partici-

346

CHAPTER 11

pants under performance-criterion conditions were told to continue turning the cards and reading the behaviors until they had enough information to form an impression. In contrast, participants under enjoyment-criterion conditions were told to continue as long as they were enjoying what they were doing. The first group of participants should implicitly ask themselves as they read the behaviors if they feel they have enough information. Therefore, they should be more likely to answer this question affirmatively, and to stop reading sooner, if they are happy than if they are sad. The second group of participants, however, should ask themselves if they are enjoying the task, again responding more affirmatively if they are happy than if they are sad. In this case, therefore, happy participants should continue reading behaviors longer than sad participants. Results confirmed these predictions. That is, happy participants read fewer behaviors than sad participants in performance-criterion conditions (24 vs. 43, respectively), but more behaviors than sad participants in enjoyment-criterion conditions (47 vs. 36, respectively). A second study replicated these findings using a different task. In this case, participants were asked to generate a list of birds. Some participants were told to continue until they thought it was a good time to stop, whereas others were told to continue as long as they were enjoying themselves. Happy participants persevered less long on the task than unhappy participants when they were told to use a performance criterion (140 s vs. 218 s, respectively), but longer than unhappy participants when they were encouraged to employ an enjoyment criterion (240 s vs. 145 s, respectively). Individual and Situational Differences in Goal Orientation In the conditions just described, the criterion for stopping or continuing was explicitly stated by the experimenter. When the criterion is unspecified, both individual and situational differences could influence the criteria that participants use. Persons with high need for cognition (Cacioppo & Petty, 1982), for example, typically find intellectual activity to be intrinsically enjoyable. They may therefore be likely to adopt an enjoyment criterion, whereas persons with low need for cognition may be guided to a greater extent by performance criteria. If this is so, the experience of positive affect might increase the persistence of high need-for-cognition individuals on response-generation tasks, but might decrease the persistence of low need-for-cognition individuals on such tasks. In fact, additional data obtained by Martin et al. (1993) confirmed this prediction. The relative emphasis on performance and enjoyment is also likely to vary with the nature of the task and the context in which it is performed. In laboratory studies, participants are often given a particular task objective

THE ROLE OF AFFECT

347

by the experimenter. Under these conditions, participants may spontaneously employ a performance criterion. That is, they may base their decision to persevere on whether they feel that their responses fulfill these externally imposed task requirements. This hypothesis was supported in a third condition of Martin et al.’s second experiment in which participants were simply told to stop generating birds when “they wanted to.” The results in this condition were very similar to those obtained when participants were given an explicit performance criterion; that is, happy participants stopped sooner than unhappy ones. These results confirm the speculation that participants in many experimental situations spontaneously focus their attention on whether they have fulfilled the implicit task objectives they are given. Nevertheless, when an experimental task is particularly interesting, or when achievement goals are de-emphasized, an enjoyment criterion might be applied as well. Some evidence in support of the latter possibility was reported by Murray et al. (1990) in a study mentioned earlier in this chapter. Some participants were asked to list similarities between two popular television programs, whereas others were asked to list differences between the programs. The task was intrinsically interesting to college-age participants and, therefore, was likely to stimulate them to use an enjoyment criterion in deciding how much effort they should expend. To this extent, participants should use the affect they were experiencing as a basis for inferring their enjoyment. Consistent with this assumption, happy participants reported enjoying the task more than control participants did. Correspondingly, happy participants listed more similarities between the television programs when they were asked to focus on similarities, and listed more differences between the programs when they were focused on differences, than control participants did. These results would be hard to explain on the basis of either the assumption that positive affect generally increases the accessibility of global concepts in memory. Nor is it consistent with the assumption that positive affect reduces cognitive resources. It seems more reasonable to suppose that happy participants inferred from the affect they were experiencing that they were enjoying the task, and therefore devoted more effort to an identification of both similarities and differences in the stimuli they were judging than other participants did. Murray et al.’s finding that happy participants reported greater enjoyment of the task than other participants raises an additional consideration. That is, when several goals are available in a situation, the affect that persons happen to be experiencing could influence the goal they choose to pursue as well as their inference that they have or have not attained it (for a discussion of this possibility, see Schwarz & Clore, 1996). For example, suppose happy individuals are generally more inclined to adopt an enjoyment objective than unhappy individuals are (Isen, 1987). If this is so, happy

348

CHAPTER 11

persons may persist on a task because they infer they are enjoying it, whereas unhappy persons, who adopt a performance criterion, may persist because they feel they have not yet attained an adequate level of achievement. To this extent, happy and unhappy persons might persist to an equal extent, albeit for different reasons. Despite this ambiguity, the influence of affect on behavior in a large number of situations both in and outside the laboratory can be conceptualized in terms of its mediating influence on the answers to questions that persons implicitly ask themselves about how they feel about their behavior and its outcomes. The broader implications of this conceptualization can be summarized in two postulates: Postulate 11.3. When people engage in behavior for the purpose of attaining an external goal and objective criteria for inferring the attainment of this goal do not exist, they use the affect they are experiencing to infer whether they feel the strategy they have used in pursuit of this goal, and the outputs they have generated by employing it, are appropriate. Therefore, they are more likely to infer they have met the criterion for satisfactory performance if they are experiencing positive affect than if they are not. Postulate 11.4. When people’s primary reason for engaging in behavior is enjoyment, they use the affect they are experiencing as a basis for inferring this enjoyment. Therefore, they will continue the activity longer if they are experiencing positive affect than if they are not.

An application of these postulates calls attention to several contingencies in the impact of affect that have not been investigated in the research performed to date. Several areas of research are worth reviewing from this perspective. Creativity A creativity task often employed in the laboratory requires participants to generate instances of a concept. This task is similar to that constructed by Martin et al. (1993, Experiment 2) except that the concept involved has unclear boundaries, thus allowing participants to generate novel responses. Thus, for example, participants might be asked to generate uses of a brick or things they could use to sit on. In this situation, like that constructed by Martin et al., participants might assume they should stop when they have generated enough good responses to fulfill the demands of the task. However, the task itself might often be intrinsically enjoyable, as it requires a degree of imaginativeness. Thus, many participants might spontaneously apply an enjoyment criterion as well. Consistent with this speculation, Isen and Daubman (1984) found that happy participants generated more re-

THE ROLE OF AFFECT

349

sponses when performing such a task than control participants did. This suggests that an enjoyment criterion predominated in their study. A somewhat different interpretation of Isen and Daubman’s results is suggested by the fact that happy participants also tend to generate more novel responses than other participants do. When the criterion for an appropriate response is unclear, participants may inhibit reporting responses they feel are poor exemplars of the concept they are considering. Specifically, they may subjectively generate a response (e.g., a potential use for a brick), ask themselves if they feel their response is a good one, and only report it if they answer this question affirmatively. Thus, they are more likely to record their response if they are happy than if they are not. Evidence that happy persons perform better then others on the Remote Associates Test (Isen, Daubman, & Nowicki, 1987) is consistent with this interpretation. Furthermore, the interpretation could explain the different results obtained by Isen and Daubman’s (1984) study and those reported by Martin et al. (1993). That is, the criterion for membership in the category that participants considered in Martin et al.’s experiment (birds) was clearly defined, and so the use of affect to evaluate the appropriateness of individual exemplars was less likely. Impression Formation In impression-formation research (for a review, see Wyer & Carlston, 1994), participants typically receive information about several behaviors a person has performed with instructions to form an impression of the person. A large amount of information is typically presented (e.g., 25–30 items; see Wyer & Martin, 1986), and so it is difficult to form an impression that takes all of it into account. Consequently, participants in these experiments may often ask themselves at various points if the information they have received is a sufficient basis for a concept of what the individual is like. If they answer this selfgenerated question affirmatively, they may base their concept of the person on this information without taking into account the implications of the information they receive later (see Postulate 3.1). If this is so, and if happy participants are more likely than unhappy participants to respond affirmatively to such a question, they should use relatively less information to form their impressions. Moreover, the information they use should typically occur at the beginning of the sequence rather than at the end. Martin et al.’s (1993, Experiment 1) results confirmed the assumption that happy participants with an impression-formation objective use less information to form an impression than sad participants do. The prediction that happy persons are more likely than unhappy ones to base their impression on the first information they receive was confirmed in research by Isbell (2000; Isbell, Clore, and Wyer (1998). Participants who were feeling ei-

350

CHAPTER 11

ther happy or sad were asked to form an impression of someone who was described at the outset as either extraverted or introverted. This general description was followed by a series of behaviors, an equal number of which exemplified each of these traits. After receiving the information, participants judged the target’s extraversion and then recalled the behaviors they had received. Isbell expected that the general description of the target that participants received at the outset would activate a concept of either an extravert or an introvert, and that participants would implicitly ask themselves if they felt this concept was an appropriate basis for their impression. Happy participants, who should answer this question affirmatively, should be more inclined than unhappy participants to apply the concept to the target and to base the judgments they report later on this concept. This was in fact the case, as indicated earlier. That is, the initial target description had a substantial effect on happy participants’ judgments of the target’s extraversion but had very little impact on unhappy participants’ judgments. The fact that happy participants based their judgments on the initial description of the target does not mean that they failed to think about the behavioral information they received later. To the contrary, persons who apply this concept to the target should think about the behavioral information they receive with reference to it (Srull & Wyer, 1989). Moreover, they should think more extensively about behaviors that are inconsistent with the concept in order to reconcile their occurrence (Hastie, 1980; Srull & Wyer, 1989; Wyer & Srull, 1989). As implied by this reasoning, these participants recalled more inconsistent behaviors than consistent ones. Unhappy participants, who did not apply the concept activated by the initial trait description to the target, did not engage in this inconsistency resolution. Consequently, they recalled behaviors equally well regardless of their consistency with the initial trait description. In many cases, people’s initial impressions of someone are based largely on their feelings toward this person. This is particularly likely if their first encounter with the person elicits a spontaneous appraisal on the basis of his or her physical appearance (cf. Yeung & Wyer, in press). In such cases, the affect they experience at the time they judge the person could potentially have a dual function. First, it may convey judgment-relevant information about the individual being judged (Schwarz & Clore, 1983). Second, it could be used as an indication of whether this affect is a sufficient basis for making the judgment. Thus, people who experience positive affect at the time they are asked to judge someone might not only infer they feel favorably toward the target, but also that these feelings are a sufficient basis for the judgment they are asked to make. In contrast, people who experience negative affect may infer they feel negatively toward the target but also may conclude that these feelings are not a sufficient basis for the judgment.

THE ROLE OF AFFECT

351

This implies that the effects of positive affect on impression judgments may often be greater than the effects of negative effect. There is a contingency in these predictions. That is, the contention that happy people are inclined to decide that the initial information they receive is a sufficient basis for judgment assumes that the implications of this initial information, considered in isolation, are unambiguous. If these implications are mixed or unclear, they may not be considered sufficient bases for judgments by either happy or unhappy individuals. A recent study by Isbell (2003) bears on this contingency. She found that happy participants were more likely than unhappy participants to base their evaluations of the target on a general trait characterization of him even when this characterization was presented after the target’s behaviors were described. In this study, however, the behavior descriptions were both favorable and unfavorable, and so a clear initial impression of the target was difficult to form on the basis of these descriptions alone. Therefore, participants were unlikely to consider the behaviors to be a sufficient basis for judgment regardless of their affective state. The greater use of categorical information by happy participants under these conditions could reflect their generally greater tendency to apply broad categories in interpreting information, as suggested by Adaval (2003) and Bless et al. (1996) and noted earlier.

Stereotyping Our interpretation of Isbell et al.’s (1998) findings also has implications for the effects of affect on the use of stereotypes. In a study by Bodenhausen (1993; see also Bodenhausen et al., 1994), for example, happy and sad participants were asked to judge the culpability of a defendant on the basis of information about the defendant’s background and the conditions surrounding the crime. The description of the crime was preceded by the target’s name and hometown, which either identified him as Hispanic or provided no clue as to his ethnicity. It seems reasonable to suppose that the defendant’s ethnicity spontaneously activated a stereotype, and that participants implicitly asked themselves if they felt this was an appropriate basis for their judgment, answering affirmatively if they were happy and negatively if they were sad. Consistent with this reasoning, the stereotype had more impact on happy participants’ judgments than on unhappy participants’. Note that if the stereotype had not been activated until after the detailed crimerelated information was presented, participants would not be in a position to ask themselves at the outset whether the stereotype was a sufficient basis for judgment. If anything, they should ask themselves if they felt the crime-related information was sufficient. To this latter extent, happy participants in this condition might be more influenced by the crime-related infor-

352

CHAPTER 11

mation, and less influenced by the stereotype, than sad participants are. To my knowledge, this possibility has not been examined. Self-Evaluation Processes The determinants and effects of self-judgments have been a major focus of social-psychological research for many years (cf. Greenwald & Pratkanis, 1984; Klein & Loftus, 1993; McGuire & McGuire, 1988). Because of the general interest in these matters, the implications of the performance-feedback model for self-judgment processes are worth considering in some detail. The influence of affect on four different types of self-judgments—judgments of global life satisfaction, self-worth, specific trait judgments, and estimates of personal risk—are therefore considered from the perspective of this model. Several alternative cognitive processes could underlie these judgments, each of which can involve the use of a different type of self-knowledge. The effects of these procedures can be conceptualized in terms of the general processing model outlined at the beginning of this volume. Specifically, when a self-judgment is required, the goal specification may resonate with a number of different cognitive procedures for generating this judgment. These processes could include (a) the search for a previously formed judgment-relevant trait description, (b) the identification of a judgment-related past experience and a construal of its implications for the judgment to be made, and (c) an assessment of the implications of the affective reactions one is experiencing. Situational or individual difference factors that make one process easier to use than another will increase the likelihood that this process is applied. An additional consideration, however, is implied by the performance-feedback conceptualization. That is, participants who have generated a response on the basis of the “winning” procedure may spontaneously ask themselves if they feel that this response is a sufficient basis for judgment. participants who are experiencing positive affect may answer this question affirmatively, whereas those who are experiencing negative affect may not. Therefore, happy persons are more likely than unhappy persons to accept and apply the first criterion they are able to generate. The following discussion is based on this assumption. 1. Judgments of Life Satisfaction A request to judge one’s life satisfaction could activate three different processes. First, respondents might search for a previously formed concept of their life as a whole that is directly applicable to the judgment. Second, they might recall one or more previous life events that are relevant to the judgment and base their response on the affect elicited by these memories. Finally, they might assess the affect they happen to be experiencing at the

THE ROLE OF AFFECT

353

FIG. 11.2. Processing stages involved in making (A) global judgments of life satisfaction and self-esteem, and (b) specific trait judgments. Based on Wyer et al. (1999).

time and use this as an indication of their feelings about life in general without conducting a search of memory at all. These processes, which could occur simultaneously, are summarized in Fig. 11.2a. On a priori grounds, it is unclear which of these processes will predominate. For example, if participants in an experiment have a previously

354

CHAPTER 11

formed concept of their life as a whole, a request to estimate their life satisfaction might spontaneously activate this concept and so a judgment can be quickly computed on the basis of the affect it elicits. On the other hand, people are infrequently called upon to make this type of judgment in daily life. Consequently, they usually do not have a previously formed concept of their life as a whole stored in memory, and must compute a judgment at the time they are asked to report it. In this case, it would normally take less time to compute a judgment on the basis of the transitory affect one is experiencing at the time of judgment than to do so on the basis of other, less easily accessible criteria, and so judgments are more likely to be based on the former computation. Evidence that people use transitory affect as a basis for life satisfaction judgments independently of any specific self-knowledge they have available (Schwarz & Clore, 1983; Strack, Schwarz, & Gschneidinger, 1985) is consistent with this hypothesis. However, the performance-feedback model we have proposed raises a further consideration. People who experience positive affect for reasons that are objectively unrelated to their life situation might not only use these feelings as a basis for judgment but also consider this criterion to be sufficient. In contrast, people who experience negative affect might be disposed to question the sufficiency of this criterion. Therefore, they might either seek other judgment-relevant information or, if no alternative criteria are readily available, might adjust for the bias that results from their use of feelings as information. Schwarz and Clore’s (1983, Experiment 1) findings support this conjecture. To reiterate, happy persons appeared to use their feelings as a basis for judging their life satisfaction regardless of the situational factors that had implications for their validity as a basis for judgment. In contrast, unhappy participants adjusted their judgments to compensate for the effects of these factors. Perhaps both happy and unhappy participants spontaneously considered the affect they were experiencing as a potential basis for judgment. However, whereas happy participants concluded that these feelings were a sufficient basis for judging without thinking about situational factors that might have influenced them, unhappy participants considered that an assessment of their feelings per se was insufficient and took situational factors into account as well. 2. Judgments of Self-Worth Estimates of self-worth, like estimates of life satisfaction, may typically be based on feelings. These feelings are likely to be estimated by either (a) retrieving a previously formed concept of oneself with which affective reactions have become associated through learning, (b) recalling descriptive self-knowledge to which affective reactions have become condi-

THE ROLE OF AFFECT

355

tioned and combining the implications of these reactions to arrive at an overall estimate of one’s feelings; and (c) assessing the implications of the transitory affect one happens to be experiencing for reasons that may or may not actually be relevant. If a previously formed concept of “self” exists and is activated by the request to evaluate one’s self-worth, the affective reactions elicited by this concept might be used as a basis for judgment. If such a concept has not previously been formed, however, and if the affective reactions based on descriptive self-knowledge take time to compute, the transitory affect one happens to be experiencing for other reasons should normally be activated more quickly and, therefore, is more likely to be used. The aforementioned study by Levine et al. (1994) bears on this possibility. In this study, the descriptive implications of participants’ recalled experiences only affected their judgments of competence in the domain to which these experiences were relevant. However, participants’ estimates of their global self-esteem were influenced by the affect elicited by the life events they had recalled regardless of its descriptive implications. In terms of our conceptualization, this suggests that participants in Levine et al.’s study did not have a previously formed concept of themselves easily accessible in memory. Therefore, they based their judgments on the affect they were experiencing at the time, regardless of whether or not the event that elicited the affect was self relevant. 3. Trait Judgments Judgments of the specific traits one possesses are more likely to be based on descriptive rather than affective criteria. The processes that underlie these judgments are summarized in Fig. 11.3b. Specifically, people who are asked if they have a specific trait may recall and report a previously formed concept of themselves with respect to this trait. Alternatively, they may infer the trait from the descriptive implications of trait-relevant self-knowledge that they have acquired (e.g., behaviors that exemplify the trait, descriptions of oneself by others, etc.) For affective reactions to have an informational influence on trait judgments, mediating computations must be performed in order to transform these reactions into descriptive terms. For example, people who experience negative affect might judge themselves to be incompetent based on the implicational molecule [I am incompetent; I feel badly about myself].

The retrieval and use of this molecule, however, is likely to take longer than computations based on descriptive criteria. This is particularly true if a trait-relevant experience is easily accessible in memory at the time of judg-

356

CHAPTER 11

ment. Levine et al.’s (1994) findings, described earlier, are consistent with this hypothesis. That is, the descriptive implications of participants’ recalled life experiences were used as a basis for judgments of attributes to which they were relevant. The affective implications of recalled experiences only influenced judgments of attributes in a domain to which the experiences were unrelated. The implications of the performance-feedback model become of interest in this context. If trait judgments are computed more quickly on the basis of descriptive criteria than on the basis of affective criteria, people who experience positive affect should be likely to infer that these criteria are sufficient, whereas people who experience negative effect are likely to consider additional (e.g., affective) criteria. Thus, in contrast to judgments of life satisfaction and self-worth, trait inferences are more likely to be influenced by negative affect than by positive affect. Communication and Persuasion Suppose people read a persuasive communication for the purpose of deciding if the position it advocates is valid. They are likely to have an initial impression of the position’s validity position before they evaluate the communication’s contents. This impression could be based on the source of the message, on surface features of the communication itself (e.g., writing style), or on a previously formed attitude toward the position in question. If people who have made formed this impression ask themselves if it is an appropriate basis for accepting the position advocated, they should be more inclined to answer this question affirmatively if they are happy than if they are sad. This has two implications. First, happy persons are more likely than unhappy persons to rely on the source of a message as a basis for accepting the position advocated. Second, happy persons are less inclined than sad persons to make a careful evaluation of the arguments contained in a message. Consequently, their acceptance of the position advocated is less likely to be influenced by the quality of these arguments. Support for the first hypothesis has been obtained by Roselli (1995), and the second hypothesis is consistent with results of several studies described previously (Bless et al., 1990; Mackie &Worth, 1989). A performance-feedback conceptualization of these results suggests two contingencies, however. First, the different cognitive activities performed by happy and unhappy persons should only occur if these persons perceive their feelings to have implications for the validity of their initial impression. If participants attribute their feelings to other, impression-irrelevant factors, the difference in processing should theoretically not occur. Results reported by Sinclair, Mark, and Clore (1994) confirmed this prediction. They found that the effect of argument strength on participants’ accep-

THE ROLE OF AFFECT

357

tance of a persuasive message was less on sunny days (when participants were happy) than on rainy days (when they were unhappy). However, calling participants’ attention to the weather (thereby making salient the true source of their feelings) eliminated this difference. Second, our analysis assumes that recipients’ objective at the time they read the message is to evaluate the validity of the position being advocated. In contrast, suppose persons have an enjoyment objective. Then, precisely the opposite influence of affect might occur. The aforementioned study by Wegener et al. (1995), suggesting that happy persons are more influenced by the quality of arguments in a message than other participants when they are led to believe that the message will be enjoyable, confirms this possibility. Accuracy The performance-feedback conceptualization can apply to judgments of physical stimuli as well as social ones. People who are called upon to judge a stimulus may spontaneously make a holistic appraisal of the stimulus at the time they are first exposed to it before they scrutinize its individual features. They may then ask themselves if this appraisal provides a sufficient basis for judgments. To this extent, happy individuals should be more likely than unhappy ones to base their judgments on this initial appraisal alone. One implication of this is that when accurate judgments are likely to require a careful analysis of the individual stimulus features happy participants are less likely to be accurate than unhappy ones. Evidence reported by Sinclair and Mark (1992), showing that happy participants were less accurate than sad participants in estimating the correlation between two variables on the basis of scatter plots, confirms this speculation. Behavior in Nonlaboratory Situations Many situations arise outside the laboratory in which persons’ behavioral decisions are likely to be mediated by their feelings about the behavior and its implications. Students who prepare for an exam must decide if they have studied long enough. A person who wants to ask someone for a date must decide if it is the right time to do so. Or, a person who wishes to buy a car must decide if he or she has enough information about it to make a decision. In such instances, people should stop studying earlier, ask persons for a date sooner, and make a purchasing decision more quickly, if they feel happy than if they do not. For similar reasons, happy persons may be more inclined to behave impulsively, or in ways that upon reflection they might consider undesirable. A series of studies by Forgas (1998) are interesting to consider in this con-

358

CHAPTER 11

text. In three studies, participants were asked to indicate how they would phrase a verbal request in a hypothetical social interaction. It seems reasonable to suppose that people who want another person to do something (e.g., to repay a loan) will spontaneously generate a verbal phrasing of the request that corresponds directly to the desired action (e.g., “Give me the money you owe me”) and may implicitly ask themselves if it is appropriate to express the request in these terms. If they answer this question positively, they are likely to alter the phrasing in a way that corresponds to normative standards of politeness. Otherwise, they might not. If this is so, the performance-feedback conceptualization we propose suggests that happy persons are more likely to express their requests impolitely than unhappy persons are. This, in fact, is what Forgas (1998) found. Although his studies were performed in the laboratory, they seem likely to generalize to other situations as well. It would nevertheless be hazardous to conclude that happy persons are generally less likely to engage in extensive information processing than unhappy persons are. This difference should theoretically exist only when individuals have an implicit performance objective. In nonlaboratory situations, information processing is often performed for enjoyment. In these circumstances, the performance-feedback implies that happy persons will persevere longer, and will process information more analytically, than sad persons will. Thus, for example, happy individuals might work harder than sad individuals on crossword puzzles, and may read news articles in relatively greater depth. Moreover, they might make less hasty decisions about other persons on the basis of their initial impressions of these persons, in contrast to results observed in the laboratory (Isbell et al., 1998). The possibilities, which are not implied by other conceptualizations of the influence of affect on performance, should be kept in mind when drawing conclusions on the basis of experimental research in which performance objectives are implicitly emphasized.

AFFECT, ATTITUDES, AND BEHAVIOR As Thurstone (1959) assumed, attitudes are often expressions of the affective reactions we experience toward a person, object, event, or social policy. To this extent, the situational and informational factors that influence the use of affect as information have direct implications for the determinants of attitude formation and change. However, much of the theoretical and empirical interest in this topic derives from the assumption that attitudes provide the basis for behavioral decisions. In fact, there are many contingencies on the validity of this assumption (cf. Fazio, 1986; Fishbein & Ajzen, 1975; Miniard & Cohen, 1981). In light of the centrality of the attitude

359

THE ROLE OF AFFECT

construct in social psychological research, however, the influence of affect in attitude formation and change deserves special attention. In fact, the role of affect in attitude formation and change has already been discussed in several earlier sections of this chapter, and does not require reiteration. Here, we attempt to outline more formally the processes that underlie the construction of an attitude and the situational and informational factors that influence it. Empirical evidence bearing on this analysis is then described. General Considerations The ultimate objective of most persuasive communications is to influence recipients to behave in a way that the communicator considers to be desirable (to purchase a particular product, to endorse a social policy, or to stop smoking, etc.). However, it is useful to distinguish between two types of communications that could be constructed for this purpose. One type attempts to alter a person’s beliefs, or estimates of the likelihood that an event will occur, or that a particular proposition is true (e.g., that American’s bombing of Afghanistan has resulted in the deaths of thousands of innocent victims, or that air and water pollution will extinguish life on earth by the end of the 21st century). As suggested in chapter 5 (Wyer & Hartwick, 1984), communications that are intended to focus on beliefs in a proposition are likely to emphasize antecedents, that is, propositions about events that, if they are true, have implications for the truth of the target proposition. Affective reactions are likely to influence the impact of such a communication only indirectly, through the extent to which they influence the ability or motivation to evaluate the arguments conveyed in the communication or their implications for the validity of the target proposition (for evidence that affect can have this influence, see Fiedler, 1988.) The other type of message, which is of primary concern in the present context, is intended to influence an attitude, or evaluation of a person, object, or event (Americas bombing of Afghanistan, recycling, etc.). These messages are likely to focus on consequences of the event or state of affairs toward which the attitude is directed. The cognitions that typically underlie the construction of an attitude are captured by the model proposed by Fishbein (1963) or, more recently, Fishbein and Ajzen (1975). One’s attitude toward a behavior, for example (Ab), is captured by the equation Ab = Sbi ei ,

[11.1]

Where bi is the belief that the behavior will have a consequence i, and ei is the evaluation of this consequence. (Not all possible consequences of the behavior are taken into account, of course, but only a subset that is most

360

CHAPTER 11

“salient” at the time the attitude is formed.) To this extent, a communication should be effective in changing an attitude if it either increases beliefs that the behavior has consequences that the recipient considers to be desirable or, alternatively, increases the evaluation of consequences that the recipient already believes are likely to occur. Affective reactions are likely to play a quite different role in the processing of attitude-focused communications than in the processing of belieffocused messages. Moreover, this role may depend in part on whether the attitude in question is based on hedonic or utilitarian criteria (Zanna & Rempel, 1988). When the criterion is hedonic, the role of affect could depend on whether (a) recipients think sufficiently about the consequences described in a message for the affect associated with these consequences to be elicited, (b) whether they can effectively combine the implications of these consequences into a single attitudinal response in the manner specified in Equation 11.1, and (c) whether they can distinguish this affect from the feelings they are experiencing for other, unrelated reasons. Once an attitude is formed in this manner it may be reactivated at the time the course of action is actually contemplated (see Fazio, 1986, for a discussion of the factors that influence this activation). As noted earlier, however, the specific consequences of the position advocated in a communication may not be the only determinants of the attitude that recipients report. Affect could also be stimulated by the intrinsic pleasantness or unpleasantness of the behavior being advocated, independently of any specific information about it. (The thought of taking comprehensive examinations, for example, might spontaneously elicit negative affect independently of any specific consequences the activity might have.) Finally, extraneous affect that recipients happen to be experiencing may be misattributed to their feelings about this activity. Affect from all of these sources could contribute to people’s perceptions of their attitude toward the behavior being advocated. Experimental Evidence The role of affect in responses to attitude-focused persuasive messages was evaluated in a series of studies by Albarracin and Wyer (2001). We told participants we were concerned with how people process information in natural settings, and that to do this, we would like them both to transmit and receive information in a simulated coffee shop situation. On this pretense, we induced them to feel happy or unhappy by asking them to write a letter to a friend describing a positive or negative life experience. Then, we asked participants to read a communication, ostensibly taken from a local newspaper, which summarized either strong or weak arguments in support of instituting comprehensive examinations and urging recipients to vote in favor

THE ROLE OF AFFECT

361

of the exams in a forthcoming referendum. Participants read the message in the presence of background noise similar to that encountered in a coffee shop. In low-distraction conditions, the noise was of low volume and not particularly disturbing. In high-distraction conditions, however, it consisted of a high volume get-acquainted conversation between two students of the sort that might take place in a coffee shop situation. Participants after reading the message reported their attitudes toward instituting the exams and their intentions to vote for them. Then, they reported their beliefs that the various consequences described in the message would occur and the desirability of these consequences. 1. Effects of Distraction on the Processing of Message Content We expected that the influence of the persuasive message would result from cognitive activities at four different stages of processing: the computation of beliefs that the consequences specified in the message would occur, evaluations of the desirability of these consequences, an integration of these beliefs and evaluations to form an attitude in the manner described by Equation 11.2, and the transformation of this attitude into a behavioral intention. We further expected that the effects of distraction would be more evident at later stages of processing (e.g., the computation of an attitude or behavior intention) than at earlier ones. This was in fact the case. The effect of argument strength on judgments was inferred from the difference between judgments when arguments were strong and judgments when arguments were weak. These differences are shown in the top half of Table 11.5 for judgments that were presumably formed at each stage of processing. Distraction had a significant effect on the impact of argument strength on both attitudes and behavioral intentions, which were formed at later stages of processing. However, it had a much smaller effect on the impact of argument strength on evaluations of individual consequences (ei ), and had no effect at all on its impact of argument strength on beliefs that the consequences would occur (bi ). 2. Impact of Distraction on the Influence of Extraneous Affect If participants’ evaluations of the consequences mentioned in the message are based on the affective reactions that were elicited by thinking about them, Equation 11.1 should describe the results of integrating the implications of these reactions. However, distraction could not only prevent participants from performing this integration but could also prevent them from distinguishing between their affective reactions to the consequences and the affect they are experiencing for other reasons. If this is, so, the ex-

362

CHAPTER 11 TABLE 11.5 Effects of Argument Strength and Extraneous Affect on Behavioral Intentions, Attitudes, and Beliefs and Evaluations of Specific Behavioral Outcomes as a Function of Distraction

Effect of argument Strength High distraction Low distraction F(1, 73) Effect of affect High distraction Low distraction F(1, 73)

Outcome Beliefs (bi )

Outcome Evaluations (ei )

Attitudes (Ab)

Behavior Intentions

2.10a 2.08