Advances in Experimental Social Psychology, Volume 6

ADVANCES IN

EXPERIMENTAL SOCIAL PSYCHOLOGY

VOLUME 6

CONTRIBUTORS TO VOLUME 6

DARYL J. BEM

LEONARD BERKOWITZ CLAUDE FAUCHEUX

ALBERTJ. L o n BERNICE E. L o n SERGEMOSCOVICI CHARLAN NEMETH

SEYMOUR ROSENBERG

ANDREA SEDLAK

ADVANCES IN

Experimental Social Psychology EDITED BY

Leonard Berkowitz DEPARTMENT OF PSYCHOLOGY UNIVERSITY OF WISCONSIN MADISON, WISCONSIN

VOLUME 6

ACADEMIC PRESS

New York and London

1972

COPYRIGHT 0 1972, BY ACADEMIC PRESS, INC. ALL RIGHTS RESERVED NO PART OF THIS BOOK MAY BE REPRODUCED m ANY FORM, BY PHOTOSTAT, MICROFILM, RETRIEVAL SYSTEM, OR A N Y OTHER MEANS, WITHOUT WRIlTEN PERMISSION PROM THE PUBLISHERS.

ACADEMIC PRESS INC.

111 Fifth Avenue, New gork, New York 10003

United Kingdom Edition published by ACADEMIC PRESS, INC. (LONDON) LTD.

24/20 Oval Road, London M y 1

LIBRARY OF

PRINTED

CONGRESS CATALOO CARD

m TRE mrrm

NUMBER:64-23452

STATE^ OF AMERICA

CONTENTS CONTRIBUTORS . . . . CONTENTS OF OTHERVOLUMES

. .

. . . . . . . . . . . . . . .

. .

.

ix

xi

Self -Perception Theory Daryl J. Bern

.

I. Introduction . . . . . . . . . . . 11. Experimental Evidence for the Theory . . . . . . . 111. The Reinterpretation of Cognitive Dissonance Phenomena . . . IV. Other Self-Perception Phenomena . . . . . . . . V. Some Differences between Self-perception and Interpersonal Perception VI. The Shift of Paradigm in Social Psychology . . . . . . VII. Some Unsolved Problems . . . . . . . . . References . . . . . . . . . . . .

. . . .

. . . .

2 8 15 33

40 42

45

57

Social Norms, Feelings, and Other Factors Affecting Helping and Altruism Leonard Berkowitz

.

. .

I. Social Exchange and Helpfulness , . . . . 11. Reactions to Another Person’s Dependency . . . . . . 111. Moods and Feelings . . . . . . . . IV. Qualities of the Dependent Person Affecting Willingness to Help Him V. Conclusions . . . . . . . . . References . . . . . . . . . .

.

.

. .

.

.

.

. . .

. .

.

63 68 80 94 105

106

The Power of Liking: Consequences of Interpersonal Attitudes Derived from a Liberalized View of Secondary Reinforcement Albert J. Lott and Bernice E. Lott I. Introduction . . . . , . 11. Propositions Relevant to Consequences V

. .

. .

. . . . . . . .

.

.

109 112

vi

CONTENTS

111. Specific Predictions and Supporting Data IV. Areas of Applicability . . . . References . . . . .

.

. . . . . . . . . . . . . . . . . . . . .

116 136 141

Social Influence, Conformity Bias, and the Study of Active Minorities Serge Moscovici and Claude Faucheux I. The Conformity Bias and the Reasons to Challenge It . . . 11. The Source of Social Influence: Dependency or Style of Behavior . 111. Conformity, Normalization, Innovation, and Forms of Social Conflict IV. Six Experiments in Which Consistency Is Considered . . . i. . . Directly as a Source of Influence . V. Summary and Conclusion . . . . . . . . . References . . . . . . . . . . .

. . . . . . .

183 198 199

Overview and Scope . . . . . . . . . . . Definitions . . . . . . . . . . . . . Historical Perspectives . . . . . . . . . . . A Comparison of Research on the Prisoner’s Dilemma and on Reciprocity . . . . . . . . . . . General Conclusions References . . . . . . . . . . . . .

203 206 208 215 228 231

.

150 156 165

A Critical Analysis of Research Utilizing the Prisoner’s Dilemma Paradigm for the Study of Bargaining Charlan Neineth I. 11. 111. IV. V.

Structural Representations of Implicit Personality Theory Seymour Roseliberg and Andrea Sedlak I. 11. 111. IV. V.

Introduction . . . . . . . . . . . Scaling Methods and the Interpretation of Scaling Solutions . Laboratory Analogs of Trait Influence and Personality Description Naturalistic Personality Descriptions . . . . . . Individual Differences . . . . . . . . .

. . . . . . . . . .

235 238 247 259 214

vii

CONTENTS

VI. Personality Perception in Children VII . Some Future Directions . . . . . . . . References

. . .

. . .

. . . . . . . . . . . . . . . . . .

287 291 293

AUTHORINDEX .

. .

. .

. .

299

SUBJECTINDEX.

. .

. .

. .

. .

. .

. .

. . . . . . . .

307

This Page Intentionally Left Blank

CONTRIBUTORS Numbers in parentheses indicate the pages on which the authors’ contributions begin.

DARYL J. BEM,Stanford University, Stanford, California ( 1 ) LEONARD BERKOWITZ, University of Wisconsin, Madison, Wisconsin ( 63) CLAUDEFAUCHEUX,Ecole Pratique des Hautes Etudes, Paris, France (149)

ALBERTJ. L o n , University of Rhode Island, Kingston, Rhode Island (109) BERNICE E. L o n , University of Rho& Island, Kingston, Rhode Island (109) SERGE MOSCOVICI, Ecole Pratique &s Hautes Etudes, Paris, France (149) CHARLAN NEMETH,University of Chicago, Chicago, Illinois (203) SEYMOUR ROSENBERG,Rutgers University, New Brunswick, New Jersey ( 235 ) ANDREASEDLAK,Rutgers University, New Brunswick, New Jersey ( 2 3 5 )

ix

This Page Intentionally Left Blank

CONTENTS OF OTHER VOLUMES

Volume 1 Cultural Influences upon Cognitive Processes Harry C . Triandis The Interaction of Cognitive and Physiological Determinants of Emotional State Stanley Schachter Experimental Studies of Coalition Formation William A. Gamson Communication Networks Marvin E . Shaio A Contingency Model of Leadership Effectiveness Fred E . Fiedler Inducing Resistance to Persuasion: Some Contemporary Approaches William 1. McGuire Social Motivation, Dependency, and Susceptibility to Social Influence Richard H . Walters and Ross D. Parke Sociability and Social Organization in Monkeys and Apes William A. Mason AUTHOR INDEX-SUB

J ECT INDEX

Volume 2 Vicarious Processes: A Case of No-Trial Learning Albert Banclura Selective Exposure Jonathan L. Freedman and David 0 . Sears Group Problem Solving L. Richard Hoffman Situational Factors in Conformity Vernon L. Allen Social Power John Schopler From Acts to Dispositions: The Attribution Process in Person Perception Edward E . Jones and Keith E . Davis xi

xii

CONTENTS OF OTHER VOLUMES

Inequity in Social Exchange J. Stacy A d u m The Concept of Aggressive Drive: Some Additional Considerations Leonard Berkowitz AUTHOR INDEX-SUB

JECT INDEX

Volume 3

Mathematical Models in Social Psychology Robert P . Abelson The Experimental Analysis of Social Performance Michuel Argyle and Adam Kendon A Structural Balance Approach to the Analysis of Communication Effects N . T . Feather Effects of Fear Arousal on Attitude Change: Recent Developments in Theory and Experimental Research Zrving L. Janis Communication Processes and the Properties of Language Serge Moscovici The Congruity Principle Revisited: Studies in the Reduction, Induction, and Generalization of Persuasion Percy H . Tannenbaum AUTHOR INDEX-SUB

JECT INDEX

Volume 4

The Theory of Cognitive Dissonance: A Current Perspective Elliot Aronson Attitudes and Attraction Donn Byrne Sociolinguistics Swan M. Eroin-Tripp Recognition of Emotion Nico H . Frijah Studies of Status Congruence Edward E. Sampson Exploratory Investigations of Empathy Ezra Stotland The Personal Reference Scale: An Approach to Social Judgment Hary S. Upshaw AUTHOR INDEX-SUB

JECT INDEX

CONTENTS OF OTHER VOLUMES

xiii

Volume 5 Media Violence and Aggressive Behavior: A Review of Experimental Research

Richard E . Goranson Studies in Leader Legitimacy, Influence, and Innovation Edwin P . Hollander and James W. fulian Experimental Studies of Negro-White Relationships

Irwin Katz Findings and Theory in the Study of Fear Communications

Howard Leuenthal Perceived Freedom

Ivan D. Steiner Experimental Studies of Families Nancy E . Waxler and Elliot G. Mishler Why Do Groups Make Riskier Decisions than Individuals? Kenneth L. Dion, Robert S . Baron, and Norman Miller AUTHOR INDEX-SUB

JECrr INDEX

This Page Intentionally Left Blank

SELF-PERCEPTION THEORY'

.

Daryl J Bern STANFORD UNIVERSITY STANFORD. CALIFORNIA

I. Introduction . . . . . . . . . . . . A . The Ontogeny of Self-Attributions . . . . . . . B. The Self-Perception Postulates . . . . . . . . I1. Experimental Evidence for the Theory . . . . . . . A . The Cartoon Experiment . . . . . . . . . B. The False Confession Experiment . . . . . . . C. The Pain Perception Experiment . . . . . . . I11. The Reinterpretation of Cognitive Dissonance Phenomena A. The Forced-Compliance Studies: The Psychology of Insufficient . . . . . . . . . . . Justifhation B. The Free Choice Studies . . . . . . . . . C . The Interpersonal Simulations of Cognitive Dissonance Studies . D . The Controversy over the Interpersonal Simulations . . . E . The Possibility of a Crucial Experiment . . . . . . IV. Other Self-perception Phenomena . . . . . . . . A . Misattribution Effects . . . . . . . . . B. The Self-Attribution of Dispositional Properties . . . . C . Overjustification Effects . . . . . . . . . V . Some Differences between Self-perception and Interpersonal Perception VI . The Shift of Paradigm in Social Psychology . . . . . . VII . Some Unsolved Problems . . . . . . . . . A. The Conceptual Status of Noncognitive Response Classes . . B. Do Attributions Mediate Behavior? . . . . . . . C. The Strategy of Functional Analysis . . . . . . References . . . . . . . . . . . .

. .

2 2 4 8

. . . .

9 10

. . . .

12 15

. .

22

. . . . . .

23 27 31 33 33 37 39

.

.

.

. .

.

. .

.

16

40

42 45 45 50 54 57

'Development of self-perception theory was supported primarily by a grant from the National Science Foundation (GS 1452) awarded to the author during his tenure at Carnegie-Mellon University.

1

2

DARYL J. BEM

1. Introduction

Individuals come to “know” their own attitudes, emotions, and other internal states partially by inferring them from observations of their own overt behavior and/ or the circumstances in which this behavior occurs. Thus, to the extent that internal cues are weak, ambiguous, or uninterpretable, the individual is functionally in the same position as an outside observer, an observer who must necessarily rely upon those same external cues to infer the individual‘s inner states. These two propositions constitute the heart of the author’s selfperception theory and, accordingly, the central topic of this review. This article will trace the conceptual antecedents and empirical consequences of these propositions, attempt to place the theory in a slightly enlarged frame of reference, and, hopefully, clarify just what phenomena the theory can and cannot account for in the rapidly growing experimental literature of self-attribution phenomena. Self-perception theory was initially formulated, in part, to address empirically certain questions in the “philosophy of mind” (e.g., Chappell, 1962; Ryle, 1949). When an individual asserts, “I am hungry,” how does he know? Is it an observation? An inference? Direct knowledge? Can he be in error, or is that impossible by definition? How does the evidential basis for such a first-person statement ( or self-attribution) differ from the evidential basis for the third-person attribution, “He is hungry”? Such questions have traditionally been subjected to purely analytic rather than empirical analyses, and psychologists have generally been willing to leave such exercises to the philosophers. In earlier times, when debates about introspection were in vogue, psychologists did discuss such matters, but those discussions, too, were primarily philosophical rather than empirical in tone. In fact, it appears that the only discussion in the literature which treats such questions as substantive psychological problems is B. F. Skinner’s “radical-behavioral” analysis of “private events” and their role in a science of human behavior (Skinner, 1945, 1953, 1957). It was Skinner’s analysis which inspired the present “selfperception theory”; accordingly, it is with Skinner’s analysis that this review begins. A. THEONTOCENY OF SELF-ATTRIBUTIONS In order to identify and label things in his environment, a child must initially have someone else around who will play the “original word game” of pointing and naming, someone who will teach the child to distinguish between objects and events that appear similar and to label

SELF-PERCEPTION THEORY

3

them with different descriptors. The skills of self-description would appear to emerge from the same procedure, both with respect to overt behavior (“I seem to be eating more today”) and to the effects of stimuli on the individual (“It gives me goose pimples”). The problem, of course, arises when the stimuli or events to be described are “private” internal states to which nobody but the individual himself has access, for it then becomes difficult for the verbal community to make differential reinforcement of the appropriate descriptive response directly contingent upon the presence or absence of the stimuli which are to be labeled. What, then, can be done? As Skinner has noted, there are a few cases when an appropriate descriptor can still be acquired without explicit training. For example, some private stimuli are generated from covert behavior which was at one time overt or which accompanied parallel overt behavior to which descriptors could be attached. Thinking sometimes has this property: “I said to myself said I . . .” A second resource which is sometimes available is metaphor or stimulus generalization. An individual, for example, can easily learn to identify “butterflies in the stomach,” and at least one child has generated his own metaphor-by describing a foot which had fallen asleep as “feeling like gingerale when I hold the glass to my face.” Some descriptors of emotional states (e.g., ”feeling blue” or “down in the mouth”) may have similar historical origins. But these special cases cover a very limited domain of self-descriptive statements, and most of the time a child must still be explicitly taught how to describe his internal states in the same way he is taught to describe his outer environment: someone must be able to “point and name.” In training a child to describe pain, for example, an observer, at some point, must teach him the correct response at the critical time when the appropriate private stimuli are impinging upon him. But the observer himself must necessarily identify the “critical time” on the basis of observable stimuli or responses and implicitly assume that the private stimuli are, in fact, accompanying these public events. The description “it hurts” may thus be established in a child’s repertoire by saying: “Don’t cry; I know it hurts to bump your head.” The descriptor itself can then generalize to a large class of private “painful” stimuli even though it was originally an observable response (crying) and a public stimulus variable (bumping the head) which set the occasion for the observer’s inference that the child was experiencing pain. If the resources available to the community for setting up descriptions of private events seem inadequateand it is proposed that this list of resources is exhaustive-it should be recalled that the result is also often inadequate. Not only does the community remain ignorant

4

DARYL J. B E M

of whether the complaint of a headache is valid, for example, but the same deficiencies which generate public mistrust lead the individual himself to faulty self-knowledge. “There appears to be no way in which the individual may sharpen the reference of his own verbal repertoire in this respect [Skinner, 1953, p. 2811.’’ Thus, far from having direct and unerring knowledge of our internal states, Skinner’s analysis implies that we have virtually no knowledge at all until we have been explicitly trained. Internal identifications that we have not been taught remain internal identifications that we cannot make. In everyday life we are spared from confronting our incompetence in this regard only because people know better than to call upon us to make internal discriminations which we are typically not taught (“Is it your spleen or your liver which is causing you the discomfort, Mr. Jones?”). It should be noted that Skinner’s analysis thus reverses the usual practice in psychology of assuming the existence of awareness a priori and then treating any residual unawareness as problematic. For Skinner, awareness, not unawareness, is the phenomenon which normally requires analysis, and the uniqueness of his conceptual contribution resides precisely in his explicit recognition of this fact.

B. THESELF-PERCEPTION POSTULATES It was the above analysis which first suggested that many of the self-descriptive statements which appear to be exclusively under the control of private stimuli may in fact still be partially controlled by the same accompanying public events used by the training community to infer the individual’s inner states in the first place (Bern, 1964, 1965). Meanwhile, the underlying corollary that private stimuli probably play a smaller role in self-description than we have come to believeeither as self-observers ourselves or as psychologists-was already receiving support from the experimental work on emotional states by Schachter and his colleagues ( Schachter, 1964). For example, in their now classic experiment, Schachter and Singer (19f32)manipulated the external cues of the situation and were able to evoke self-descriptions of emotional states as disparate as euphoria and anger from subjects in whom operationally identical states of physiological arousal had been induced. These subjects required the internal stimuli of arousal in order to make the gross discrimination that they were emotional, but the more subtle discrimination of which emotion they were experiencing was under the control of the external cues, the “emotional” behavior of a stooge. In another experiment conducted within Schachter’s theoretical framework, Valins ( 1966) was able to manipulate attitudes toward stimulus pictures of seminude females by

SELF-PERCEPTION THEORY

5

giving his male subjects false auditory feedback which they could interpret as their heartbeat, thus showing that any internal stimulus control of attitude statements could be overridden by external cues, The controlling external cues in these and similar experiments have usually resided in the social or physical setting in which the individual was placed or in verbaI instructions given to him by the experimenter, But that is not the only possible source of such cues. To us as observers, the most important clues to an individual’s inner states are found in his behavior. When we want to know how a person feels, we look to see how he acts. Accordingly, it seemed possible that when an individual himself wants to know how he feels, he may look to see how he acts, a possibility suggested anecdotally by such statements as, “I guess I’m hungrier than I first thought.” It was from this line of reasoning that the first postulate of self-perception theory was derived: Individuals come to “know” their own attitudes, emotions, and other internal states partially by inferring them from observations of their own overt behavior and/ or the circumstances in which this behavior occurs. The second postulate of self-perception theory suggests a partial identity between self- and interpersonal perception: To the extent that internal cues are weak, ambiguous, or uninterpretable, the individual is functionally in the same position as an outside observer, an observer who must necessarily rely upon those same external cues to infer the individual‘s inner states. Because self-perception theory is conceived of as a “behaviorist’s” theory, it is important to emphasize that neither the interpersonal observer nor the individual himself is confined to inferences based upon overt actions only. Social psychologists (e.g., Asch, 1952) have long been critical of behavioral analyses of social interaction (e.g., the MillerDollard analysis of imitation; see Miller & Dollard, 1941) precisely because they feel that there is something more to interpersonal perception than responding to the overt behavior of another individual. In particular, so the criticism goes, behavioral analyses fail to explicate how it is that individuals are able to take account of one another’s meanings, motives, intentions, and the like. This criticism is often illustrated by citing cases in which identical behaviors may have different “meanings,” meanings which observers have no difficulty discerning. For example, an individual might attribute “suddenly aroused determination” to his friend were he to see him chasing a mouse into the room with a raised broom. But he would attribute “fear” if an identical hasty entrance were to be followed rather than preceded by the mouse. Were no mouse present, he might well classify his friend’s action as anger directed at him. In all three cases

6

DARYL J . BEM

the overt behaviors observed are the same, and it is not particularly illuminating simply to say that the individual is responding to the “intent” of his friend or to the “meaning” of the action, since it is precisely the intent and meaning which require explication. In this example, it is clear that the meaning of the action resides in the mouse, that is, the intent or meaning is inferred from the stimulus conditions that appear to be controlling the observed behavior. To a radical behaviorist, this is the “intent” or “meaning” of the behavior. This, then, is the “something more” of interpersonal perception: the ability to respond not only to the overt behavior of others, but to respond as well to the controlling variables of which their behavior appears to be a function. The first postulate of self-perception theory embraces this observation in its proposal that self-attributions are made from the individual’s observations of his “own overt behavior and/or the circumstances in which it occurs,” for most important among those “circumstances” are the apparent controlling variables of that behavior. A less whimsical, more pertinent example of interpersonal perception in which the apparent controlling variables of the behavior can provide a basis of inference is provided by the individual who attempts to infer the actual beliefs or attitudes of a persuasive communicator. Is the communicator being paid? If so, how much? Did the communicator have a free choice in what to say, or was he coerced? To the extent that the communicator appears to be free from the control of these kinds of explicit reinforcement contingencies, that is, to the extent that he does not appear to be “manding” reinforcement (Skinner, 1957), he will be judged to be a credible communicator, and his statements will typically be described as his “actual” beliefs and attitudes. If one now applies the postulates of self-perception theory to this same example, one arrives at the hypothesis that the communicator himself might infer his own beliefs and attitudes from his behavior if that behavior appears to be free from the control of explicit reinforcement contingencies. This hypothesis was, in fact, the first prediction to be derived from self-perception theory (Bem, 1964, 1965), and tentative evidence for its validity appeared to reside already in the forcedcompliance experiments conducted within Festinger’s ( 1957) theory of cognitive dissonance. Consider, for example, the classic experiment by Festinger and Carlsmith ( 1959). In this experiment, 60 undergraduates were randomly assigned to one of three experimental conditions. In the $1 condition, the subject was first required to perform long repetitive laboratory tasks in an individual experimental session. He was then hired by the experimenter as an “assistant” and paid $1to tell a waiting fellow student ( a

SELF-PERCEPTTON THEORY

7

stooge) that the tasks were enjoyable and interesting. In the $20 condition, each subject was hired for $20 to do the same thing. Control subjects simply engaged in the repetitive tasks. After the experiment each subject indicated how much he had enjoyed the tasks. The results show that the subjects paid $1 evaluated the tasks as significantly more enjoyable than did subjects who had been paid $20. Subjects paid $20 did not express attitudes significantly different from those expressed by the control subjects. This phenomenon is now known as the reverseincentive effect. As implied above, self-perception theory interprets such results by considering the viewpoint of an outside observer who hears the individual making favorable statements about the tasks to a fellow student, and who further knows that the individual was paid $1 [$20] to do so. This hypothetical observer is then asked to state the actual attitude of the individual he has heard. If the observer had seen an individual making such statements for little compensation ($l), he can rule out financial incentive as a motivating factor and infer something about the individual’s attitudes. He can use an implicit self-selection rule and ask: “What must this man’s attitude be if he is willing to behave in this fashion in this situation?” Accordingly, he can conclude that the individual holds an attitude consistent with the view that is expressed in the behavior: He must have actually enjoyed the tasks. On the other hand, if an observer sees an individual making such statements for a large compensation (e.g., $20), he can infer little or nothing about the actual attitude of that individual, because such an incentive appears sufficient to evoke the behavior regardless of the individual’s private views. The subject paid $20 is not credible in the sense that his behavior cannot be used as a guide for inferring his private views. The observer’s best guess, then, is to suppose that the individual’s attitude is similar to that which would be expressed by anybody who was selected at random and asked for his opinion-the attitude of a control subject, in other words. Self-perception theory asserts that subjects in the Festinger-Carlsmith experiment ( and other experiments utilizing this paradigm ) are themselves behaving just like these hypothetical observers. They survey their own behavior of making favorable statements about the task and then essentially ask themselves (implicitly) : “What must my attitude be if I am willing to behave in this fashion in this situation?” Accordingly, they produce the same pattern of results as the outside observers: low-compensation subjects infer that they must agree with the arguments in their communication, whereas high-compensation subjects discard their behavior as a relevant guide to their “actual” attitudes and express

8

DARYL J. BEM

the same attitudes as the control subjects. In short, if one places the hypothetical observer and the communicator into the same skin, the findings obtained by Festinger and Carlsmith are the result. The final attitudes of the actual subjects in the experiment are thus viewed as a set of self-attributions made by the individual on the basis of his o w n behavior in the light of the contextual constraints in which this behavior appeared to be occurring. Judges made blind ratings of the persuasive communications delivered by subjects in the Festinger-Carlsmith experiment and found that $1communications were no more persuasive than $20 communications; in fact, the trend was in the other direction. This illustrates again that it is not the behavior per se,but the behavior in conjunction with its apparent controlling variables which provides the crucial information for either interpersonal or self-attributions. The crux of the self-perception interpretation in such studies is that the individual’s own behavior will be used by him as a source of evidence for his beliefs and attitudes to the extent that the contingencies of reinforcement for engaging in the behavior are made more subtle or less discriminable. Monetary inducement is not the only way of manipulating this parameter of self-credibility. For example, several cognitive dissonance studies have manipulated the degree to which the individual appears to have the choice of engaging in the behavior or refusing to do so. Thus if an outside observer sees an individual freely choosing to make opinion statements (e.g., “the tasks were fun and interesting”), he is more likely to take such statements at face value, to infer that they reflect the individual’s true opinions, than if the individual was required or coerced into making such statements. The same analysis applies to the amount of justification given a subject for engaging in the counter-attitudinal behavior: The more justification given, the less likely it is that the behavior will be used by an observer or the individual himself to infer what he believes. When the operations of forced-compliance studies are submitted to a careful analysis in terms of the discriminative stimuli presented to the subject, then those conditions in which outside observers would base an inference of the individual’s attitudes on the overt behavior are found to be the same conditions in which the individual’s own attitudes are affected. 11. Experimental Evidence for the Theory As we shall see in Section 111, several other experiments and paradigms from the cognitive dissonance literature are amenable to self-perception interpretations. But precisely because such experiments

SELF-PERCEPTION THEORY

9

are subject to alternative interpretations, they cannot be used as unequivocal evidence for self-perception theory. In particular, the stimulus operations in such studies have several functional properties. For example, monetary compensation not only manipulates the parameter of self-credibility (according to self-perception theory), but that of incentive and reinforcement as well, a fact which has led to replication failures, replication reversals, and a profusion of interpretations. [For several discussions, see Abelson, Aronson, McGuire, Newcomb, Rosenberg, and Tannenbaum ( 1968, pp. 801-833) .] Similarly, manipulations of justification, choice, commitment, and so forth involve a veritable jungle of complex and ambiguous stimulus operations, a feature of dissonance studies which has been a major target of criticism (e.g., Chapanis & Chapanis, 1964). It thus seemed desirable to design a “self-credibility” experiment in which the controlling stimuli would be “raised from birth” in the laboratory so that they would have no functional properties other than those postulated by the theory to be relevant to the self-credibility of the induced behavior. Secondly, it seemed desirable to eliminate the variance in the overt behavior which arises from permitting the subject to compose his own persuasive communications. Finally, reflecting the author’s Skinnerian bias, it was decided to design experiments for testing selfperception theory in which every subject is his own control and provides a complete replication of the entire design. The cartoon experiment (Bem, 1964, 1965) was the first of these studies. A. THECARTOON EXPERIMENT The dependent variable in the cartoon study was the subjects’ attitudes toward a series of magazine cartoons. The induced behavior which served as the source of self-attribution was the overt statement, “This cartoon is very funny” or, “This cartoon is very unfunny.” The selfcredibility parameter was manipulated with two colored lights whose functional properties were established through the preliminary training procedure described below. The experimental session was disguised as a tape-recording session for the preparation of experimental stimulus materials. Each subject first underwent a training procedure in which he answered simple questions about himself. After each question was asked, a tape recorder was turned on, automatically illuminating a colored light. The subject was instructed to answer the question truthfully whenever the light was amber, Whenever the light was green, he was to make up a false answer to the question and say it aloud into the tape recorder. In this way, the subject learned that he could believe himself whenever he spoke in the presence of the amber light, but could not believe himself in the presence

10

DARYL J. BEM

of the green light, The lights were reversed for half the subjects. After this training procedure, the subject was shown a series of magazine cartoons which he had rated as “neutral,” that is, as neither funny nor unfunny in a previous session. For each cartoon, he was instructed by the experimenter either to say, “This cartoon is very funny,” or, ‘This cartoon is very unfunny.” The tape recorder was turned on just before he made each statement so that one of the colored lights was on while he spoke, even though he was not instructed to attend to the lights in this portion of the session. Sometimes the “truth” light was illuminated, other times the “lie” light. After the subject had made each statement and the recorder (and light) had been turned off, he was asked to indicate his actual attitude toward the cartoon on an attitude scale. An awareness questionnaire was administered at the end of the session. As the self-perception hypothesis predicted, the subjects changed their attitudes significantly more when they made their statements in the presence of the “truth” light than when they made their statements in the presence of the “lie” light. For example, if a subject had said “This is a very funny cartoon” in the presence of the “truth” light, then he subsequently rated this cartoon to be funnier than if he had made the statement in the presence of the “lie” light. As a parallel to the Festinger-Carlsmith dissonance experiment, the “truth light replicated the controlling functional property of the $1 inducement, signifying to the individual that his behavior could be used as a guide to his “true” attitude; the “lie” light abstracted the function of the $20 inducement, signifying to the individual that his behavior was irrelevant to his true attitude. It should also be noted that no subject was aware of any attitude change nor of any effects of the two lights on his subsequent cartoon ratings.

B. THEFALSE CONFESSION EXPERIMENT The cartoon study demonstrated that the self-attributions known as attitude statements could be brought directly under the controI of an individual’s own verbal behavior and the accompanying stimulus conditions in which that behavior occurs. The false confession experiment (Bem, 1966) was an attempt to retain the same independent variables, but to extend the evidence for the theory to a different kind of dependent variable: the recall of prior events. In particular, the false confession study attempted to verify the possibility that a false confession can effectively distort an individual’s recall of his past behavior if the confession is made in the presence of cues previously associated with telling the truth. A second possibility tested was that cues pre-

SELF-PERCEPTION THEORY

11

viously associated with lying can create self-disbelief in true statements, leading again to distortions in recall of the actual behavior. The experiment utilized the same experimental setting as the cartoon experiment, but instead of rating cartoons prior to the experimental session, subjects were given a list of 100 common nouns and an alphabetical list containing 50 of those nouns. Their task was to cross out each word on the master list which also appeared on the alphabetical list. Thus each subject was required to scan the alphabetical list for each of 100 words and then either to cross out the word or to not cross out the word. This was the behavior which the subject was later asked to recall. Each subject then went through the preliminary training session, described above, in which he learned to make true statements whenever the amber recording light was on and false statements whenever the green recording light was on. He was then required to make statements about the 100 nouns. Sometimes he was required to state aloud that he had crossed out a word and sometimes to state that he had not crossed out a word (e.g., “I did not cross out the word tree”). Half the statements he had to make were true, and half were false. Again the colored lights were connected to the tape recorder so that either the amber or the green light was illuminated as he made his “confession.” After each statement was made and the recorder and light turned off, the subject indicated whether he recalled crossing out the word or not crossing it out. In addition, he indicated on a five-point scale how coddent he was in the accuracy of his recall. The complete design for each subject was thus a 2 X 2 X 2 in which the word had either been crossed out or not, the statement was either true or false, and it was made in the presence of either the “truth light or the “lie” light. In addition, each subject was asked to recall his behavior on a set of control words about which no overt statements had been made. The results confirmed the expectations of the self-perception hypothesis: In the presence of the “lie” light, the false confessions had no effect. Subjects were as able to recall their previous behavior just as accurately as they were for the control words not appearing in their confessions. But in the presence of the “truth” light, false confessions did produce significantly more errors of recall and less confidence in recall accuracy than either false confessions in the presence of the “lie light” or no confession at all. In addition, there was weak support for the complementary possibility that more recall errors would be produced by the “lie” light when the overt statement was true than would be produced by the “truth” light-where the light confirms rather than contradicts the validity of the statement.

12

DARYL J. BEM

In a replication of this experiment, Maslach (1971) reconfirmed the major finding that the truth light produced more errors of recall following false confessions, but failed to find that the “lie” light distorted recall for true statements. Instead, she reports that the “truth” light produced more errors of recall in both the false and true confession conditions, although the effect was stronger for false confessions. This opens up the possibility that the subjects are exercising more care in responding after a ‘lie” light statement, an hypothesis supported by auxiliary data. Unfortunately, because Maslach‘s design did not contain a set of control words, it is not possible to know which cells are deviating significantly from a recall baseline and hence which combination of independent variables is producing the effects. Nevertheless, her pattern of results casts a cloud over the false confession experiment as a pure demonstration of a self-perception effect.

C. THEPAINPERCEPTION EXPERIMENT Both the cartoon and false confession studies employed the overt behavior of self-persuasive statements and self-credibility cues as the independent variables. The pain perception study ( Bandler, Madaras, & Bem, 1968) was designed to move beyond this conceptual paradigm and also to make contact with an extensive literature on pain perception, It has long been known that an individual’s perception of pain is only partially a function of the pain-producing stimulus. This is apparent from the wide cultural differences in labeling stimuli as painful (e.g., childbirth; Melzack, 1961), from research on placebo effects (Beecher, 1959, l W ) , and from the phenomena of hypnotic analgesia (Barber, 1959, 1963) and masochism (Brown, 1965). Recent research has also revealed a number of “cognitive” operations which can influence an individual’s inference that a particular stimulus is painful. For example, Nisbett and Schachter (1966) showed that the judged intensity of shock-produced pain and the willingness to tolerate such pain can be manipulated by supplying the individual with an alternative explanation for the physiological arousal he is experiencing. And, working within a cognitive dissonance framework, Zimbardo, Cohen, Weisenberg, Dworkin, and Firestone ( 1969) demonstrated that individuals who have volunteered to continue participation in an experiment using painful electric shocks and who were given little justification for volunteering to do so reported the shocks to be less painful and were physiologically (GSR) less responsive than individuals who were given no choice about continuing or who were given much justification for volunteering to continue. As noted earlier, these parameters of choice and justification can

SELF-PERCEPTION THEORY

13

be embraced within the self-perception framework. If an outside observer sees an individual choosing freely to continue enduring shocks with little justification, the observer might well assume that the shocks are not very painful. If on the other hand, the individual appears to have been given little or no option to endure the shocks, the observer could infer nothing about their painfulness from the individual’s behavior. Again, if one places this hypothetical observer in the same skin as the subject, then the pattern of results displayed in the experiment by Zimbardo et al. (1969) is the result. There is another kind of behavior which an observer can use to judge the painfulness of the shock: Does the individual attempt to escape from the shock or does he endure it? It seems reasonable to suppose that an observer would judge shocks from which the individual tried to escape as more uncomfortable than shocks the individual was willing to endure. Self-perception theory, then, predicts that the individual who observes himself freely choosing to escape a shock should rate it as more painful than a shock he chooses to endure. On the other hand, if he is not permitted to choose the action, but is simply instructed by the experimenter to escape the shock, he should not rate it as any more painful than a control shock. The study by Bandler et al. (1968) attempted to verify this hypothesis. Each subject in the experiment received a series of shocks to his hand. Just after the onset of each shock, one of three colored lights was illuminated on a box in front of him. He was told that if the red light came on, he should press a button which he held in his hand and this would terminate the shock. In order to enhance feelings of choice, he was also told, “However, if the shock is not uncomfortable, you may elect to not depress the button. The choice is up to you.” This condition was labeled the escape condition. In the green-light condition, subjects were instructed to not depress the button (the no-escape condition) unless the “shock is so uncomfortable that you feel you must . . . Again the choice is up to you.” In the yellow-light condition, subjects were told that “we are interested in recording only the time that it takes you to press the button once the yellow light comes on. Therefore, please press the button as soon as the yellow light is illuminated. Your depression of the button m a y or may not turn off the shock.” Following each shock the subject rated the discomfort it produced on a 7-point rating scale. The 30 critical shocks were all of equal intensity, and, if not terminated by the subject, had a duration of 2 seconds. Thus, for 10 shocks paired with the escape light, the subject pressed a button and terminated the shock. For the 10 shocks paired with the no-escape light, the subject did not press the button. (Despite

14

DARYL J. BEM

the implied choice, subjects overwhelmingly complied with the instructions.) And, finally, for the 10 shocks paired with the “reaction-time” trials, the subject pressed the button as soon as he could after the light was illuminated, For five of these trials, pressing the button terminated the shock. For the remaining five reaction-time trials, pressing the button had no effect on the shock. It should be noted that the subject’s overt behavior was the same in the reaction-time condition as it was in the escape condition: He pressed the button when the light was illuminated. But the subject was not given the implied choice of pressing or not pressing and, as the instructions make clear, pressing the button did not necessarily terminate the shock. Thus in the reaction-time condition, the button press should no longer be seen by the subject as a self-determined escape response, and he should not infer his discomfort from it, Discomfort ratings should therefore be significantly lower in the reactiontime condition than in the escape condition. The results supported the self-perception hypothesis. Subjects rated shocks as significantly more uncomfortable when they escaped them than when they endured them, the same inference an outside observer might have drawn. This was true even though the endured shocks were necessarily longer than the escaped shocks. (Within the reaction-time condition, the endured shocks were rated as slightly more uncomfortable than the briefer terminated shocks.) Moreover, the subject had to perceive that he had some choice in the matter. Pushing the button in the reaction-time condition was not used as a guide for inferring the discomfort of the shock. Ratings of shocks in the reaction-time condition were significantly lower than those in the escape condition and were not significantly different from ratings of shocks in the no-escape condition. This experiment has since been replicated by Corah and Boffa (1970) using white noise as the aversive stimulus and a between-subjects design for the choice parameter. That is, half of their subjects were given the implied choice of escaping or not escaping, just as Bandler et al. had done, but the remaining subjects were not given such a choice. The results of Bandler et al. were confirmed. When given choice, subjects rated bursts of noise they escaped as significantly more aversive than bursts of noise they endured. When the choice was omitted, endured noise was slightly more aversive than the escaped noise, a replication of the Bandler et al. finding within the reaction-time condition. A second replication of the Bandler et al. experiment was conducted by Klemp and Leventhal (1972), who controlled for stimulus duration by contrasting the shocks the subject “chose” to escape with the reaction-time shocks he escaped (i.e., without choice) and the shocks he

SELF-PERCEPTION THEORY

15

“chose” to endure with the reaction-time shocks which were not tenninated. These investigators found that the self-perception eff ect-chosen escaped shocks rated more painful than unchosen escaped shocks-held only for subjects who had a high tolerance for shock. An opposite pattern was found for low tolerance subjects: After choosing to escape a shock, they rated it as less uncomfortable than a shock which is escaped without choice. Klemp and Leventhal suggest, among other possibilities, that low-tolerance subjects may be more frightened, i.e., they may show more of the bodily components of fear, and fear might intenslfy the judged discomfort of shock. If being able to choose is critical in reducing this fear (as other research implies), then lower discomfort ratings in the choice conditions would be expected. This finding is also consistent with an earlier finding by Nisbett and Schachter (1966), who actually manipulated fear in a self-attribution experiment involving electric shock. The predicted self-perception effect occurred only in the low-fear condition; there were no significant differences between conditions in the high-fear treatment. It will be recalled that the basic self-perception postulate states that self-perception effects based upon observations of overt behavior occur “to the extent that internal cues are weak, ambiguous, or uninterpretable.” The findings of Klemp and Leventhal and of Nisbett and Schachter would appear to exemplify cases in which the internal stimuli of fear override any information potentially available to the individual from external sources, including overt behavior. 111. The Reinterpretation of Cognitive Dissonance Phenomena

If a person holds two cognitions that are inconsistent with one another, he will experience the pressure of an aversive motivational state called cognitive dissonance, a pressure which he will seek to remooe, among other ways, by altering one of the two “dissonant” cognitions, This proposition is the heart of Festinger‘s theory of cognitive dissonance ( Festinger, 1957), a theory which received more widespread attention from personality and social psychologists during the decade of the Sixties than any other contemporary statement about human behavior. In 1965, it was suggested that self-perception theory might be able to account for the major phenomena of cognitive dissonance theory (Bem, 19sS), a suggestion which was elaborated more systematically in Bem (1967b). The basic theme of that reinterpretation has already been spelled out in Section I, B, with respect to the Festinger and Carlsmith (1959) experiment. Now that the experiments providing independent evidence for the self-perception postulates have been de-

16

DARYL J. BEM

scribed, it is appropriate to return to the conceptual reanalysis of the various dissonance paradigms. A. THEFORCED-COMPLIANCE STUDIES:THEPSYCHOLOGY OF

INSUFFICIENTJUSTIFICATION The Festinger-Carlsmith study exemplifies the experimentaI procedure known as the forced-compliance paradigm. In all of these experiments, an individual is induced to engage in some behavior that would imply his endorsement of a particular set of beliefs or attitudes. Following his behavior, his “actual” attitude or belief is assessed to see if it is a function of the behavior in which he has engaged and of the manipulated stimulus conditions under which it was evoked. Thus in the Festinger-Carlsmith experiment, subjects were induced for either $1or $20 to tell a waiting fellow student that the repetitive tasks were fun and interesting, and then their own attitudes were assessed. As noted earlier, the low-compensation subjects expressed significantly more favorable attitudes toward the tasks than the high-compensation subjects or control subjects, the reverse-incentive effect. Dissonance theory interprets these results by noting that all subjects initially hold the cognition that the tasks are dull and boring. In addition, however, the experimental subjects have the cognition that they have expressed favorable attitudes toward the tasks to a fellow student. These two cognitions are dissonant for subjects in the $1 condition because their overt behavior does not “follow from” their cognition about the task, nor does it follow from the small compensation they are receiving. To reduce the resulting dissonance pressure, they change their cognition about the task so that it is consistent with their overt behavior: they become more favorable toward the tasks. The subjects in the $20 condition, however, experience little or no dissonance because engaging in such behavior “follows from” the large compensation they are receiving. Hence, their final attitude ratings do not differ from those of the control group, It is the motivational force provided by the drive toward consistency or dissonance reduction which changes the attitudes of the lowcompensation subjects. As noted earlier, self-perception theory considers the subject in such an experiment as simply an observer of his own behavior. just as an outside observer might ask himself, “What must this man’s attitude be if he is willing to behave in this fashion in this situation?” so too, the subject implicitly asks himself, “What must my attitude be if I am willing to behave in this fashion in this situation?” Thus the subject who receives $1 discards the monetary inducement as the major motivating factor for his behavior and infers that it must reflect his actual

SELF-PERCEPTION THEORY

17

attitude; he infers that he must have actually enjoyed the tasks. The subject who receives $20 notes that his behavior is adequately accounted for by the monetary inducement, and hence he cannot extract from the behavior any information relevant to his actual opinions; he is in the same situation as a control subject insofar as information about his attitude is concerned. Thus in the self-perception explanation, there is no aversive motivational pressure postulated. The dependent variable is viewed simply as a self-attribution based on the available evidence, which includes the overt behavior of the communication and the apparent controlling variables of that behavior. A similar analysis can be applied to other experiments conducted within the forced-compliance paradigm. The advantages of the self-perception explanation emerge more clearly in forced-compliance experiments in which dissonance theory does not provide an applicable account of the findings. Three such experiments can be mentioned here.

I. Pro-Attitudiml Advocacy Kiesler, Nisbett, and Zanna ( 1969) conducted a forced-compliance study specifically designed to rule out the dissonance explanation by inducing subjects into committing themselves to argue a position with which they initially agreed. Such a situation presumably does not arouse the “dissonance” produced by counterattitudinal advocacy. In addition, Kiesler et al. manipulated the “meaning” of the behavior by having a confederate subject express his willingness to argue the issue assigned to him because ( a ) he believed strongly in it (belief-relevant condition), or ( b ) the experiment was scientifically valuable (belief-irrelevant condition). The confederate’s issue was not the same one assigned to the actual subject. The results produced a self-perception effect: Subjects in the belief-relevant condition were found to be more favorable to the position they were to argue than either belief-irrelevant subjects or control subjects who were not committed to the behavior, 2. Rejection of Alternative Action A second study in which subjects were to give a speech advocating a position they already endorsed was conducted by Zanna (1970). This study was specifically designed to show that the individual’s inferences about his beliefs may be based not only on acts he performs but also on alternative acts he rejects. After committing themselves to giving a speech which advocated a moderate position on an issue, subjects were given the opportunity to alter the speech so that it advocated a position more extreme, an opportunity which was arranged so that all subjects

18

DARYL J. BEM

would nevertheless choose to reject it. Control subjects were not given the opportunity to alter the speech. A belief-relevance manipulation similar to that in the Kiesler et al. (1969) study was introduced for all subjects. The results supported the self-perception analysis: Subjects who had observed themselves decline a chance to make their speech more extreme attributed to themselves an attitude toward the issue less extreme than the self-attributions made by control subjects. In further support of the self-perception postulate that such self-attributions are like interpersonal attributions, Zanna also conducted a successful parallel experiment which showed that observers generate the same pattern of results as the actual subjects. This kind of study is now known as an interpersonal simulation (Bem, 1968c), and will be discussed at length in Section 111, C. Harvey and Mills ( 1971) conducted an experiment conceptually similar to the Zanna study in which subjects were given the opportunity to substitute a different speech for the one they had initially given. As in the Zanna study, the opportunity was offered in such a way that all subjects rejected it. Again the results showed that subjects who were given the opportunity and rejected it expressed final attitudes more in line with the speech they had given than subjects who were not given that opportunity. However, because the speech in this study was counterattitudinal, both dissonance theory and self-perception theory can predict this result (notwithstanding the contention of Harvey and Mills that this study favors dissonance theory over self-perception theory). A dissonance-theoretic prediction that attitude change would be greater if the subject’s initial attitude were made more salient for him was not confirmed; in fact, the differences were in the opposite direction. As we shall see in Section 111, D and E, self-perception theory suggests that making an initial attitude more salient should diminish the degree to which selfattributions can be altered by observations of overt behavior.

3. The Effectsof Repeated Advocacy Cohen (Brehm & Cohen, 1962, pp. 97-104) conducted an experiment to see what the effects of repeated “dissonances” might be on attitude change. Although dissonance theory might predict that dissonance pressure would build up, yielding greater final attitude change, Brehm and Cohen note that it is also possible for a “tolerance for dissonance” to build up at the same time. Thus they ventured no specific predictions. In this study, subjects in the predissonance conditions were required to write four separate essays against positions they currently held; subjects in the preconsonance conditions wrote four essays in favor of positions they currently held. All subjects were then induced to write a fifth essay which

SELF-PERCEPTION “ImORY

19

was counter-attitudinal. The fifth essay was preceded by a manipulation of justification. Half of the subjects were given little justification for writing the fifth essay; half were given a great deal of justification. Following the writing of the final essay, attitudes were assessed on that one issue. The pattern of results did not lend itself to any simple dissonance or “tolerance” of dissonance interpretation. Within the predissonance conditions, there was no attitude change at all. But within the preconsonance conditions, the usual effect of justification was found: significantly greater attitude change when little justification was given than when a great deal of justification was given. Another way of stating the results is that attitude change occurred only in the preconsonance-low justification condition. The self-perception analysis of this pattern of findings is straightforward. In fact, the experiment is remarkably like the cartoon and false confession experiments described in Section 11. Each subject receives pretraining which informs him either that he can believe what he says ( preconsonance condition) or that he cannot ( predissonance condition). Thus preconsonance subjects arrive at the fifth essay with their selfcredibility intact, a credibility which can be maintained (low justification) or destroyed (high justification) for the crucial fifth essay. The predissonance subjects arrive at the fifth essay with their self-credibility already destroyed, and hence further manipulations have no effect, and no attitude change is observed.

4. Combining Sources of Credibility Information The experiment on repeated “dissonances” makes explicit an underlying assumption of self-perception theory concerning the way in which different sources of information combine. Self-perception theory states that if external contingencies seem sufficient to account for the behavior, then the individual will not be led into using the behavior as a source of evidence for his self-attributions. This suggests that if any elements of the situation imply that the behavior is not credible or is not otherwise relevant to his private views, then the presence of other cues implying credibility does not help. Thus, if either predissonance training or high justification was present in the repeated dissonance study, the behavior was not used as the basis for self-attribution. Another way of saying this is that the sources of self-credibility combine multiplicatively: If any source implies irrelevance or low credibility, then self-attributions will not occur. More direct support for this combination rule is provided in an experiment by Linder and Jones (1969) which is a clever hybrid between a

20

DARYL J. BEM

self-perception study and a dissonance study. Their subjects were trained to make true and false statements in the presence of two colored lights following exactly the procedures used in the cartoon and false-confession studies. Thus subjects learned that they could believe what they said in the presence of €he “truth” light, and that they could not believe themselves in the presence of the “lie” light. All subjects then had to read counter-attitudinal essays in the presence of one of the two lights. Half of the subjects were permitted a “free” choice whether or not to comply; the remaining subjects were given no choice. Final attitudes were then assessed. The results show that significantly greater agreement with the position advocated in the essay appeared only when essays were read under free choice conditions in the presence of the “truth” light. In other words, if either the ‘lie” light was on or if the subject had no choice, selfattributions based upon the behavior did not occur. The rule for combining conflicting sources of credibility information appears to be multiplicative.

5. The Forbidden Toy Studies One variation of the forced-compliance paradigm which has been employed with increasing frequency is the forbidden toy procedure. In this situation, children are asked not to pIay with an attractive toy and are given either a mild or a severe threat of punishment for disobeying. Following a period in which all the children do resist the temptation, children under mild threat have been shown to devalue the prohibited toy ( Aronson & Carlsmith, 1963). Unlike the studies of counter-attitudinal advocacy, which frequently fail to replicate, the forbidden toy finding has proved extremely reliable, producing both verbal derogation of the previously forbidden toy ( Aronson & Carlsmith, 1963; Carlsmith, Ebbesen, Lepper, Zanna, Joncas, & Abelson, 1969; Lepper, Zanna, & Abelson, 1970; Ostfeld & Katz, 1969; Turner & Wright, 1965) and actual behavioral avoidance of the toy up to 6 weeks after the initial session ( Freedman, 1965; Pepitone, McCauley, & Hammond, 1967). The dissonance theory account of this effect is that a severe threat provides high justification for the child not to play with the toy, rendering the temptation period nondissonant; the mild threat, however, is insufficient to justify his not playing with the toy, producing dissonance which can be resolved by convincing himself he really doesn’t like the toy. The self-perception account is also straightforward. The child, when asked his toy preferences after the temptation period reviews his behavior toward the toy and the apparent controlling variables of his avoiding the toy. If he has refrained from playing with the toy under severe threat, he can still infer that he may like the toy, but if he has

SELF-PERCEPTION THEORY

21

refrained under mild threat, then he could conclude that he must not like the toy. Again, it is as if subjects use the implicit self-selection rule, ‘What must my attitude be if I am willing to behave in this fashion in this situation.” Both theories can also handle certain variations of the procedure. For example, Lepper et al. (1970) show that if the child is told prior to the temptation period that other children did not play with the toy when asked, then mild-threat subjects no longer devalue the toy. Within the dissonance theory framework, this consensual information gives the child a “consonant” reason for not playing with the toy, and hence no dissonance is aroused. Similarly, in self-perception theory, the information prevents the subject from using the implicit self-selection rule; because everyone behaves the same way, the child cannot use such normative behavior to infer something unique about his attitudes from it. [Kelley (1967) has pointed out the importance to the self-attribution process of having subjects believe their behavior is distinctive if it is to be used as the basis of inference.] But both theories require added assumptions to account for the second finding of Lepper et al.: Mild-threat subjects still do devalue the toy if the consensual information is given after rather than before the temptation period. This implies that the attitudinal attribution takes place during the temptation period and is irreversible. Dissonance theory does not commit itself with regard to when dissonance reduction takes place, but the implication that it is a continuing process which begins immediately and which, once accomplished, is resistant to reversal is consistent with the spirit of the theory. In self-perception theory, however, it has always been implicitly assumed that the attribution probably occurs when the experimenter asks the individual for his final opinion, at which time he reviews the immediate past for the answer. The Lepper et al. experiment implies that the child seeks an inferential account for his behavior earlier, during the temptation period. Something prompts him to ask, “How come I’m not playing with this toy?” But even if this assumption is added to the self-perception account, there is still nothing within the theory which would predict that the child’s attribution should resist change when new information comes in (e.g., the consensual information). In this instance, then, self-perception theory requires more “patching up” to account for the results than does dissonance theory. Interestingly, however, the forbidden toy paradigm has simultaneously become the vehicle for demonstrating two nondissonance, self-perception phenomena (Lepper, 1971). These will be discussed in detail in Sections IV, A and B. These, then, are the major phenomena which have emerged from the

22

DARYL J. BEM

forced-compliance paradigm of cognitive dissonance theory. As we shall see in Section VII, there are hidden intricacies in these phenomena which outrun all current theories, including the present analysis. At this juncture, however, this section may be taken to constitute the psychology of insufficient justification as seen from the perspective of self-perception theory.

B. THEFREECHOICESTUDIES After the forced-compliance paradigm, it is the free choice paradigm which has received the most attention within dissonance theory (Brehm & Cohen, 1962). In these studies, a subject is permitted to make a selection from a set of objects or courses of actions. The dependent variable is his subsequent attitude rating of the chosen and rejected alternatives. Dissonance theory reasons that any unfavorable aspects of the chosen alternative and any favorable aspects of the rejected alternatives provide cognitions that are dissonant with the cognition that the individual has chosen as he did. To reduce the resulting dissonance pressure, the individual exaggerates the favorable features of the chosen alternative and plays down its unfavorable aspects. This leads him to enhance his rating of the chosen alternative. Similar reasoning predicts that he will lower his rating of the rejected alternatives. These predictions are confirmed in a number of studies (see Brehm & Cohen, 1962, p. 303; Festinger, 1964). A number of secondary predictions concerning parameters of the choice have also been confirmed. In an experiment by Brehm and Cohen (1959),school children were permitted to select a toy from either two or four alternatives. Some children chose from qualitatively similar toys; others chose from qualitatively dissimilar alternatives. The children’s postchoice ratings of the toys on a set of rating scales were then compared to initial ratings obtained a week before the experiment. The main displacement effect appeared as predicted: Chosen toys were displaced in the more favorable direction; rejected toys were generally displaced in the unfavorable direction. In addition, however, the displacement effect was larger when the choice was made from the larger number of alternatives. This is so, according to dissonance theory, because “the greater the number of alternatives from which one must choose, the more one must give up and consequently the greater the magnitude of dissonance [Brehm & Cohen, 1959, p. 3731.” similarly, the displacement effect was larger when the choice was made from dissimilar rather than similar alternatives because “what one has to give up relative to what one gains increases [p. 3731,” again increasing the magnitude of the dissonance experienced.

SELF-PERCEPTION THEORY

23

Again, self-perception approaches the phenomenon by considering an observer trying to estimate a child’s ratings of toys; the observer has not seen the child engage in any behavior with the toys. Now compare this observer with one who has just seen the child select one of the toys as a gift for himself. This comparison parallels, respectively, the prechoice and the postchoice ratings made by the children themselves. It seems likely that the latter observer would displace the estimated ratings of the chosen and rejected alternatives further from one another simply because he has some behavioral evidence upon which to base differential ratings of these toys. This is the effect displayed in the children’s &a1 ratings. The positive relation between the number of alternatives and the displacement can be similarly analyzed. If an observer had seen the selected toy “win out” over more competing alternatives, it seems reasonable that he might increase the estimated displacement between the “exceptional” toy and the group of rejected alternatives. Finally, the fact that the displacement effect is larger when the alternatives are dissimilar would appear to be an instance of simple stimulus generalization. That is, to the extent that the chosen and rejected alternatives are similar to one another, they will be rated closer together on a scale by any rater, outside observer, or the child himself. In sum, if one regards the children as observers of their own choice behavior and their subsequent ratings as inferences from that behavior, the dissonance findings appear to follow. C. THE INTERPERSONAL SIMULATIONSOF COGNITIVE DISSONANCE STUDIES As we have seen, the reinterpretation of dissonance studies proceeds by utilizing the second postulate of self-perception theory: To the extent that internal cues are weak, ambiguous, or uninterpretable, the individual is functionally in the same position as an outside observer. In order to bolster the purely conceptual use of this postulate in the reinterpretation of the dissonance studies, an experimental methodology was devised for testing this notion more experimentally. The resulting experiments within this methodology have now become known as interpersonal simulations (Bem, 1965, 1967n, 1967b, 1967c, 1968a, 1968b, 1968~;Bem & McConnell, 1970). In these experiments, an observersubject is either given a description of one of the conditions of a dissonance experiment or actually permitted to observe one of these conditions and then asked to estimate the attitude of the subject whose behavior is either described or observed. The prediction is that such observer-subjects should be able to reproduce the patterns of results

24

DARYL J. BEM

generated by actual subjects in the original experiments. It should be noted that observer-subjects are not asked to play amateur psychologist and predict the results of the experiment; rather, each observer-subject attempts to infer the attitude of a single “other.” In the simulation, he “stands in” for the actual subject. For example, the free choice study described in the prior section was simulated by giving college students a sheet of paper which informed them that an ll-year-old boy in a psychology experiment was permitted to select one of several toys to keep for himself (Bem, 1967b). The sheet informed the observer-subject which toy the child had chosen and from which alternatives he was permitted to choose. Each observersubject then estimated the child’s ratings of the chosen and rejected toys. These observer-subjects were randomly assigned to one of four conditions corresponding to the combinations of number of alternatives (two or four) and similarity of alternatives (similar or dissimilar), The toys listed were selected from the list reported in the original experiment (Brehm & Cohen, 1959) and were rated on the same rating scales. A group of control observer-subjects simply estimated toy ratings of a typical ll-year-old boy. The results showed that observer-subjects not only replicated the main displacement effect (chosen toy enhanced, rejected toy devalued), but also replicated the effects of the number of toys ( displacement effect enhanced with more rejected alternatives) and the effects of the similarity of the toys (displacement effect diminished with similar alternatives). A somewhat more elaborate simulation was conducted for the Festinger and Carlsmith (1959) study in which, it will be recalled, subjects who had received $1for telling a stooge that a series of tasks were fun and interesting subsequently gave the tasks higher ratings than either subjects paid $20 or control subjects. Seventy-five college undergraduates participated in an experiment designed to “determine how accurately people can judge another person.” Twenty-five subjects each served in a $1, a $20, or a control condition. All subjects listened to a tape recording which described a college sophomore named Bob Downing, who had participated in an experiment involving two motor tasks. The tasks were described in detail, but nonevaluatively; the alleged purpose of the experiment was also described. At this point, the control subjects were asked to evaluate Bob’s attitudes toward the tasks. The experimental subjects were further told that Bob had accepted an offer of $1 ($20) to go into the waiting room, tell the next subject that the tasks were fun, and to be prepared to do this again in the future if they needed him. The subjects then listened to a brief conversation which they were told was an actual recording of Bob and the girl who was in the waiting room. Bob was heard to argue rather imagi-

SELF-PERCEPTION THEORY

25

natively that the tasks were fun and enjoyable, while the girl responded very little except for the comments that Festinger and Carlsmith‘s stooge was instructed to make. The recorded conversation was identical for both experimental conditions in order to remain true to the original study in which no differences in persuasiveness were found between the $1 and the $20 communications. In sum, the situation attempted to duplicate on tape the situation actually experienced by Festinger and Carlsmith’s subjects. All subjects estimated Bob‘s responses to the same set of questions employed in the original study. The results of this simulation showed that interpersonal observers were able to replicate the reverse-incentive effect. Observers of $1 subjects estimated “Bob‘s” attitude to be significantly more favorable toward the tasks than did observers of either $20 subjects or control subjects. Observers of $20 subjects did not differ in their estimates of Bob‘s attitude from observers of control subjects. An extended version of this same simulation (Bem, 1967b) replicated an interaction effect between the monetary compensation and the length of the communication found in the original study. Similar simulations conducted by the author (Bern, 1965) have replicated results from a forced-compliance experiment which utilized small compensations of 504 and $1 (Brehm & Cohen, 1962, p. 73) and results from an experiment on hunger ratings by Brehm and Crocker (Brehm & Cohen, 1962, pp. 133-136). A particularly persuasive simulation was performed by Alexander and Knight (1971) in order to replicate a complex pattern of results found in a forced-compliance study by Carlsmith, Collins, and Helmreich (1966). In their experiment, Carlsmith et al. had shown that one obtains the classical reverse-incentive effect when the subject actually has a faceto-face encounter with the waiting stooge, but that an incentive effect emerges (more money, more attitude change) if the subject simply writes an anonymous essay stating that the tasks were fun and interesting. Alexander and Knight were able to simulate this interaction effect between the monetary inducement and the mode of counter-attitudinal behavior by modifying and extending Bem’s use of tapes about “Bob Downing.”? There have also been several simulations which have failed to

’The successful interpersonal simulation of the Carlsmith et al. ( 1966) findings is actually the least important aspect of the Alexander-Knight article, and it trivializes their contribution to present it in this context. The simulation was actually the first step toward demonstrating their theory that the results of many experiments can be predicted from a knowledge of the most socially desirable response in the situation. Their interpretation of the forced-compliance studies in particular is an alternative to both the dissonance and self-perception analyses. For both methodological and theoretical reasons, the Alexander-Knight article is an important document for social psychologists.

26

DARYL J. BEM

replicate the original studies. Some of these failures have come from simulations specifically modified in ways designed to disconfirm the selfperception analysis (e.g., R. A. Jones, Linder, Kiesler, Zanna, & Brehm, 1988; Piliavin, Piliavin, Loewenton, McCauley, & Hammond, 1969). For reasons to be discussed below, such failures are not, in fact, informative with respect to the validity of the theory. Several other failures are also not informative in this regard because they yield only unsystematic noise rather than systematic differences between observer-subjects and actual involved subjects. (Most of these appear in unpublished manuscripts sent to the author over the years.) The major reason such failures are uninformative is the unreliability of the dissonance phenomena themselves, particularly the reverse-incentive effect, the design most commonly simulated. Indeed, one has learned to fear that an experimenter’s unintended belch might destroy or reverse the effect (for discussions, see Abelson et al., 1968, pp. 801-833). It is only semifacetious to suggest that serious doubts about the self-perception analysis could be raised if the simulations were too reliable or more robust than the phenomena themselves, that is, if the simulations were not also sensitive to unknown and uncontrolled parameters. For example, Harris and Tamler (1971a) have shown that one of the simulations fails if it is conducted the way most of the theory’s critics have conducted their simulations, but it succeeds if the observer knows that the assessment of final attitudes was made by an experimenter different from the one who manipulated the independent variable, hardly a predictable contingency. Such variability of outcome suggests that simulations should probably always be conducted directly in conjunction with the experiment being simulated, either simultaneously or with stimulus materials and consequent behaviors actually generated from the experiment itself. Only a few simulations have been conducted in this way. For example, Harris and Tamler (1971a, 1971b) actually conducted the dissonance experiment which separated the attitude assessment from the initial experimental setting and confirmed that the dissonance effect could be obtained in this way. Materials for their successful simulation were then carefully patterned after the experiment itself. As noted above, they demonstrated that the simulation failed if this single item of information was omitted. A second example is provided by Zanna (1970) who conducted both the original experiment reported in Section 111, A, 2, on the rejection of alternative actions as a source of self-attribution and an associated interpersonal simulation, The simulation successfully replicated his finding that rejection of the opportunity to make a more extreme speech results

SELF-PERCEPTION THEORY

27

in the attribution that the subject is more moderate. He also found, as predicted, that observer-subjects who were told that a subject rejected the opportunity to make the speech more moderate judged him to be more extreme than did control observer-subjects, a prediction that was not actually tested in the original experiment because of time limitations. R. G . Jones (1966) has reported a study in which observer-subjects listened to tapes of the original sessions. Again observers’ attributions replicated the inverse functional relation between monetary compensation and final attitude found with involved subjects. The strategy of conducting both the original experiment and simulation simultaneously pays off even when the simulation fails. For example, in one study of this type, Wolosin (1969) uncovered some important systematic differences between observers and actors in their perceptions of the actor’s freedom to comply with the experimenter’s request. As we shall see in Section V, such a finding has important implications well beyond the controversy over the simulations, and Wolosin has wisely decided to follow up this auxiliary finding (Wolosin, 1971; Wolosin & Denner, 1970). At the same time, however, it should be noted that even the care and rigor of Wolosin’s study did not guarantee that negative results on the simulation would provide an unequivocal disconfirmation of self-perception theory, because none of the main effects or predicted interactions on the major dependent variables (ratings of thirst, willingness to undergo water deprivation, and actual water consumption) were significant for the involved subjects themselves. Thus most-albeit not all-of the comparisons that Wolosin hoped to make between observers and involved subjects required him to argue that the columns of random noise generated by two types of observers do not fluctuate in harmony with the column of random noise generated by involved subjects.

D.THECONTROVERSY OVER

THE

INTERPERSONAL SIMULATIONS

Although there have been some simulations which have failed to reproduce the pattern of results found in the original dissonance experiments, the controversy surrounding the simuIation methodology has focused on the simulations which are successful. In particular, the controversy has centered around the information that the observer-subject ought or ought not to be given concerning the original situation (Bem, 1967a, 1968c; Elms, 1967; R. A. Jones et al., 1968; Mills, 1967; Piliavin et al., 1969). Most of the critics have objected specifically to the fact that observer-subjects are not told the original subject’s premanipulation attitude. The conceptual reason for this deliberate omission can be understood by considering some of the epistemological features of the simulation methodology.

28

DARYL J. BEM

Self-perception theory asserts that an individual’s attitude statements and an observer’s judgments about the individual’s attitudes are “output statements” from the same internal “program.” Both the individual and observer are assumed to use a self-selection rule: ‘What must my [this man’s] attitude be if I am [he is] willing to behave in this fashion in this situation?” To test this isomorphism, we run a simulation of a selfjudgment situation, the dissonance experiment, but instead of writing our own program, we plug in an interpersonal judgment program that the culture has written for us. This program is embodied in our interpersonal observer, who “stands in” for the original subject. But before we can actually run such a simulation, we must first abstract the relevant “input statements” from the situation being simulated: we must decide how to describe the situation to the observer. This requires some theoretically guided assumptions. For example, if the dissonance experiment subject actually arrives at his final attitude by using the self-selection rule, as the theory implies, then it follows that any conflicting initial attitude he may have had prior to the experiment must no longer be very salient for him. That is, the self-perception analysis implies that the data of his incoming behavior “update” his information on his attitude, replacing any prior information to the contrary. Insofar as the individual himself is concerned, his postmanipulation attitude is, in fact, the same attitude which motivated him to comply in the first place; phenomenologically, there is no attitude “change” as such. If an interpersonal simulation is to constitute a valid test of the isomorphism between the subject and an observer, then the theory dictates that a conflicting “initial” attitude of the original subject must not be part of the “input” description for the observer any more than it is for the subject himself. The observer, too, is postulated to be using the self-selection rule to infer the original subject’s postmanipulation attitude. It should be noted that this set of assumptions about what input information an observer-subject in the simulations should receive is selfcorrecting. If the wrong input statements are selected, then the simulation will not succeed in producing output statements which match the output of the original experiment. Thus, R. A. Jones et al. (1988) reconfirmed that the simulations produce the “dissonance effect” outputs when the inputs dictated by the self-perception model are employed, but they found that the simulations fail when a conflicting “initial” attitude is introduced into the description given to the observer-subject. After an intensive analysis of observer-subject’s inferential processes under several variations of the simulations, Piliavin et al. (1969) reported that when the “Bem” inputs are utilized, observer-subjects do, in fact, utilize the self-selection rule and replicate his results. But when additional infor-

29

SELF-PERCEPTION THEORY

mation, including a reference to the subject’s initial attitude, is introduced into the description, the observer-subjects become amateur psychologists and revert to hypotheses about attitude change. They are no longer stand-ins for the original subjects and, accordingly, they fail to reproduce the dissonance effects. Such results underscore the importance of conducting simulations directly in conjunction with the experiment being simulated. For example, in the combined experiment and simulations of Harris and Tamler (1971a, 1971b) mentioned above, these investigators first verified that they could obtain the reverse-incentive effect even if the subject’s initial attitude were made salient for him just prior to the forced-compliance manipulation (reinstatement condition) and if the final attitude was assessed by a different experimenter in a “different experiment” ( separation condition). They then conducted a 2 X 2 X 2 simulation in which two levels of monetary incentive were crossed with the reinstatement us. nonreinstatement variable and the separation us. nonseparation variable. Under nonseparation conditions [which parallel the R. A. Jones et al. (1968) and Piliavin et al. (1969) designs], these investigators obtained an incentive effect when reinstatement was employed, and a reverse-incentive effect when it was not. But under separation conditions, both reinstatement and nonreinstatement conditions generated the reverse-incentive effect predicted by self-perception theory. When the actual experiment was simulated exactly, the simulation worked. It is important to note just what a successful simulation means. It implies the same thing that a successful computer simulation implies, namely, that the process model embodied in the “program” is functionally equivalent to the process being simulated, and, further, that the selection of the input statements was not in error. The simulation becomes a plausibility demonstration, a sufficiency test: The process model embodied in the program is sufficient-but not necessarily “true” or unique-for generating the output statements observed in the situation being modeled by the simulation. But there are weaknesses in this methodology when it is applied to this context. Abelson (1968) has noted some of the validation problems connected with the simulation methodology in general, and his discussion of the “degrees of freedom” problem is particularly relevant to interpersonal simulations. Abelson stated the problem this way: If a simulation could be “right for the wrong reasons,” that is, fit the data by virtue of compensating errors, then in what sense can a good fit be regarded as support for the theory underlying the simulation model? . Most cognitive simulations are so rich in qualitative detail that it is very easy for them to fail . Because it is so hard to obtain good data fits, anything which conies close is impressive, and any

. .

. .

30

DARYL J. B E M

cognitive model yielding an apparently perfect fit to a wide range of data would indeed deserve serious theoretical recognition. With social simulations, however, the issue is probably more cogent. If the outcome variables of the model are few while the number of parameters to be juggled is great, there can always be the lingering suspicion that a good fit was too easy to achieve and thus not strongly supportive of the model [pp. 343-3441,

This, then, is the main reason why the interpersonal simulations provide only weak support for the self-perception theory. There are too many “degrees of freedom,” input variables that can be “juggled” (like the initial attitude), while the complexity of the output predictions rarely exceeds a prediction about the ordering of two or three means. Abelson suggests some of the paths opcn for strengthening simulation arguments, however. The most obvious remedy, of course, is “to design the simulation so as to generate as large a number of outcome variables as possible. The more outcomes that can be validated, the merrier-and the more convincing the underlying theory [Abelson, 1968, p. 3441.” This path toward strengthening the self-perception theory of dissonance phenomena was followed in three interpersonal simulations (discussed in Section 111, C ) which not only replicated main effects of dissonance experiments but reproduced either an interaction effect which dissonance theoiy itself had not anticipated or the secondary effects of additional parameters in the experiments (Alexander & Knight, 1971; Bem, 196%). Although these extended simulations do not go very far toward eliminating the “degrees of freedom” problem, they are illustrative of the method. A second remedy is “to show, if possible, that the fit was not so easy by changing the model in various ways and demonstrating consistent lack of fit [Abelson, 1968, p. 3431.” In effect, this is what some of the critics have done. They have altered some of the input assumptions of the model and demonstrated that the simulations fail (R. A. Jones et al., 1968; Piliavin et al., 1969). As Abelson has remarked: Ironically, what [Bern’s] detractors should now really be doing if they must still simulate is to replicate [his] outcome with clearly bad descriptions to the observer, rather than to reverse [his] outcome with purportedly good descriptions [R. P. Abelson, personal communication, January 22, 19691.

The third and most important way of strengthening simulation arguments which suffer from the “degrees of freedom” problem, however, is to return to the original situation and demonstrate that the assumed isomorphism between the inputs of the original situation and the inputs of the simulation actually exists. This is what Harris and Tamler (1971a, 1971b) have done. This too, is what was done in an experiment by Ben1 and McConnell (1970) which demonstrated that

SELF-PERCEPTION THEORY

31

subjects in dissonance experiments cannot, in fact, recall their initial attitudes at the time of the final attitude assessment, that they see their postmanipulation attitudes as the same attitudes which motivated them to comply in the first place, and that they do not experience any attitude change phenomenologically. Hence, initial attitudes are not salient for involved subjects, and should thus not be made salient for observers in the simulations. This process of moving back and forth between the simulation and the actual situation is precisely the one which cognitive theorists have attempted to follow and, in fact, it is this interaction between simulation and direct experimentation which constitutes the heuristic utility of the simulation methodology. A simulation reveals an underlying assumption or implication of the model which was not originally observed or even anticipated. The theorist can then return to the original situation armed with a new hypothesis. Thus, the hypothesis that subjects in dissonance experiments would perceive their postexperimental attitudes to be identical to the pre-experimental attitudes came to light through the debate between the author and his critics over the simulations. Even though the hypothesis was logically implied by the original analysis, it remained unarticulated until the “countersimulations” of R. A. Jones et uZ. (1968) raised it explicitly. It is important to note, however, that once one of the isomorphisms is questioned, the issue can be resolved only by returning to the original situation. A simulation will not suffice: “No ‘as if‘ methodology, including the technique of interpersonal simulation, is an adequate substitute for the intensive study of the actual situation being modeled [Bem, 1968c, p. 2731.”

E. THE POSSIBILITY OF A CRUCIAL EXPERIMENT At the end of their experiment, Bem and McConnell (1970) state that “If the past history of controversies like this is any guide, it seems unlikely that a ‘crucial’ experiment for discriminating between [dissonance theory and self-perception theory] will ever be executed. At this juncture each theory appears capable of claiming some territory not claimed by the other, and one’s choice of theory in areas of overlap is diminishing to a matter of loyalty or aesthetics (p. 30).” Admittedly, this was partially a ploy to bring the whole thing to a halt. But it didn’t work, for the controversy over the salience of initial attitudes itself suggested a possible “crucial” test to Snyder and Ebbesen. In the typical forced-compliance experiment, the subject’s initial attitude is in conflict with the induced behavior. What should happen, Snyder and Ebbesen wondered, if the initial attitude is made more salient to the subject? Self-perception theory predicts that this will

32

DARYL J. BEM

diminish the degree to which the final attitude attribution will be based upon the induced behavior and hence will diminish the amount of attitude change observed. Dissonance theory, however, predicts just the opposite: Making the initial attitude salient makes one of the two “dissonant” cognitions salient, thus increasing the amount of dissonance aroused. This in turn should produce more attitude change. If, on the other hand, the behavior is made more salient, then both theories predict more attitude change. For self-perception theory, this would make salient the very source of evidence upon which the final attribution is to be based; for dissonance theory, this is simply making the second of two dissonant cognitions more salient, again arousing more dissonance. Finally, consider the possibility that both the initial attitude and the behavior were to be made more salient. Under the multiplicative rule for combining different sources of credibility information ( Section 111, A, 4 ) , self-perception theory predicts that this will be equivalent to making the initial attitude alone salient; hence, attitude change will diminish. Dissonance theory here predicts the greatest attitude change of all since both “dissonant” elements are made salient, thus maximizing the amount of dissonance aroused. With these differential predictions in hand, Snyder and Ebbesen attempted the crucial test by duplicating exactly the experiment by Bem and McConnell (1970), a standard forced-compliance experiment in which subjects were given a choice or no choice to write a counter-attitudinal essay. First they verified that they could replicate the standard effect as Bem and McConnell had done: Subjects given their choice expressed final attitudes more in line with their counter-attitudinal essays than those expressed either by no-choice subjects or control subjects who wrote no essays at all. When initial attitudes alone were made salient, choice subjects did not differ from no-choice subjects, and neither group differed from the no-essay control group. Considered alone then, this column of results supported self-perception theory and runs counter to the dissonance theory prediction. When the behavior alone was made salient-where both theories predict enhanced attitude change-the column of results looked like the standard condition: There was more attitude change in the choice than in the no-choice condition which, in turn, did not differ from the control condition, but the effect was not significantly larger than in the standard condition. Neither theory is supported strongly by this condition, but neither was this a condition designed to discriminate between the two theories. Finally, no attitude change was observed in either condition when both the attitude and the behavior were made salient. Again, this supports self-perception theory (multiplicative rule) and not dissonance theory.

SELF-PERCEPTION THEORY

33

A contrast applied to the entire pattern of results shows that selfperception theory accounts for more of the between-cells variance than dissonance theory. But on the other hand, both theories leave a significant amount of the variance unexplained. In particular, there were differential amounts of attitude change among the several no-choice conditions which neither theory anticipates. Snyder and Ebbesen propose a self-perception theory of their own to account for the pattern of results, a theory which proposes that the subject’s perception of how counter-attitudinal his behavior appeared to him mediates the attitude change effects. Whatever the merits of the Snyder-Ebbesen theory turn out to be, the “crucial” test between dissonance theory and the original self-perception interpretation of forced-compliance effects remains equivocal. Nevertheless, contrary to the prediction of Bem and McConnell (1970), somebody did manage to force the two theories to confront each other. It is particularly unfortunate that the Snyder-Ebbesen experiment is equivocal, for the remaining Zeigarnik is likely to tempt others to go forth and do likewise. There are many more interesting paths off the self-attribution road to be explored, and the demise of the precious controversy between dissonance theory and self-perception theory is a consummation devoutly to be wished. It is resolved that, henceforth, nothing more on this subject shall be heard from this quarter. IV. Other Self-Perception Phenomena In addition to the cognitive dissonance experiments and the initial studies specifically designed to provide support for self-perception theory, there are a number of other effects which have begun to emerge which fit more or less into the self-perception explanatory framework. Three of these will be reviewed here. A. MISATTRIBUTION EFFECTS Before self-perception theory had been enunciated, Schachter’s (1964) work on emotional states was already providing evidence that a person’s attribution of his internal states was a joint function of both internal physiological cues and external factors of the situation (e.g., Schachter & Singer, 1962). It follows from this fact that one can manipulate an individual’s self-attributions by manipulating those external factors. Valins (1966) took this possibility a step further when he showed that an individual’s self-attributions could also be influenced by giving him false feedback about his autonomic arousal. Thus, as noted earlier, Valins was able to manipulate attitudes toward stimulus pictures of

34

DARYL J. BEM

seminude females by giving his male subjects false auditory feedback which they could interpret as their heartbeat. Since then, a number of studies have adopted this general procedure. For example, Valins and Ray (1967) asked snake-phobic subjects to look at slides picturing snakes; subjects were given false feedback about their heartbeat designed to imply that they were not afraid of snakes. Subsequently, these subjects were able to approach the snake more closely than control subjects. Berkowitz, Lepinski, and Angulo (1969) and Berkowitz and Turner (1972) have showed that they could influence the amount of instrumental aggression displayed by subjects by first giving those subjects false feedback about how angry they were. And finally, extinction of the GSR can be facilitated (Loftis & Ross, 1971) or retarded (Koenig & Henriksen, 1972) by giving subjects false information about their arousal. Although there are some intracacies in these studies to be discussed later, the general point to be made at this juncture is that false feedback concerning arousal can lead individuals to misattribute emotional states to themselves. Misattribution can also be created by manipulating not only the apparent degree of arousal, but the apparent sources or causes of arousal (cf. Nisbett & Valins, 1971). For example, Nisbett and Schachter (1966) were able to obtain higher shock tolerance from subjects who believed that their arousal was produced by a pill rather than the shock. Using a very similar technique Ross, Rodin, and Zimbardo (1969) reduced fear of electric shock by persuading their subjects to attribute their arousal to loud noise. And finally, Stoims and Nisbett (1970) gave insomniacs placebo pills and told them that the pills would produce either arousal or relaxation. As predicted, “arousal” subjects reported that they got to sleep more quickly than they had on nights without the pills-presumably because they attributed their arousal to the pills and, as a consequence, worried less about their insomnia, a worry which seems to exacerbate insomnia. Similarly, “relaxed subjects reported that they got to sleep less quickly than usual-presumably because they worried that their emotions were unusually intense since their arousal level was high even after taking an arousal-reducing agent. As Nisbett and Valins (1971) point out, all such studies can be seen as special cases of the underlying assumptions of self-perception theory even though the source of cues for the self-attributions are not the individual’s overt behavior per se. But even this gap was closed in a study by Davison and Valins (1969), who actually manipulated the subjects’ behavior and its apparent controlling variables to produce misattribution. In their study, Davison and Valins asked subjects to

SELF-PERCEPTION THEORY

35

take a series of electric shocks of steadily increasing intensity and told them to report when the shocks first became painful and when they could no longer tolerate them. Following this series, subjects were given a placebo which they were told might change their skin sensitivity, A second shock series was then administered, but the intensity of all shocks was surrepititiously halved so that subjects ended up taking nearly double the number of shocks than in the first series before reporting pain or asking the experimenter to terminate the series. All subjects were then told that the experiment was over. Some subjects were merely thanked and told that in a few minutes, after the drug had worn off, they would participate in another experiment. Other subjects were told that the drug was really a placebo. Davison and Valins reasoned that the placebo subjects would thus have to attribute their high-shock tolerance on the second series to themselves, their actual ability to withstand shock, since they now knew the drug could not have been responsible, whereas the subjects who had not been “debriefed” would make no such attribution of ability to themselves. This hypothesis was confirmed when a third shock series was administered as part of a “different” experiment. Placebo subjects took significantly more shock on the third series than they had on the first, whereas subjects who continued to believe that their performance on the second series was due to the drug did not. Bowers ( 1971) conducted an experiment conceptually quite similar to the Davison-Valins experiment, except that he altered picture preference rather than shock tolerances through misattribution of behavior, His subjects were shown pairs of postcard pictures and asked to state their preference for one of the two pictures in each pair, Each picture also had a four-digit code number printed on its face, and each paired comparison required the subject to choose between a landscape and a portrait. Immediately prior to the series of trials, subjects in two of the experimental conditions were hypnotized and told they should always select pictures whose code numbers contained the digit 7. They were then given amnesia for the suggestion and for the fact that they had been hypnotized. The series of paired comparisons was then begun. As in the Davison-Valins experiment, three series of trials were given. The first 20 trials determined the subject’s baseline for preferring either portraits or landscapes. None of the pictures in these trials had a code number containing the digit 7. During the next 90 (treatment) trials, 60 of the pictures contained the digit 7 paired with the type of picture (portrait or landscape) the subject had least preferred in the baseline trials. Thus, all previously hypnotized subjects saw themselves choosing either portraits or landscapes for at least two-thirds of the

36

DARYL J. BEM

treatment trials with no explanation for their behavior. Bowers reasoned that these subjects would be led to infer that they must, in fact, now prefer their previously nonpreferred type of picture. But for half of these subjects, obvious explicit verbal reinforcement was administered during the treatment trials each time the subject expressed his preference for the nonpreferred type of picture. This, according to Bowers, should provide a suitable explanation to the subject for his preference change and thus prevent the self-attribution from taking place. Several types of control groups who had not been hypnotized were also run. The test of this self-attribution hypothesis was sought in the subjects’ preferences on a final set of 40 trials in which none of the pictures contained the digit 7. As predicted, subjects whose behavior had been manipulated through the hypnotic suggestion and who were left with no alternative explanation for their preference change persisted in choosing the initially nonpreferred type of picture significantly more than the reinforced group or any of several control groups. It is clear why these various phenomena have earned the label of misattribution effects. On the other hand, it is easily overlooked that every effect discussed so far in this review is, without exception, a “misattribution” effect. If individuals actually accurately discriminated the variables controlling their behavior, then none of the predicted selfattributions would have occurred. For example, as Kelley (1967) has pointed out about forced-compliance studies, subjects in low-justification conditions must have an “illusion of freedom,” must fail to apprehend the forces which induced them to comply, if they are to draw the predicted self-attributions from the behavior in which they engage, [The perception of freedom has itself now become a major research topic. For a review and analysis, see Steiner (1970).] Self-perception theory may thus appear deficient because it does not attempt to account for this pervasive unawareness of the actual controlling variables. Or, as a Stanford colleague is fond of saying, selfperception theory appears to explain everything about why induced compliance leads to attitude change except why induced compliance leads to attitude change. But this apparent deficiency emerges only if one assumes that awareness is the normal state of things, and that unawareness is the phenomenon to be explained. As noted in Section I, A, the unique contribution of the radical-behavioral analysis of selfreferring statements resides precisely in its explicit assumption that unawareness is the given and awareness the problematic. From this perspective, what needs to be explained-and what Skinner’s analysis explains-is how individuals learn to respond to the apparent controlling variables we have purposely made salient in the laboratory, not why

SELF-PERCEPTION THEORY

37

they fail to discriminate the actual controlling variables we have intentionally obscured. Under such conditions, the radical behaviorist’s answer to the question, ‘Why do subjects have this ‘illusion’ of freedom?” is “Why not?” It is consistent with Western man’s fascination with the unconscious that misattribution seems somehow “sexier” and more mysterious than veridical attribution. There are, of course, problematic cases of unawareness involving repression and motivated distortion. But these are challenge enough without adding the extra theoretical burden of having to explain the pseudo-problem of why individuals are not also perfect information processers.

B. THESELF-ATTRIBUTION OF DISPOSITIONAL PROPERTIES All of the studies reviewed so far in this article demonstrate that external sources of stimuli can exercise control over an individual’s attributions of his transitory states or his attitudes. Some studies are beginning to be reported, however, which suggest that it might be possible to change longer-standing attributions that the individual might make about himself by manipulating his behavior and apparent controlling variables appropriately. Interestingly, the first real clue that this might be possible was discovered almost accidentally by Freedman and Fraser (1966), who were investigating the so called “foot-in-thedoor” phenomenon in which a person who can be induced to comply with an initial small request is then more likely to comply subsequently with a larger and more substantial demand. In their study, Freedman and Fraser had two undergraduate experimenters contact suburban housewives in their homes, first with a small request and later with a larger more consequential request. The housewives were first asked either to place a small sign in their window or to sign a petition on the issue of either safe driving or keeping California beautiful. Two weeks later, a second experimenter returned to each home, asking all subjects to place a large and rather unattractive billboard pronioting auto safety on their front lawn for several weeks. The second request thus involved both an action and an issue which were either similar to or different from those involved in the first request. A control group was contacted only about the second request. The results showcd a very strong foot-in-the-door effect. Subjects who had complied with the earlier trivial request were much more likely to coniply with the larger one 2 weeks later. The remarkable finding, however, was the striking generality of the effect. It did not matter which issue or action had been involved in the initial request; the compliance generalized cqually in all four conditions to the later

38

DARYL J. BEM

larger request. Thus, even the small action of signing an innocuous petition to “Keep California Beautiful” increased the subsequent probability of the subject’s agreeing to place a large billboard reading “Drive Safely” on her lawn when asked to do so 2 weeks later by a second unrelated experimenter. In fact, the failure to find a generalization gradient as a function of similarity between the initial and final requests tends to rule out several plausible theoretical explanations of the effect. Thus, Freedman and Fraser (1966) arrive post hoc at what is essentially a self-perception explanation: What may occur is a change in the person’s feelings about getting involved or about taking action. Once he has agreed to a request, his attitude may change. He may become, in his own eyes, the kind of person who does this sort of thing, who agrees to requests made by strangers, who takes action on things he believes in, who cooperates with good causes [p. 2011.

In thinking about the Freedman-Frascr study and self-perception theory, it occurred to Lepper (1971) that attributions of this kind might also be taking place in other experimental settings which had previously measured attitudiiial attributions only. For example, in the forbidden toy paradigm it is found that children who comply under mild threat conditions devalue the forbidden toy. But the attribution, “I don’t like that toy” is only one possible inference that a child in the mild-threat condition might draw. Lepper suggests that another inference might be that he is a particularly “good” boy, one who is able to resist temptation. Moreover, this is not an inference which would be drawn under conditions of severe threat. Such an attribution might then generalize so that the child would display increased resistance to temptation in a different situation, an exact analogue to the Freedman-Fraser finding. Lepper tested this hypothesis in the classical forbidden toy setting. Two groups of second-grade children were forbidden from playing with an attractive toy under mild or severe threat of punishment, while a third, control group received no initial prohibition. Three weeks later, a second experimenter asked these subjects to play a game in which they could obtain attractive prizes only by falsifying their scores, As predicted from the self-perception analysis, subjects who complied with the initial prohibition under mild threat showed more resistance to temptation in this second situation than control subjects or subjects who had initially complied under severe threat. Lepper also reports that subjects who had initially complied under severe threat tended to show even less resistance to temptation than control subjects. Although this finding is open to more than one interpretation, Lepper suggests that it might be an “overjustification” effect, wherein the child under threat of severe punishment infers that

SELF-PERCEPTION THEORY

39

he resists temptation only because of strong external forces. Hence, when such pressures are subsequentIy withdrawn, he is even less resistant to temptation than before. We turn now to a more detailed consideration of such overjustification effects. JUSTIFICATION EFFECTS C. OVER

The self-perception analysis of insufficient justification essentially states that a person will infer that he was intrinsically motivated to execute the induced behavior to the extent that external contingencies of reinforcement appeared to be absent. Thus he infers that he “wanted” to do the activity, that he believes in it, or that it reflects his true opinions. An overjustification effect is predicted if one is willing to assume that to the extent that external contingencies of reinforcement are strongly apparent, the individual infers that he did not want to perform the activity, that he does not believe in it, or that it does not reflect his true opinions. Because behavior is “consonant” with the initial attitudes in most overjustification studies, dissonance theory does not apply, and some writers (e.g., Nisbett & Valins, 1972) have looked to such effects as clearer instances of self-perception phenomena. If overjustification effects do occur, then they provide a possible affirmative answer to the old question of whether or not extrinsic reinforcement for an activity reduces the intrinsic motivation to engage in that activity. Performing the activity under strong contingencies of reinforcement leads to the attribution that the activity must not be enjoyable in itself and then perhaps to decreased motivation to engage in that activity (decharms, 1968; Deci, 1971, 1972; Deci & Cascio, 1971; Lepper, Greene, & Nisbett, 1971). For a number of reasons, early experiments with monkeys do not provide an affirmative answer to this question, discussions in elementary textbooks, and elsewhere ( e.g., deCharms, 1968) notwithstanding. Accordingly, new attempts to confirm this overjustification effect have now begun to appear. For example, a series of studies by Deci (1971, 1972; Deci & Cascio, 1971) does suggest that intrinsic motivation to solve puzzles is reduced when the activity is executed under either monetary reinforcement or, possibly, threat of punishment, but not if positive verbal feedback is the reinforcer. It is clear that the “meaning” of the individual’s self-observed behavior is going to be a function of his past history with the particular reinforcement contingencies used. Thus, money has probably acquired the discriminative property of “buying” compliance more than verbal praise. The definitive experiment in this area, however, appears to be an elegant study by Lepper et al. (1971), who carefully divorced the measurement of intrinsic interest from the situation in which the rewards

40

DARYL J. BEM

were administered in order to rule out alternative interpretations of the anticipated results. Children who met a criterion of intrinsic interest in a play activity during baseline observations in their classrooms were randomly assigned to one of three conditions. In separate individual sessions 2 weeks later, children in the Expected-Award condition engaged in the activity in the anticipation of receiving an extrinsic reward. Children in the Unexpected-Award condition engaged in the activity only for its own sake, but subsequently received the same reward. The No-Award control subjects neither expected nor received the reward, but otherwise duplicated the experience of the subjects in the other conditions. One to two weeks after the experimental sessions, the target activity was again introduced into the children’s classrooms, and unobtrusive measures of intrinsic interest were again obtained. The results confirmed that children in the Expected-Reward condition now freely engaged in the activity less than did either children in the UnexpectedAward or No-Award conditions. Although these effects have here been interpreted in terms of selfperception theory, there are some problems involved in doing so which will be discussed in Section VII where the role of other explanations (e.g., decharms, 1968) will also be considered.

V. Some Differences between Self-perception and Interpersonal Perception Self-perception theory asserts that self- and interpersonal perception are similar in two ways. First, the processes of inference involved in attribution are the same, and second, both actors and observers share certain sources of evidence-overt behavior and its apparent controlling variables-upon which those attributions can be based. This leaves open at least four ways in which the self-attributions and interpersonal attributions can still differ. The first difference is what might be called the Znsider 0s. Outsider difference. All of us have approximately 3-4 ft3 of potential stimuli inside of us which are unavailable to others but which are available to us for self-attributions. The thrust of the Skinnerian analysis of selfattributions is not that we can make no discriminations among internal stimuli, but only that we are far more severely limited than we suppose in this regard because the verbal community is limited in how extensively it can train us to make such discriminations. Nevertheless, the Insider can often detect, for example, that he is “trying hard” to solve a problem, and can infer that the problem is difficult, whereas an Out-

SELF-PERCEPTION THEORY

41

sider lacking such internal information, might infer laziness and suppose the problem to be easy. A closely related difference is the Intimate us. Stranger distinction. Here it is our knowledge of our past behavior which guides our attributions, whereas others typically lack such historical information. If past experience has convinced the Intimate that he is intellectually capable, then he will dismiss an experimental task as unfair, irrelevant, or both when he fails it, but as fair and pertinent when he succeeds. The Stranger, however, might well infer that the individual is stupid if he fails but capable if he succeeds. The difference between Intimate and Stranger is that the Stranger does not have any past performance upon which to “anchor” a dispositional attribution, and hence, he is more likely to permit task performance to determine a dispositional attribution than is the Intimate for whom the present task is but a single datum point in a familiar history of intellectual competence. The individual himself has already achieved a relatively stable dispositional inference about his ability, and hence fluctuations in performance can more plausibly be attributed to the task. [This is essentially a partial restatement of Kelley’s (1967) “analysis of variance” model for attributions,] A third difference between self-perception and interpersonal perception stems from the Self us. Other difference. It is here that motivational effects may enter as the Self seeks to protect his esteem or defend against threat. Presumably the several Freudian defense mechanisms are prototypic of processes which distort self-attributions so that they differ from interpersonal attributions. On the other hand, the jump to motivational explanations is probably made too hastily. For example, when subjects and observers differ in evaluating intellectual competence in success and failure conditions, it is tempting to infer that the subject is defensively trying to maintain his self-esteem. But he may simply be veridical. As the example in the above paragraph illustrates, such attribution differences can be frequently explained by looking at the Self‘s knowledge of his past history. In other words, the motivational Self vs. Other difference is probably too often invoked when it is the Zntimute us. Stranger difference which is operative. Moreover, the evidence for esteem-maintenance processes in selfattribution is not nearly so strong as is often supposed. For example, Ross, Bierbrauer, and Polly (1971) conducted a study in which professional teachers and college students attempted to teach an ll-year-old boy the spelling of a list of commonly misspelled words. Contrary to theories of self-esteem maintenance or ego-defensiveness, participants tended to rate “teacher factors” as being more important when the child failed than when he succeeded, and “student factors” more important

42

DARYL J. B E M

in success than in failure conditions. Moreover, this pattern of attribution was considerably more pronounced for professional teachers than for college students, a result which seems to challenge the frequent assertion that esteem maintenance and ego-defensiveness become factors when outcomes are important or central to one’s self concept. This is not to say that motivational distortions do not occur in self-attribution, But Self should be innocent until proved guilty. There is, finally, the possibility that there exists an actual difference in perspective between Actor us. Observer, in which different features of the situation are differentially salient to them, In an excellent article which seems likely to become highIy influential, E. E. Jones and Nisbett (1972) suggest that an actor’s attention is focused outward toward situational cues rather than inward on his own behavior. For the observer, however, the actor’s behavior is the figural stimulus against the ground of the situation. E. E. Jones and Nisbett (1972) present a wellreasoned argument for the existence of this perspective difference and some preliminary findings relevant to it. It is clear, however, that some ingenuity will be required to isolate a pure perspective difference empirically, uncontaminated by the other differences discussed above. At the time of this writing, there appears to be only one study which comes close to accomplishing this feat (Storms, 1971). The major thesis of the Jones-Nisbett article is that the several differences mentioned above, including the perspective difference, conspire to create a pervasive tendency for actors to attribute their actions to situational requirements, whereas observers tend to attribute the same actions to stable personal dispositions of the actor. It is, of course, too early to evaluate the validity of this proposition; but it is sufficiently rich in its implications, and it is SO likely that the various actor-observer differences will pull in opposite directions in some situations, that the €LIU exploration of this single hypothesis is likely to set the direction of research in this area for the next few years.

VI. The Shift of Paradigm in Social Psychology During the Sixties, it will be recalled, all thinking beings were characterized b y chronic drives toward consistency and uncertainty reduction, vigilant forces which coaxed us all toward cognitive quiescence. Our affects, cognitions, and behaviors were held in homeostatic harmony, and our “evaluative needs” initiated emergency information searches whenever any internal state broke through threshold without clear identification or certified cause. In contrast, we are emerging into the Seventies as less driven, more contemplative creatures, thoughtful men and women whose only motivation is the willingness to answer

SELF-PERCEPTION THEORY

43

the question, “How do you feel?” as honestly and as carefully as possible after calmly surveying the available internal and external evidence. There is, in short, a shift of paradigm taking place within social psychology, a shift from motivational/drive models of cognitions, behaviors, and internal states to information processinglattribution models of such phenomena. Self-perception theory is only one element in that shift, and thus it is appropriate at this juncture to place it within this larger context. We begin with a brief history of this transformation and the four separate lines of research which mark it. First, of course, are the various cognitive consistency theories, whose formal history is usually traced to Heider’s (1946) article, “Attitudes and Cognitive Organizations” ( McGuire, 1966). The system Heider proposed employed the motivational constructs of Gestalt psychology, and he elaborated the theory in the well-known The Psychology of Interpersonal Relations ( Heider, 1958), and explored the motivational aspects in a 1960 contribution to the Nebraska Symposium on Motiuaticm ( Heider, 1960). During the decade of the Fifties, several other consistency forniulations were developed, and collectively they set the dominant tone of the motivation/drive paradigm during the early Sixties. This era appeared to culminate with the publication of the massive source book of such theories in 1968 ( Abelson et al., 1968). Cognitive dissonance theory is the most prominent example of this paradigm. The second research tradition involved in the paradigm shift is Schachter’s ( 1964) work on the cognitive and physiological foundations of emotional states. Although Schachter’s theorizing has not been associated with the cognitive consistency paradigm as such, it is rooted in the same tradition, and the major motivational concept “evaluative needs” rests upon Festinger’s (1954) earlier theory of social comparison processes. This is the motivation which leads individuals to seek out an appropriate explanation and label for otherwise ambiguous internal states. It is illustrative of the paradigm shift that this prominent motivational feature of Schachter’s (1959) earlier work on affiliation and the initial research on emotional states (e.g., Schachter & Singer, 1962) has now receded very much into the background, and investigators who trained under Schachter tend increasingly to employ the vocabulary of self-attribution in their studies. Some convergence between the cognitive consistency research and Schachter’s work was achieved when dissonance-theoretic operations began to be applied to the manipulation of emotional and motivational states ( e.g., Zimbardo, 1989). The third relevant conceptual development has been self-perception

44

DARYL J. BEM

theory. As noted in this review, the theory derives from the very different tradition of radical behaviorism. However, as the paradigm shift has developed, the Skinnerian parentage of the theory has been increasingly muted in successive translations. Indeed, one purpose of this review has been to repay homage to the origins of self-perception theory and, in Section VII, C, below, to restate the need for some of the heuristic advantages that the stubborn functional orientation can bestow when behavioral mysteries threaten to become behavioral-science muddles. The ways in which self-perception theory has attempted to assimilate both the Schachter tradition and cognitive dissonance theory have already been reviewed in detail. It should be noted that self-perception theory lacks any motivational construct other than an implicit assumption that individuaIs are willing to answer inquiries concerning their internal states. The fourth major development in the move to the attribution paradigm is attribution theory itself. Once again history begins with Heider, who stated the major ideas in The Psychology of Interpersonal Relutiom (Heider, 1958). During the “consistency” era, this book was cited primarily for its formal balance theory, while Heider’s rich but less formalized observations about person perception and attribution were relatively ignored. This was remedied by E. E. Jones and Davis (1965), who added a number of testable propositions and explicated some specific empirical consequences of the attribution hypotheses contained within the book. The resulting research tended to focus on an observer’s attribution of an actor’s intentions and attitudes (e.g., E. E. Jones & Harris, 1M7) and would probably have proceeded independently of the other three traditions discussed above had it not been for the influential essay, “Attribution Theory in Social Psychology” by Harold Kelley in the 1967 Nebraska Symposium on Motiuation. This essay integrated the Jones and Davis formulation and self-perception theory into a single theoretical framework along with some propositions about attributional biases, errors, and illusions. These latter considerations also afforded Kelley the opportunity to make his observations about the “illusion of freedom” found in dissonance experiments, thus providing an added flourish to the convergence of these several distinct lines of research and theory. If Kelley can be seen as a final step in this shift from drive models to information-processing models, as this brief intellectual history implies, then it is somewhat ironic that his essay appears in a symposium on motivation. For despite Kelley’s valiant try, the motivational flavor is very bland indeed:

SELF-PERCEPTION THEORY

45

Consideration of attribution theory is relevant for a symposium on motivation in several respects. The theory describes processes that operate as if the individual were motivated to attain a cognitive mastery of the causal structure of his environment. Indeed, Heider explicitly assumes that “we try to make sense out of the manifold of proximal stimuli . . . ” And Jones and Davis state, ‘The perceiver seeks to find sufficient reason why the person acted and why the act took on a particular form.” The broad motivational assumption makes little difference in the development and application of the theory, but it gives the theory a definite functionalistic flavor . . . and affords whatever motivational basis might seem necessary to support the complex cognitive processes entailed in attribution. More important for the student of motivation are the specific processes and their consequences. Attribution processes are assumed to instigate, under certain conditions, such activities as information-seeking, communication, and persuasion. Thus attribution theory plays an important role in describing the motivational conditions for these significant classes of social behavior. Equally important is the relevance of attribution theory to the perception of motivation, both in others and in one’s self [From Kelley, 1967, p. 193, by permission of The University of Nebraska, Lincoln, Nebraska]. [Emphasis in the original.]

It is an admirable attempt, but the strongest motivation to emerge from this quotation appears to be Kelley’s need to understand why he was there. Presumably his fellow participants were thus provided with “sufficient reason why the person acted and why the act took on a particular form.” It is, finally, Kelley’s article which has now set the stage for the analysis of the differences between self- and interpersonal perception, discussed in the previous section, and this appears to be the next phase of research in this area. This, then, is where the paradigm shift currently stands and where it appears to be going. But there are still some sticky problems left in what has gone before, us we shall now see.

VJI. Some Unsolved Problems Up to this point, this review has attempted to fit as many phenomena as possible into a single framework with a minimum number of loose threads. This pedagogically motivated elegance, however, must now come to an end, for it has been purchased at the price of some fairly glib legerdemain. It is time to sneak backstage and see what the performance looks like from the wings. A. THECONCEPTUAL STATUSOF NONCOGNITIVE RESPONSECLASSES

If one has managed to alter an individual’s attitude or self-attribution, it is not unreasonable to expect that this will induce consequent changes in other response systems. For example, if one has increased a

48

D A R n J. BEM

person’s favorability toward a dull task, he might be expected to work at the task more assiduously. Induce him to believe that he doesn’t fear snakes and he will approach snakes more cIosely. Convince the child he is obedient, and he will resist temptation. Change a man’s perception of his hunger, thirst, or pain, and he should consume more or less food or drink, or endure more or less aversive stimulation. Nor should such expectations be confined to instrumental or consummatory behaviors only, for there is a long history of evidence that beliefs, attitudes, and selfattributions can exercise influence over physiological responses as well (for reviews, see Frank, 1961; Zimbardo, 1969). It is therefore not unreasonable to expect physiological changes to follow upon induced selfattributions of internal states. Happily, the experimental laboratory has blessed such expectations with some striking confirmations. Dissonance manipulations designed to enhance the perceived attractiveness of dull tasks do produce greater intensity of behavior on the task itself (for a review, see Weick, 1967). Behavioral observations of subjects in Schachter’s experiments reveals them to be behaving “appropriately” in accord with their induced emotional states (e.g., Schachter & Singer, 1962; Schachter & Wheeler, 1962); and, dissonance manipulations designed to alter self-attributions of drive states like hunger, thirst, and pain do alter overt behaviors with respect to their respective stimuli and do produce striking physiological changes (Zimbardo, 1969)In the present review, we have seen that false feedback designed to imply that subjects are not snake-phobic leads them to approach the snake more closely (Valins & Ray, 1967); feedback designed to manipulate self-attributions of anger produces changes in overt instrumental aggression (Berkowitz & Turner, 1972); and feedback designed to create misattributions concerning autonomic arousal alters resistance to extinction of the GSR (Koenig & Henriksen, 1972; Loftis h Ross, 1971). Attempts to encourage subjects to attribute fear-induced arousal to a pill or loud noise rather than shock produce greater shock tolerance (Nisbett & Schachter, 1966; Ross et al., 1969), as does self-observed behavior designed to imply that the subject has higher shock tolerance than he initially thought ( Davison & Valins, 1969). Induced obedience which implies to the child that he is “good produces greater resistance to temptation ( Lepper, 1971) ; and finally, overjustification for performing intrinsically interesting activities diminishes subsequent engagement in those activities (Deci, 1971, 1972; Lepper et al., 1971). But precisely because it has been “not unreasonable to expect” these phenomena to occur, and precisely because they have in fact occurred, the problematic nature of their conceptual status within the

SELF-PERCEPTION THEORY

47

various theories has been insufficiently appreciated. Thus, the “theoretical” predictions or explanations of these phenomena that one finds in the literature are rarely more sophisticated than the “it-is-not-unreasonable-to-expect . . .” statement with which this section opened three paragraphs above. The lucky theory within which the particular investigator is working then gets gratuitous credit for the ‘derivation.” A related practice, also encouraged by the fact that the response classes seem to “hang together,” is to treat the response classes interchangeably as if they were functionally equivalent; the self-attributions and the noncognitive responses are simply grouped together as the “effects” of the stimulus manipulations. Such practices are unfortunate for they can easily obscure important gaps in our understanding by causing us to pretend to knowledge that we do not in fact possess. It is thus important to explore how the various theories account for these noncognitive response classes. In attribution models generally-and in self-perception theory in particular-cognitions or self-attributions are the dependent variables. Instrumental behaviors, consummatory responses, and physiological responses (real or falsified) are among the variables which can serve as antecedent or independent variables, the stimuli from which selfattributions of beliefs, attitudes, or internal states can be partially inferred by the individual. Attribution models are thus very explicit about the direction of the causal arrow, and they remain mute about any phenomenon in which the noncognitive response classes play the dependent variable role; as dependent variables, such response classes are extratheoretical. To state this another way, attribution models do not treat cognitions, overt behaviors, and physiological responses as functionally equivalent response classes, but rather, spell out in detail the mechanisms through which the cognitive response class can be under the partial functional control of the other two. How do attribution models account for noncognitive response classes? They don’t! Selfperception theory can get us from the stimulus manipulation to the attribution. It cannot get us from the attribution to anything beyond that. The consistency paradigm is in much the same position as the attribution paradigm with regard to the physiological response class. Thus an early prediction that physiological phenomena might emerge from dissonance-theory settings was precisely a speculation in the spirit of “it would not be unreasonable to expect . . . [Brehm & Cohen, 1962, pp. 151-1551.’’ The positive empirical results which followed and confirmed that early hunch (Zimbardo, 1969) in no way altered the theoretical status of the hypothesis within the formal theory itself, nor does the invocation of dissonance reduction as a motivational explanatory

48

DARYL J. BEM

concept bridge the gap from the attribution changes to the physiological effects. It is, for example, no “explanation” to assert that an individual lowers his GSR in order to reduce dissonance until it is explained just how the individual goes about doing just that. Again, this gap is not to be confused with the prior gap from the stimulus operations to the attribution changes, a link with which the theory is prepared to deal. This is, of course, the same position in which self-perception theory finds itself; as noted above it, too, has a theory about the first link, but is reduced to handwaving about the second. Similarly, the effects of false feedback on CSR extinction reported by Koenig and Henriksen (1972) are not accounted for by any of the three theories they mentionmodeling, Schachter’s theory, or self-perception theory-and for the same reason: None of these theories contains the theoretical machinery for explaining physiological changes in a nontrivial way. For example, the “explanation” borrowed from Schachter‘s formulation, that “a state of arousal will be perceived as positive or negative depending upon the label which a person attaches to that state, and that he wiIl then behave accordingly” (i.e., show higher resistance to extinction of the GSR) is, at best, a restatement of the data. It is in no sense an explanation of that second link. ( See also Bern, 1972a.) It is thus an important step forward simply to recognize that a detailed theoretical model is still needed to account for the cognitive controI of physiological responses. One of the criteria for a successful theory in this domain will almost certainly be its ability to account simultaneously for the related physiological effects of placebo medication, hypnotically or cognitively induced anesthesia, and the associated phenomena of the “mind-body” problem. A start in this direction is provided by Zimbardo (1969, pp. 269-283) whose theoretical discussion at least outdistances the dissonance theory framework which guided the choice of stimulus manipulations. When one turns from the physiological to the behavioral variables associated with self-attributions, the consistency paradigm appears to be on firmer ground. For example, although the theory of cognitive dissonance is, in literal terms, a theory about cognitions (like the attribution models), the concept of a general drive toward consistency extends itself more easily to instrumental and consummatory behaviors than to physiological responses. Thus, if an individual suffers inconsistency between something he believes and the cognition that he is not behaving in accord with that belief, a purely cognitive conflict, then it follows from the basic postulate of the consistency model that he can achieve drive reduction by altering either the belief or the behavior. The motivational, construct within the theory provides a built-in “motor” force

SELF-PERCEPTION THEORY

49

behind a change in overt behavior.” If a dissonance manipulation makes an individual inore favorable toward a dull task, a higher rate of performance on the task is a reasonably legitimate prediction from the theory. It is important to note that the behavior in this formulation is necessarily mediated by a prior belief, attitude, cognition, or attribution with which the behavior is brought into harmony. A similar kind of consistency principle is also invoked to explain behavioral effects by many investigators who employ Schachter’s theory of emotional states. As alrendy noted above, Koenig and Henriksen (1972) remark that Schachter postulates that an individual “will then behave accordingly” after hc has labeIed his emotional state. Similarly, Berkowitz and Turner (1972) interpret Schachter as saying that an individual interprets his state and “then acts in a manner consistent with this interpretation.” And although Berkowitz and Turner go beyond the Schachter formulation in their own analysis of the stimulus variables leading to instrumental aggression, they too come back to the same mechanism in order to get from the self-attribution of anger to the act of aggression: “Looked at from a larger perspective, the findings also provide yet another demonstration of the search for cognitive consistency. We want our actions to be in accord with our emotions, as we understand them . . .” Interestingly, however, Schachter himself does not invoke such a principle of consistency in his own writings on the topic [including Schachter (1964), which other writers most frequently cite in this connection]. Rather, he treats seIf-attributions and overt behaviors as separate “indices” of the underlying “mood” he set about to produce; that is, Schachter’s conceptual analysis treats the two response classes as functionally equivalent. Thus, from the very beginning, Schachter and his colleagues have routinely collected behavioral observations along with, or even in lieu of, self-report data of emotional state (e.g., ’ I t will be noted that we thus grant legitimacy to a motivational concept for explaining cognitive and behavioral responses, but deny it legitimacy in accounting for physiological responses. The distinction, however, is not based upon response class membership per se, but upon the individual’s ability to directly control the response in question. A motivational construct is still, at bottom, a way of saying that the individual “wants to” perform some response, even if unconsciously; being motivated is not sufficient, however, if he “doesn’t know which string to pull.” As recent work demonstrates, physiological responses can be brought under direct conscious control; presumably they then fall subject to motivational explanations in the same way that instrumental responses do. Nevertheless, we are still avoiding here the deeper epistemological problems concerning the explanatory legitimacy of motivational constructs generally. As a sometimes radical behaviorist, I am inclined to the view that their explanatory power is, in general, illusory.

50

DARYL J. BEM

Schachter & Singer, 1962; Schachter & Wheeler, 1962). Similarly, selfattributions of hunger appear in some of the Schachter obesity studies ( eg., Goldman, Jaffa, & Schachter, KISS), but the dependent variable has now become eating behavior per se, and the word “hunger” has faded quietly from view (e.g., Nisbett & Storms, 1972). Although several new conceptual distinctions have been introduced into this important research to keep abreast of the new findings, there has been no comparable conceptual distinction introduced to parallel or accompany the sub rosa shifts from one response class to another. In sum, pure attribution models presume only to deal with the cognitive response class; additional machinery must be added if they are to deal with behavioral or physiological responses as dependmt variables. Schachter’s model, hovering somewhere between the information-processing/attribution paradigm and the motivational/drive paradigm, does not distinguish on the dependent variable side between the selfattributions and the “emotional” behavior. Just as the attribution models do, Schachter’s model places physiological responses in the role of independent variables only; they are stimuli which partially determine the individual’s perception of his emotional state. Finally, theories within the motivational/drive paradigm, particularly the theory of cognitive dissonance, cannot handle the physiological response class in anything other than a trivial fashion, but they do have a conceptual device for predicting or explaining any overt behavioral changes that are mediated by prior cognitions, attitudes, or attributions. We turn now to a closer examination of this proposed sequence of events from stimulus manipulation to attribution change to behavior change.

B. Do ATTRIBUTIONSMEDIATEBEHAVIOR? Increase a person’s favorability toward a dull task, and he will work at it more assiduously. Make him think he is angry, and he will act more aggressively. Change his perception of hunger, thirst, or pain, and he should consume more or less food or drink, or endure more or less aversive stimulation. Alter the attribution, according to the theory, and “consistent’’ overt behavior will follow. There seems to be only one snag: It appears not to be true. It is not that the behavioral effects sometimes fail to occur as predicted; that kind of negative evidence rarely embarrasses anyone. It is that they occur more easily, more strongly, more reliably, and more persuasively than the attribution changes that are, theoreticalIy, supposed to be mediating them. For example, in a well-controlled study by Grinker (1969) on eyelid conditioning, it was predicted that ‘‘. . . the dissonance aroused by

SELF-PERCEPTION THEORY

51

voluntary commitment to a painful stimulus will be reduced by lowering pain-avoidance motivation-that is, by perceiving the UCS to be less threatening or painful-and thus the conditioning level [will be less].” The study did obtain the predicted effects in conditioning, but “there were no significant differences between any groups on self-report measures of perceived pain, irritability, eye tearing, or apprehension, or on other questionnaire items designed to measure subjective response to the aversive aspects of the situation [Grinker, 1969, p. 1321.” And in a closely related experiment, Zimbardo et al. (1969) were able to obtain predicted changes in learning performance, physiological measures, and pain perception. But the attributions of pain showed the weakest effects, and furthermore, the correlations between these cognitive attributions and the behavioral measures of learning they were supposed to be mediating were -.01 for one group and +.11for the other. Similarly, in the Davison and Valins (1969) shock study, experimental subjects were willing to take more shock than control subjects, as predicted, but they did not rate a set of sample shocks as any less painful than did controls. The snake study by Valins and Ray (1967) gives similar results: Experimental subjects were able to approach the snake more closely than control subjects, but they did not report themselves to be any less frightened of snakes than did the controls. Finally, Weick (1967) reviewed all the studies designed to increase an individual’s favorability toward a dull task. He found that increased effort on the task often occurred in the absence of the attitude change toward the task which was supposed to cause the ,increased effort. Weick concludes that initial cognitive enhancement of the task followed by increased effort simply does not occur often enough for us to be convinced that this is a reasonable explanation. Instead, it appears that the phenomenon in which we are interested may involve just the opposite sequence of events, namely behavioral change followed by occasional attempts to summarize the experience evaluatively.

What is one to make of such failures? One possible explanation for them is that the measures of attributions are not well designed or appropriate to the self-attribution which actually mediates the behavior. Another possibility is that subjects are hesitant to admit to some states like anger (Schachter & Singer, 1962). Although these methodological explanations may account for some of the negative findings, the same pattern of results-behavior changes in the absence of equally strong attribution changes-is found in some of the best designed and carefully executed experiments in the field ( e.g., Grinker, 1969). Another possibility is that the attributions do change as predicted and do mediate the behaviors, but that the attributions themselves are

52

DARYL J. BEM

unconscious (Brock & Grant, 1963; Zimbardo, 1969, p. 76). There have been arguments, of course, that inconsistency itself and the process of inconsistency reduction need not be represented in awareness (e.g., Tannenbaum, 1968), and the author himself has made a parallel claim that individuals need not be able to verbalize the cues they use in arriving at self-attributions (Bern, 1965, 1968b). But such claims can edge dangerously close to metaphysics, and the next retreat into invisibility-that one of the “dissonant” cognitions itself is unknown to the individual-should surely be resisted mightily until all other alternatives, save angels perhaps, have been eliminated. A related but more plausible explanation involving defensive denial processes has been proposed by Zimbardo ( 1969, pp. 269-273); his version of an unconscious cognition at least generates some empirical consequences, and there are some suggestive supporting data for such a process within the settings explored in that body of research. A final lead is provided by Weick‘s suggestion, quoted above, that the attributions or attitudes may follow upon rather than precede the behaviors. This is, of course, the major postulate of self-perception theory and a phenomenon well known to dissonance theory. If this is, in fact, the sequence involved, then it would explain why the measured attribution changes are often less reliable and weaker than the behavioral changes since they are the third, rather than the second link in the chain, as originally assumed. For example, the self-reports of euphoria and anger in the classic Schachter and Singer experiment (1962) were obtained after the behavior had occurred and were, in fact, less reliable than the behavioral observations themselves. A similar instance appears in the Berkowitz and Turner study ( 1972). The self-reports of anger were retrospective measures in which the subjects had to recall how angry they had been prior to engaging in the aggressive behavior. A study by Bem and McConnell (1970) would imply that such “recall” measures would be more highly correlated with the subject’s current attribution (as altered by the intervening behavior) than they would be with the actual previous state he is attempting to recall. And in fact, the Berkowitz-Turner self-report data-designed to check the false feedback manipulation of anger-do appear to parallel the overt aggression displayed by the subjects more closely than they correspond to the meter readings themselves. Prophetically, Berkowitz (1968) himself has said elsewhere: ‘We generally assume as a matter of course that the human being acts as he does because of wants arising from his understanding of his environment. In some cases, however, this understanding may develop after stimuli have evoked the action so that the understanding justifies but has not caused the behavior [p. 3081.’’

SELF-PERCEPTION THEORY

53

Is, then, the Berkowitz-Turner study one of these cases of reverse sequence? Not necessarily, for it is still possible that the appropriate attributions were actually present prior to the behavior and did mediate it. All that can be said is that self-report measures collected after the behavior has occurred may not be a valid index of those attributions. The same holds true for the Schachter and Singer study (1962).and several other studies which have collected attribution data confounded by intervening behavior. This analysis, then, implies that some of the failures to find attribution changes may simply reflect the methodological practice of collecting the self-repoi-ts after other stimulus events, including overt behaviors, have intervened. But like the other explanations offered above, this cannot account for all the failures, for again some of the best studies are not subject to this criticism (e.g., Zimbardo et al., 1969), and they still do not find attribution effects as strong as the behavioral ones they are supposed to be mediating. If we are thus forced to the conclusion that, at least in some settings, attribution changes do not mediate the observed behavioral effects, then we find that a phenomenon which had been previously accounted for within the consistency paradigm has become “unsolved.” That is, we are still left with the task of accounting for the behavior changes themselves. Several attempts to do so are already under way. For example, Nisbett and Valins (1972) have proposed that the stimulus manipulations themselves may be insufficient to actually alter the attribution, but that they do cause the individual to question his current attribution sufficiently so that he “tests” the new attributional hypothesis by engaging in behavior to find out. As a result of engaging in the behavior, the hypothesis about his attribution may be confirmed, and he will accept the attribution as valid. Or the hypothesis may be disconfirmed, leading him to reject the attribution, and leaving the investigator with a set of results showing a behavioral effect and no attribution effect. Thus, false feedback implying that one is not afraid of snakes is not sufficient to create a stable attribution of “I’m not afraid,” but it is sufficient to motivate a test of this possibility by approaching the snake. The process of handling the snake can then stabilize the new attribution via the self-perception process. This intriguing scenario is spelled out in greater detail in Nisbett and Valins ( 1972). With regard to the task enhancement studies, Weick (1967) has suggested that the behavioral effects in these experiments might be accounted for by propositions drawn from frustration theory and cueutilization theory; again the attribution effects-when they do occurcan be handled as postbehavior phenomena by either self-perception theory or dissonance theory.

54

DARYL J. BEM

It may be that still other cases of behavioral effects such as the overjustification phenomenon will be “re-solved by using variations of motivational constructs like the need to be in control of one’s self and environment ( cf. decharms, 1968; Zimbardo, 1969). And, as suggested earlier, the physiological effects should probably be split off and reunited with other physiological phenomena under cognitive influence rather than being grouped according to their independent-variable manipulations as dissonance or attribution phenomena. It is clear that the door has now been opened for many minitheories, for it is unlikely that any single process will account for the diverse phenomena which found themselves grouped together when consistency and attribution models converged. It is, of course, painful to have to deny to self-perception theory some of the effects with which it has been gratuitously credited by other investigators. At least its heuristic value for such phenomena remains intact even if its explanatory power is more limited than its friends had realized. Similarly, it may seem a shame to abandon the parsimony which obtained during the reign of the consistency theories, but it is now clear that some of that parsimony was illusory and was purchased at the cost of obscuring some important gaps in our knowledge. The fact that everything seems to be falling apart should probably be taken as an index of scientific advance.

C. THESTRATEGYOF FUXCTIONAL ANALYSIS If there is an underlying moral here, it is that response classes should be given independent conceptual statuses from one another and analyzed separately for the stimulus variables which control them. If they are observed to covary, one should first inspect the stimulus manipulations for overlapping functional properties which produce that covariation. Any theory which assumes that one response class should vary as a function of another ought to spell out in detail the mechanism of control. What are the stimulus properties of the response class which is presumed to exert functional control over the other? Finally, response classes should not be treated as functionally equivalent unless the theory explicitly dictates that they can be and/or experimentation vindicates the merger. These prescriptions form a part of a more general strategy known as functional analysis, the strategy associated with the radical behaviorism within which self-perception theory was initially enunciated, A functional analysis of a complex behavioral phenomenon proceeds by first inquiring into the ontogenetic origins of the observed dependent variables-treated as response classes in their own right rather than as reflections of underlying structures, processes, or internal states-and

SELF-PERCEPTION THEORY

55

then attempts to ascertain the controlling or independent stimulus variables of which those observable behaviors are a function. In its purest form, such a strategy also assumes that the principles of behavior, that is, the most general orderly functional relations between stimuli and responses, are relatively simple and few in number. The complexity of human social behavior derives, it is assumed, from the complexity and variety of the environmental conditions under which these principles have been operative in the individual’s past history. Thus, the radical behaviorist does not begin with the a priori expectation that he will discover new principles of behavior through the study of social behavior in its terminal complexity. Rather, he attempts to establish a complex behavioral phenomenon as a special case or a compound of previously substantiated functional relations discovered in the experimental analysis of simpler behaviors. The re-emergence from the analysis of the previously established functional relations becomes the vindication for the extrapolation into new domains and for the network of assumptions upon which the extrapolation rests. The spirit of the analysis, therefore, is frankly inductive, not only in its experimental execution, but in its formal presentation. The radical behaviorist’s usual insistence that the analysis also eschew any reference to internal physiological or conceptual processes, real or hypothetical is, of course, its most celebrated (and misunderstood) prescription. Whatever the heuristic value of such a restriction may be in other psychological areas, the tactic carries especial probative force in the analysis of self-perception precisely because the socializing community itself must necessarily train the individual’s self-descriptive skills on the basis of observable stimuli and response^.^ Not all functional analyses comprise every one of these tactics, and orthodox radical behaviorists are not the only psychologists whose It is probably the relaxation of this restriction which has robbed latter-day self-perception theory of its radical behavioristic flavor. One does not remain a behaviorist in good standing with repeated references to “inferential processes” and hypothetical inner dialogues (“What must my attitude be if I am willing to behave . . .”). In order to reclaim membership, therefore, it should probably be said that such concessions to expositional clarity do not, in my view, add anything to the explanatory power of the theory; it remains formally equivalent to its earlier, albeit nearly incomprehensible, incarnation in the more rigid and and vocabulary of radical behaviorism (Bem, 1964). But as this section is attempting to demonstrate, a choice of language is not without heuristic consequences. For private “thinking” purposes, functional analysis remains my preference; but for exposition purposes, English prose does not seem overly risky. Roman and arabic numeral systems are also formally equivalent, but performing long division in the former is reputed to be unwieldy.

56

DARYL J. BEM

analyses are informed by particular elements of the approach. For example, Leventhal (1970) has recently employed a similar strategy in analyzing the attitudinal and behavioral effects of fear-arousing communications, thereby bringing elegant order out of the chaotic and conflicting findings in this area. (In addition, Leventhal’s analysis is another instance of a drive theory being replaced by an information processing orientation.) Berkowitz’s (1965) analysis of aggression has a similar spirit and strategy behind it. If it had been employed earlier, the functional approach would have led to very different kinds of analyses within the domain of self-attributions. For example, Schachter’s (1964) own review of the literature on emotion reveals that physiological cues should have more functional control over the cognitive attributions than over “emotional” behaviors per se. Thus, sympathectomized animals continue to show emotional behaviors, and human “subjects with cervical lesions described themselves as acting emotional but not feeling emotional [p. 741.’’ These findings would seem to have implications for which response class is being used as the “index” of mood in the Schachter experiments. But as noted earlier, Schachter and his colleagues interchange the two response classes repeatedly and nowhere acknowledge a functional distinction between them. A similar confounding of these same response classes appears in the insomnia experiment by Storms and Nisbett (1970), an experiment also conducted within the Schachter framework. As described in Section IV, A, insomniacs who thought the placebo pill would arouse them reported that they got to sleep more quickly than they had on nights without the pills, whereas subjects who thought the placebo to be a relaxant reported that they got to sleep less quickly than usual. The important point to note here is that the dependent variable is the subject’s report of how much time had passed before he fell asleep. But when this experiment is cited, and even in the abstract of the article itself, it is reported that “arousal” subjects got to sleep more quickly and “relaxation” subjects got to sleep less quickly. And that’s not the same thing! Estimates of time passage are themselves attributions which are subject to manipulation (cf. London & Monello, 1972). Perhaps the implied state of arousal is more interesting to introspect than the implied state of relaxation, making time appear to pass more quickly. If true, then “arousal” subjects would report getting to sleep more quickly even if it weren’t the case. This alternative explanation is admittedly less plausible than the original, but the point to be made here is that the time it takes to fall asleep is a different response measure from the self-report of the time it takes to fall asleep, and both are subject to cognitive manipulations. Indeed, the therapeutic implica-

SELF-PERCEPTION THEORY

57

tions of this experiment could be quite different from those suggested by the authors unless we are prepared to assume that getting insomniacs to think they are falling asleep faster is the same as curing the insomnia. For years, personality theorizing has been dominated by the “trait” assumption that there are pervasive cross-situational consistencies in an individual’s behavior. After reviewing the literature, Mischel ( 1968) concludes that the empirical search for such consistencies or traits rarely generates a correlation above +.30, a finding of some disappointment if one’s theory of human behavior anticipates +1.00. He, too, suggests a learning-theoretic functional analysis in which covariance of responses is sought in the overlap of situational conditions which evoke and maintain particular response classes. Under such a strategy, one constructs the consistencies from the ground up rather than assuniing them a priori, and any increment over zero in the magnitude of the cross-situational correlations becomes a matter for some rejoicing (for discussion, see Bem, 197213). There is a parallel in social psychology. The decade of the consistency theories was dominated by the assumption that everything was glued tvgether until proved otherwise (cf. Bem, 1970, p. 34). Since it is now proving otherwise, it is suggested that we try the opposite assumption that nothing is glued together until proved otherwise. It is a question of whether we should begin with expectations of +1.00 correlations or .OO correlations. The heuristic advantage of this strategy is not guaranteed, of course. But the difference in morale if +.30 correlations continue to come is in itself worth considering. REFERENCES Abelson, R. P. Simulation of social behavior. In G . Lindzey & E. Aronson (Eds.), Handbook of social psychology. Vol. 2 (2nd ed.). Reading, Mass.: Addison-Wesley, 1968. Abelson, R. P., Aronson, E., McGuire, W. J., Newcomb, T. M., Rosenberg, M. J., & Tannenbaum, P. H. (Eds.), Theories of cognitice consistency: A sourcebook. Chicago: Rand McNally, 1968. Alexander, N. C., & Knight, G. W. Situated identities and social psychological experimentation. Sociometry, 1971, 34, 65-82. Aronson, E., & Carlsmith, J. M. Effect of severity of threat on the valuation of forbidden behavior. Journal of Abnormal and Social Psychology, 1963, 66, 584-588. Asch, S. E. Social psychology. Englewood Cliffs, N. J.: Prentice-Hall, 1952. Bandler, R. J., Madaras, G . R., & Bem, D. J. Self-observation as a source of pain perception. Journal of Personality and Social Psychology, 1968, 9, 205-209. Barber, T. X. Toward a theory of pain: Relief of chronic pain by pre-frontal leucotomy, opiates, placebos, and hypnosis. PsychoZogicaZ Bulletin, 1959, 59, 430-460. Barber, T. X. The effects of hypnosis on pain. Psychosomatic Medicine, 1963, 25, 303-333.

58

DARYL J. BEM

Beecher, H. K. Measu~ementof subjective responses: Quantitatiue effects of drugs. London and New York: Oxford University Press, 1959. Beecher, H. K. Increased stress and effectiveness of placebos and “active” drugs. Science, 1960, 132, 91-92. Bern, D. J. An experimental analysis of beliefs and attitudes. (Doctoral dissertation, University of Michigan) Ann Arbor, Mich.: University Microfilms, 1964. No. 6412,588. Bern, D. J, An experimental analysis of self-persuasion. Journal of Experimental Social Psychology, 1965, 1, 199-218. Bern, D. J. Inducing belief in false confessions. Journal of Personality and Social P S Y C ~ O ~1966, O ~ Y3, , 707-710. Bern, D. J. Reply to Judson Mills. Psychological Review, 1967, 74, 536537. ( a ) Bern, D. J. Self-perception: An alternative interpretation of cognitive dissonance phenomena. Psychological Review, 1967, 74, 183-200. ( b ) Bern, D. J. Self-perception: The dependent variable of human performance. Organizational Behauior and Human Perfomnce, 1967, 2, 105-121. ( c ) Bern, D. J. Attitudes as self-descriptions: Another look at the attitude-behavior link. In A. G. Greenwald, T. C. Brock, & T. M. Ostrom (Eds.), Psychological foundutions of attitudes. New York: Academic Press, 1968. ( a ) Bern, D. J . Dissonance reduction in the behaviorist. In R. P. Abelson, E. Aronson, W. J. McGuire, T. M. Newcomb, M. J. Rosenberg, & P. H. Tannenbaum (Eds.), Theories of cognitive consisteticy: A sourcebook. Chicago: Rand McNally, 1968. Pp. 246-256. ( b ) Bern, D. 3. The epistemological status of interpersonal simulations: A reply to Jones, Linder, Kiesler, Zanna, and Brehrn. Journal of Experimental Social Psychology, 1968, 4, 270-274. ( c ) Bern, D. J. Beliefs, attitudes, and human affairs. Monterey, Calif.: Brooks/Cole, 1970. Bern, D. J . The cognitive alteration of feeling states: A discussion. In H. London & R. E. Nisbett (Eds.), Cognitive alteration of feeling state$. Chicago: Aldine, 1972. (a) Bem, D. J, Constructing cross-situational consistencies in behavior: Some thoughts on Alker’s critique of Mischel. Journal of Personality, 1972, in press. ( b ) Bem, D. J., & McConnell, H. K. Testing the self-perception explanation of dissonance phenomena: On the salience of premanipulation attitudes. Journal of Personality and Social Psychology, 1970, 14, 23-31. Berkowitz, L. The concept of aggressive drive: Some additional considerations. In L. Berkowitz (Ed.), Advances in experimental social psychology. Vol. 2. New York: Academic Press, 1965. Pp. 301329. Berkowitz, L. The motivational status of cognitive consistency theorizing. In R. P. Abelson, E. Aronson, W. J. McGuire, T. M. Newcomb, M. J. Rosenberg, & P. H. Tannenbaurn ( Eds. ), Theories of cognitiue consistency: A sourcebook. Chicago: Rand McNally, 1968. Pp. 303-310. Berkowitz, L., Lepinski, J., & Angulo, E. Awareness of own anger level and subsequent aggression. Journal of Personality and Social Psychology, 1969, 11, 293300. Berkowitz, L., & Turner, C. Perceived anger level, instigating agent, and aggression. In H. London & R. E. Nisbett (Eds.), Cognitive alteration of feeling states. Chicago: Aldine, 1972. Bowers, K. S. An attributional analysis of operant conditioning: The problem of behavioral persistence. Unpublished manuscript, University of Waterloo, 1971.

SELF-PERCEPTION THEORY

59

Brehm, J. W., & Cohen, A. R. Re-evaluation of choice alternatives as a function of their number and qualitative similarity. Journal of Abnormal and Socia2 Psychology, 1959, 58, 373-378. Brehm, J. W., & Cohen, A. R. Explorations in cognitiue dissonance. New York: Wiley, 1962. Brock, T. C., & Grant, L. D. Dissonance, awareness, and motivation. Journal of Abnormal and Social Psychology, 1963, 67, 53-60. Brown, J. S. A behavioral analysis of masochism. Journal of Experimental Research in Personality, 1965, 5, 65-70. Carlsmith, J. M., Collins, B. E., & Helmreich, R. L. Studies in forced compliance: I. The effect of pressure for compliance on attitude change produced by face-to-face role playing and anonymous essay writing. JournaE of Personality and Social Psychology, 1966, 4, 1-13. Carlsmith, J. M., Ebbesen, E. B., Lepper, M. R., Zanna, M. P., Jonas, A. J., & Abelson, R. P. Dissonance reduction following forced attention to the dissonance. Proceedings of the American Psychological Association, 1969, 321-322. Chapanis, N. P., & Chapanis, A. Cognitive dissonance: Five years later. Psychological Bulletin, 1964, 61, 1-22. Chappell, V. C. (Ed.) The philosophy of mind. Englewood Cliffs, N. J.: PrenticeHall, 1962. Corah, N. L., & Boffa, J. Perceived control, self-observation, and response to aversive stimulation. Journal of Personality and Social Psychology, 1970, 16, 1 4 . Davison, G . C., & Valins, S. Maintenance of self-attributed and drug-attributed behavior change. Joumal of Personality and Social Psychology, 1969, l l , 2 5 3 3 . decharms, R. Personal causation: The internal affectioe determinants of behaoi0.r. New York: Academic Press, 1968. Deci, E. L. Effects of externally mediated rewards on intrinsic motivation. Journal of Personality and Social Psychology, 1971, 18, 105-115. Deci, E. L. Intrinsic motivation, extrinsic reinforcement, and inequity. Journal of Personality and Social Psychology, 1972, in press. Deci, E. L., & Cascio, W. F. Changes in intrinsic motivation as a function of negative feedback and threats. Unpublished manuscript, University of Rochester, 1971. Elms, A. C. Role playing, incentive, and dissonance. Psychological Bulletin, 1967, 68, 132-148. Festinger, L. A theory of social comparison processes. Human Relations, 1954, 7, 117140. Festinger, L. A theory of cognitive dissonance. Stanford: Stanford University Press, 1957. Festinger, L. Conflict, decision and dissonance. Stanford: Stanford University Press, 1964. Festinger, L., & Carlsmith, J. M. Cognitive consequences of forced compliance. Journal of Abnormal and Social PsrJchoZogy, 1959, 58, 203-210. Frank, J. Persuasion and healing. Baltimore: Johns Hopkins Press, 1961. Freedman, J. L. Long-term behavioral effects of cognitive dissonance. Journal of Experimental Social Psychology, 1965, 1, 1Oa3-l20. Freedman, J . L., & Fraser, S. C. Compliance without pressure: The foot-in-the-door technique. Journal of Personality and Social Psychology, 1966, 4, 195-202. Goldman, R., Jaffa, M., & Schachter, S . Yom Kippur, Air France, dormitory food, and

60

DARYL J. BEM

the eating behavior of obese and normal persons. Journal of Personality and Social Psychology, 1968, 10, 117-123. Grinker, J. Cognitive control of classical eyelid conditioning. In P. G. Zimbardo (Ed.), The cognitive control of motiuation. Glenview, 111.: Scott, Foresman, 1969. Harris, V. A., & Tamler, H. The effects of attitude reinstatement in bystander replications. Unpublished manuscript, State University of New York at Buffalo, 1971. (8)

Harris, V. A., & Tamler, H. Reinstatement of initial attitude and forced-compliance attitude change. Journal of Soclal Psychology, 1971, 84, 127-134. ( 6 ) Harvey, J., & Mills, J. Effect of an opportunity to revoke a counterattitudinal action upon attitude change. Journal of Personulity and Social Psychology, 1971, 18, 201209. Heider, F. Attitudes and cognitive organizations. Journal of Psychology, 1946, 21, 107-112. Heider, F. The psychology of interpersonal relations. New York: Wiley, 1958. Heider, F. The gestalt theory of motivation. In M. R. Jones (Ed.), Nebraska symposium on motiuation. Vol. 8. Lincoln: University of Nebraska Press, 1960. Pp. 145-172. Jones, E. E., & Davis, K. E. From acts to dispositions. In L. Berkowitz (Ed.), Advances in experimental social psychology. Vol. 2. New York: Academic Press, 1965. Pp. 219-266. Jones, E. E., & Harris, V. A. The attribution of attitudes. Journal of Experimental Social Psychology, 1967, 3, 1-24. Jones, E. E., & Nisbett, R. E. The actor and the observer: Divergent perceptions of the causes of behavior. In E. E. Jones, D. Kanouse, H. H. Kelley, R. E. Nisbett, S. Valins, & B. Weiner (Eds.), Attribution: Perceioing the causes of behauior. New York: General Learning Press, 1972. Jones, R. A., Linder, D. E., Kiesler, C. A., Zanna, M., & Brehm, J. W. Internal states or external stimuli: Observers’ attitude judgments and the dissonance theory-self-persuasion controversy. Journal of Experimental Social Psychology, 1968, 4, 247-269. Jones, R. G. Forced compliance dissonance predictions: obvious, non-obvious, or non-sense? Paper presented at the meeting of the American Psychological Association, New York, September 1966. Kelley, H. H. Attribution theory in social psychology. In D. Levine (Ed.), Nebraska symposium on motioation. Vol. 15. Lincoln: University of Nebraska Press, 1967. Pp. 192-238. Kiesler, C. A., Nisbett, R. E., & Zanna, M. P. On inferring one’s beliefs from one’s behavior. Journal of Personality and Social Psychology, 1969, 11, 321-327. Klemp, G . O., & Leventhal, H. Self-persuasion and fear reduction from escape behavior. In H. London & R. E. Nisbett (Eds.), Cognitive alteration of feeling states. Chicago: Aldine, 1972. Koenig, K. P., & Henriksen, K. Cognitive manipulation of GSR extinction: Analogues for conditioning therapies. In H. London & R. E. Nisbett (Eds.), Cognitive alteration o f feeling states. Chicago: Aldine, 1972. Lepper, M. R. Dissonance, self perception, and honesty in children. Unpublished manuscript, Stanford University, 1971. Lepper, M. R., Creene, D., & Nisbett, R. E. Undermining children’s intrinsic interest with extrinsic reward: A test of the “overjustification” hypothesis. Unpublished manuscript, Stanford University, 1971.

SELF-PERCEPTION THEORY

61

Lepper, M. R., Zanna, M. P., & Ahelson, R. P. Cognitive irreversibility in a dissonance reduction situation. Journal of Personality and Social Psychology, 1970, 16, 191-198. Leventhal, H. Findings and theory in the study of fear communications. In L. Berkowitz (Ed.), Advances in experimental social psychology. Vol. 5. New York: Academic Press, 1970. Pp. 120-186. Linder, D. E., & Jones, R. A. Discriminative stimuli as determinants of consonance and dissonance. Journal of Experimental Social Psychology, 1969, 5, 467-482. Loftis, J., & Ross, L. Facilitation of GSR extinction through misattribution. Unpublished manuscript, Stanford University, 1971. London, H., & Monello, L. Cognitive manipulation of boredom. In H. London & R. E. Nisbett (Eds.), cognitive aZteration of feeling states. Chicago: Aldine, 1972. Maslach, C. The “ t r u t h about false confessions. Journal of Personality and Social Psychology, 1971, 20, 141-146. McGuire, W. J. The current status of cognitive consistency theories. In S. Feldman ( Ed. ) , Cognitive consistency: Motivational antecedents and behavioral comequents. New York: Academic Press, 1966. Pp. 1-46. Melzack, R. The perception of pain. Scientific American, 1961, 204, 4 1 4 9 . Miller, N., & Dollard, J. Social learning and imitation. New Haven: Yale University Press, 1941. Mills, J. Comment on Bem’s “Self-perception: An alternative interpretation of cognitive dissonance phenomena.” Psychological Review, 1967, 74, 535. Mischel, W. Personality and assessment. New York: Wiley, 1968. Nisbett, R. E., & Schachter, S. Cognitive manipulation of pain. Journal of Experimental Social Psychology, 1966, 2, 227-236. Nisbett, R. E., & Storms, M. D. Cognitive and social determinants of food intake. In H. London & R. E. Nisbett (Eds.), Cognitive alteration of feeling states. Chicago: Aldine, 1972. Nisbett, R. E., & Valins, S. Perceiving the causes of one’s own behavior. In E. E. Jones, D. E. Kanouse, H. H. Kelley, R. E. Nisbett, S. Valins, & B. Weiner (Eds.), Attribution: Perceioing the causes of behavior. New York: General Learning Press, 1972. Ostfeld, B., & Katz, P. A. The effect of threat severity in children of varying socioeconomic levels. Deoelopmental Psychology, 1969, 1, 205-210. Pepitone, A., McCauley, C., & Hammond, P. Change in attractiveness of forbidden toys as a function of severity of threat. Journal of Experimental Social Psychology, 1967, 3, 221-229. Piliavin, J. A., Piliavin, I. M., Loewenton, E. P., McCauley, C., & Hammond, P. On observers’ reproductions of dissonance effects: The right answers for the wrong reasons? Journal of Personality and Social Psychology, 1969, 13, 98-106. Ross, L., Bierbrauer, G. A., & Polly, S. The attribution of success and failure in student-teacher interaction. Unpublished manuscript, Stanford University, 1971. Ross, L., Rodin, J., & Zimbardo, P. G. Toward an attribution therapy: The reduction of fear through induced cognitive-emotional misattribution. Journal of Personality and Social Psychology, 1969, 4, 279-288. Ryle, G. The concept of mind. London: Hutchinson, 1949. Schachter, S. The psychology of atfiliation. Stanford: Stanford University Press, 1959. Schachter, S. The interaction of cognitive and physiological determinants of emotional state. In L. Berkowitz (Ed.), Advances in experimental social psychology. Vol. 1. New York: Academic Press, 1964. Pp. 49-80.

62

DARYL J. BEM

Schachter, S., & Singer, J. E. Cognitive, social, and physiological determinants of eniotional state. Psychological Review, 1962, 69, 379-399. Schachter, S., & Wheeler, L. Epinephrine, chlorproinazine and amusement. Journal of Abnormal and Social Psychology, 1962, 65, 121-128. Skinner, B. F. The operational analysis of psychological terms. Psychological Review, 1945, 52, 270-277, 291-294. Skinner, B. F. Science and human behavior. New York: Macmillan, 1953. Skinner, B. F. Verbal behavior. New York: Appleton, 1957. Snyder, M., & Ebbesen, E. B. Dissonance awareness: A test of dissonance theory versus self-perception theory. In press. Steiner, I. D. Perceived freedom. In L. Berkowitz (Ed.), Advances in experimental social psychology. Vol. 5. New York: Academic Press, 1970. Pp. 187-248. Storms, M. D. Video tape and the attribution process: Changing actors’ and observers’ points of view. Unpublished doctoral dissertation, Yale University, 1971. Storms, M. D., & Nisbett, R . E. Insomnia and the attribution process. journal of Personality and Social Psychology, 1970, 2, 319-328. Tannenbaum, P. H. The congruity principle: Retrospective reflections and recent research. In R. P. Abelson, E. Aronson, W. J. McGuire, T. M. Newcomb, M. J. Rosenberg, & P. H. Tannenbauni (Eds.), Theories of cognitive consistency: A sourcebook. Chicago: Rand McNally, 1968. Pp. 52-72. Turner, E. A., & Wright, J. Effects of severity of threat and perceived availability on the attractiveness of objects. Journal of Personulity and Social Psychology, 1965, 2, 128-132, Valins, S. Cognitive effects of false heart-rate feedback. Journal of Personality and Social Psychology, 1966, 4 , 400-408. Valins, S., & Ray, A. A. Effects of cognitive desensitization on avoidance behavior, Journal of Personality and Social Psychology, 1967, 7 , 345-350. Weick, K. E. Dissonance and task enhancement: A problem for compensation theory? Organizational Behavior and Huinan Performance, 1967, 2, 175-216. Wolosin, R. J. Self- and social perception and the attribution of internal states. ( Doctoral dissertation, University of Michigan ) Ann Arbor, Mich. : University Microfilms, 1969. No. 69-12,276. Wolosin, R. J. Attribution of freedom to the self and others. Paper presented at the meeting of the Midwestern Psychological Association, Detroit, May 1971. Wolosin, R. J., & Denner, B. Three studies of the attribution of freedom to the self and to others. Unpublished manuscript, Indiana University, 1970. Zanna, M. P. Inference of belief from rejection of an alternative action. Unpublished manuscript, Princeton University, 1970. Zimbardo, P. G. The cognitive control of motivation: The consequences of choice and dissonance. Glenview, 111.: Scott, Foresman, 1969. Zimbardo, P. G., Cohen, A., Weisenberg, M., Dworkin, L., & Firestone, I. The control of experimental pain. In P. G. Zimbardo (Ed.), The cognitive control of motivation. Glenview, 111.: Scott, Foresman, 1969. Pp. 100-125.

SOCIAL NORIMS, FEELINGS, AND OTHER FACTORS AFFECTING HELPING AND ALTRUISM’

Leonard Berkowitz UNIVERSITY OF WISCONSIN

I. Social Exchange and Helpfulness . . . . . . . . . Examples of Helping and Altruism . . . . . . . 11. Reactions to Another Person’s Dependency . . . . . . . A. Moral Norms and Helpfulness . . . . . . . . B. A Direct Test of the Role of Moral Norms . . . . . . C. Observational Influences Presumably Increasing the Salience of the . . . . . . . . . . . Helping Ideal . 111. Moods and Feelings . . . . . . . . . . A. Good and Bad Moods as Determinants of Helpfulness . . . B. Guilt . . . . . . . . . . . . . . C. Empathy . . . . . . . . . . . . D. Self-concern . . . . . . . . . . . . IV. Qualities of the Dependent Person Affecting Willingness to Help Him . A. The Other’s Attractiveness . . . . . . . . . B. The Other’s Previous Helpfulness . . . . . . . C. Some Social Class Differences in Reciprocity Orientation . . D. Reactance and the Legitimacy of the Other’s Dependency . . . V. Conclusions . . . . . . . . . . . . References . . . . . . . . . . . . .

. .

. . . . . . .

63 65 68 68 74

77 80 80 83 86 89 94 94 95 98

103

105 106

I. Social Exchange and Helpfulness

Blau (1968) has spelled out both the fundamental premise and the boundaries of social exchange analyses of human interaction. This viewpoint basically contends, he notes, that men form social associations because of the rewards they think these relationships will bring, and continue interacting with people because of the rewards these others provide. Obvious though the statement is, it highlights the exchange ‘The author’s research reported in this paper was carried out under grants from the National Science Foundation. 63

64

LEONARD BEXKOWITZ

theory focus on benefits derived from dealing with other persons. Here, as elsewhere in his essay, Blau describes man as thinking primarily of what he can get from others; people are instrumental to the satisfaction of his wants. Homans’ (1961) discussion is grounded on the same conception. Just as the Skinnerian pigeon continues to press the bar in bis cage as long as the action brings rewards, so will the human per@ in those social encounters that provide him with greater gratifications than punishments from others. Of course, rules and regulations may insist that certain pleasures be postponed until some future time or that the individuaI should incur greater costs than rewards on certain occasions. Here too, however, according to the exchange theorists, adherence to these standards usually comes about because conformity is instrumental to the long-range attainment of external rewards. Blau agrees with Gouldner (1960) in this regard. Both say that reciprocity is a moral principle. A person who obtains benefits from his associates is obligated to repay them in some way. But in following this rule, the individual supposedly carries out his responsibilities because it is profitable to do so: “Both parties profit from the association, and their exchange of rewarding experiences fortifies the social bond between them [Blau, 1968, p. 4531.’’True, we sometimes go out of our way to help friends or even strangers, yet even this seeming altruism is supposedly often only instrumental behavior. We know those we help are obligated to pay us back; their gratitude and appreciation promise future rewards. Some years earlier, Blau had offered a much broader exchange analysis of moral behavior. People act in socially proper ways partly in order to avoid the costs that might arise from immoral behavior, but these costs then included guilt feelings. “The guilt feelings and social disapproval . . . constitute costs that are expected to outweigh the rewards this behavior could bring him [Blau, 1964, p. 2581.” Having acknowledged internally motivated adherence to moral standards, however, Blau turned quickly from such internal determinants even in this earlier formulation. Most of his discussion of conformity to social ideals emphasized the benefits the individual would obtain from other people by living up to these moral principles. Blau’s brief glance away from externally derived rewards had apparently made him uncomfortable. He saw the danger of an overextended formulation. It was tempting to try to apply this notion to all social conduct, but the concept would “lose its distinctive meaning and become tautological,” he correctly pointed out, if it was used to explain all social behavior. Much of social conduct “is oriented toward expected returns from others,” but “not all of it is [Blau, 1968, p. 4531.”

FACXORS AFFECTING HELPING AND ALTRUISM

65

Blau’s warning should be addressed to those social psychologists who would stretch the exchange conception well beyond recognition into something diffuse and imprccise. Theorists who talk of one part of the personality exchanging rewards and costs with other parts may believe they provide a useful formulation, but actually describe only a mist before their eyes, covering everything but defining nothing. Blau (1968, p. 457) recognized this and explicitly placed many forms of moral behavior outside the scope of exchange theorizing. These nonexchange actions, nevertheless, were relatively rare, Blau insisted. Men who work selflessly for others “without thought of reward and even without expecting gratitude” are “virtually saints,” and there are very few saints. Most people presumably require a direct incentive if they are to help others, even if that incentive consists of only social approval (Blau, 1968, p. 453). I shall take quite a different position in this paper. Externally derived incentives are undoubtedly major determinants of behavior, perhaps more important than internalized ideals for many persons and in many situations. But there is probably a far greater incidence of selfless action in behalf of others-even in the absence of reciprocal or anticipated benefits-than the L I S L I ~form ~ of exchange theory would have LIS believe. Some of this behavior, although not all, is iduenced by the operation of social rules, internalized standards of conduct, People sometimes act altruistically because this is the right thing to do in a given situation. On other occasions, however, they might help someone else because they empathize with him. We will also consider another complication: a person may deviate from social regulations in some instances even though he is firmly convinced of their propriety and has attempted to adhere to these standards in other situations; these occasional deviations do not mean that he does not believe in these rules or that they do not frequently govern his behavior. Other factors obviously may become potent determinants of behavior in some situations, and we will look at what some of these influences might be.

EXAMPLES OF HELPING AND ALTRUISM When Blau said “saintly” altruism was rare, he probably was thinking of heroic actions involving grave dangers or serious hardships to the helper. Less dramatic occurrences of altruism are fairly commonplace, however, according to a number of recent investigations. In most of these cases, the helpers incurred comparatively little cost, but they did help other people, apparently even though they did not expect any form of direct or material payment in return. Latan6 and Darley sent their college students, mostly males,

66

LEONARD BERKOWITZ

throughout Manhattan to make a variety of requests. The students’ success varied greatly depending on what they wanted, as we would expect: 85%of the passers-by told the interrogator the time when asked, 73%made change for a quarter, and 34%gave the student a dime when asked for this amount of money. Interestingly, a higher proportion of Manhattanites gave the dime when the student supplied an acceptabIe explanation: 38% when the asker said only that he had spent all his money, 64% when the student explained that he wanted to make a telephone call, and 72% when he said his wallet had been stolen (Latan6 & Darley, 1970, pp. 10-11). In an unpublished investigation, Jacqueline Macaulay and I found that somewhat over 50% of the women shoppers studied in Madison, Wisconsin gave a male university student 40 cents for bus fare when the student explained that his wallet had “disappeared.” Feldman (1968) also provided information about the incidence of “low cost” helping. In this experiment, male subway riders at major subway stops in Paris, Athens, and Boston were asked by an obvious foreigner to mail a letter. Only 44%of the Parisians refused, even though the envelope lacked stamps, while 52%of the Athenians and 60%of the Bostonians said “no.” In another aspect of this investigation, cashiers in pastry shops had an opportunity to help a customer by refunding money after the customer had deliberately overpaid. Forty-six % of the Parisian cashiers gave this type of help, compared with 4% of the Athenians and fully 82%of the Bostonians. Latan6 and Darley looked at other forms of helping besides giving money. For example, one of their students acted as if he had a serious knee injury and fell to the ground, apparently in need of assistance. This was done in two locations in New York City, at LaGuardia Airport and in a subway station. The location made a considerable difference. Where only 41%of the passers-by in the airport offered help, 83%of the people in the subway came to the student’s aid (Latank & Darley, 1970, p. 118). Another field experiment in the New York subway also indicates that many people are disposed to help someone who seems ill. When Piliavin, Rodin, and Piliavin (1969) had one of their confederates repeatedly fall down in a subway car, pretending he was either physically infirm or drunk, he was assisted by some bystander every single time when he seemed physically disabled and on over 70%of the trials when he appeared drunk. Disasters can lead to a remarkable amount of helpfulness. Contrary to the widespread impression that unexpectcd catastrophes lead only

FACTORS AFFECTING HELPING AND ALTRUISM

67

to disorganization and cause people to wander around in a daze, a report of reactions to a sudden disaster in Worcester, Massachusetts (Wallace, 1956, cited in Midlarsky, 1968) noted that 10 of the 15 people who had not suffered any injury themselves went to the assistance of other persons. A substantial proportion of all of the people in this sample helped others outside their families. Many of the injured were “self-sacrificing and willing to let others be cared for first [Midlarsky, 1968, p. 2331.’’ In all of the studies cited so far, the assistance given was not particularly costly to the helper. We would, of course, expect a decline in the number of persons rendering aid when the consequences for providing this assistance become more unpleasant ( Wheeler & Wagner, 1968); still, Schwartz (1970a) has shown that a fairly high percentage of people are willing to donate their bone marrow even though they would suffer some physical discomfort for several days. In his experiment, 144 persons (mostly men and the greatest proportion holding white-collar jobs) were asked to donate bone marrow to a young woman. Fifty-nine % said they were willing to join the pool of possible donors. The exact proportion of help-givers in these studies does not really matter, of course; with the possible exception of the shop cashiers in the Feldman investigation, we do not know if the samples were even roughly representative of the people in the given community. What is important about the figures I have cited is that they are far from zero. Many persons do help their fellow men even when there are no obvious material benefits to be gained from this action. But can we rule out all external rewards? After all, the people in each study mentioned so far could have been influenced by the hope of immediate approval. Other persons, if only the help recipients, were aware of their actions and could thank them, shake their hands, pat them on the back, or otherwise praise them. The helpers might have given aid in anticipation of this good will. The approval would sometimes come because the helpers had conformed to a widely shared ideal, but it could have been the hope of this reward rather than the ideal per se that led to this helpfulness. Because of this ambiguity, then, we turn to laboratory experiments which sought to minimize anticipations of approval from onlookers. Before doing so, however, let me acknowledge a point that ninny readers are bound to make: The studies do not entirely eliminate the possibility that the subjects had some hope of approval from other persons. These anticipations were probably minimized, nevertheless, perhaps to the extent where they were not a major detenninant of the help that was given.

68

LEONARD BERKOWI’IZ

II. Reactions to Another Person’s Dependency A. MORALNORMS AND HELPFULNESS My early experiments in this area were based on the notion that many people in our society, especially middle-class university students, have acquired strong standards of conduct which prescribe that they aid those who are dependent upon them. These persons will therefore often help someone in need (with the likelihood of such socially responsible behavior probably being in inverse proportion to the anticipated costs of the action), not for tangible gains or social approval, but supposedly primarily for approval from themselves. They presumably act unselfishly for the good feeling they anticipate and in order to avoid the guilt that would arise if they violated their moral standard. The basic assumption of this research was a simple one. For many people in our society, the perception that other persons are greatly dependent upon them for their rewards theoretically arouses feelings of responsibility to these others, This felt obligation generally produces a heightened motivation to help the dependent persons achieve their goals, with the level of this motivation being in inverse proportion to the anticipated cost of the help. While this reasoning is in accord with much conventional theorizing in the social sciences, later evidence (we will see below) suggests that the potency of the conjectured “social responsibility norm” was greatly exaggerated in this early formulation. 1. Early Experimental Evidence Other writers have also advanced what are essentially normative analyses of helpfulness, and several early studies provided suggestive findings. However, these investigations did not always rule out the possibility that the help was given in hope of return benefits. Thomas (1957) proposed, for example, that the magnitude of the “responsibility forces” within a group is in direct proportion to the degree of “facilitative role interdependence” among the members. In his experiment, women group members who were highly dependent upon each other for the performance of a task exhibited greater task motivation than subjects who could work independently of the others in their group. However, his results may be attributable, in part at least, to a desire for reciprocity rather than to the subjects’ feeling of responsibility to dependent others. The interdependent subjects could have become highly motivated to help the others in the group in order not to endanger their own chances of gaining the seward. They gave in order to receive. Much the same explanation can be offered for the results in other experiments. Deutsch ( 1949) demonstrated that cooperative groups

FACTORS AFFECTING HELPING AND ALTRUISM

69

were more highly task motivated than competitive groups, probably because the people in the former groups were (in Deutsch‘s words) promotively interdependent; any member’s behavior could advance d members toward the goal, and hence, each felt the others in the group were dependent upon him. But again, reciprocity may have been in operation. Each member could have worked hard for his partners because he wanted them to work hard for him. Similarly, in an experiment by Berkowitz and Levy (1956),groups evaluated as a unit became more highly motivated than groups whose members were evaluated separately; the high task motivation presumably resulted from the perceived member interdependence created by applying rewards to the group as a whole. The interdependent subjects probably were concerned with helping both their fellow group members and themselves. Because of the reciprocity involved in such interdependent groups, I conducted another experiment ( Berkowitz, 1957) designed to assess the motivational effects of both interdependency and dependency relationships. Using pairs of men engaged in a joint endeavor, I found that there was the greatest motivation toward the assigned task when both the subject and his partner could gain valuable prizes for effective performance. The subjects in such highly interdependent pairs worked harder than the subjects assembled in pairs in which only one member could achieve the valued goal. But, consistent with the “social responsibility” thesis, the results also indicated that the individual subject was as highly motivated when only his partner was eligible for the prize as when only he could attain this reward. In both of these latter conditions, further, the subjects worked harder than when no prize was offered, Thus, the subjects’ task motivation was strengthened by telling them that their partners’ (but not their own) chances of attaining a valued goal were dependent upon their performance in the situation. 2. Did the Subjects Work for the Approval of Others?

These subjects may have anticipated some benefits, however. The work situation was so structured in this experiment that the partner was quickly aware of the subjects’ performance. Realizing this, they might have worked hard in order to gain their partner’s approval. Later experiments (Berkowitz & Daniels, 1963) sought to overcome this difficulty by informing the subjects that their partner would not learn of the help given him for several weeks, if not months. This delay presumably lessened the subjects’ belief that they would win their partner’s thanks if they worked hard for him. Each person in two separate studies was told that he would have the role of a “worker” performing under the indirect guidance of a peer ( the “supervisor”), and that the supervisor’s

70

LEONARD BERKOWITL

likelihood of winning a prize would be either greatly dependent or less dependent upon thc subject’s productivity. Both experiments agreed in showing that the subjects worked hardest when their partner was greatly dependent upon them, even though he would not find out about the subjccts’ effort until the end of the experiment. Table I summarizes the findings obtained in the second study. Other findings to be reported shortly indicate that these results are not due to a desire to win approval from the watching experimenter. The men apparently assisted their dependent peer even though there was little in it for them. Nevertheless, there were some indications of a barely submerged desire for distributivc justice in the second of the Berkowitz and Daniels (1963) experiments. The people in this study were told that their “supervisor” was eligible to receive a cash prize. When asked at the end of the session what changes they would have desired in the work situation, it was primarily the men in the high dependency condition-those who had expended a good deal of effort TABLE I RR.\CTION TO PARTNER’S DEPENDENCY A N D How SOON HE WOULD LEARNOF SUBJECT’S WORK I N HIS BEHALF Parher’s dependency on subject” High

Low

P would find out

P would find out

Measures

Right away

Lateld

Right away

Increase in work for partner from practice periodc Satisfaction with partnerd

19. Oa

19.5,

14.6,

18.8b

14.5&

17.5ab

Late+ ~

7.1b

16.5ab

The subjects had been told that their performance would determine, to a greater or lesser extent, whether or not their partner won a cash prize. b In this condition the subject was informed that the partner would not find out if he had won, and hence, if the subject had helped him sufficiently, until the end of the semester several weeks or months from then. The partner was supposedly also leaving the University a t that time. c The help measure was the inciease in the number of paper boxes made above the number made in the practice period. d This question asked: “To what extent would your work have been better if your partner had been replaced?” The higher the score the greater the satisfaction with the partner. Cells having a subscript in common are not significantly different a t the .05 level by Duncan Multiple Range Test. The main effect for dependency is significant a t the .01level. Data from Berkowitz and Daniels (1963, Experiment II). (1

FACTORS AFFECTING HELPING AND ALTRUISM

71

for their peer-who expressed a wish for a personal incentive, such as a prize, for themselves as well. Moreover, as is shown in the second line of Table I, the students operating under the pressure of their partner’s need, and who believed that he would not learn of their performance until the end of the semester, also generally showed the strongest resentment against this person on a final questionnaire. They had worked hard in adhering to the conjectured ideal even though they had gained little-not even the approval he might give them if he knew of their work-and apparently had been a bit unhappy about not getting any rewards at all. We will return to this matter later in this paper. But if the subjects did not appear to be influenced by the prospect of winning quick approval from the person who was dependent upon them, they still might have been affected by the experimenter’s presence. The experimenter would know how well they had worked, and so they might have been concerned with gaining her approbation. (In this case, it should be noted, we would have the following interpretation of the dependency results: The subjects were aware that the “right” thing to do was to help the person who was dependent upon them, and they lived up to this expectation in order to win the experimenter’s approval. ) An experiment by Berkowitz, Klanderman, and Harris (1964) sought to determine whether the subjects’ effort in behalf of their partner would vary with how soon the experimenter would become aware of their performance. The performance situation was the same as that employed in our other studies. Experimenter awareness was varied by telling half of the subjects that the experimenter would keep track of their productivity throughout the session ( High Experimenter Awareness), while the remaining subjects were told to place their work output in a closed carton that supposedly would not be opened until the end of the semester more than a month away (Low Experimenter Awareness). Questionnaire responses made at the end of the session indicated that the subjects accepted this information. The sex of the subject as well as that of the experimenter was also varied in a 2 X 2 X 2 factorial design. Analyses of the increase-in-productivity scores (above that in the practice period) revealed a highly significant main effect for the dependency manipulation, and showed that none of the other variables interacted significantly with the dependency variable. Thus, regardless of the level of the experimenter’s awareness, or the sex of the experimenter or of the subject, the people who believed that their partner was highly dependent upon their performance worked harder for him than the subjects who believed that their partner was not too dependent upon them.

72

LEONARD BERXOWITZ

A statistically significant interaction indicated, however, that the other experimental manipulations did influence the students’ effort to some extent. When the experimenter and subject were of different sexes, the subjects worked harder for their partner the sooner the experimenter would see their work; they wanted to gain the opposite-sex experimenter’s approval if he (or she) would learn of their effort right away. By contrast, if the experimenter was of the same sex, they actually expended less effort if the experimenter would soon see their output. It is as if the subjects said to themselves, “The experimenter (of my sex) expects me to work hard in behalf of my partner, but I’m going to show the experimenter that I am not so easily pushed around.” Needless to say, such an interpretation must be checked by further research. It does suggest, nonetheless, that people may resent the implied expectation that they should go out of their way to help others, Indeed, in accord with Brehm’s (1986) reactance theory, I shall propose later that these perceived expectations at times give rise to a dislike for those seen as creating the unwelcome pressure and a resistance to these felt demands, especially if there is a readily available excuse for not assisting the dependent persons. At any rate, in order to minimize the subject’s desire to gain the experimenter’s approval, all of our later experiments employed investigators of the same sex as the subject. Any reactance aroused in this research should oppose the subject’s inclination to be helpful. 3. Personal Responsibility and the Other’s Dependency Schwartz ( 1970b) has advanced a decision-making analysis of moral behavior that is closely related to the present conception of dependency and helpfulness. He noted that an individual will not necessarily aid others even if he strongly subscribes to the kind of ideal I have been discussing. Other things must also happen if this social standard is to be translated into behavior on any given occasion. First, Schwartz (1970b) maintained, the individlial must become aware that certain actions in this particular situation have “consequences for the welfarephysical or psychological-of the other person [p. 1281.” Second, the individual must then decide that he has a personal responsibility “for these actions and consequences to which the norm refers [p. 1301.” This second step clearly involves the individual’s recognition that other people are dependent upon him if their well-being is to be preserved or enhanced. In one test of his formulation, Schwartz (197Ob) demonstrated that male university students’ verbalized acceptance of various moral

FACTORS AFFECTING HELPING AND ALTRUISM

73

norms was most strongly related to their peers’ descriptions of how they actually behaved when (1) they were generally sensitive to the feelings and thoughts of other persons and ( 2 ) believed they had a personal rcsponsibility for actions having interpersonal consequences. This latter measure assessing the individual’s readiness to accept personal responsibility was used i n a second study (Schwartz & Clausen, 1970) in which the subjects heard a peer in a n adjoining room have an epileptic seizure. A much greater proportion of the high than low scorers on this scale werc overtly affected by this emergency and took direct action to aid the victim. Latank and Darley (1970) also analyzed bystander reactions to emergency situations as a series of decisions involving, in part, the feeling of personal responsibility. Brfore someone will intervene to assist the victim, they pointed out, he must ( a ) notice the incident, (11) interpret it as an emergency, ( c ) decide he has a personal responsibility to act, and ( d ) believe he can carry out the required behavior. Rephrasing this analysis, we can say the person who recognizes the emergency realizes Something is happening that affects another’s welfare (Schwartz’s first step), and also thinks the victim is highly dependent upon him for aid. According to several experiments reported by Latank and Darley, people are less apt to intervene in emergencies if strangers are nearby than if no one else is present: The strangers might be distracting so that the event isn’t noticed. Also, we often try to act in a cool, calm, and collected fashion when strangers are present so that our unemotional or restrained behavior in the face of the emergency might convince another onlooker that maybe nothing really bad is happening after all. Perhaps most important of all, any one bystander might think his own personal responsibility to act is lessened if there arc other pcople around who might also do something. In our present terms, this bystander believes the victim is not highly dependent upon him alone; thc diffuscd responsibility also implies a lesser perceived dependency. The evidence is quite consistent even though the nature of the dependency and the kind of help that may be given is not always the same (see Midlarsky, 1971). According to many different studies, information that another person’s well-being is dependent on our efforts in his behalf often motivates many of us to help him [also see Krebs (1970, pp. 277-279) for a summary of relevant research]. It is far from certain, however, that this dependency effect arises because of the operation of moral norms. There may be a widespread ideal in our society calling for aid to people in need, but research has not shown that it is this ideal that prompts whatever help is given.

74

LEONARD BERKOWITZ

B. A DIRECTTESTOF THE ROLEOF MORALNORMS The normative analysis of reactions to another’s dependency rested on two kinds of evidence. One, as was just indicated, many subjects exerted considerable effort for a peer who needed their assistance even when there was little “in it for them.” Two, the subjects typically verbalized this kind of social standard at the conclusion of the work session when they indicated, in response to questionnaire items, that they believed one ought to help the person who was dependent upon them, and that “people would expect them to aid this individual.” As Latan6 and Darley ( 1970) have suggested, however, these normative statements could have been advanced after the fact in an attempt to place the behavior in the best possible light. Many people have learned to give normative explanation for conduct that actually is determined by a complex interplay of many influences. Given this type of difficulty, direct experimental tests are required in which normative conditions are deliberately varied so that we can then observe the effects of this manipulation upon dependent measures of helpfulness. Berkowitz and Macaulay attempted such a test by varying the salience of the conjectured social responsibility, or helpfulness norm. In the course of an interview at one of the Madison shopping centers, women were individually asked questions either about (1) their own and other persons’ helpfulness, or ( 2 ) the extent to which they were “smart” consumers. The people in the former condition presumably would be reminded of the social responsibility norm and, as a consequence, should be more helpful soon afterwards than the women in the other condition. In addition, half of the subjects in each of these two conditions were given frequent approval by the interviewer (for either help-related or smart-consumer answers, depending upon the condition) in order to see if rewarded helpful ideas would lead to the greatest amount of subsequent helpfulness. Finally, a fifth control condition was also established in which the women were not interviewed. Within a minute of the end of the interview a male college student, who had been waiting about 50 yxds away, approached the respondent without knowledge of the particular interview condition imposed on her, He said his wallet had disappeared and asked if he could borrow 40 cents for a bus ticket to a nearby community. Table I1 shows the number of respondents in each condition and the percentage of these who gave money to the supplicant. As the table indicates, the initial expectation was not upheld; there was no significant effect attributable to the nature of the immediately preceding interview, and the interview type did not interact with the experimenter’s approval.

75

FACTORS AFFECTING HELPING AND ALTRUISM

TABLE I1 PERCENT GIVINGMONEYTO SUPPLICANT IN CONDITIONS VARYING IN TERMS OF INTERVIEW CONTENT AND INTERVIEWER’S APPROVAL Helpfulness interview Rewarded Total N Proportion giving

Not rewarded

17 657, X*

20

30%

Consumer interview Rewarded

Not rewarded

NO interview

16 56%

16 31%

20 55%

Analysis of Experimental Conditions

~

Source

df

X2

Interview type Reward-no reward Interaction

1 1 1

ns. 6 . 2ga n.s.

“ p = .02. Data from Berkowitz and Macaulay (unpublished experiment).

However, regardless of the interview topic, the women who had been rewarded with the interviewer’s approval were much more apt to donate money than were the nonrewarded subjects. Comparing the Not-rewarded groups with the No-interview control condition, it appears that the interviews had generaIIy lessened many of the women’s willingness to help the student ( p < .lo). The interview was apparently an imposition that a good proportion of these people resented. This feeling of annoyance could have dampened their inclination to aid the supplicant. The interviewer’s approval may have counteracted this resentment, probably by making them feel good, so that they returned at least to their baseline level of helpfulness. While this particular investigation did not obtain any support for the normative conception of helpfulness, a second field experiment by Macaulay and Berkowitz indicated rewarded ideas can affect later aid. The subjects again were women shoppers at local shopping centers. Only the helpfulness interview was used, but the nature of the responses receiving approval differed: In one condition, only helpful answers were rewarded by the interviewer, and the interviewer tried to promote expression of ideals of responsibility and charity. In yet another condition, by contrast, only selfish and cynical answers were given approval, and the interviewer tried to elicit expression of the importance of avoiding social involvement, of minding one’s own business, of not interfering

76

LEONARD BERKOWITZ

in other people’s troubles, and so on. A Not-Rewarded but interviewed group served as the control condition. Since the two interviewers employed in this experiment often found it difficult to elicit a sufficient number of selfish remarks from their respondents, they rated how successful they were at getting the appropriate kinds of statements from the subjects. We kept separate those women who, because they had expressed relatively few of the desired remarks, were judged to have been failures. Table I11 shows the number of women in each group, and the percentage of subjects who gave money to the student asking for the 40 cents bus fare. Even though the women did not connect the preceding interview with the supplicant’s request, the interview condition influenced their donation rate. As the table indicates, the respondents were least likely to help the student if the interviewer had rewarded selfish, noninvolvement statements just before the request was made. The approval for help-related remarks raised the donation rate somewhat above that in the control condition, but not to a significant extent. Again, we have no support for the normative conception. There is also a question as to why the Reward-Noninvolvement treatment had lowered the women’s readiness to aid the supplicant. One obvious possibility, of course, is that the interviewer had made the subjects more willing to act selfishly when she had approved selfish ideas. She had indicated, essentially, that it was permissible not to become invoIved with other people and help them when they were in trouble. Whatever merit there is to this possibility, other findings suggest that the experimental treatment had also affected the subjects’ mood. The supplicants’ postrequest ratings of the women’s manner toward them indicated that the donors were less pleasant in the Reward-Noninvolvement condition than in the Reward-Helpfulness and Not-Rewarded groups. The women who TABLE 111 PERCENT GIVINQMONEYTO SUPPLICANT AS A FUNCTION OF INTRRVIEWBR’S AND TYPEOF IDEA REWARDED APPROVAL Reward helpful ideas

Total N Proportion giving

Good subjects 29 66

Reward noninvolvement ideas

Failures

Good subjects

Failures

Interviewed-no t rewarded

2 50

21 28

18

28

56

50

Good subjects only: Chi-square = 6.64, 2 df,p Berkowitz (unpublished experiment).

< .05. Data

from Macaulay and

FACTORS AFFECTING HELPING AND ALTRUISM

77

had been induced to express selfish, noninvolved ideas were evidently less pleasant to the college student afterwards than were most of the other women, even though all had helped him. When the former had given money, they had apparently done so only grudgingly. All in all, these results suggest that the respondents’ rewarded verbalizations had evoked ideas and feelings within them which influenced their behavior soon afterwards. The internal reactions interfered with semantically incompatible behavior (as in the case of the approved selfish statements lessening helpfulness ) and possibly facilitated actions having the same general meaning ( rewarded helpful remarks apparently leading to greater helpfulness ). Nonetheless, the findings do not provide any clear-cut support for the normative analysis of help-giving. Reminding our adult women of the presumed cultural ideal calling for aid to those requiring assistance did not in itself produce a significant increment in the number of helpful subjects. I am reluctant to rule out the operation of moral norms on the basis of these negative results, however. Rather than being nonexistent, the conjectured helpfulness norm is probably only a relatively weak determinant of help-giving in most people in many situations. As Schwartz (1970b) had noted, few persons will expend much effort in behalf of others solely because our society maintains that help ought to be given to those in need, even though they generally accept this norm. In the case of this particular moral standard at least, other factors apparently also have to operate in the same direction if the norm is to influence behavior successfully.

C. OBSERVATIONAL INFLUENCES PRESUMABLY INCREASING THE SALIENCE OF THE HELPING IDEAL If awareness of someone’s need for assistance can activate an ideal prescribing that he be helped, might not the sight of a helpful action also heighten awareness of the moral standard? Seeing this help, we might be reminded that we ought to aid those who are dependent upon us. Daniels and I (Berkowitz & Daniels, 1964) followed this reasoning in trying to show that a prior experience of being helped can enhance motivation to help a highly dependent person. The subjects, female college students, wcre first given a preliminary job altogether different from our standard task, with a peer (actually the experimenter’s confederate) either “voluntarily” helping or not aiding each person during this first phase. After this, the subjects were assigned our usual “worker’s” role, supposedly under the supervision of another person (not the confederate) who was said to be either highly or less dependent upon them. The students exhibited the greatest effort

78

LEONARD BERKOWITZ

for the highly dependent supervisor when they previously had been helped by the confederate. Encouraging though these results might be, they can be explained in at least two different ways. First, the prior help could have increased the subjects’ awareness of the helping ideal, as we had hypothesized. But, second, the subjects’ experiences in the first phase might have made them feel either pleasant or unpleasant-pleasant if they had been helped, and unpleasant if they had not been assisted. These feelings could have influenced their subsequent willingness to aid their dependent supervisor. We will consider the first possibility here and touch on the latter alternative later. An increasing number of social psychological investigations have demonstrated that the sight of a person (the model) carrying out a particular action can indeed heighten the onlookers’ own inclination to behave this way themselves (see Krebs, 1970, pp. 267-277). As an example, Blake, Mouton, and Hain (1956) and Helson, Mouton, and Blake (1958) showed that a confederate’s compliance with a request to sign a petition increased the number of onlookers who were also willing to sign petitions. Similarly, Rosenbaum and Blake ( 1955) and Rosenbaum (1956) found that a model who volunteered for an experiment increased volunteering among people who saw the model. In the helping realm, Bryan and Test (1967) reported that a donating model raised the rate of donations to a Salvation Army bellringer, while Wheeler and Wagner (1968) showed that a generous model increased contributions to a charitable cause, and that a selfish iuodcl lowered such donations, particularly when the cost of the donations was relatively slight. The model’s behavior in these instances probably served two somewhat related functions: The action reminded the observers of what they themselves ought to do in the given situation, and also helped define what was proper behavior. This last function occurs when people are uncertain as to what is the “right” thing to do. In the Wheeler and Wagner (1968) study, to illustrate, the sailors approached to give money to the charitable cause may not have been sure that they ought to donate, and in any case, probably were reluctant to give their money away. The generous (donating) model’s actions then made them more conscious of the ideal that “people should help those in need,” and also told the men, “This is a worthy cause, one that falls under the scope of this inoral standard.” The selfish model who refused to donate, on the other hand, cast doubt on the propriety of giving money to the particular cause (“This may not be a worthy charity.”) and also gave the sailors a legitimate excuse for avoiding what they had wanted to avoid-giving away money.

FACTORS AFFECTING HELPING AND ALTRUISM

79

Many times, however, there is little ambiguity in the situation. Most people may believe that only a certain course of action is correct on this type of occasion. Jacqueline Macaulay (1970) reasoned that a model’s behavior in such relatively clear-cut situations would generally call attention to the social ideal even when the model deviated from it. In her experiment, carried out as part of the Wisconsin research program, R model either placed money in a Volunteers of American Christmas season box or loudly insisted that she would not donate to this charity. The donating model produced the greatest number of subsequent donations, but the refusing model also led to an increased number of donations in comparison to the number occurring in appropriate control periods. Interviews with nongivers who had seen the model also indicated that the model’s behavior had made them more aware of the ideal, “one should give to Christmas charities.” Most of the women shoppers involved in this experiment evidently believed in Christmas giving. The model’s action called attention to this ideal leading generally to increased adherence to it, although the model’s refusal apparently also legitimized nongiving for a small proportion of people.

1. Ideals and Self-lmuges The discussion so far has been based-admittedly loosely-on the interrelated concepts of “group norms,” “social ideals,” and “moral standards.” However termed, our reasoning has assumed that many persons have adopted a widely shared rule of conduct. People who give help even when there are no external rewards presumably adhere to this type of social code in order to avoid feeling guilty. Other processes may also operate, however, in addition to or instead of these moral norms. Bryan and Test (1967) suggested one possibility. The people who saw the model drop money into the Salvation Army kettle may have been reminded not of a general rule, but of how they wanted to think of themselves. Assuming that they wanted to look at themselves as “good,” “kind,” and “helpful,” they became fully aware of what they expected of thcniselves, and then donated in order to be consistent with this self-image (or self-expectation). Social scientists have long employed this kind of reasoning in accounting for man’s actions toward his fellows. Whether their thinking is traced to Cooley, Festinger, or Rogers, many writers have proposed that people often want to be true to themselves-as they see themselves. According to Reckless, Dinitz, and Kay (1957), as an example, boys living in high delinquency areas but who were regarded as probable nondelinquents by their teachers generally had self-concepts which “insulated” them from criminal actions. More recently, Aronson and

80

LEONARD BERKOWITZ

Mettee (1968) reported that female college students who were induced to think ill of themselves were more apt to cheat in a later card game than other students who were led to have high self-esteem or whose self-concepts were not modified. The cheating supposedly was consistent with the former girls’ unfavorable picture of themselves. They cheated, or did not resist the immoral temptation, because this behavior was not incompatible with their expectations for themselves. I do not question the basic validity of this type of analysis. Some purported demonstrations of this phenomenon provide only equivocal evidence, however. They fail to consider other factors which could have led to the apparently %onsistent” behavior. To illustrate, take the previously mentioned Aronson and Mettee (1968) experiment. The treatment creating the temporarily low self-esteem-a fake personality evaluation-undoubtedly also made the girls unhappy. Feeling bad, they may have been relatively unable to withstand the temptation to cheat. It may not have been their poor self-image per se that had led to the cheating, but the unhappiness produced by the esteem-lowering information. 111. Moods and Feelings

A. GOODAND BADMOODSAS DETERMINANTS OF HELPFULNESS

The previously reported field experiment by Berkowitz and Macaulay had obtained some indications that a person’s mood can affect his readiness to aid others. In that study, you will recall, women who had just been interviewed were much more likely to donate the money later asked of them if the interviewer had approved the respondent’s answers than if she had not given this approbation. The interviewer’s rewards had evidently put the women in a good mood which then heightened their. willingness to assist the supplicant. One of our laboratory experiments also demonstrated that prior success or failure on an assigned task can also influence.helpfulness, presumably through altering the person’s mood. In this investigation (Berkowitz & Connor, 1966) 108 male college students were assigned to the nine conditions of a 3 X 3 factorial design. Two thirds of the men were individually made to succeed or fail on a preliminary task (thereby either earning or not receiving a dollar prize), while a control group did not have either experience. Immediately after this, each subject was led to believe that he would be a “worker” whose task would be to make envelopes under the guidance of a peer operating as the “supervisor.” The supervisor was eligible to win a $5.00 prize,

81

FACTORS AFFECTING HELPING AND ALTRUISM

but there would be no prize for the subject. In order to vary the supervisor’s dependency on the subject, the subject was informed that either 20%, SO%,or 80% of the supervisor’s grade would be based on his productivity. The motivation measure is the number of envelopes made by the subject in his peer’s behalf during the 10-minute work period, adjusting this score (by covariance) for the productivity during a 5-minute practice period given before the experimental inanipulations were established, Both the prior-experience and dependency-levels treatments yielded significant effects by anaIysis of covariance: (1) the people given the success experience generally worked harder for their supervisor than did the control subjects, and ( 2 ) greater effort was expended for this person as his dependency upon the subjcct increased. Inspection of the condition means given in Table IV suggests, furthermore, that both the success and failure experiences affected the magnitude of effort exerted in behalf of the person requiring aid: The differences among the three dependency levels were somewhat less after the failure experience than after the Success or Control treatments. Thus, where the failed subjects working for a highly dependent partner were only slightly more productive than the Failed-Low Dependency group, the difference between the High and Low Dependency conditions was much greater for both the successful and control subjects. Moreover, the people who had failed on the preliminary tasks worked significantly less hard for the supcrvisor who was very highly dependent upon them than did the successful subjects operating under the same dependency level, but there was no reliable difference between this latter successful group and the comparable control condition. In general, the prior failure TABLE IV MEANSIN CONDITIONS D I F F E R I N G IN TERMS OF P.\RTNI:R’S DEGREE O F DEPENDENCY ON SUBJECT AND SUBJECT’S PRIOR EXPERIKNCF.

ADJUSTED

PRODUCTIVITY

Dependency level of partner Subject’s previous experience

20% (Low)

50%

80% (High)

Total

Success Failure Control

24. Gnbo 21 . 7 u b 20. I,

26.01, 22. &b 22.2ab

29.2, 23. 2ai, 24. 6.bo

26.6, 22.5, 22.3,

Total

22.1,

23.7,,

25.7,

Cells having a subscript in common are not significantly different at the .05 level by Duncan test. Separate tests were made for the column totals and for the row totals. Data from Berkowitz and Connor (1966).

82

LEONARD BERKOWITZ

apparently lowered the present subjects’ willingness to help the person who was greatly in need of their assistance. Questionnaire results suggest that the High Dependency condition created resentment after the frustration experience. In one item, the subjects were asked to indicate the extent to which they liked or disliked participating in the experiment. The failed men expressed a reliably greater dislike for the experiment the more their supervisor could have benefited from their effort. Consequently, in the High Dependency group, the failed subjects were significantly more unfavorable to the experiment that either the Success or Control subjects, although there were no such reliable differences at the 20%or 50% dependency levels. The high dependency presumably aroused strong feelings of obligation which were annoying to the frustrated subjects. Some of this resentment was directed against the highly dependent supervisor. When asked to give their first impression of their supervisor, the previously frustrated men who believed that their peer was very highly dependent upon them generally expressed a stronger dislike for this person than either the failed subjects in the 50%dependency group or the successful subjects whose supervisor was only slightly in need of their help (Le., the 20% group ) Probably because of the resentment arising from the relatively strong pressures they felt, the failed subjects were somewhat more likely to deny their obligation to work for their supervisor the more dependent he actually was upon them (in their responses to the question, “How hard should a worker work for a supervisor in this situation?”). By contrast, in both the Control and Success conditions, there was a nonsignificant tendency for the subjects to report a stronger obligation the higher the dependency. Thus, under both 50% and 80% dependency, but not in the case of the 20%dependency groups, the successful men differed significantly from the faded subjects in the obligation they believed the worker should have toward his supervisor. All in all, the general pattern of findings suggests that the frustrated subjects were more resentful of the felt obligations, and that the successful men were also somewhat less resentful of these pressures than were the control subjects. Isen (1970) extended the Berkowitz and Connor findings to a broader variety of helpful actions (donating money to a school fund and assisting a young woman with an armload of books and cartons). In three separate experiments having school teachers and university students as subjects, she also demonstrated that people who had just succeeded on an assigned task were niore inclined to aid others than were persons who had not done well. Also in general agreement with the Berkowitz-Connor results, Isen’s data indicate that the success experience had exerted the a

F A a O R S AFFECTING HELPING AND ALTRUISM

83

strongest effect on the subsequent helpfulness. She therefore concluded that the people who had succeeded on their assignment had experienced a “warm glow of success” which made them more generous and helpful toward others. However, we still do not know why this “warm glow” leads to greater helpfulness. One good possibility is that the earlier experience affects the potential helper’s frustration tolerance or willingness to accept restrictions on his freedom of action. He is more tolerant of the demands the help request imposes upon him than he otherwise would be if he is happy, and probably is less accepting of these pressures if he had just had an unpleasant failure. The individual’s mood might also affect his empathic capacities. As we will see later, a person is relatively likely to assist someone requiring aid if he empathizes with the distress felt by the victim. Success might enlarge his ability to empathize with others, while the unhappy failure could conceivably restrict it. B. GUILT

1. Compensating for One’s Own Wrongdoing All unpleasant feelings do not dampen helpfulness. The guilt produced by the violation of one’s own moral standards sometimes motivates aid to other people even though we might feel bad [see Freedman (1970) for a summary of some relevant research]. In one of the first experiments dealing with moral transgressions and helpfulness, Darlington and Macker (1966) found that persons who believed they had harmed someone else were relatively quick to agree to donate blood to a local hospital. The blood donation would apparently compensate (to some extent at least) for the injury the subjects had just done, even though it was highly unlikely that the person they had harmed would benefit in any way from their gift. Similarly, when college students in another study were asked to make telephone calls in behalf of a fictitious committee to save the California redwood trees, those people who had just given a peer electric shocks in an experiment agreed to do more work than did other men who had given their peer only loud buzzes (Carlsmith & Gross, 1969). The more the students thought they had hurt their partner earlier, the more they evidently tried to make up for this by acceding to the request to aid the ecology group. But even though the guilt feelings had evidently prompted helpfulness in these studies, other findings suggest the victim of the injustice is not necessarily the preferred beneficiary for the expiatory assistance. Freedman, Wallington, and Bless (1967) reported that a moral transgression generally heightens willingness to help someone other than the injured individual but not necessarily the one sinned against. As the

84

LEONARD BERKOWI’IZ

source of the unpleasant guilt feelings, the victim might become unpleasant himself so that he is aided only very grudgingly. Also pointing to the aversive nature of the victim, the girls in the Freedman et al. study were more willing to help the person they had accidently injured if they thought they would not meet him than if they believed they would see him again. However, another possible explanation for the reluctance to aid the injured person is suggested by the procedure Freedman et al. followed in arousing guilt: The victim could have easily learned that the subject had harmed her. As a consequence, the guilty subject might have refused to comply with the request for help primarily in order to avoid an unpleasant scene in which she would be reprimanded or even insulted, rather than because the guilt arousal in itself had made the victim an aversive stimulus. Data recently collected by Susanna Neumann at the University of Wisconsin supports this latter explanation of the Freedman et al. findings. Neumann led her subjects to think they were participating in a memory experiment, and then aroused guilt in half of these women by having them “overhear” the experimenter explain to someone else how to obtain a good score. The subjects in this condition did not voluntarily confess to the experimenter, probably because they wanted to impress her by doing well on the memory task. At the conclusion of this manipulation the subject was given an opportunity to aid someone-either the psychologist sponsoring the memory experiment or someone else-and could do this by agreeing to make 50 telephone calls for an ecology survey (the same help measure used by Carlsmith and Gross). In Neumann’s procedure, the experimenter would not know, at least for sure, that the subject in the Guilt Arousal condition had wrongly hurt her, although the subject, of course, was aware of her transgression. Being less afraid of reprisal, the guilty women were more apt to assist their victim than someone else (57.5%of the guilty subjects agreeing to help the victim as against 40%complying when someone else wanted aid). The experiment-sponsor was also more likely to be helped by guilty than by nonguilty subjects (only 37.5% of this latter group agreed to help). In general, then, while guilt arousal in this study did not produce a heightened acquiescence to any request for aid, it apparently did prompt many of the subjects to compensate the person they had hurt-as long as this encounter was not unpleasant,

2. Seeing Others Zniured It is altogether likely, as I have suggested, that the help was given in the above studies because the subjects thought they had injured

FACTORS AFFECTING HELPING AND ALTRUISM

85

someone. Working within the Wisconsin program, Rawlings ( 1970) hypothesized that people might feel bad, and consequently aid others, merely as a result of seeing someone hurt, even when they themselves were not responsible for the injury. The perception of someone being seriously injured can increase the salience of the moral standard saying that people should not be hurt. Female university students in Rawlings’ study were told that every time they made an error on a preliminary task their partner would receive an electric shock. A second group of girls was led to believe that there was no relationship between the subject’s initial performance and the shocks administered to the partner. In both conditions, the partner presumably was shocked on 10 of the 16 trials. For the subjects in a third condition, no other person was injured. They were informed whenever they made an error on these preliminary trials and, as in the first group, they supposedly made 10 errors, but their partner was never shocked. Following this initial treatment, each subject was paired with another person and now could regulate the duration of shock each pair member-the subject and her new partner-would receive on each of 16 trials on an entirely different task. In controlling these shock durations, the subject knew the longer was the shock she accepted for herself, the shorter would be the shock given to her partner. Rawlings found that awareness of injury being inflicted on one person resulted in increased altruism toward a second individual, even when the subject had not been responsible for the first partner’s being hurt, and the second partner was not the person who had been shocked earlier. Both the group responsible for the first partner’s injury and the group told only that their first partner had received shocks accepted longer shocks for themselves on the second task than did the control subjects. In interpreting her results, Rawlings invoked the notion of “anticipatory guilt.” Seeing another girl being hurt in the first phase of the study had presumably lowered the subjects’ threshold for guilt arousal; their earlier experience had supposedly made them highly conscious of the ideal that people should not be deliberately injured, so that even thoughts of violating this now highly salient moral standard-by sending the second-phase shocks to the new partner rather than to themselvesaroused fairly strong guilt feelings. The guilt was produced by anticipation of a “ b a d action. Goranson has obtained comparable results in an experiment dealing with filmed violence. After being deliberately angered by the experimenter’s confederate, the subjects, male university students this time, watched a prize fight scene from the movie “Champion” in which an actor received a bad beating. Different conditions were established by orally giving the subjects different kinds of information as to what

86

LEONARD BERKOWI’IZ

supposcdly had happened next in the movie. After this information was provided, each subject had an opportunity to give electric shocks to the confederate. The angered men told that the protagonist died in the film administered fewer shocks to their tormentor than did the control group, even when this death was described as entirely unrelated to the witncssed beating. Thc information about the film character’s death evidently had evoked inhibitions against aggression, perhaps because this information had aroused the subjects’ ideal the people should not be hurt. At this point we should recognize an important limitation. Learning that someone-anyone-has been injured may not always lead to this injury-inhibiting effect. Thc first person’s perceived injury probably must be regarded as being unwarranted or improper before it will produce the inhibition of other injurious acts. In other words, if the subject believes that Person A has been seriously harmed and does not deserve this deep hurt, he will then be inclined to avoid hurting Person B. The first unjustified injury prcsumably defines the anticipated injury to B as “bad.”

C. EMPATHY To return to Rawlings’ (1970) experiment, in addition to postulating an anticipatory guilt mechanism, she also suggested that sympathy or empathy might have been in operation. “As a result of having watched their first partner being shocked, subjects might have anticipated the pain that their new partner would feel and thus, out of empathy, tried to minimize her pain [p. 1731.” Several psychologists have placed particular stress on the role of empathy in helpfulness and altruism. Aronfreed (1970) has even contended that the term “altruism” should be confined to those actions that are controlled by empathic processes. He insisted (as I do) that in talking about altruistic behavior it is important to show the given action is not impelled by the hope of benefits from others. However, where I suggest that altruism can bc prompted by internalized moral standards and the desire to lessen or avoid guilt, Aronfreed would restrict this concept solely to those instances in which the individual’s behavior is affected by his experience of another person’s emotions. More specifically, according to Aronfreed, an altruistically-inclined individual has learned to associate his own emotions with cues signaling someone else’s feelings. He will try to alleviate another person’s distress because the victim’s pleasure at being helped makes him happy as well. However altruism is defined theoretically, a number of experiments by Aronfreed and others indicate that empathy contributes to helpfulness. But where Aronfreed had emphasized the helper’s experience of the

FACTORS AFFECTING HELPING AND ALTRUISM

87

uictim’s feelings, other research suggests that the potential benefactor can be affected by empathy with the person in need OT with other helpful people. Hornstein ( 1970) has reported an intriguing experiment carried out in midtown Manhattan which illustrates this latter influence. He and his colleagues dropped envelopes containing a man’s wallet ( containing a check and $2.00 in cash) and an accompanying letter, and then ascertained whether the letter’s contents influenced people finding the envelope to return it to the ostensible wallet owner. In all cases the letter showed that someone had found the wallet and was about to send it back to the owner (but, unfortunately, had evidently now dropped the package). Sometimes, however, the letter writer (who serves as the “helpful model”) said he was a foreigner, while in other cases he apparently was an American. In other words, the helpful model was either different from or similar to the great majority of the envelope finders. Within each of these two conditions, the letter also indicated that the model had either been happy or unhappy about returning the wallet or expressed no feelings at all about this. The model’s feeIings affected the finders’ subsequent behavior, but primarily when he was similar to them. Of the 105 wallets dropped, Hornstein tells us, 40%were returned intact. The greatest proportion of these were in the Similar Model-Happy Mood group (70%) while the wallets were least likely to be returned if the helpful model had also been similar to the typical finder but expressed unhappiness about, having to send the wallet back (10%).The model’s similarity to the finders had apparently facilitated their acceptance of his feelings. So, if he was unhappy about sending the wallet back, they were too, and many didn’t even bother to mail the envelope to the owner. In the Hornstein research the potential benefactors were only able to observe the helpful model and were not given any cues about the victim’s feelings. However, on many occasions we can observe the person in need as well as the actions of the other possible helpers in the situation. According to Aderman and Berkowitz ( 1970), what we then do, whether we ourselves provide help or not, depends partly on the focus of our attention. Attending to the potential helper facilitates empathy with him, while focusing on the victim enhances the acceptance of his feelings. In this experiment male university students, run in pairs, were told they were participating in a preliminary investigation of “social observation.” The subjects supposedly were only to give their reactions to a brief tape-recorded conversation between two men so that the investigators could work the bugs out of the procedure for later research, Just before they heard the recording, written instructions (with the experimenter being blind to the content) asked the men to concentrate on one

88

LEONARD BERKOWI’IZ

of the two students involved in the encounter. His name was given, and this later turned out to be either the student in need or his potential benefactor. They then heard one college student tell another about his library assignment, and it became evident that the other man could help his friend. Variations in the conversation created three types of outcomes: ( a ) the potential helper refused to assist the other student, ( b ) the potential benefactor offered to help for a while, and ( c ) the potential helper volunteered to aid his friend and was thanked for his offer. At the end of the tape the experimenter asked the subject to fill out a questionnaire, indicating his mood at that moment (and not the supposed mood of the individual in the recording). Then after this, supposedly to complete the study, the experimenter said he would interview each subject about the tape recording in order to see how people reacted to it, and each person went to a separate room. Once they were settled in these rooms, the experimenter briefly visited each subject, “explaining” he was going to interview the other man first. While he waited, the experimenter went on, would the subject do him a favor and score data sheets that had been collected in another study? Our measure of the subject’s helpfulness is the number of pages he completed in the experimenter’s behalf. The results are given in Table V. As the first line indicates, the subjects exerted the greatest effort for the experimenter after ( a ) they had attended to the student needing assistance and this person had not been aided, or ( b ) they had concentrated on the potential helper, and this individual had volunteered to TABLE V HELP AND MOODSCORESIN CONDITIONS DIFFERINQ IN Focus AND OUTCOME OF SCENE ATTENTION Attending to person in need Measure

No help outcome

Help S C O W Sadness moodb Angry mood0 Pleasant moodd

72.8b 2.5b 2.8. 2.1.

66.4-b 1.6. 1.7&b 2.6b

SUBJECT’S

Attending to potential helper

Help Help and No help outcome thanks outcome 65.2,b 1.6. 1.3, 2.5.b

OF

55.5, 1.9. 1.8b 2.8b

Help Help and outcome thanks 62.8.h 1.9. 1.8b 2.3,b

73.8b 1.5, 1.6~13 2.8. _____

The help measure is the number of pages the subject scored in doing the experimenter a “favor.” * Self-ratings on the Nowlis Mood Questionnaire items: blue, sad, and lonely. c Self-ratings on the items: rebellious, angry, defiant. d Self-ratings on the items: refreshed, pleasant, light-hearted. Cells having different subscripts are significantly different at the .05 level by Duncan range test. Data from Aderman and Berkowitz (1970). 4

FACTORS AFFECTING HELPING AND ALTRUISM

89

help and had been thanked. These conditions had evidently facilitated the empathic reactions which spurred the listeners to be helpful to another person soon afterwards. Looking at the mood ratings in Table V, we can see that the subjects had felt relatively sad, angry, and unpleasant if they had attended to the student in need, but generally described themselves as feeling relatively good if they had focused on the helpful model who had been thanked. Experiencing the victim’s distress or the pleasant feelings of a rewarded benefactor had heightened the subjects’ motivation to aid others.

D. SELF-CONCERN Altruism is often contrasted with selfishness or self-preoccupation, as if the altruistic individual were not concerned about his own desires and his self-image. According to this commonplace analysis, egocentrism, a preoccupation with one’s value as a person, lessens the individual’s inclination to help others unless he believes it is to his benefit. There are several reasons why this conception might be correct, it seems to me. For one, self-concern might interfere with empathy; the person who thinks only of himself, his hopes, his fears and uncertainties, is not likely to imagine or experience the joys and anxieties other people might be feeling. Then too, the self-concern could be decidedly unpleasant. As a result, the individual .who is unsurc of his self-worth might be in too bad a mood to help others (unless he thinks there is something in it for him). The self-preoccupation might also interfere with other thoughts, preventing the individual from remembering the helpfulness norm. Wrapped up in his doubts, he temporarily forgets that he ought to aid people in need.

1. Failure and Self-Concern Whatever the exact cause, the self-concerned person is apt to resent the obligation to assist others. One of our previously mentioned studies suggests this. If wc assume the failed subjects in the Berkowitz and Connor (1966) experiment had become somewhat unsure of their worth, their relatively low output for the highly dependent supervisor could have stemmed from a self-concern-generated resistance to the pressures implicitly placed on them. Recall also that the failed subjects operating under the pressure of their supervisor’s great dependency on them were reliably more unfavorable to the experiment than either the successful or control subjects who were also in this High Dependency condition, The other person’s great need for their work had evidently aroused strong feelings of obligation to him which were particularly annoying to the failed and probably self-preoccupied subjects.

90

LEONARD BERKOWITZ

2. Eualuation-Generated Self-Concern An essentially similar phenomenon arises when the individual thinks he is being evaluated along some dimension that is important to him. This evaluation might also make him relatively self-concerned, leading to an enhanced resistance to felt obligations. An experiment by Brehm and Cole (1966) growing out of Brehni’s reactance theory is illustrative. The male undergraduates in this investigation were first given either an innocuous introduction to the study that played down its impoitance or were told that their chances of being a success in life would soon be assessed. This latter statement presumably increased the men’s selfconcern. Following this introduction, and supposedly before the experiment proper got underway, another person (the experimenter’s confederate) did the subject a favor; he gave the subject a soft drink. No such favor was given to the other half of the subjects. The experiment then supposedly began, and the confederate was assigned a dull and arduous task, The subject could help him if he desired. The investigators found that 93% of the men reciprocated the favor done them by the confederate and aided that person when the experimental situation was said to be relatively unimportant. Whcn the subjects’ chances of success in life would be measured later, however, only about 131%of the men helped the person who had done them a favor earlier. All in all, the men who were induced to be self-concerned evidently resented the pressure they felt to return the confederate’s favor. As n consequence, they became even more insistent on going against thc implied demand, and thus were least likely to help the confederate. Several of our Wisconsin experiments also point to the help-dampening effects of evaluation-generatcd self-concern. In one of the first of these, by Aderman, Johnson, and Berkowitz, each subject was first asked to make judgments of photographs taken from the University yearbook. Self-concern was aroused in half of the cases by telling the students that these judgments would serve as a reliable indicator of their “social intelligence.” In the remaining cases, intended to be low in self-concern, the experimenter was ostensibly only interested in student preferences. When the judgments had bcen made, the experimenter said they could not proceed with the main part of the study since the person who was supposed to be the subject’s partner had failed to show up. The subject would get credit for participating in the study, the experimenter went on, but since he had the time, would he help the experimenter by scoi-ing data from another study? As they walked to the location where the work was to be done, the experimenter introduced the second variation by saying that he was either an undergraduatc like the subject (low status

FACTORS AFFECTING HELPING AND ALTRUISM

91

help recipient) or a graduate student about to receive his Ph.D. (high status help recipient). ( I n order to keep the perceived importance of the task relatively constant, all subjects were told their work would ultimately benefit a graduate student.) The help measure here was the number of pages scored by the subject in the 15-minute period. The data from the male subjects (but not from the women) yielded a significant main effect for the experimenter’s status; the higher his status, the greater the amount of help given him. But more interestingly, self-concern evidently lowered the subject’s motivation to aid the highstatus person. The men made to be concerned about their worth exerted almost as little effort for the high-status experimenter as for the low-status student, and did reliably less work for the high-status person than the nonconcerned subjects in this condition. The self-concern had counteracted the motivational effect of the help recipient’s higher status. Yet another experiment, by Johnson, Hildebrand, and Berkowitz, inquired further into the effects of self-concern on men and women. We wondered if the women in the previous study might not have been insufficiently self-preoccupied, and if a more sensitive performance measure might not have yielded better results for the female subjects. In establishing this improved help-giving index, a baseline work rate was first determined for each subject by asking him to score data for the experimenter for a few minutes until his (supposed) partner arrived. After 5 minutes, the experimenter stopped the subject, saying he guessed the other subject was not going to come, but that they would proceed with part of the experiment nevertheless. The subject was then shown a set of 10 pictures, each portraying a group of people. These were employed for the experimental variations. In two conditions, the subject was asked to identify the group leader and the group attitude toward this person. One of the two conditions was informed that these pictures comprised a test of “social sensitivity,” while the other condition was told that the picture test was a measure of the individual’s “supervisory ability.” Finally, in a third, control condition, the subjects were only to indicate their liking for the pictures. Our expectation here was that the female subjects would become more selfpreoccupied if they thought thc test was a measure of social sensitivity rather than supervisory ability, while the opposite would be true for the men. Then, to make the subjects still more self-concerned, they were told that the experimenter would go off to score the test results and that they would learn how they did in a few minutes. Before departing, however, the experimenter again asked the subjects to help him by scoring the other data he had collected. The influence of the presumed self-concern is assessed in terms of the amount of work done for the experimenter in

92

LEONARD BERKOWITZ

TABLE VI ADJUSTED

MEANAMOUNT OF WORK CARRIED O U T FOR EXPERIMENTER SELF-CONCERN VARIATION, BY SEXOF SUBJECT

AFTER

Nature of picture test

Males

Females

Social sensitivity Supervisory ability Personal preference (control)

8.2.b 6.6, 8.2.b

7.7b 8.9, 8.8,

~

Cells having differentsubscriptsare significantly different at the .05 level by Duncan rmge test. Data from Johnson, Hildebrand, and Berkowitz (unpublished study).

this final 10-minute work period, adjusting (by covariance analysis) for the work rate in the initial, baseline interval. The results are given in Table VI. The analysis of covariance yielded the expected Sex by Experimental Variation interaction. The men increased their output for the experimenter after taking either of the uninvolving tests, the social sensitivity, or personal preference tests. Their work rate was much less, by contrast (and was actually below that in the baseline period) when they thought the test had measured something important about them; their supervisory ability. In the case of the women, as we had predicted, the least amount of work also done after they thought they had taken the test that presumably was important to them, the assessment of their social sensitivity. All in all, the men and women were concerned about somewhat different things, but whatever made them self-preoccupied, the arousal of selfconcern evidently lowered their willingness to help the experimenter. 3. Self-Concern Enhancing Helpfulness Preoccupation with one’s worth doesn’t always interfere with helpfulness. I noted earlier that the self-concerned individual might not behave altruistically but still could go out of his way to assist others if he could gain approval or other rewards by aiding other people. He actually might be fairly slow to recognize this. As I suggested above in discussing the possible effects of self-concern, the individual might be so wrapped up in the question of his value as a person that he does not think of the relevant social ideals at that time, and momentarily forgets that helpfulness could bring approval. However, should he be reminded of the helpfulness norm and its possible desirable consequences for him, his self-concern might now make him very willing to assist others. He now recognizes that he can gain a great deal by helping others, and he strongly wants these rewards from other people in order to reduce his doubts about himself.

.

FACTORS AFFECTING HELPING AND ALTRUISM

93

An experiment by Hildebrand and Berkowitz shows how selfconcerned people can be affected by clear reminders of the helpfulness norm. The women University students in this investigation were run in pairs. They were first seated in separate booths, and were given a questionnaire designed to create differences in their self-concern. The items were the same, but for half of the women a brief introductory statement identified the questionnaire as a test of social sensitivity and social intelligence. For the remaining subjects, on the other hand, the questionnaire was said to be assessing only their personal opinions and preferences. When this form was completed the supposed major part of the study was introduced. As in the Aderman and Berkowitz (1970) study summarized above, the girls were told they were participating in an experiment in social observation. Following the procedure we had employed earlier, they were informed they would listen to a brief, tape-recorded conversation between two students, and then wouId be interviewed about their reactions. The recording was then played. Two female voices were heard which said exactly the same things as the male voices in the Aderman-Berkowitz study; one person told the other about her burdensome library assignment. This time there were only two outcomes to the conversation: For half of the subjects the potential helper refused to aid her friend, while in the remaining cases she volunteered to help and was thanked. When the tape had ended the women were ushered into separate rooms, ostensibly for their interviews. Then, telling each of them the other person would be interviewed first, the experimenter asked the girl to do her a “favor” and score the data sheets from “the other study.” The experimenter returned in I0 minutes. Analysis of the work scores yielded a significant interaction between the conversation outcome and self-concern variables. As Table VII indicates, the girls presumably experiencing little self -concern were no more productive after listening to the “help-plus-thanks” ending to the tape than after hearing the “no-help” conclusion. If anything, they did somewhat (but not significantly) less work for the experimenter upon being reminded of the desirability of aiding others. The subjects who were more strongly concerned about themselves, on the other hand, expended greater effort on behalf of the experimenter after hearing the potential helper volunteer her assistance and then be thanked for her offer. Wanting some sort of assurance of their worth, they became strongly motivated to aid the experimenter after they were reminded that helpfulness was “good.” Self-concern apparently enhances only a selfishly motivated helpfulness, not true altruism.

94

LEONARD BERKOWI‘IZ

TABLE VII MEANNUMBER OF PAGES SCORED FOR THE EXPERIMENTER DURINQ 10-MINUTEWORKPERIOD Self-concern level0 Overheard conversation outcome

High

Low

Help and thanks No help

75.4

54.6

56.8

57.8

~~

~a Self-concern was manipulated by varying the supposed purpose of the preceding questionnaire, with high self-concern being created by identifying the form as a test of social sensitivity and intelligence. In this significant interaction the mean, 75.4, is reliably different from all other means and there are no other significant differences. Data from Hildebrand and Berkowitz (unpublished experiment).

IV. Qualities of the Dependent Person Affecting WilIingness to Help Him

A. THEOTHER’S ATTRACTIVENESS Recognition of an obligation to help someone does not necessarily produce an ungrudging adherence to it, as I pointed out earlier. Many of the subjects in our Wisconsin experiments seemed to resent having to work for the other person without reciprocity or compensation; they acted altruistically but didn’t necessarily like it. This lack of enthusiasm may increase to definite resistance to giving help if the dependent person is unattractive. In one of our experiments cited above (Daniels & Berkowitz, 1963 ), as an illustration, the subjects exerted considerably less effort for a peer when he was less rather than highly attractive to them-even though this individual was greatly dependent upon them if he was to reach his goal. Pairs of subjects in a 2 X 2 X 2 factorial design were given much the same experimental treatments as in the earlier studies; they were told that they were “workers” operating under the supervision of the other peer member, and that this person’s chances of winning a cash prize were either highly or less dependent upon their performance. The subjects also believed that the “supervisor” would learn of their productivity either right away or more than 2 months later at the end of the semester. Finally, the men were induced to have either high or low liking for this person. Statistical analysis of the productivity measures revealed (as was also found in our earlier research) that the immediacy of “feedback” to the supervisor significantly affected performance under low but not high

FACTORS AFFECTING HELPING AND ALTRUISM

95

dependency; the subjects worked harder when they believed that their peer would learn of their performance quickly-but only when he was not highly dependent upon them. More important for our present purposes, degree of liking for the supervisor also was related to the subjects’ willingness to work for him. The men exerted the greatest effort for the supervisor when he was highly dependent upon them and when they had a relatively strong liking for him. This effect upon productivity held even when the supervisor supposedly would not be told of the subjects’ performance for several months. Under low dependency, liking for the supervisor was not related to performance. These results suggest that people are less annoyed by the pressure they feel to help the dependent other when this person is highly rather than less attractive. Even when the help-giving ideal was presumably clear, the subjects were less inclined to conform to it when they didn’t like the person needing their aid.

B. THEOTHER’S PREVIOUS HELPFULNESS The dependent person’s attractiveness (when it exists) could contribute to the often-observed tendency to reciprocate for earlier assistance. As an example, Pruitt (1968) reported that subjects in an experimental gaming situation behave in accord with the reciprocity principle: The more their partner had given to them previously, the more they tended to reward him in return. But more interestingly, the subjects reciprocated to a greater extent if the partner had given them a certain amount from a small pool of resources he commanded than if he had given the same amount from a larger pool. Clearly, the partner seemed more generous the greater the proportion of the rewards he provided relutive to the amount of rewurds he possessed. This apparent generosity could have heightened his attractiveness so that the subjects responded to this quality of their partner as well as to the economic aspects of their interaction with him. Frisch and Greenberg ( 1968) obtained essentially similar results. Their subjects were more likely to reciprocate a favor given them voluntarily than unintentional favors. Again, the partner might have become more attractive if he had done them a favor of his own free will so that they were more willing to help him in return. An earlier experiment by Goranson and Berkowitz (1966) in the present program yielded comparable findings. The subjects, college women, were first given a preliminary task to perform. A peer, actually the experimenter’s confederate, “voluntarily” assisted one third of the subjects on this task, while she was ostensibly “required” to give aid in another third of the cases (compulsory help), and deliberately refused

96

LEONARD BERKOWI’I7,

to help the remaining women. After this, all subjects were placed in our usual work situation in which they were supposedly workers operating under the supervision of someone who was highly dependent upon their work in order to obtain a $5.00 prize. In order to test an hypothesis that need not concern us here, half of the subjects within each of the three help conditions were toId, before the second phase of the study, that the confederate was the supervisor for whom they were working, while the remaining women were informed that their supervisor was yet another person (not the confederate). The effort measure was the number of paper boxes produced by each subject after a brief period. Analysis of the productivity scores (actually, the increase in productivity from the practice period level) revealed that the effects of the prior help variable ( voluntary-compulso-refused ) differed with the identity of the dependent supervisor (whether she was the same person they had encountered earlier or someone else). As is shown in line 1 of Table VIII, of the girls who believed that they were working for the previously encountered confederate, those who thought this person had voluntarily helped them earlier now expended reliably greater effort in TABLE VIII HELP A N D QUESTIONNAIRE SCORESIN CONDITIONS DIFFERINQIN MOTIVATION FOR PRIORHELP AND NATUREOF LATERDEPENDENT OTHER Later dependent other was: ~

Same person

Different person

Prior help was given: Measure

Voluntarily

Compulsorily Refused

Voluntarily ~~~

1. Increase in number of boxes made: 2. Other people’s expectations:” 3. Dislike for dependent other:b

Compulsorily Refused ~~

16.6.

12.8b

8.9.

11.7b

11.2b

10.6&

10.2.

8.1b

8.6b

9.lab

8.9ab

7.6b

4.7b

5 .l b

7.8s

5.9b

4.9b

5.4b

, Questionnaire item: “To what extent would most people have expected you to work hard to help your supervisor win the prize?” A high score indicates that subjects felt very much that moat people would have expected them to work hard. b Questionnaire item: “On the basis of whatever impressions you might have, would you want your supervisor as a roommate (assuming you were looking for a roommate)?” A high score indicata a low liking for the supervisor. Cells having different subscripts are significantlydifferent at the .05 level by Duncan range test. Data from Goranson and Berkowitz (1966).

FACTORS AFFECTING HELPING AND ALTRUISM

97

her behalf than any of the other five groups. The subjects working for the person who had been compelled to aid them exhibited the next highest level of effort, significantly greater than that displayed by those now working for the supervisor who earlier had refused to help them. The reciprocity exhibited by girls working for the same person who had aided them earlier is not entirely (or perhaps even largely) due to this individual’s attractiveness. Although we generally tend to like those people who go out of their way to assist us when there is nothing in it for them, many of us also recognize an obligation to pay something back for the benefits we have received. The subjects in the GoransonBerkowitz experiment were apparently aware of such normative pressures. Look at the second line of Table VIII. The girls aiding the same individual who had voluntarily helped them earlier typically said, in response to a questionnaire item, that most people definitely expect them to work hard for their supervisor. The normative expectations were evidently strongest for them. By contrast, the women working for a person who had been compelled to aid them earlier, or who had refused to assist them, seemed to feel under somewhat less pressure to work hard for her. The data summarized on line 3 of Table VIII shows that the intentionality of the prior help had affected the subjects’ liking for their supervisor. If this individual was the same person they had encountered earlier, she generally was most attractive if she had voluntarily assisted them during the first phase, and was least liked if she had refused to help them. The earlier refusal seemed to have been taken as a direct attack on them. All of these studies, I have suggested, indicate that an individual’s personal qualities govern our desire to reciprocate for the benefits we have received from him. The other fellow is more apt to be paid back in some way if he seems to be “generous,” “kind,” or generally “a nice person.” Schopler (1970) has argued in a similar vein; he contends that motives attributed to the helper are all-important. Reciprocation will be maximal if the help-recipient believes that his benefactor had been thinking primarily of him (the recipient) and had not been motivated by a self-centered hope of gain. The same principle can probably be applied to parties, at least in the academic world. People invited to relatively small dinner parties generally feel a greater obligation to reciprocate than do persons invited to large cocktail parties. The dinner party guest is more apt to be flattered; he feels that he has been singled out as an individual. The guest at the large cocktail party, on the other hand, is only one of many that the host wanted to entertain for some reason, and is less likely to think that he is appreciated as a unique

98

LEONARD BERKOWITZ

person. The first situation is more rewarding, leading to a greater liking for the host (other things being equal).

C. SOME SOCIALCLASSDIFFERENCES IN RECIPROCITY ORIENTATION Some groups are apparently more likely to follow a reciprocity principle than are others. Simmel has suggested that the conception of social dealings as an exchange, in which the parties involved consider the ratio between the rewards and costs of their interaction, is an outgrowth of experience with the money economy. Middle-class entrepreneurs, whose family breadwinner works for himself and/or for a commission, should therefore be more likely to behave in terms of the social exchange conception than middle-class bureaucratic people. As Miller and Swanson (1958) have pointed out in drawing attention to the importance of distinguishing between entrepreneurial and bureaucratic personalities, the bureaucratic middle-class individual should be most concerned with achieving accommodative relationships with others. In order to investigate the effects of social class differences, Friedman and I (Berkowitz & Friedman, 1967) recruited adolescent boys between 13 and 16 years of age from three Madison, Wisconsin high schools, supposedly for a study of supervisory ability. They came from three kinds of social class backgrounds (defined in terms of the fathers’ or guardians’ occupation and education) : bureaucratic middle-class, entrepreneurial middle-class, and working-class. When each boy arrived at the laboratory, he was told that he was to write a note to the boy in the next room explaining how that boy should perform a given task. If he, the subject, earned a good grade, he could win a cash prize, but 80% of his grade depended upon the other person’s productivity. Half of the subjects were then led to bclievc that the other person had worked hard and had therefore helped them n great deal, or had not worked hard and had given them little help. The roles were then reversed, and each subject was supposedly to be the worker operating under the guidance of a supervisor. Half of the people in each condition were informed that this supervisor would be the same person who had worked for them earlier, while the remaining subjects were told that the supervisor would be someone else. Finally, before they were put to work for this supervisor (70%of whose grade would depend upon the number of units they produced), the subjects werc led to believe that the supervisor (and the other person if there had been another) came either from a middle-class or working-class background. Not counting the boys used in the pilot studics or those who were discarded for one reason or another, 345 youngsters served in the 24 different conditions of this experiment. The same experiment also was

FAmORS AFFECTING HELPING AND ALTRUISM

99

carried out in Oxford, England (Berkowitz, 1968). Oxford and Madison are somewhat similar; both have relatively large universities, and both cities have a population of between 100,000 and 200,000. Madison, however, is not as heavily industrialized as Oxford, and it has a greater proportion of its people in middle-class occupations. All of the English boys, moreover, came from state schools, and consequently, while the Madison boys had a wide range of different family backgrounds, assuming a single socioeconomic continuum, there was a much greater restriction in the Oxford sample; the Oxford middle-class group can be more accurately characterized as lower middle-class. Finally, it proved to be exceedingly difficult to recruit youngsters from middle-class entrepreneurial families in Oxford. The English study therefore contrasted boys from middle-class bureaucratic backgrounds with their working-class counterparts. There were 192 subjects in the 16 conditions of this experiment. In the Wisconsin study, our expectation was that the middle-class entrepreneurial boys would have a stronger “reciprocity” orientation than their bureaucratic counterparts; their entrepreneurial families’ experience in a money economy presumably would cause the adolescents to regard social interactions as being akin to an economic exchange in which people “gave” in order to “get.” This emphasis upon reciprocity theoretically is much weaker among middle-class bureaucratic people who are less inclined to think in terms of immediate payment for services rendered. The American results, summarized in Table IX, were consistent with this reasoning. As is shown in line 1, the middle-class boys having the lowest level of productivity for their fictitious supervisors, and thus displaying the lowest level of motivation to help this other person, were the middle-class entrepreneurial youngsters who had received little help themselves earlier. Their productivity was highest, on the other hand, after they had received considerable help. Apparently, they gave effort only to the degree that they themselves had gotten benefits beforehand. As far as their productivity was concerned, it mattered relatively little whether they were working for the same person or a different one. By contrast, the performance of the middle-class bureaucratic and workingclass boys was much less affected by the level of help they had received earlier, with the working-class group generally displaying lower motivations to help their supervisor than the bureaucratic middle-class group. Questionnaire responses made at the conclusion of the experimental session were in accord with the performance differences. One question, for example, was designed to determine whether the boys recognized the existence of a cultural value prescribing that they should help the person

100

LEONARD BERKOWITZ

TABLE IX WORKSCORES AND PERCEIVED WORKEXPECTATIONS OF OTHERS IN CONDITIONS IN SUBJECT’S CLASSAND AMOUNTOF HELP RECEIVED DIFFERING (MADISON EXPERIMENT) Subject’s clam Bureaucratic middle-class

Entrepreneurial middle-class

Working class

High Low High Low High Law help help help help help help received received received received received received

Measures 1. Number of units produced 2. Other people’s expectationso

17.2.b 3.0b

16.9.bo

17.9.

14.8d

15.5bd

15.44

3.7.b

2.$

4.4.

3.5.b

3.6,

~~

The questionnaire item asked: “How much would other people have expected you to work hard for your supervisor?” This was an 11-point scale ranging from 1 (a great deal) to 11 (very little). Cells having different subscripts are significantly different at the .05 level by Duncan range test. Data from Berkowitz and Friedman (1967). 0

who needed their aid. This question asked, “How much would other people have expected you to work hard for your supervisor?” The conditions means are given in line 2 of Table IX. The strongest expectation was reported by the entrepreneurial middle-class adolescents who had received a good deal of help and whose supervisor supposedly was the person who had helped them; these boys evidentIy felt that they would be expected to return benefits to the boys who had worked hard for them. But similarly, governed by the same reciprocity orientation, the entrepreneurial boys who had been given little help earlier and who were working for this same person who had not helped them very much reported lowest expectations; more than any other group also getting little help, they said that people would not expect them to help this other person. Unfortunately, as noted earlier, relatively few entrepreneurial middle-class youngsters volunteered for the experiment in Oxford, and this social class grouping had to be eliminated from the study. As can be seen in Table X,the English working-class boys, however, responded more sharply to the experimental treatments given them than did their American counterparts. In particular, the former group exhibited a somewhat stronger reciprocity orientation. Look at line 1, showing the work scores in each condition. Where the middle-class bureaucratic

TABLE X WORKSCORESAND PERCEIVED WORKEXPECTATIONS OF OTHERSIN CONDITIONS DIFFERING IN SUBJECT'S CLASS, AMOUNTOF HELP RECEIVED, AND PERSONBEING HELPED (OXFORD EXPERIMENT) Same partner Bureaucratic middle-class

Different partner Working class

Bureaucratic middle-class

Working class

Measures

High help received

Low help received

High help received

Low help received

High help received

Low help received

High help received

1. Number of units made 2. Other people's expectations"

15.5.b 4 . Obe

13.8b 4.2b

17.6. 3.6b

12.6, 6.4.

16.7, 3.7b

15.h 4.4b

16.o.b 2.7.

Low help received 14. gab 3.8,

a This is the same item used in the Madison study, and as before, the lower the score the more definitely other people expected the subject to work hard for his supervisor. Cells having different subscripts are significantly different a t the .05 level by Duncan range test. Data from Berkowitz (1968).

102

LEONARD BERKOWITZ

adolescents from Oxford, like the American bureaucrats, were relatively unaffected by the level of help they had received earlier, the workingclass boys given a good deal of help worked very hard for the same boy who had helped them, and displayed the lowest level of productivity for the supervisor who had given them little help earlier. Again, the questionnaire responses, suinmarized in line 2 of Table X, were consistent with the productivity data. The level of expectation that they should help their supervisor perceived by the middle-class bureaucrats was fairly consistent across conditions, regardless of the treatment they themselves had received. The working-class youngsters, on the other hand, believed that other people would expect them to work hard in behalf of the other boy when they were working for the same boy who had given them a good deal of help, and said that they would not be expected to work hard when their supervisor was the same person who had given them very little help. All in all, our results appear to be in accord with the findings obtained in other investigations into social class differences in moral standards. In the Anglo-American society, at least, middle-class boys from a bureaucratic background seem to have a more strongly internalized set of abstract ideals prescribing that they should help people who are dependent upon them, regardless of benefits received. The ideals possessed by working-class boys, on the other hand, especially those in Oxford, are evidently much more concrete in nature and emphasize reciprocity to a much greater extent. Both studies, finally, also yielded information regarding the kind of person most likely to elicit reciprocity reactions. In the Wisconsin experiment, the youngsters exhibiting the strongest reciprocity tendencies-the entrepreneurial middle-class subjects-generally reciprocated most when they were dealing with someone from a diflerent social-class level, i.e., when the person they could help was from the working class rather than from the middle class. Essentially comparable results were obtained in the Oxford study. There, the boys displaying the strongest reciprocity orientation, those from the working class, also followed the reciprocity principle primarily when they were dealing with someone from a different social level. Thus, when the second-phase supervisor was the same person who had worked for them earlier, the working-class boys expended the most effort for a middle-class person who had given them high help in the first phase, and exhibited the lowest productivity for a middle-class supervisor who had given them little help earlier. These results make sense both sociologically and psychologically. Societies possessing a strong reciprocity orientation often call for re-

FACTORS AFFECTING HELPING AND ALTRUISM

103

sponsibility behavior only towards members of one’s own family group or people who are fairly close (cf. Banfield, 1958). Interactions with persons outside the family or the intimate group are in terms of reciprocity. Thus, the more distant the other people are, socially speaking, the more likely it is that the reciprocator will help them only to the cxtent that he has benefited from the situation himself. This phenomenon might arise from the reciprocator’s uncertainty about the “distant” people. Not knowing what to expect from them, the reciprocator might be very surprised when he is given benefits by persons who are socially quite distant from him. He is greatly pleased and feels obligated to provide benefits to them in return. On the other hand, if these “distant” others fail to benefit him when they could have done so, the reciprocator might interpret this failure as hostility directed toward him, He feels slighted, if not insulted, and reciprocates by not helping them when the opportunity arises. One implication of this reasoning is that the readiness to help someone does not necessarily parallel liking for him. The Daniels and Berkowitz ( 1963 ) study showed that middle-class university students typically work harder for someone who is greatly attractive to them than for someone who is less attractive. It may be, however, that there are individual differences in this relationship between help-giving and liking. People who are reciprocity oriented should be most inclined to help someone they like and should be less motivated to help a less-liked person. Responsibility-oriented people, on the other hand, such as people from bureaucratic middle-class families, could well become highly motivated to help someone else whom they did not care for personally. Data from the Wisconsin-Oxford experiments are consistent with this analysis. D. REACTANCEAND

LEGITIMACY OF THE OTHER’S DEPENDENCY On various occasions throughout this paper I suggested that people often resent the demands imposed upon them by the dependent individual’s need for their assistance. The request for help, and even the felt obligation to aid another, may generate reactance (Brehm, 1966). The potential helper feels somewhat annoyed and may even be inclined to resist the pressure he perceives. An interesting experiment reported by Latank and Darley (1970, pp. 81-85) can be interpreted in this manner. Male college students who were waiting to be interviewed for a supposed survey of consumer opinions individually overheard a fight between two children in the adjoining room. Judging from the voices, a bully seemed to be picking on a much younger child. In one condition the subjects were led to THE

104

LEONARD BERKOWITZ

think the children were alone, while in the other group they had been informed another adult was in the room with the youngsters. The researchers assumed that the subject would feel some responsibility to stop the fight if he was the only adult around, and would think himself less obliged to intervene if the other adult in the room with the children could stop the fight. The felt pressure to intervene in the former condition apparently was unpleasant and motivated the men in this group to deny their obligation. When they were later asked to say what had been going on in the next room, fully 88% of the men in the Low Personal Responsibility condition (another adult in the room with the children) said the youngsters had been having a real fight. On the other hand, if they had been the only adult present, so their responsibility was high, only 25% of them said the fight was real. The felt responsibility was an unwelcome burden, leading most of them to deny it. If there is unhappiness at giving help when the implied obligation seems legitimate, imagine the resistance that would exist if the demand appeared unwarranted and improper. The research evidence is clear on this point. People are much more willing to aid someone requiring assistance because of factors beyond his control than a person whose dependency is his own fault (Schopler & Matthews, 1965). We can also see this in charitable donations, according to observations reported by Bryan and Davenport ( unpublished study, 1968). Upon analyzing the contributions made to the New York Times’ 100 Neediest Cases, the investigators found that child abuse cases drew, proportionately, the greatest number of contributors; these victims were obviously in need because of circumstances beyond their control, and this “external locus of dependency” evoked many charitable donations. Similarly, cases requiring medical aid tended to draw a comparatively great number of contributions, probably because the physical illness was also regarded as beyond the victim’s control. Psychological illness and cases involving moral transgressions, on the other hand, tended to attract proportionately few contributions, perhaps because these misfortunes were attributed to the individual’s own deficiencies. These cases were sponsored by a respected institution, the New York Times, but the implied demand on the readers evidently was seen as improper when the need was the victim’s own fault, and this provoked comparatively strong resistance. In sum, many people appear to be unhappy about obligations imposed on them-or obligations they feel-but may not actively resist the felt demand unless they regard the demand as unwarranted, The “illegitimate” demand might even lead to a boomerang effect, One of our experiments (Berkowitz, 1968) found that there were different relationships between the degree of a partner’s dependency on the

FACTORS AFFECTING HELPING AND ALTRUISM

105

subject and how much work the subject did for his partner under externally and internally caused dependency. When the partner required help because of circumstances beyond his control, the greater his dependency on the subject, the more work the subject tended to turn out on his behalf. When the partner’s own deficiency had made him dependent on the subject, however, increased levels of dependency were associated with a somewhat lower level of work in his behalf. Consistent with Brehm’s (1966) analysis of psychological reactance, it is as if the subjects displayed a stronger resistance the greater the demand upon them when the demand was not entirely proper. Also in accord with this argument, another experiment ( Horowitz, 1968) has highlighted the importance of the feeling of behavioral freedom. College students tended to be more willing to help a person requiring assistance because of his own inadequacies if they thought they had a freedom of choice than if they believed they were required to help their dependent peer.

V. Conclusions Clearly, the usual exchange formulations provide only limited tools for the analysis of human social interaction. They are useful, but only for a quite restricted range of situations. The research background of the leading proponents of exchange theorizing, Homans and Blau, is instructive in this regard: Both men have devoted considerable attention to the study of organizations in the world of business and/or government. Organizational conditions might well promote, in many people at least, the kinds of concerns stressed by exchange analyses. The required work is often routine and uninteresting; it is carried out not for intrinsic values but for extrinsic benefits. Persons involved in this world are understandably frequently preoccupied with monetary returns for the effort and time they expend. They are embedded in conditions which heighten economic or economiclike thoughts of giving and getting. Organizations, furthermore, generally emphasize status considerations and ideas of status congruence and distributive justice. The people at the top of the hierarchy are given greater benefits-executive dining rooms, keys to the executive washrooms, multiple pens on their desks, titles on their office doors, and carpets on their office floors. The organizations also insist that they rightfully deserve their rewards and privileges. Social-exchange conceptions apply to the organizational world because this setting promotes exchange concerns. Outside this milieu,

106

LEONARD BERKOWITZ

however, exchange ideas may be less influential as other motives, interests, and values come into play. Nonetheless, the findings reviewed in this paper also indicate that many normative conceptions of social behavior are vastly oversimplified. These latter analyses generally think of people as regularly following the rules they learned in early childhood in order to avoid guilt feelings (as well as social disapproval), and attribute individual differences in moral behavior primarily to differences in the degree to which the relevant moral standards have been internalized. The action is supposedly governed largely by internal determinants. As we have seen, however, situational conditions play a very important role in helpfulness (and in other forms of moral conduct as well). Among other things, these conditions affect the extent to which the individual is aware that someone is dependent upon him, recalls the pertinent social ideals, believes that this dependency is proper or improper, and is willing to accept the psychological costs of being helpful. Normative analyses obviously should not be confined to static, internal influences, and must recognize the complex interplay between the individual's previously acquired predispositions and such external factors. Having long been focused on situational determinants of behavior, social psychology can contribute much to these interactional conceptions of moral conduct. REFERENCES Aderman, D., & Berkowitz, L. Observational set, empathy, and helping. Journal of Personality and Social Psychology, 1970, 14, 14 1-148. Aronfreed, J. The socialization of altruistic and sympathetic behavior: Some theoretical and experimental analyses. In J. R. Macaulay & L. Berkowitz (Eds.), Altruism and helping behaoior. New York: Academic Press, 1970. Pp. 103-126. Aronson, E., & Mettee, D. R. Dishonest behavior as a function of different levels of induced self-esteem. Journal of Personality and Social Psychology, 1968, 9, 121127. Banfield, E. C. The moral basis of a backward society. Glencoe, Ill.: Free Press, 1958. Berkowitz, L. Effects of perceived dependency relationships upon conformity to group expectations. Journal of Abnormal and Social Psychology, 1957, 55, 350354. Berkowitz, L. Responsibility, reciprocity and social distance in help-giving. Journal of Experimental Social Psychology, 1968, 4, 4 6 6 3 . Berkowitz, L., & Connor, W. H. Success, failure and social responsibility. Journal of Personality ond Social Psychology, 1966, 4, 664-669. Berkowitz, L., & Daniels, L. R. Responsibility and dependency. Journal of Abnornwl and Social Psychology, 1963, 66, 429-437. Berkowitz, L., & Daniels, L. R. Affecting the salience of the social responsibility norm: Effects of past help on the response to dependency relationships. JournaZ of Abnormal and Social Psychology, 1964, 68, 275-281. Berkowitz, L., & Friedman, P. Some social class differences in helping behavior. Journal of Personality and Social Psychology, 1967, 5, 217-225.

FACTORS AFFECTING HELPING AND ALTRUISM

107

Berkowitz, L., Klanderman, S. B., & Harris, R. Effects of experimenter awareness and sex of subject and experimenter on reactions to dependency relationships. Sociometry, 1964, 27, 327-337. Berkowitz, L., & Levy, B. I. Pride in group performance and group-task motivation. Journul of Abnormal and Social Psychology, 1956, 53, 300406. Blake, R. R., Mouton, J. S., & Hain, J. D. Social forces in petition signing. Southwestern Social Science Quarterly, 1956, 36, 385-390. Blau, P. M. Exchange and power in social life. New York: Wiley, 1964. Blau, P. M. Social exchange. In D. L. Sills (Ed.), International encyclopedia of the social sciences. Vol. 7. New York: Macmillan, 1968. Pp. 452-457. Brehm, J. W. A theory of psychological reactance. New York: Academic Press, 1966. Brehm, J. W., & Cole, A. H. Effect of a favor which reduces freedom. journal o f Personality and Social Psychology, 1966, 3, 420426. Bryan, J., & Test, M. A. Models and helping: Naturalistic studies in aiding behavior. Journal of Personality and Social Psychology, 1967, 6, 400-407. Carlsmith, J. M., & Gross, A. E. Some effects of guilt on compliance. Journal of Personality and Social Psychology, 1969, 11, 232-239. Daniels, L. R., & Berkowitz, L. Liking and response to dependency relationships. Human Relations, 1963, 16, 141-148. Darlington, R. B., & Macker, C. F. Displacement of guilt-produced altruistic behavior. Journal of Personality and Social Psychology, 1966, 4, 442443. Deutsch, M. The effects of cooperation and competition upon group process. Human Relations, 1949, 2, 129-152, 199-231. Feldnian, R. E. Response to compatriot and foreigner who seeks assistance. Journal of Personality and Social Psychology, 1968, 10, 203-214. Freedman, J. L. Transgression, compliance, and guilt. In J. R. Macaulay and L. Berkowitz (Eds. ), A h i s n i and helping behavior. New York: Academic Press, 1970. Pp. 155-161. Freedman, J. L., Wallington, S. A., & Bless, E. Compliance without pressure: The effects of guilt. Journal of Personality and Social Psychology, 1967, 7, 117-124. Frisch, D. M., & Greenberg, M. S, Reciprocity and intentionality in the giving of help. Paper presented at the meeting of the American Psychological Association, Washington, D. C., 1968. Goranson, R. E., & Berkowitz, L. Reciprocity and responsibility reactions to prior help. Journal of Personality and Social Psychology, 1966, 3, 227-232. Gouldner, A. W. The norm of reciprocity: A preliminary statement. American Sociological Reoiew, 1960, 25, 161-178. Helson, H., Mouton, J. S., & Blake, R. R. Petition signing as adjustment to situational and personality factors. Journal of Social Psychology, 1958, 48, 3-10. Homans, G. C. Social behaoior: Its elementary forms. New York: Harcourt, 1961. Hornstein, H. A. The influence of social models on helping. In J. R. Macaulay & L. Berkowitz ( Eds.), Altruism and helping behavior. New York: Academic Press, 1970. Pp. 29-41. IIorowitz, I. A. Effect of choice and locus of dependence on helping behavior. Journal of Personality and Social Psychology, 1968, 8, 373-375. Isen, A. M. Success, failure, attention and reactions to others: The warm glow of success. Journal of Personality and Social Psychology, 1970, 15, 294-301. Krebs, D. L. Altruism-an examination of the concept and a review of the literature. Psychological Bulletin, 1970, 73, 258-302. Latank, B., & Darley, J. M. The unresponsive bystander: W h y doesn’t he help? New York: Appleton, 1970.

108

LEONARD BERKOWI'IZ

Macaulay, J. R. A skill for charity. In J. R. Macaulay & L. Berkowitz (Eds.), Altruism and helping behavior. New York: Academic Press, 1970. Pp. 43-59. Midlarsky, E. Aiding responses: An analysis and review. Merrill-Palmer Quarterly, 1968, 14, 229-260. Midlarsky, E. Aiding under stress: The effects of competence, dependency, visibility and fatalism. Journal of Personality, 1971,39, 132-149, Miller, D. R., & Swanson, G. E. The changing American parent. New York: WiIey,

1958. Piliavin, 1. M., Rodin, J., & Piliavin, J. A. Good Samaritanism: An underground phenomenon? Journal of Personality and Social Psychology, 1969, 13, 289-299. Pruitt, D. G. Reciprocity and credit building in a laboratory dyad. Journal of Personality and Social Psychohgy, 1968, 8, 143-147. Rawlings, E. I. Reactive guilt and anticipatory guilt in altruistic behavior. In J. R. Macaulay & L. Berkowitz (Ens.), Altruism and helping behauior. New York: Academic Press, 1970.Pp. 163-177. Reckless, W. C., Dinitz, S., & Kay, B. The self component in potential delinquency and potential nondelinquency. American Sociological Review, 1957, 22, 56f3-570. Rosenbaum, M. The effect of stimulus and background factors on the volunteering response. Joumnl of Abnormal and Social Psychology, 1956, 53, 118-121. Rosenbaum, M., & Blake, R. R. Volunteering as a function of field structure. Joupnal of Abnormal and Sociul Psychology, 1955,50, 193-196. Schopler, J. An attribution analysis of some determinants of reciprocating a benefit. In J. R. Macaulay & L. Berkowitz (Eds.), Altruism and helping behaotor. New York: Academic Press, 1970. Pp. 231-238, Schopler, J., & Matthews, M. W. The iduence of the perceived causal locus of partner's dependence on the use of interpersonal power. ]mnaal of Personality and Social Psychology, 1965,2, 609-612. Schwartz, S. H. Elicitation of moral obligation and self-sacrificing behavior: An experimental study of volunteering to be a bone marrow donor. Journal of Personality and Sock1 Psychology, 1970, 15, 283-293. ( a ) Schwartz, S. H. Moral decision making and behavior. In J. R. Macaulay and L. Berkowitz ( Eds.), Altruism and helping behaoior. New York: Academic Press, 1970. Pp. 127-141. ( b ) Schwartz, S. H., & Clausen, G. T. Responsibility, norms, and helping in an emergency. Journal of Personality and Socid Psycho~ogy,1970,16,299310. Thomas, E. Effects of facilitative role interdependence on group functioning. Human Relations, 1957, 10,347-366. Wheeler, L.. & Wagner, C. M. The contagion of generosity. Paper presented at the meeting of the Eastern Psychological Association, Washington, D. C., 1968.

THE POWER OF LIKING: CONSEQUENCES OF INTERPERSONAL ATTITUDES DERIVED FROM A LIBERALIZED VIEW OF SECONDARY REINFORCEMENT

Albert J. Lott Bernice E. Lott UNIVERSITY OF RHODE ISLAND

I. Introduction . . . . . . . . . 11. Propositions Relevant to Consequences . . . . A. Cue Properties of Persons . . . . . B. Motivational Properties of Persons . . . . C. Reward Properties of Persons . . . . . D. Generalization . . . . . . . . 111. Specific Predictions and Supporting Data . . . A. Approach and Avoidance Responses . . . B. Descriptive Responses . . . . . . C. Evaluative Responses . . . . . . D. Conformity, Modeling, and Acquiescent Responses . . . . . . E. Vicarious Responses . F. Learning . . . . . . . . . G. Responses to Stress . . . . . . . IV. Areas of Applicability . . . . . . . References , . . . . . . . .

. . . . . . . . . . . . .

. . .

. . . . , . . . . . . . . . .

.

. . . . . . . . . . . . . .

. .

.

. . .

.

. . .

.

. . .

.

. . .

109 112 112 113 114 115 116 116 119 123 126 129 130 134 136 141

I. Introduction This paper focuses on expected consequences of interpersonal attitudes derived from a theoretical position in which persons who evoke attitudes are conceptualized as secondary reinforcers. The intent is to show ( a ) that interrelated hypotheses regarding consequences can be generated by placing the concept of attitude within a learning theory framework, ( b ) that a substantial number of empirical relationships can be interpreted as supporting these hypotheses, and ( c ) that the be109

110

ALBERT J. LOTT AND BERNICE E. LO’IT

havioral effects of interpersonal attraction have far-reaching social implications, making the potential application of verified propositions an enticing possibility. Systematic attention to the subject of attraction, or liking and disliking, is comparatively recent. Attitudes toward issues or events, or specialized related questions such as prejudice are, of course, traditional areas of social psychological concern, but theories and research focusing specifically on the attitudes people have toward one another are relatively new to the field. Despite this recency of attention, there is a considerable accumulation of knowledge which can be examined for empirical consistency and for its potential applicability to social problems. Our primary concern here is with theory. We propose that certain general principles of behavior can be shown to be relevant and can be extended to interpret the data at hand and to provide creative hypotheses testable in natural social situations. Most of the ongoing work on interpersonal attitudes is concerned with delineating the determinants of positive attraction; few researchers have directly explored the more difficult problem of the acquisition of dislike. The sizeable literature clearly indicates that there is a positive relationship between liking for a person and his or her previous association with some desirable condition (cf. A. J. Lott & Lott, 1965). Berscheid and Walster (1969), in a monograph on the antecedents of attraction, devote five out of eight chapters to reviewing research on liking as a function of the “rewards others provide,” and they note that the ‘‘psychological principle which is most frequently used to predict interpersonal attraction is the principle of reinforcement [p. 291 .” Byrne and Clore (1970), for example, have proposed a reinforcement model of evaluative responses, utilizing a conditioning paradigm, to account for a widely documented linear relationship between the proportion of attitude statements by a stranger which are similar to those of a subject, and the subject’s subsequent attraction toward the stranger. Attitude similarity is viewed as an unconditioned stimulus, the positive reinforcement value of which derives from its satisfaction of the need for effectance (Byrne & Clore, 1987). The theoretical position of the present authors makes explicit use of general principles from the behavior theory of Hull, as modified and extended primarily by Spence, Miller, and Mowrer (the ”Yale-Iowa viewpoint;” see Cofer & Appley, 1964). The primary proposition of this “learning theory approach to interpersonal attitudes” (A. J. Lott & Lott, 1968; B. E. Lott, 1961; B. E. Lott & Lott, 1960) is that a discriminable neutral person-stimuIus who is consistently associated with a reinforcing state of affairs will become able to evoke the implicit anticipatory com-

THE POWER OF LIKING

111

ponent of the response made to reward. This anticipatory response (r,-s,), which has the logical status of an intervening variable, is identified as a positive attitude (as suggested by Doob, 1947) which can be inferred from a variety of overt measurable behaviors [Eisman (B. E. Lott), 19551. Stimuli which can evoke implicit anticipatory goal responses are considered, thereby, to have acquired reinforcement properties of their own and to be able to function as secondary rewards (Hull, 1951). It follows, then, that we can say of person A who is liked by person B, that A evokes a positive attitude ( r,-s,) in B and can serve as a secondary reinforcer for him or for her. Analogously, neutral personstimuli who are consistently associated with aversive conditions, such as frustration or punishment, become able to evoke the implicit, anticipatory components of responses to these conditions ( rc-sfor rD-sp);these particular implicit response-produced stimuli we identify with the concept of negative attitude. Any person-stimulus evoking a negative attitude (dislike) can function as a secondary negative reinforcer. In extending behavior theory principles to the learning of interpersonal attraction, it has been necessary to clarify and define in social terms the concepts of reinforcement (or reward), response to reward, and implicit anticipatory responses. A reward is simply any stimulus which a person judges to be desirable and pursues as a goal, whether the need it serves to satisfy is physiological or psychological and learned. Whether or not a stimulus constitutes a reward for a particular child, for example, will depend upon his or her past experience with that stimulus, the social setting in which it is presented (which may provide it with additional meaning), and his or her level of drive to attain it. Some events, like attention from others, agreement, or approval, seem to be generally reinforcing (desirable) across most situations in our culture. Other potential reinforcers like the verbalization, “right,” or a gold star, may vary in effectiveness for children from different socioeconomic classes or from homes in which parents have or have not used such stimuli in the socialization of their children. The concept of goal response ( R G ) , we have noted elsewhere (A. J. Lott & Lott, 1968, p. 69) implies not just consummatory activity, but the more complex combination of evaluative, verbal, autonomic, and central reactions which are only partially overt, but which we assume human beings reliably make to rewarding objects or situations. When a child, for example, receives praise (or some other desirable event) the Ro will almost certainly include a smile of pleasure. When any individual is rewarded, positive affect is assumed to accompany such observable responses as a laugh, an exclamation, physical approach to the goal, etc.

Similarly, when a person is punished, negative affect is assumed to ac-

112

ALBERT J. LOTT AND BERNICE E. L O l T

company such observable responses as cries, facial grimaces, flight, and so on. An implicit anticipatory response, the r,-s, mechanism, is a theoretical concept, an intervening variable with no necessary implication regarding its peripheral or central locus (Black, 1965; Miller, 1963). It connotes positive affect, “expectation” (as in its usage by Hull, 1952) or “hope” (following Mowrer, 1960), whereas rI-sf or rp-s, connotes negative affect, “fear,” or “anxiety” (Amsel, 1962). Being in the presence of a discriminable person (or some symbolic representation of him) when one attains satisfaction of any drive or succeeds in reaching any desirable goal, regardless of whether the discriminable person has any instrumental relationship to this state of affairs, is a sufficient condition for the acquisition of liking for the contiguous person (B. E. Lott & Lott, 1960). Being in the presence of a discriminable person when one experiences pain or frustration is a sufficient condition to produce dislike for the contiguous person. Specific hypotheses dealing with other antecedent variables have been derived from reinforcement learning theory, and it has been shown, for example, that frequency of reward (James & Lott, 1964), vicarious reinforcement (A. J. Lott, Lott, & Matthews, 1969), immediate us. delayed reward (A. J. Lott, Aponte, Lott, & McGinley, 1969), high us. low drive, and quality of reward (A. J. Lott, Bright, Weinstein, & Lott, 1970a) all significantly affect the development of interpersonal attitudes in a theoretically predictable way.

11. Propositions Relevant to Consequences Liked persons are stimuli which evoke expectations of reward, and disliked persons are stimuli which evoke expectations of pain or frustration. Liked and disliked persons can, therefore, function as positive and negative secondary reinforcers, respectively, From this basic assumption there follows a number of expectations for the behavior of individuals when liked or disliked persons (or their symbolic representations) are present either prior to, during, or contingent upon the behavior. In considering the consequences of interpersonal attraction, we have found it useful to think in terms of three major, and logically separable, properties which secondary reinforcing stimuli in general should be expected to possess: cue properties, motivational properties, and reward properties. We will consider each, in turn.

A. CUEPROPERTIES OF PERSONS 1. A liked person will evoke a wide variety of overt and covert responses, classifiable as approach, some of which are assumed to be

THE POWER OF LIKING

113

reflexive and unlearned in nature, while most have been learned because they successfully mediate the attainment of a variety of positive goals. A disliked person will evoke a wide variety of responses classifiable as avoidance, some of which are reflexive and innate, while most have been learned as successful ways of evading sources of pain or frustration. 2. One special class of approach and avoidance responses may be considered perceptual, and it is therefore expected that liked personstimuli will be maximally attention-directing, salient, or distinctive (i.e., highly discriminable). A disliked person should be less distinctive than a liked one, but because successful escape often depends upon attending to, or recognizing an aversive stimulus, vigilant perceptual reactions are learned. A disliked person should therefore be more distinctive and salient than a neutral person (one who evokes no affective anticipatory responses) . The position that perceptual reactions follow the same laws as other kinds of responses has been taken by others. D. T. Campbell (1967), for example, has found Hullian theory useful in explaining the influence of motivation, value, and familiarity on both perceptual thresholds (the probability of a stimulus being noticed) and “covaluant errors” ( perceiving the stimulus to be present when it is not) by identifying the independent variables with the theoretical concepts of D, K, and H, respectively.

B. MOTIVATIONAL PROPERTIES OF PERSONS A liked person can’ function as an incentive and raise general drive level in the individual who responds with liking; this should be reflected in improved performance on simple tasks, and in more vigorous behavior in the presence of the liked person. A disliked person will function as a negative incentive and also raise general drive level in the individual who responds with disliking. Miller’s (1944) descriptions of the relative steepness of approach and avoidance gradients provide some basis for the expectation that the increment to drive level will be greater in the presence of a disliked person than in the presence of a liked one (assuming equal intensities of positive and negative attitude or affect), since under the condition of least distance from (or greatest proximity to) positive and negative goals, avoidance tendencies are stronger than approach. One might, therefore, expect, on the basis of drive considerations alone, that performance on a simple task, for example, will be more enhanced in the presence of a disliked than a liked person, However, a disliked person is also evoking avoidance behavior which should serve effectively to compete with and to disrupt any task-relevant responses which must be performed in the disliked person’s presence.

114

ALBERT J. LOTT AND BERNICE E. LOTT

C. REWARD PROPERTIES OF PERSONS Also to be expected of liked and disliked persons is that they can function directly as positive and negative reinforcers of the behavior of individuals who are positively or negatively attracted to them. If a liked person or some symbolic representation of him or her (e.g., a photograph or name) appears, contingent upon some behavior, this behavior will be strengthened, and its probability of occurrence will increase. This constitutes a “learning test’’ of secondary reinforcement. If a response has been followed by the presence of a disliked person, it is expected that its probability of occurrence will decrease. An “extinction test” of a positive secondary reinforcer would require that an individual continue to perform some response, after a personstimulus has been repeatedly paired with reward, under conditions where only the person-stimulus and not the original reward follows the response. Related to this expectation is that the presence of a secondary reinforcer will be able to mediate between performance of some behavior and delayed reward. Spence (1947) proposed that this positive role in delayed reward learning was one of the major functions of learned reinforcement. Is it contradictory to maintain that a positive secondary reinforcer ( a liked person) can both increase general drive, thereby functioning as an incentive or energizer or motivator, and also strengthen new behavior by functioning as a reward, and presumably reducing drive? There appears, at the moment, to be no completely satisfactory way of responding to this question, the paradox having been “built-in” by Hull and more or less accepted on empirical grounds by others. The problem of the “same stimulus object” acquiring “the capacity to serve as either an incentive or a reward was recognized by Miller and Dollard (1941, Chapter 4), and although it continues to be noted (e.g., the “doublefunction paradox”; see Cofer & Appley, 1964), it remains essentially unresolved. That the implicit cue component of the fractional goal response can function as a drive stimulus was assumed by Doob (1947) mainly because, by the very nature of its association with goal attainment, the s, is most likely to be strong and persistent, and thereby to meet the criteria for drive suggested by Miller and Dollard (1941). It was Spence ( 1956), however, who strongly emphasized the motivational nature of the fractional anticipatory goal response and suggested that its motivational property could be represented by variable K, incentive. He also suggested that rK-sgincreased D as a result of the conflict produced when it is evoked in the absence of the original goal or that K (representing the motivational property of rg-s,) could function independently and analogously to D as a response energizer.

THE POWER OF LIKING

115

Trying to explain how secondary reinforcers reinforce is another matter. Wike (1966) in a book devoted exclusively to secondary reinforcement, notes at the very end of his final chapter, that he has preferred "to avoid the question of why secondary reinforcers work" [p. 4801 and to look instead for empirically based principles. Hull (1951) proposed that a secondary reinforcer can evoke a response like relaxation, which has accompanied past goal attainment, and thereby produce reinforcement. Perhaps it is simpIy the case that once a stimulus has acquired, through association with reward, the power to evoke anticipation of reward, it becomes a desirable goal to attain in and of itself. Striving-toattain can be viewed in drive terms, and attainment can be interpreted as drive reducing. When a positively valued (attractive, liked) person is near enough to see, touch, or interact with, this is a satisfying state of affairs, equivalent to other such states of affairs which are assumed to be produced by the satisfaction of primary or secondary needs, and the consequent reduction of general drive level.

D. GENERALIZATION As a resuIt of stimulus generalization, persons who resemble, or are similar to, liked and disliked individuals on salient characteristics (e.g., facial features, skin color, mode of dress, etc.) can also function as positive and negative secondary reinforcing stimuli, but to a lesser extent. Mediated response generalization should also extend secondary reinforcing properties to other persons who evoke, for example, the same verbal response (such as a name or other identifying label) as persons who are liked and disliked. The foregoing analysis of interpersonal attitudes has focused on their theoretical consequences for behavior as a result of the cue, motivational, and reward properties of persons derived from their status as secondary reinforcers. Staats ( 1968) has discussed attitudes in general in a similar manner, but within a Skinnerian framework, suggesting that they function as affect eliciters, reinforcers, and discriminative stimuli. Each approach has generated complementary, but different types of research programs ( cf. Greenwald, 1968). From the general propositions presented above, a number of more specific predictions can be made regarding how individuals will behave in the presence of (or relative to) persons to whom they are differentially attracted. Some of the predictions will be seen to be consistent with derivations from other theoretical approaches to interpersonal processes, especially the congruity models developed by Heider (1958) and Newcomb (1956). What we propose is that attraction between persons can be understood in general behavior theory terms, thus enabling us to examine, organize, and empirically test a wide range of

116

ALBERT J. LOlT AND BERNICE E. LOIT

behavioral phenomena, In addition, by attempting to specify the kinds of behavior from which positive and negative attitudes toward persons can theoretically be inferred, the specific predictions also serve to indicate ways in which interpersonal attraction can be measured, Some of the predicted behavioral consequences of attraction can thus provide “unobtrusive” attitude indices permitting investigators to move beyond their dependence upon verbal self-report measures. The particdar behaviors to which we refer, and which will be elaborated and illustrated below, are as follows: approach and avoidance responses; descriptive responses; evaluative responses; conformity, modeling, and acquiescent responses; vicarious responses; learning; and responses to stress.

III. Specific Predictions and Supporting Data This section of the paper is not intended to be a review of current literature, but rather to illustrate, by reference to data on human behavior ( a ) that findings relevant to the effects of differential liking for persons by one another can be systematized and explained by the general behavior principles relating to secondary reinforcers outlined above and ( b ) that interpersonal attitudes affect a wide variety of socially significant behaviors.

A. APPROACHAND AVOIDANCE RESPONSES It is predicted that people will move toward those they like and away from those they dislike. This “movement” may be overt and physically involve the entire body or only portions of the body; it may be a verbal statement of intention or implied intention; it may be a symbolic or representational act, or an act of imagery or fantasy. 1. Visual behavior. Hess (1965) has proposed that dilation and constriction of the pupils reflect not only changes in Iight intensity, but also interest, emotion, and attitudes. He has summarized the data obtained in his laboratory, as follows:

..

the evidence suggests that . with respect to visual material there is a continuum of responses that ranges from extreme dilation for interesting or pleasing stimuli to extreme constriction for material that is unpleasant or distasteful to the viewer. In the presence of uninteresting or boring pictures we find only slight random variations in pupil size.

Illustrative data with respect to person stimuli are reported by Hess, Seltzer, and Shlein (1965). Among five heterosexual males, the pupillary response (dilation) was greater when they were looking at pictures of women than of men, whereas five homosexual males responded in the opposite direction, producing a clear difTerence between

THE POWER OF LIKING

117

the two groups. Woodmansee (1970), on the other hand, has reported failure to find reliable evidence of pupillary constriction associated with negative affect, and he has also pointed to several sources of possible methodological error in the work by Hess. The dilation reaction, however, he concludes, may be used to indicate arousal, attentiveness, or interest. A relationship between liking for persons (where differences in liking are either known or assumed on the basis of experimental manipulations) and amount of looking at them has been demonstrated in a number of investigations. Varying, but related measures have been employed, e.g., frequency and duration of glances at interviewers who asked either innocuous or very personal and embarrassing questions (Exline, Gray, & Schuette, 1965); amount of time gazing at an interviewer’s eyes (Goldberg, Kiesler, & Collins, 1969); eye contact with a physically present partner (Efran & Broughton, 1966); and eye contact with an imaginary person ( Mehrabian, 1968). A. J. Lott and Lott (1970) made use of a ‘looking box” technique adapted from Nunnally, Duchnowski, and Parker (1965), with children from two independent samples, taken from different natural group situations, a summer dramatics program and a residential camp for boys. It was found that the children looked reliably more at photographs of highly liked fellow group members than at less liked ones, i.e., they pressed a button to illuminate a colored slide (for .3 seconds) more frequently when the slide was of a liked than of a less-liked peer. Subjects were permitted to press the button for viewing each slide as many times as they wished. 2. Interpersonal distance. It is to be expected that persons will attempt to achieve a physical closeness with those they like, and distance from those they dislike. Clore (1969), using a procedure which worked against spatial effects, found that subjects who liked a “stranger” most were more likely to place their chairs so as to face his, regardless of the distance between them. Some studies have made use of the technique of having subjects place symbolic figures relative to one another. Using diagrams of seating arrangements, Kiesler and Goldberg (1968) found that subjects’ placement of their own chairs relative to another person was Significantly related to their liking of him, the greater the liking, the less the distance between chairs. Little (1965) asked subjects to place pairs of cardboard mounted figures on various backgrounds. When pairs of figures were designated to be friends they were placed significantly closer to one another than if they were labeled acquaintances, and figures designated as strangers were placed farthest apart. The same effect occurred when subjects, instructed to “direct”

118

ALBERT J. LOTT AND BERNICE E. LOTT

scenes, placed real actresses on a stage. Kleck (1967), using photographs of classmates previously ranked for liking, and Levinger and Moreland ( 1969), using tape-figures representing the subjects, their friends, and strangers, obtained similar, confirming results. In the previously mentioned study by Mehrabian (1968) in which subjects assumed postures relative to imaginary persons, distance from the person’s imagined location was found to be a decreasing linear function of positive attitude toward him. Outside of the laboratory, too, encouraging data have been reported. Willis (1966) had students measure the distance, from nose to nose, between themselves and any individuals who began to converse with them, at the moment the conversation began. The students found that they were approached more closely by their friends than by any other category of person. Byrne, Ervin, and Lamberth (1970), in a study of computer dating, measured the distance between couples as each stood in front of the experimenter’s desk right after returning from a date. They, too, found that interpersonal attraction correlated significantly with physical closeness. Relevant findings within a different context come from a study by Moss and Robson (cf. Mussen, Conger, & Kagan, 1969, p. 86 f.). Mothers who had been interviewed during the last 3 months of their pregnancy with a first child regarding their attitudes toward the expected baby were subsequently visited in their homes when the babies were l-month old. Those mothers who had had the most positive attitudes toward their unborn infants were observed to spend the most time in face-to-face contact with them, The present authors, using a doll-play technique, recently studied a group of elementary school age girls attending a summer day camp. Each girl was observed in several situations, in which a doll representing herself interacted with a doll who represented either a highly liked fellow day-camper in her particular group or a disliked one, In one situation the subject was instructed to imagine that she was on an overnight hike with her group, and that the self-doll and the peer-doll had their sleeping bags adjacent to one another. The actual distance between the two dolls, after the subject placed them “sleeping” on a table, was measured by counting preplaced adhesive markers and was found to significantly differentiate disliked from liked peers ( Mann-Whitney U test, l-tailed, p < .025), as predicted. The subjects who were playing with a disliked peer-doll placed their self-dolls farther away from it ( M = 4.14) than did the subjects who were playing with a liked peer doll ( M = 1.54). 3. Communication. Both Newcomb ( 1953) and Festinger (1950) proposed that frequency of communication between persons should vary

THE POWER OF LIKING

119

directly with the degree to which they are attracted to one another (Newcomb) or to the group (Festinger), but no clear tests of these relationships appear in the early cohesiveness literature. That the degree of liking among group members is predictive of how much they talk to one another, however, has been shown by later investigators. A. J. Lott and Lott (1961) measured the frequency of communication within 15 ongoing natural voluntary college groups as each engaged in a 30-minute discussion of “student apathy.” The amount of communication was found to be positively and significantly correlated with the degree of intennember liking. This relationship was demonstrated again by Moran (1966) who worked with industrial training groups in the Netherlands. Data from a study by Taylor, Altman, and Sorrentino (1969) are also relevant. Among sailors who interacted with a confederate of the investigators via an intercom system, those for whom the interaction was positive talked about more aspects of themselves, in greater depth, and spoke longer than did those in the negative interaction condition. Interesting results have been reported on accuracy of interpreting or understanding communication as a function of interpersonal attraction. M. G. Hornstein (1967) had pairs of freshman roommates in a woman’s college, on whom compatibility scores had been obtained, communicate emotional states to each other by vocalizing consecutive letters of the alphabet instead of words. After 3 months of rooming together, compatibility and the degree to which the communications were accurately understood, were related significantly. These findings are congruent with data obtained by Eisman (B. E. Lott) and Levy (1980) in a different kind of situation, but one in which ambiguous communication symbols were also utilized. Members of college study groups were each given an opportunity to speak sentences to the others in their group behind a one-way vision screen; the elimination of audibility required the viewers to read the lips of the communicator. Those persons who were ranked as “most preferred group members were lip read with significantly greater accuracy than those who were ranked “least preferred.”

B. DESCRIPTIVE RESPONSES Since liked, disliked, and neutral persons are assumed to differ, in that order, in salience and distinctiveness, it should be the case that the degree to which a person is liked will influence the speed and accuracy with which he is recognized, and the richness of detail with which he will be described. The predicted differences between liked and neutral persons are based upon the expectation that the former will

120

ALBERT J. LOlT AND BERNICE E. L o n

evoke positive affect, implicit anticipatory goal responses, to which a variety of approach responses have been learned. Prediction is more difficult for stimuli which evoke negative attitudes. Evidence for perceptual “vigilance” (e.g., Bootzin & Stephens, 1967) suggests that such stimuli are high in salience, whereas evidence for perceptual “defense” (e.g., Bootzin & Natsoulas, 1965) suggests the opposite. Since we expect some form of avoidance behavior to be typically evoked by negatively evaluated stimuli, greater salience of such stimuli should, under some circumstances, increase the probability that effective overt avoidance can take place, whereas inaccurate or slow recognition is itself interpretable as an avoidance response. 1. Recognition. Despite the fact that much of the early work by the “new look” researchers in perception was subjected to careful criticism on methodological grounds ( cf. Natsoulas, 1965) the proposition that perceptual responses such as recognition may be influenced by functional factors has by no means been discarded. Park and Smith ( 1966), for example, have reported that recognition time for valuerelated words, all of infrequent usage, correlated positively with the values of subjects, and that this was especially so for those subjects characterized as “inner-directed.” Johnson, Thomson, and Frincke ( 1980) tachistoscopically presented lists of words that were matched in frequency, but varied in “goodness,” as previously determined by semantic differential scale ratings and found that the “good words were reported at significantly lower thresholds than the matched “bad” ones. With respect to persons, Vernon (1966) suggests that in everyday life “we perceive most clearly and remember in most detail the faces of those in whom we are interested, and notice less clearly the faces of those to whom we are indiffeient.” Compelling as this general observation may be, relevant experimental evidence remains equivocal (cf. Tajfel, 1969). Supporting data have been reported by Chance, Goldstein, and Schicht (1967) who presented to children photographs of their classmates, under masks, so that only a portion of each face could be seen. Among kindergarteners, first graders, and second graders, “friends” were correctly identified significantly more often than “nonfriends.” Only within a sample of older sixth-grade children was the “enhancing effect of friendship on accuracy of identification” not obtained. In an investigation by A. J. Lott and Lott (1970), first-grade children were shown slides of some of their classmates. All subjects viewed six different slides, of a liked, neutral, and disliked same-sex peer, and of a liked, neutral, and disliked opposite-sex peer. Each slide was presented, in a predetermined random order, twice at each of four different

THE POWER OF LIKING

121

brief exposure times. The particular slides viewed by each subject were selected on the basis of responses to previously administered liking scales. Each subject was instructed to select from a sheet of photographs of the entire class, the child whose face had been recognized. The procedure of having a subject point to one face among a group of faces was used to minimize verbal response bias. Because most of the slides were accurately identified, the exposure time variable was ignored in analyzing the data, and errors were examined only in terms of whether they were associated with a child's attitude toward the person pictured in the slide. It was found that the boys behaved as predicted, and were reliably less accurate in recognizing disliked than liked faces, while the girls behaved in an opposite manner. In each case the difference between proportion of errors for liked and disliked photos was significant. The number of errors made on neutral peer slides was the same for boys and girls and intermediate between the like and dislike categories. A provocative possibility raised by these findings is that, at least in children, there may be sex-related differences in the perceptual responses learned to anxiety-evoking stimuli. Since the children in this study made few errors of recognition, the data obtained must be regarded as only suggestive. Evidence from two studies on binocular rivalry also seem relevant to our general recognition hypothesis. A condition of binocular rivalry can be produced by presenting different visual material to the left and right eyes via stereoscopic slides. When one of the components stands out more clearly than the other, this can be related not only to the relative structural characteristics of the two stimuli such as brightness and contrast, but also to familiarity or past experience with them (e.g., Bagby, 1957; Engel, 1961; Moore, 1966). The stereoscopic technique was utilized by Pettigrew, Allport, and Barnett (1958) with four different racial-ethnic groups in South Africa. Pairs of faces (equated for pleasantness of expression) were presented so that each racial group was paired with every other an equal number of times. Among the findings reported was that the white South Africans, especially the Afrikaners, tended more often than persons from other groups to report racially mixed photographs as either pure European or pure African, rather than as racially mixed Coloreds or Indians. Murray (1968), in a Master's thesis study, directly tested the hypothesis that the face of a liked person will predominate over that of a neutral person. She stereoscopically presented to law students faces of classmates whom they had previously rated on a liking scale, and which were equated for familiarity. Although the usable sample of subjects was small, the data obtained on them supported the hypothesis. In a different perceptual situation, one in which aniseikonic lenses

122

ALBERT J. LOTI AND BERNICE E. LO’IT

were used to induce optical distortions, Asthana (1960) found that subjects who viewed self-images in a mirror, liked persons, strangers, disliked persons, and physical objects, experienced different degrees of distortion, from least to most, in that order. In addition, a latency measure, i.e., how long it took for distortion to begin after putting on the lenses, indicated that distortion began soonest for disliked persons, while the latencies for self-images and liked persons were longer. 2. Detail. The distinctiveness, or salience, of a stimulus can be operationalized by a measure of the amount or degree of detail with which it is described. Maccoby and Wilson (1957), in a study directed at an overtly different problem, obtained results which can be interpreted as relevant to the present discussion. They showed movies to seventh-grade children, obtained measures of “identification” with the characters, and tested the children a week later for knowledge of the movies’ contents. Memory for actions and words was found to be better for the identified-with characters than for the others. We obtained evidence of a more direct nature (A. J. Lott & Lott, 1970). Each child in a first-grade class was asked to draw one same-sex classmate in each of three interpersonal attraction categories; liked, neutral, and disliked, Six adult judges, working independently without knowing the affect categories of the drawings, ranked each set made by the same child with respect to degree of detail. Drawings of liked peers were judged to be significantly greater in detail than those of disliked peers and of neutral ones. A different kind of measure, number of adjectives used to describe persons, was obtained in another investigation (Lott, Lott, Reed, & Crow, 1970b). Two samples of students from widely separated geographical areas, Kentucky and California, were asked to describe liked, neutral, and disliked persons with whom they were well acquainted by choosing words from a list of 200 personality-trait adjectives taken from Anderson (1968). In both samples, subjects used the greatest number of words to describe liked persons, the least number of words to describe neutrally regarded ones, and an intermediate number to describe disliked ones. [Very similar results have been reported by Supnick (cf. Crockett, 1965).] The tendency for liked persons to generate more descriptive words or properties than disliked and neutral persons may lead to less discrimination among a group of such persons (as found by Irwin, Tripodi, & Bieri, 1967), since liked persons may be seen as sharing a large number of characteristics. Disliked persons, described by fewer adjectives or properties, could provide less basis for stimulus generalization and thus be more easily distinguishable from one another. An additional finding in our word-description study was that of a

THE POWER OF LIKING

123

greater variability in the number of words used to describe disliked as compared with other persons. This lends support to the view that aversive stimuli such as disliked persons may evoke defensive reactions, ie., fewer descriptive words in some individuals (or under certain conditions) and to evoke vigilant reactions, i.e., more descriptive words in other individuals (or under certain other conditions). 3. Judgments of size. A number of studies in the 1940’s demonstrated that size estimation of stimuli such as coins (Bruner & Goodman, 1947), dollar signs, and swastikas (Bruner & Postman, 1948), and poker chips associated with candy (Lambert, Solomon, & Watson, 1949) varied with the “value” of the stimuli to the perceiver. That highly valued (liked) persons, and anxiety-provoking ( disliked) persons will also be seen as larger than neutral ones who evoke a minimum of affect is suggested by three investigations which have dealt with mythical person-stimuli. Solley and Haigh ( 1957) found that children’s drawings of Santa Claus became increasingly larger as Christmas approached and smaller shortly thereafter; Sechrest and Wallace (1964) had children draw Santa, a man, and a house, and found that the Santa drawings, relative to the other two, were larger during the Christmas season; and Craddick ( 1S63), who had children draw a “woman” and a “witch,” reported that shortly before Halloween drawings of the latter were larger. With respect to real persons, Ward (1967) found a positive relationship between liking for President Lyndon Johnson and his judged height. We reasoned ( A . J. Lott & Lott, 1970) that since the heights of different children are objectively different, and sometimes strikingly so, a measure of head-relative-to-body size in drawings would be more likely to reflect the value or likability of one’s peers than size alone. First-grade children were asked to draw one highly liked, one highly disliked, and one neutrally regarded same-sex classmate, and liking was found to be significantly related to the dependent measure. Size of headrelative-to-body was greatest for liked, next greatest for disliked, and least for neutral peers. C. EVALUATIVE RESPONSES It is predicted that people will tend to ( a ) made positive judgments, on a variety of dimensions, of those they like, and negative judgments of those they dislike, ( b ) judge liked others as similar to themselves ( and disliked others as dissimilar to themselves) relative to positive or neutrally evaluated traits, and ( c ) behave as though they expect positive outcomes to be associated with liked persons and negative outcomes with disliked ones.

124

ALBERT J. LOTC AND BERNICE E. LOTT

In a classic study by Zillig reported in 1928 (cf. Krech & Crutchfield, 1948, p. 105), two groups of children performed calisthenic exercises before an audience of their classmates. One of the performing groups was composed of children who were almost uniformly disliked by their classmates, and the other group, of children who were liked. The experimenter had trained the liked group to make mistakes deliberately and the disliked group to perform the calisthenics letter perfect. At the end of the two performances the experimenter discovered that the audience had “seen’’ the disliked group as having made the mistakes. A mistake, it appears, is much more likely to be organized together with disliked people than with liked people.

This can be further illustrated with data reported by L. Berkowitz (1957) and, more recently, by Lerner (1965). In the former investigation highly liked partners were judged to have made a higher percentage of correct target estimates in an artillery game than were less liked partners, In the Lerner study, female subjects overheard two male confederates, who differed in attractiveness, working together at anagrams. When the less attractive fellow was selected by chance for payment of a sizable amount of money, the subjects were ‘less comfortable” and “less favorably impressed with the performance of both workers” than when the more attractive work partner was paid. Punishment for transgression has also been demonstrated to be predictably affected by interpersonal attraction. Defendants accused of negligence in an automobile homicide case, the details of which were read by subjects, were sentenced by them to longer prison terms when either the victim was described as attractive or when the defendant was described as unattractive (Landy & Aronson, 1969). Grifiitt and Jackson (1970), using a variation of the same case presented on videotape, varied the defendants’ attitudinal similarity to the subjects and found that subjects “exposed to a defendant with attitudes dissimilar to their own rated the defendant as more guilty, sentenced him to a greater number of years of imprisonment, recommended lengthier terms of imprisonment prior to parole eligibility, and evaluated him less positively on measures of attraction.” Persons also tend to be seen as more or less similar to oneself, depending upon whether or not they are liked. Davitz (1955) obtained sociometric information from children in a summer camp as well as their preferences for activities, and also asked the children to predict the preferences of their highest and lowest sociometric choices, He found that the former tended to be perceived as more similar to the respondent than the latter, and that high sociometric choices were seen as being more similar than they actually were. In a different setting, Kelman and Eagly (1965) obtained evidence from high-school

THE POWER OF LIKING

125

students indicating a reliable tendency to misperceive a communicator’s message by displacing it away from one’s own position on an issue, or toward one’s position, in direct relation to the strength of one’s negative or positive feelings toward the communicator. The results of a number of investigations indicate that people expect “good outcomes from association with liked persons (and “bad” consequences from association with disliked persons). In studies of “interpersonal bargaining,” for example, individuals who have positive attitudes toward one another are found to prefer cooperative strategies to competitive ones (Krauss, 1966; Schoeninger & Wood, 1969; Tornatzky & Geiwitz, 1968). In a study by Byrne and Andres (1964) of expressed expectancies about possible positive and negative reinforcements given by a hypothetical black person to a hypothetical white one interacting with him, significant differences were found between highprejudiced and low-prejudiced whites, with the ratio of positive to negative expectations being greater for the latter. Also directly relevant are data reported by Grif€itt (1969). Subjects were exposed to attitude scales presumably filled out by strangers, which varied in the proportions of attitudes similar to the subjects’. Three dependent measures were found to relate significantly to proportion of similarity: attraction toward the strangers, anticipated positiveness of contact with the strangers, and estimated success in solving problems with the strangers as partners. That liked persons are assigned generally more favorable personal characteristics than neutrally regarded persons and that disliked persons are assigned least favorable characteristics has been shown by A. J. Lott et al. (1970b). This study, mentioned previously, was concerned with the words which people use to describe one another. Subjects listed three persons with whom they were equally familiar and well acquainted, but toward whom they were differentially attracted, and then chose descriptive adjectives for each from a list of 200 trait words. The particular words chosen to describe liked, neutrally regarded, and disliked acquaintances differed significantly in their average ‘likability” indices (obtained from the trait words by Anderson, 1968). The results were the same for two independent samples. In addition it was found that persons who ore liked, neutral, and disliked seem to share with others who are similarly regarded, certain particular personality traits. Highly discriminating traits were: for liked persons-energetic, considerate, happy, intelligent, and truthful; for neutral persons-eautious, gullible, nervous, orderly, and quiet; and for disliked persons-complaining, insincere, narrow-minded, quarrelsome, and self-centered. L. Berkowitz and Knurek (1969) have recently shown that personal

126

ALBERT J. LOTT AND BERNICE E. LO’IT

evaluations can be mediated through names. A negative attitude toward one of two men’s names was classically conditioned. Then, after subjects were angered by the experimenter, they interacted with two confederates, one of whom had the critical name. Subjects not only behaved with greater overt hostility toward this confederate, but also made more negative judgments of his personality.

D. CONFORMITY, MODELING,AND ACQUIESCENTRESPONSES Liked persons should have the most, and disliked persons the least, influence on the behavior of other people, Influence can be inferred when one individual imitates the behavior of another (modeling), changes his behavior to approximate that of another (conformity) and/ or responds positively to a direct or indirect request for help. Each of the above illustrates what generally may be considered to be acquiescent behavior. People apparently believe that a person’s influence over another varies positively with the degree to which he is liked. Rittle and Cottrell (1967) asked a sample of males to make judgments about peoples’ relationships on the basis of limited information. When told that A likes B, for example, the mean subjective probability that A is influenced by B was found to be .72, and when told that A is influenced by B, the mean subjective probability that A likes B was found to be .70. In some studies of influence, attraction between persons is assumed on the basis of the degree of similarity between them, and this variable has been found to significantly affect the matching of responses to that of models, using varied behavior samples such as deep-sea diving preferences ( Burnstein, Stotland, & Zander, 1961) , nonsense-syllable preferences (Stotland, Zander, & Natsoulas, 1961), and game-playing (Rosekrans, 1967). Other variables have been shown to interact with similarity. Baron (1970), for example, has found that high attraction (based on similarity) facilitated imitation when the model was successful in performing a task, but interfered with imitation when the model was unsuccessful. In an ingenious field experiment by Brock (1965), similarity between prospective customers and their salesman (in the paint department of a large store) was manipulated by the latter’s conversation. The “similar” salesman was significantly more effective than the “dissimilar” one in influencing the customers to switch from their intended purchase to one of a different price. The warmth, “rewardingness,” or degree of nurturance in a relationship between people is also frequently utilized as the basis for inferences about liking, and is often investigated as a mediator of successful influence. Wright ( 1966) passed prearranged notes, which were

THE POWER OF LIKING

127

flattering or unflattering to the recipient, among experimental participants, and obtained evidence that influence attempts from liked persons (senders of flattering notes) are more effective than those from disliked ones. Sampson and Insko (1964) utilized trained assistants who engaged subjects in either pleasant or nasty interactions. Subjects’ judgments of light movement in an autokinetic situation were then found to shift in the direction of those made by the liked assistant (if the judgments started out as dissimilar to his), and to shift away from those made by the disliked assistant ( if the judgments started out as similar). Relevant here is the interpretation which Insko and Cialdini (1969) have offered to account for the effectiveness of the verbal reinforcer “good in changing attitudinal statements. They suggest that “good,” in contrast to a verbal reinforcer like “huh,” conveys approval of a person’s responses and, by implication, liking for the person himself which the person then reciprocates by liking the interviewer or experimenter. It is the liking for the interviewer which mediates conformity with statements he approves. In a study of observational learning by kittens (Cheder, 1969), those who observed their mothers press a lever for food in response to a visual stimulus were found to acquire that response sooner than those who observed a strange female cat. Similarly, investigations of children have found more rehearsal by them of the neutral behavior of a rewarding adult female model than a nonrewarding one (Mischel & Grusec, 1966), more adoption of a model’s criticism of their behavior when the model had previously been high in “rewardingness,” warmth, and friendliness (Grusec, 1966), and more imitation of the incidental, nontask-relevant behaviors of nurturant than nonnurturant mothers ( Mussen & Parker, 1965). Grusec and Skubiski (1970) have argued that investigations in which nurturance has not been shown to facilitate imitation are ones in which imitation necessitates the giving up of reinforcement and is consequently aversive. The secondary reinforcement to be derived from imitating a nurturant model would, under such conditions, be “neutralized.” The relationship between conformity and liking was tested by us in a study which measured the attraction between persons directly (A. J. Lott & Lott, 1961). An index of intermember liking within 15 natural college groups was obtained. Each group subsequently took part in a 30-minute discussion, at the end of which an opinion scale was administered to each subject. After the opinion scales were tabulated, results were reported which were exactly opposite to those actually expressed. On a second opinion scale, the percentage of group members who changed in the direction of the contrived group average opinion

128

ALBERT J. L O l T AND BERNICE E. LOlT

was found to be positively correlated with the level of liking among the group members. Janis and Hoffman (1970),in a very different context ( a smoking clinic for adults) also obtained a direct measure of liking between persons and related this variable to mutual influence. They found that the amount of liking which partners expressed toward each other was positively related to the amount of change in both attitude toward smoking and actual decrease in the number of cigarettes smoked, and that both liking and attitude change were greatest within pairs which maintained the greatest personal contact between weekly clinic meetings. “Helping” or altruistic behavior has been the focus of considerable recent attention, and Krebs ( 1970), in a review of the literature, suggested that interpersonal attraction has very likely “incidentally influenced the responses of benefactors in many studies [p. 2971.” In relatively direct tests of the relationship between liking and helping, Daniels and Berkowitz (1963) found more boxes made for dependent supervisors who were liked, Sawyer (1966) that more altruism was exhibited toward friends than toward strangers and antagonists, Staub and Sherk (1970) that fourth graders shared a crayon longer with liked than with disliked partners, Bryan and Test (1967) that more white shoppers contributed to a Salvation Army kettle manned by a white woman than to one manned by a black woman, and Golightly, Huffman, and Byrne (1971) that graduate business students would lend more money to similar strangers whom they Iiked than to dissidar strangers, who they Iiked less. These studies, we believe, fit within the larger framework of research concerned with influence in that one person’s need (whether verbalized or not) provides a suggestive direction for another person’s behavior. A study using the “lost letter” technique provides further confirming data (H. A. Hornstein, Masor, Sole, & Heilman, 1970). More pedestrians in a Jewish neighborhood in Brooklyn mailed a recovered questionnaire (found in a stamped, addressed envelope) filled out so as to agree with U. S. aid to Israel than mailed a recovered questionnaire indicating disagreement with such a policy. Interviews with other persons in the same neighborhood had previously shown that anonymous strangers described as having pro-aid opinions were better liked than those described as anti-aid. In the present authors’ study of doll-play behavior, previously mentioned, one situation was designed to elicit help from the self-doll for a peer-doll. Each subject was told that “Your whole group is going to see a special cartoon movie which you know is really great. Everyone in your group is already there waiting for the movie to begin, but you and (peer-dolI) are late . . . You are both running . . . all of a sudden (peer-doll) falls down . . . You look back and see her on the ground

THE POWER OF LIKING

129

. . . show

me what happens . . .” Behavior of the self-doll was rated as having provided active, physical help, e.g., running back to the peer, helping her up, etc., or as having provided verbal reassurance only, e.g., calling back “Are you O.K.? Come on.” Liked peer-dolls were actively helped significantly more (by 9 out of 12 girls) than were disliked peer dolls (by 6 out of 14 girls). The obtained Critical Ratio for differences between proportions, 1.68, has a one-tailed probability of less than .05.

E. VICARIOUS RESPONSES When positive or negative goal response behavior by one person ( a model) is observed or inferred by another, matching fractional anticipatory responses will be evoked in the observer to the extent that the model is liked. When the event is rewarding for the model and the observer’s implicit responses are therefore positive, the observer’s reaction may be called empathic; when the event is punishing or frustrating for the model and the observer’s implicit responses are therefore affectively negative, his reaction may be called sympathetic. In Berger’s (1962) analysis of vicarious instigation, both of these observer responses are labeled empathetic and are so classified because of the similarity or concordance between the affect experienced by the model (performer) and the observer. We suggest that both empathy and sympathy are concordant, with the former defined as a positive, and the latter a negative, vicarious reaction to appropriate conditions. The likelihood of each of these kinds of reactions is expected to be a positive function of the degree to which the model is liked by the observer. Some of the literature cited in the previous section as demonstrating greater imitation of the overt acts of nurturant models and greater helping of liked persons-in-need is relevant to this hypothesis. The report by Tannenbaum and Gaer (1965) of a positive relation between the degree of identification with a movie protagonist and the degree of stress induced in observers watching him in a dangerous situation also can be viewed as supporting our proposed relationship between liking and sympathy. The present authors tested this relationship directly in the doll-play study to which we have referred before. One situation presented to the subjects was designed to elicit sympathy. The peer-doll, in an imaginary situation, was accused by her counsellor of having lost the group’s ball, and the group was described as “turning against” her. The subject’s behavior, through the actions of the self-doll, was categorized either as sympathetic (e.g., offering to help find the ball, defending the accused peer, etc.) or as unsympathetic (either actively hostile or simply unconcerned). Significantly more subjects behaved positively when the accused peer-doll was liked (11 out of 14) than

130

ALBERT J. L O l T AND BERNICE E. LO lT

when the peer-doll was disliked ( 3 out of 12). The obtained Critical Ratio of 2.84 for the difference between proportions has a one-tailed probability of < .002. Data relevant to the present discussion were obtained in a study of the acquisition of interpersonal liking under conditions of vicarious reward (A. J. Lott et al., 196913). Groups composed of six boys or girls played Bingo games in which win-loss patterns were controlled. Each group contained three pairs of friends, with each pair being unacquainted with the others. One member of each pair was chosen, by chance, to play the games and be eligible for a prize, while the other member of each pair (the Helper) observed passively. After the games, all the children (Players and Helpers) filled out questionnaires in which their affective reactions to winning and losing were assessed. Among the Helpers, those who said that they had felt “very happy“ about their partner’s wins of 4, 3, or 1 game(s) reported liking the other group members more than those who answered otherwise, and those Helpers who reported being “very s a d about their partner’s losing all four games liked the other group members less than those who said they had felt “all right” or “didn’t care.” Helpers seemed to have responded in the same way as the Players (their friends), both in terms of affective reactions to wins and losses and of consequent degree of liking or disliking for the other children who played the games with them. What might be expected of the vicarious responses to another person’s success or discomfort when that other person is someone we dislike? Our analysis leads generally to the following predictions: to the extent that the model is disliked by the observer, attainment by the former of positive consequences or reward should evoke envy in the observer (rf-sf rather than rg-sg),and the attainment by the model of negative or aversive consequences should evoke in the observer sadistic pleasure. The extent to which such pleasure will accompany the observation of the suffering or discomfort of a disliked person should be strongly influenced by the observer’s previous history of successful aggressive behavior, so that another’s discomfort has already become associated with his own goal attainment and can therefore serve to evoke rg-sis. Berger (1962) used the terms envy and sadism to describe discordant vicarious reactions, i.e., responses by the observer which were dissimilar to those of the model. Our hypothesis is that such discordant responses are more likely to occur when the model is disliked.

F. LEARNING ( a ) In the actual or symbolic presence of liked persons, simple task performance (where response competition is relatively low) is expected

THE POWER OF LIKING

131

to be enhanced and behavior to be more vigorous, compared to an alone condition or being in the presence of disliked or neutral persons. Performance in the presence of a disliked person will be impaired. ( b ) Learning relative to differentially liked persons, for example, those whose names may constitute the responses to new stimuli, will be fastest for liked persons, intermediate for disliked persons, and slowest for neutral ones. ( c ) When differentially liked persons, or their representations, are made contingent upon some behavior, this behavior will be strengthened by a liked person and weakened by a disliked one. ( d ) Performance on a task typically and/or previously followed by reward can be maintained by the presence of a liked person in the absence of, or under conditions of delay in, that reward. The effect on learning of the relationship between the learner and some other person who performs some role in the learning situation (typically the experimenter) has been studied in a variety of different situations. The focus of attention has been primarily on the learner’s attitude toward the person who provides him with positive or negative consequences for his behavior. Our secondary reinforcement interpretation of the effects of being in the presence of liked or disliked persons requires a more stringent analysis of the experimental conditions than is ordinarily given to them in studies of this kind. It is necessary to know if the liked or disliked person is ( a ) present at the beginning of a learning sequence, when he can potentially affect the learner’s drive level and have a motivating effect; or ( b ) if he is present throughout the learning sequence and can, therefore, maintain drive level in the learner and also perhaps strengthen certain behaviors selectively by giving approval to them; or ( c ) if he directly dispenses positive and/or negative rewards, either verbal or material, to the learner; or ( d ) if it is only his presence subsequent to certain responses (and not others) which serves as a reward for the learner. Most investigations in this area fit the conditions of (27) and ( c ) above. Better learning (higher rates of conditioning, task persistence, and so on) is generally found with an experimenter who dispenses reinforcement after he has established a warm relationship with his subjects (e.g., Allen, Spear, & Johnson, 1969; H. Berkowitz & Zigler, 1965; McCoy & Zigler, 1965). Liking for the experimenter based on other information or manipulations typically produces the same results. Thus, mothers have been shown to be more effective than female strangers in instrumentally conditioning smiling behavior in their 3-month old infants ( Wahler, 1967), and friends have been shown to be more effective than nonfriends in changing response preferences on a simple motor task among second and third graders (Patterson & Anderson, 1964). Negative findings have also been reported, however, by Stevenson, Keen, and Knights (1963),

132

ALBERT J. LOTI’ AND BERNICE E. LOlT

on parents as compared with strangers, and by Hartup (1987) for nursery schoolers with liked peers as compared with disliked ones, suggesting that, over a period of time, consistently positive reinforcement from the same person may lead to satiation. Aronson and Linder (1965) found, for example, that an experimental confederate who first evaluated subjects negatively and then positively was liked more than a confederate who dispensed only positive feedback, although the latter was liked significantly more than a confederate who gave only negative evaluations. Attitudes toward the tester or experimenter have been manipulated on the basis of his skin color in studies by Kennedy and Vega (1965), Katz, Henchy, and Allen ( 1968), and by Smith and Dixon (1968). In the first two, Negro children performed better on an oddity discrimination and a verbal learning task, respectively, with black than with white teachers. In the latter, verbal conditioning of high prejudiced white females took place with white, but not with black experimenters. In laboratory situations where subjects have been instructed to find an experimenter highly congenial or uncongenial (Sapolsky, 1960), or where positive attitudes toward the experimenter were assumed to generalize from positive attitudes toward “medical scientists” ( Krasner, Knowles, & Ullmann, 1965), subjects performed better on a verbal conditioning task and a muscular task, respectively, when the experimenter was liked. Also relevant is a large-scale investigation by McKeachie, Lin, Milholland, and Isaacson (1966) of academic achievement under natural classroom conditions. Men with high affiliation needs were found to make better grades in classes having warm, friendly instructors than in classes where the instructors were primarily subject- or self-oriented, whereas men with low needs for affiliation behaved oppositely. In a smaller number of studies, the subjects’ known or assumed attitudes toward persons present while they are performing a task have been shown to influence that performance. Zajonc (1965), in reviewing the literature on social facilitation, proposed that the mere presence of other persons (as audience or spectators) “increases . , general arousal or drive level,” thereby facilitating the emission of dominant responses and impairing new learning where dominant responses are the incorrect ones. Cottrell (1968) and A. J. Lott and Lott (1968) have argued that the relationship between the performer and the audience may also need to be taken into account, CottrelI suggesting that the subject’s “anticipation of positive or negative outcomes” from the spectators is important and that “energizing effects” do not follow if such outcomes are not expected, and we ourselves suggesting that an audience, or co-actors, consisting of liked, neutral, or disliked persons may have differing effects.

.

THE POWER OF LIKING

133

In a study utilizing small groups of children, composed of highly liked or disliked same-sex classmates, we tested the effect of being in the presence of differentially liked persons on the learning of Spanish equivalents for English words (A. J. Lott & Lott, 1966). It was found that on measures of speed with which the groups went through the learning procedure, individual acquisition, retention over a l-week period, subsequent relearning by means of a different study procedure, and the learning of a second similar task, high IQ children did better in the presence of highly liked classmates than in the presence of less liked ones, whereas there was some tendency in the opposite direction for low IQ children. These data were interpreted as supporting the proposition that the presence of highly liked others raises individual drive level. For the low IQ children, for whom the experimental tasks were difficult, the drive increment served to decrease learning efficiency because high drive strengthens competing incorrect responses as well as correct ones. For the high IQ children, on the other hand, the tasks were relatively simple; there was thus little response competition, and increased drive served to improve performance. Data reported by Calvin, Hoffman, and Harden (1957) on problem solving tend to support the above interpretation. They found that small groups with permissive climates (which can be assumed to generate positive interpersonal attitudes) are superior to traditional ones for individuals of high intelligence, while they are a handicap for individuals of average intelligence. Matthews (1968), in R doctoral dissertation study, made a direct test of the effect of differential liking for a spectator on performance in a learning task under conditions differing in degree of response competition, She found that high-school sophomore girls, who learned a moderately difficult set of paired nonsense syllables in the presence of a highly liked peer, took significantly longer to learn the list and made significantly more errors than girls who learned the same list in the presence of a disliked or neutrally regarded peer. In another set of experimental conditions, different subjects began the learning task alone, and then a liked, neutral, or disliked peer observer was introduced into the learning situation after the subjects had almost mastered the paired associates. Under these conditions, in sharp contrast with the first, there was a significant spurt in learning and reduction in errors when the late-arriving observer was a liked peer, as opposed to a neutral or disliked one. Interpersonal attitudes have been shown to have still another effect on learning. A. J. Lott, Lott, and Walsh ( 1 9 7 0 ~ found ) that subjects who learn a list of paired associates in which the names of differentially liked real persons are paired with nonsense syllables (CVC trigrams), make fewest errors relevant to liked persons, most errors for neutrally re-

134

ALBERT J. LOTT AND BERNICE E. LOlT

garded, and an intermediate number of errors for disliked persons. These results were obtained with two independent geographically separated samples, one of which learned to associate the names of well-known public figures with trigrams, and the other the names of persons with whom they were very well acquainted. In both cases, attitudes toward the persons who were to be associated with new information made a significant difference in the speed or ease with which the learning took place. We interpret these data as supporting the theoretical assumptions that liked, disliked, and neutral persons differ in drive-arousing properties and also in salience. A different experimental paradigm is required to demonstrate the reinforcing effects of differentially liked persons, We had 100 ninth graders learn a moderately difficult discrimination problem under conditions where correct or incorrect responses were followed by photographs of differentially liked peers or blank cards ( A . J. Lott & Lott, 1969). Subjects performed in the following order, from best to poorest: those whose correct responses were reinforced with the printed word “right” and incorrect ones with a blank card; those whose correct responses were reinforced with the photograph of a liked peer and incorrect ones with a blank card; those whose correct responses were followed by a blank card, while incorrect ones were followed by the photo of a disliked peer; those whose correct responses were reinforced with a neutral peer photo; and those whose correct responses were reinforced with the photo of a disliked peer. These data indicate that liked and disliked persons (symbolically represented in this study) can eff ectiveIy function as positive and negative reinforcers, respectively.

G . RESPONSESTO STRESS

If a liked person is introduced into a stressful situation (one which contains cues for impending danger or pain or discomfort) his or her presence will reduce anxiety as inferred from a variety of typically employed behavioral and physiological indices. This prediction rests upon the assumption that the approach responses which will be evoked by a liked person will compete with the anxiety responses evoked by the threatening or distressing situation, disrupting and lessening the latter. A similar, but significantly lesser, effect should be produced even by a neutral person, whereas a disliked person will not effectively reduce anxiety responses and may increase them. Data from a number of investigations are relevant to this hypothesis. Animal studies, particularly, have shown that emotional responses to situations which typically evoke them can be reduced by the presence of

THE POWER OF LIKING

135

certain other animals, e.g., a mother goat (Liddell, 1950), a nonfearful rat (Davitz & Mason, 1955), a mother rat (Lubow & Marcuse, 1956), a terry-cloth surrogate monkey mother ( Harlow & Zimmerman, 1959),and a same-age monkey (Mason, 1960). In a review of studies of infantmother relationships in mammals and birds, King (1966) concludes that the mother can “reduce emotional distress in her offspring” because of particular stimuli associated with her which evoke pleasant feelings in the infant. With human children, reduction of emotionality in a stressful situation as a function of the presence of mothers or friendly adults has been demonstrated by Arsenian (1943) in an observational study of the behavior of nursery schoolers in a strange room. One study of adults, using a physiological measure of plasma free fatty acid (FFA), found that being with friends is less “arousing” than being with strangers (Back & Bogdonoff, 1964). “Natural” groups, each made up of four friends and “recruited groups, each made up of four strangers, participated in a conformity-inducing situation (which can be assumed to be stressful); the investigators found that the “plasma FFA levels at the time of initial puncture and at rest were lower in the natural groups than in the individually recruited subjects.” Related findings from a rather different situation have been reported by Baratz (1967) who made use of the Test Anxiety Questionnaire. Black undergraduates who were administered the test by a black examiner reported significantly less anxiety than did black students who were tested by a white examiner. Similarly, Katz and Greenbaum (1963) found that when black students were led to expect “strong nonavoidance electric shocks” while working on digit-letter substitutions, they performed better with a black administrator and co-worker than with a white administrator and co-worker. The most direct evidence for a “liking effect” on responses to stress comes from a study by Kissel (1965) who induced stress in college students by task failure on unsolvable designs and measured responses to stress by obtaining GSR readings. Subjects who experienced the failure in the presence of a friend (who was doing something different) showed significantly smaller reactions than those who experienced the failure either alone or in the presence of a stranger. Kissel reasoned that “the presence of a friend, an emot:onally pleasing stimulus, elicits positive feelings, which tend to compete with negative feelings . . , stimulated by the noxious situation . . . Generally friends tend to have attributes similar to those who in the past have been dispensers of security, affection, and protection. However, strangers . . . will not call forth competing responses and thus the stress will not be appreciably reduced or minimized [p. 3831 .”

136

ALBERT J. LOTT AND BERNICE E. LOlT

IV. Areas of Applicability Hesitantly, but with conviction, the present authors believe that the theoretical propositions and supporting data concerned with the effects of interpersonal attitudes on behavior can be scrutinized, as Walker (1969) has recently urged, for their “social relevance.” We agree with Kelman (1968) that social scientists should not only generate insights about people and society, but also, on the basis of these insights, suggest ways in which the conditions of human life may be improved. In pointing toward potential applications, our focus will be on the elementary school because: ( a ) fundamental and long-lasting attitudes are learned during the years that it is attended; and because ( b ) children of elementary school age are most available for, and potentially adaptable to, planned interventions; and because ( c ) schools are precisely where educators currently introduce various conditions, materials, and techniques designed to increase the academic and social skills of children. Within the setting of the elementary school, our concern is with two areas on which we believe positive interpersonal attitudes can have a generally constructive influence on children’s lives: educational achievement and reduction of intergroup hostility, especially between white children and black children, or those of other discriminable minorities such as Mexican or Puerto Rican. If the conditions under which children can acquire positive interpersonal attitudes are explicitly introduced into the classroom, the consequences which folIow under laboratory conditions should also manifest themselves in a natural environment by behaviors which are relevant to the specific goals noted above. Perhaps the one effect of interpersona1 attraction that has the most far-reaching potential implications is its infiuence on communication. Children who like one another should not only talk to each other more, and about more personal and important subjects, but also should be more sensitive to each other’s nonverbal communication and interpret it more accurately. Such effects may have special significance for the teaching of children from economically depressed homes where verbal communication tends to be less extensive and abstract. If the classroom is a place where children like one another and the teacher, and if it is also composed of verbally skillful as well as less skillful children, then the greater volume of talking should facilitate language learning by the latter group. Language refinement, depending, as it seems to, on feedback and reinforcement from verbally skilled persons, should proceed more effectively in a classroom where verbally skilled classmates, as well as the teacher, are available to help shape language responses. In the presence of liked others, a child will be more motivated to produce language which can be

THE POWER OF LIKING

137

responded to by others. Feedback from liked others will, in turn, have a greater effect in influencing future language behavior than feedback from disliked others. If some of the liked others are verbal experts, the quality of the language and other verbally communicated skills learned in this way should be high. It would seem to follow, then, that there should be a clear advantage in integrating classrooms, in the sense of mixing children from different socioeconomic levels. Illustrative data relevant to this point come from some analyses of the extensive nationwide study of school desegregation and academic performance by Coleman et aZ. (cf. McPartland, 1969). When family background differences of black high school students were controlled, their verbal achievement in predominantly white schools was shown to be positively related to the proportion of their classmates who were white, suggesting strongly that “it is desegregation at the classroom level” which is most important. An average difference of about one year of academic growth differentiated black ninth graders who attended majority white classes from those in entirely black classes. It is certainly to be expected that the same effect would be obtained by integrating poor black children with middle-class blacks or by integrating poor white children with either middle-class whites or middle-class blacks. The key variable is social class. The Coleman survey “showed that the social class segregation which accompanies racially segregated schooling accounted for more of the variation in the achievement of Negro high school students than did any of the differences in school facilities, curriculum, or teaching staff which were measured [ McPartland, 1969, p. 931 .” The data reported above are consistent with other studies of high school students which indicate that the characteristics of one’s friends have more influence on, for example, college plans than the characteristics of the total student body or school in general ( E . Q. Campbell & Alexander, 1965). It has also been reported (cf. Katz, 1968) that in schools having more than 50% white students, where the white students report having close friends among black children and vice versa, the black children scored better on achievement tests than black children in similar schools where the students report low levels of intergroup friendship. The degree to which positive interpersonal attitudes exist among black and white children may well be a major criterion by which to differentiate merely desegregated from truly integrated schools. Liked persons, under a variety of different conditions, appear to be more effective dispensers of social and verbal reinforcements than disliked or neutral ones, although over a period of time the receipt of only positive reinforcement from a liked source may lead to satiation. It may be expected that a pattern of positive and negative outcomes from a

138

ALBERT J. LOTT A N D BERNICE E. LOTT

liked teacher, for discriminably different and appropriately labeled “good and “ b a d behavior should be more powerful than the dispensation of rewards only. Even when liked persons are not the agents of verbal or material reward, they will have discernible effects on the learning and performance of children who are simply working in their presence because the presence of liked persons is motivating. The facilitating or interfering effects of this increase in general drive will depend, however, upon the nature of the task and the child’s abilities or previously acquired skins. Where the task is an easy one or depends upon well-established responses, the presence of liked classmates should improve a child’s performance; if, on the other hand, the task is difficult or new, being in the presence of highly liked peers will retard learning because all of the competing responses will be strengthened. Still another consequence of being in the presence of persons we like, which appears to have clear significance for a child’s school experience, is the reduction of behavior usually evoked by stressful or anxietyprovoking situations. With liked classmates present to evoke in a child the expectation of positive outcomes or pleasant experiences, behavior occurs which can successfully compete with responses to anxiety provoking stimuli like a test, a difficult assignment, a new teacher, and so on. Related to the potential effectiveness of a liked person as a fear reducer is the probability that he or she is a highly distinctive stimulus and can, thereby, introduce into a situation numerous cues which can divert attention from the stress-producing or stress-mediating cues. The greater distinctiveness, or greater number of perceived characteristics, of liked persons as compared with neutral ones should influence the interpersonal perceptions of black and white children, for example, once they have learned to like each other in the classroom. The failure of white persons in our society to make distinctions among black individuals has had profound and tragic personal and cultural consequences. We would expect that black and white children who are given the opportunity to acquire positive attitudes toward one another, and can consequently describe one another in terms of numerous qualities, will be less likely to entertain stereotypic conceptions about each other and less likely to view members of the other group as “all alike.” This prediction is supported by the findings of Yarrow, CampbelI, and Yarrow (1958) who assessed various individual and group behaviors of children attending summer camp for a 2-week period, some in segregated and others in desegregated groups. Among a large number of other findings were the following: analysis of childrens’ free descriptions of each other revealed more complex interpersonal perceptions in

THE POWER OF LIKING

139

desegregated than in segregated settings; and at the end of camp, black children were significantly more desired as friends by white children than they had been earlier in the sessions. Not only is our perception of the distinctiveness of a person i d u enced by whether we are positively, negatively, or neutrally attracted to him, but our perception of the nature of his specific attributes is also affected. High liking leads to the perception of quality performance by the liked person and high regard for his opinions and beliefs. We tend to see those we like as somewhat better than they actually are. Thus children should be less critical of classmates whom they like and more sensitive to their feelings, which they can interpret more easily, even when the available cues are ambiguous. Related to this greater sensitivity to, and perception of, positive qualities in those we like is our tendency also to regard liked others as more similar to ourselves than neutral or disliked others and as more similar than they actually are. We also tend to assume that our positive or negative attitudes toward other persons are reciprocated. Thus, in an integrated classroom, for example, where high levels of interpersonal liking have been achieved, there should be less rejection of one child by another because of feelings of strangenessfor perceptions of dissimilarity associated with differences in skin color, language, dress, and so on. Instead, liking should evoke perceptions of similarity and the expectations that one’s feelings are reciprocated. These perceptions and expectations should also be influential in shaping a child’s self-image since this will be mediated by evaluations of the child’s behavior by his or her teacher and peers. It seems likely that the cues about ourselves which come to us from the way persons we like behave toward us are received at a lower threshold, and that we receive more information about ourselves (and other issues) from those we like because we spend more time physically oriented toward them and attending to them than we do to other persons. In the previously mentioned study by Yarrow et al. (1958), the black children living in desegregated camp groups were found to change their self-esteem and view of other blacks in a more positive direction than the black children in segregated groups. Also lending support to the possibility that “interracial contact may engender black pride” are interesting recent findings reported by Hraba and Grant ( 1970). Children between the ages of 4 and 8, from interracial classrooms, were tested by means of the Clark and Clark doll-choice technique. The black children who preferred black dolls did not have all black friends; in fact, more black children who had both black and white friends preferred black

140

ALBERT J. LO’IT AND BERNICE E. LO’IT

dolls than those who had all black friends. This finding approached statistical significance. The majority of both the black and the white children preferred dolls most similar to themselves in skin coIor. The foregoing sketch of areas of appIicability has gone beyond available empirical data, but it has by no means exhausted the implications of the potential effects of liking for children in a classroom or small group setting. Our intent was to sample implications which can be explored further, Assuming, now, that a variety of behaviors does follow as a consequence of our differential attraction to other persons, and assuming, further, that we consider most of the consequences of liking to be positive ones, it is essential that practitioners understand the nature of the major variables upon which the learning of liking depends. If liking is to be used to mediate social change, we need to know how to produce it in a reliable, stable fashion. At the beginning of this paper the basic principles governing the learning of liking, from the point of view of a liberalized S-R reinforcement theory, were outlined. The central proposition is that a neutral person who is directly or indirectIy associated with rewards over time will become a conditioned stimulus for the anticipation of reward. Such a person can then be said to evoke a positive attitude and to be able to function as a secondary reinforcer. Knowledge of child development and educational procedures, when related to this proposition and to additional hypotheses, should provide concrete suggestions regarding how children can obtain rewarding outcomes or pleasant experiences in the classroom, among their classmates and teachers, in order to increase the interpersonal attraction among them. Derivable from the theory, for example, are the predictions that variables such as immediacy us. delay of reward, and the quality of reward, will influence the learning of liking, along with frequency of reward. The concept of reward, once we move beyond stimuli which reduce physiological needs, requires careful definition when applied to human beings with a learning history or culture. It is necessary to know what goal or state of affairs is considered desirable for different children or different groups of children. The incentive power of many typically used school rewards is more firmly established during socialization of a middle-class child or children from homes of particular ethnic backgrounds than during the socialization of other children. In this connection, Havighurst (1970) has recently theorized that the effective human reward-punishment system changes with age and that different subcultures “carry their children along this evolutionary path at different rates and in different ways [p. 2791.” A teacher who desires to organize a classroom so as to facilitate the

THE POWER OF LIKING

141

development of interpersonal attraction among the pupils must arrange activities in such a way that children will experience satisfaction in one another’s presence, utilizing reward stimuli known to constitute desirable events for the particular children in that classroom. Then the teacher may provide situations in which the level of interpersonal attraction among the children will be able to function most effectively to mediate those consequences considered most desirable. REFERENCES Allen, S. S., Spear, P. S., & Johnson, J. Experimenter role effects on children’s task performance and perception. Child Deoelopment, 1969, 40, 1-10. Amsel, A. Frustrative nonreward in partial reinforcement and discrimination learning: Some recent history and a theoretical extension. Psychological Reoiew, 1962, 69, 306328. Anderson, N. H. Likableness ratings of 555 personality-trait words. Journal of Personality and Social Psychology, 1968, 9, 272-279. Aronson, E., & Linder, D. Gain and loss of esteem as determinants of interpersonal attraction. Journal of Experimental Social Psychology, 1965, 1, 156-171. Arsenian, J. M. Young children in an insecure situation. Journal of Abnormal and Social Psychology, 1943, 38, 225-249. Asthana, H. S. Perceptual distortion as a function of the valence of perceived object. Journal of Social Psychology, 1960, 52, 119-125. Back, K. W., & Bogdonoff, M. D. Plasma lipid responses to leadership, conformity, and deviation. In P. H. Leiderman & D. Shapiro (Eds.), Psychobiological approaches to social behaoior. Palo Alto: Stanford University Press, 1964. Bagby, J. A cross-cultural study of perceptual predominance in binocular rivalry. Journal of Abnormal and Social Psychology, 1957, 54, 331334. Baratz, S. S. Effect of race of experimenter, instructions, and comparison popdations upon level of reported anxiety in Negro subjects. Journal of Personality and Social Psychology, 1967, 7, 194-196. Baron, R. A. Attraction toward the model and model’s competence as determinants of adult imitative behavior. Journal of Personality and Social Psychology, 1970, 14, 345-351. Berger, S. M. Conditioning through vicarious instigation. Psychological Reoiew, 1962, 69, 450-466. Berkowitz, H., & Zigler, E. Effects of preliminary positive and negative interactions and delay conditions on children’s responsiveness to social reinforcement. Journal of Personality and Social Psychology, 1965, 2, 500-505. Berkowitz, L. Liking for the group and the perceived merit of the group’s behavior. Journal of Abnormal and Social Psychology, 1957, 54, 353-357. Berkowitz, L., & Knurek, D. A. Label-mediated hostility generalization. Journal of Personality and Social Psychology, 1969, 13, 200-206. Berscheid, E., & Walster, E. H. Interpersonal attraction. Reading, Mass.: AddisonWesley, 1969. Black, R. On the combination of drive and incentive motivation. Psychological Reoiew, 1965, 72, 310-317. Bootzin, R. R., & Natsoulas, T. Evidence for perceptual defense uncontaminated by response bias. Journal of Personality and Social Psychology, 1985, 1, 481-488.

142

ALBERT J . LOTT AND BERNICE E. L O l T

Bootzin, R. R. & Stephens, M. W. Individual differences and perceptual defense in the absence of response bias. Iournal of Personality and Social Psychology, 1967, 6, 408-412. Brock, T. C. Communicator-recipient similarity and decision change. Journal O# Personality and Social Psychology, 1965, 1, 050-654. Bruner, J. S., & Goodman, C. C. Value and need as organizing factors in perception. ]ournal of Abnormal and Social Psychology, 1947, 42, 3 3 4 4 . Bruner, J. S., & Postman, L. Symbolic value as an organizing factor in perception. Journal of Social Psychology, 1948, 27, 203-208. Bryan, J. H., & Test, M. A. Models and helping: Naturalistic studies in aiding behavior. Journal of Personality and Social Psychology, 1967 6, 400407. Bumstein, E., Stotland, E., & Zander, A. Similarity to a model and self-evaluation. ]ournal of Abnormal and Social Psychology, 1961, 62, 257-264. Byrne, D., & Andres, D. Prejudice and interpersonal expectancies. Journal of Negro Education, 1964, 33, 441-455. Byme, D., & Clore, G . L. Effectance arousal and attraction. Journal of Personality and Social Psychology, 1967, 6( 4, Whole No. 638). Byrne, D., & Clore, G. L. A reinforcement model of evaluative responses. Personality: An International Journal, 1970, 1, 103-128. Byrne, D., Ervin, C. R., & Lamberth, J. Continuity between the experimental study of attraction and real-life computer dating. Journal of Personality and Social Psychology, 1970, 16, 157-165. Calvin, H. D., Hoffman, F. K., & Harden, E. L. The effect of intelligence and social atmosphere on group problem solving behavior. Journal of Social Psychology, 1957, 45, 61-74. Campbell, D. T. Stereotypes and the perception of group differences. American Psychologist, 1967, 22, 817-829. Campbell, E. Q., & Alexander, N. L. Structural effects and interpersonal relations. American Journal of Sociology, 1965, 71, 284-289. Chance, J., Goldstein, A. G., & Schicht, W. Effects of acquaintance and friendship on childrens’ recognition of classmates’ faces, Psychonomic Science, 1967, 7, 2% 224. Cheder, P. Maternal in5uence in learning by observation in kittens. Science, 1969, 166, 901-903. Clore, G. L. Attraction and interpersonal behavior. Unpublished manuscript, University of Illinois, 1969. Cofer, C . N., & Appley, M. H. Motivation: Theorv and research. New York: Wiley, 1964. Cottrell, N. B. Performance in the presence of other human beings; mere presence, audience, and affiliation effects. In E. C. Simmel, R. A. Hoppe, & G. A. Milton (Eds.), Social facilitation and imitatioe behavior. Boston: Allyn & Bacon, 1968. Craddick, R. A. Size of Halloween witch drawings prior to, on, and after Halloween. Perceptual and Motor Skills, 1963, 16, 235-238. Crockett, W. H. Cognitive complexity and impression formation. In B. A. Maher (Ed.), Progress in experimental personality research. Vol. 2. New York: Academic Press, 1965. Daniels, L. R., & Berkowitz, L. Liking and response to dependency relationships. Human Relations, 1963, 16, 141-148. Davitz, J. R. Social perception and sociometric choice of children. Journol of Abnormal and Social Psychology, 1955, 50, 173-176.

THE POWER OF LIKING

143

Davitz, J. R., & Mason, D. J. Socially facilitated reduction of a fear response in rats. Journal of Comparatioe and Physiological Psychology, 1955, 48, 149-151. Doob, L. W. The behavior of attitudes. Psychological Reoiew, 1947, 54, 135-156. Efran, J. S., & Broughton, A. Effect of expectancies for social approval on visual behavior. Journal of Personality and Social Psychology, 1966, 4, 103-107. Eisman, B. Attitude formation: The development of a color-preference response through mediated generalization. Journal of Abnormal and Social Psychology, 1955, 50, 321326. Eisman, B., & Levy, J. The influence of certain communicator characteristics on lip reading efficiency. Journal of Social Psychology, 1960, 51, 419-/125. Engel, E. Binocular methods in psychological research. In F. P. Kilpatrick (Ed.), Explorations in transactional psychology. New York: New York University Press, 1981. Exline, R., Gray, D., & Schuette, D. Visual behavior in a dyad as affected b y interview content and sex of respondent. Journal of Personality and Social Psychology, 1965, 1, u)1-209. Festinger, L. Informal social communication. Psychological Reoiew, 1950, 57, 271282. Goldberg, G. N., Kiesler, C. A., & Collins, B. E. Visual behavior and face-to-face distance during interaction. Sociometry, 1969, 3 5 43-53. Golightly, C., Huffman, D. M., & Byrne, D. Liking and loaning. Unpublished manuscript, Purdue University, 1971. Greenwald, A. G. On defining attitude and attitude theory. In A. G. Greenwald, T. C. Brock, & T. M. Ostrom (Eds.), Psychological foundations of attitudes. New York: Academic Press, 1968. Criffitt, W. Attitude evoked anticipatory responses and attraction. Psychonomic Science, 1969, 14, 153-155. Criffitt, W., & Jackson, T. Simulated jury decisions: The influence of jury-defendant attitude similarity-dissimilarity. Unpublished manuscript, Kansas State University, 1970. Grusec, J. Some antecedents of self-criticism. Journal of Personality and Social Psychology, 1966, 4, 244-252. Grusec, J., & Skubiski, S. L. Model nurturance, demand characteristics of the modeling experiment, and altruism. Journal of Personality and Social Psychology, 1970, 14, 352-359. Harlow, H. F., & Zimmerman, R. R. Affectional responses in the infant monkey. Science, 1959, 130, 421-432. Hartup, W. W. Peers as agents of social reinforcement. Young Children, 1967, 20, 214-228. Havighurst, R. L. Minority subcultures and the law of effect. In F. F. Korten, S. W. Cook, & J. I. Lacey (Eds.), Psychology and the problems of society. Washington, D. C.: American Psychological Association, 1970. Heider, F. The psychology of interpersonal relations. New York: Wiley, 1958. Hess, E. H. Attitudes and pupil size. Scientific American, 1965, 212, 46-54. Hess, E. H., Seltzer, A. L., & Shlein, J. M. Pupil response of hetero- and homosexual males to pictures of men and women: A pilot study. Joulnal of Abnormal Psychology, 1965, 70, 165-168. Hornstein, H. A., Masor, H. N., Sole, K., & Heilman, M. Effects of sentiment and completion of a helping act on observer helping: A case of socially mediated Ziegamik effects. Journal of Personality and Social Psychology, 1970, 17, 107-112.

144

ALBERT J. LO’IT AND BERNICE E. LOlT

Hornstein, M. G. Accuracy of emotional communication and interpersonal compatibility. Journal of Personality, 1967,35, 20-30. Hraba, J., & Grant, G. Black is beautiful: A re-examination of racial preference and identification. Journal of Personality and Social Psychology, 1970, 16, 398-402. Hull, C. L. Essentials of behavior. New Haven: Yale University Press, 1951. Hull, C. L. A behavior system. New Haven: Yale University Press, 1952. Insko, C. A., & Cialdini, R. B. A test of three interpretations of attitudinal reinforcement. Journal of Personality and Social Psychology, 1969, 12, 333-341. Irwin, M., Tripodi, T., & Bieri, J. Affective stimulus value and cognitive complexity. Journal of Personality and Social Psychology, 1967, 5, 444-448. James, G., & Lott, A. J. Reward frequency and the formation of positive attitudes toward group members. Journ~lof Social Psychohgy, 1964, 62, 111-115. Janis, J. L., & Hoffman, D. Facilitating effects of daily contact between partners who make a decision to cut down on smoking. Journal of Personality and Social Psychology, 1970, 17, 2-5. Johnson, R. C., Thomson, C. W., Frincke, G. Word values, word frequency, and visual duration thresholds. Psychological Review, 1960, 67, 332-342. Katz, I. Factors influencing Negro performance in the desegregated school. In M. Deutsch, I. Katz, & A. R. Jensen (Eds.), Social class, race, and psychological development. New York: Holt, 1968. Katz, I., & Greenbaum, C. Effects of anxiety, threat, and racial environment on task performance of Negro college students. Journal of Abnormul and Social Psychology, 1963, 66, 562-567. Katz, I., Henchy, T., & Allen, H. Effects of race of tester, approval-disapproval, and need on Negro children’s learning. Journal of Personality and Sochl Psychology, 1968, 8, 38-42. Kelman, H. C. A time to speak: On human values and social research. San Francisco: Jossey-Bass, 1968. Kelman, H. C., & Eagly, A. H. Attitude toward the communication content, and attitude change. Journal of Personality and Social Psychology, 1965, 1, 63-78. Kennedy, W.A., & Vega, M. Negro children’s performance on a discrimination task as a function of examiner race and verbal incentive. Journal of Personality and Social Psychology, 1965, 2, 83-43, Kiesler, C. A., & Goldberg, G. N. Multidimensional approach to the experimental study of interpersonal attraction: Effect of a blunder on the attractiveness of a competent other. Psychological Reports, 1968, 22, 693-705. King, D. L. A review and interpretation of some aspects of the infant-mother relationship in mammals and birds. Psychological Bulletin, 1966, 65, 143-155. Kissel, S. Stress-reducing properties of social stimuli. J o u w l of Personality and Social Psychology, 1965, 2, 378-384. Heck, R. The effects of interpersonal affect on errors made when reconstructing a stimulus display. Psychonomic Science, 1967, 9, 449450. Krasner, L., Knowles, J. B., & Ullmann, L. P. Effect of verbal conditioning of attitudes on subsequent motor performance. Journal of Personality and Social Psychology, 1965, 1, 407-412. Krauss, R. M. Structural and attitudinal factors in interpersonal bargaining. Journal of Experimental Social Psychology, 1966,2, 4255. Krebs, D. L. Altruism-An examination of the concept and a review of the literature, Psychological Bulletin, 1970,73, 25-02.

THE POWER OF LIKING

145

Krech, D., & Crutchfield, R. S. Theory and problems of social psychology. New York: McGraw-Hill, 1948. Lambed, W. W., Solomon, R. L., & Watson, P. D. Reinforcement and extinction as factors in size estimation. Journal of Experimental Psychology, 1949, 39, 637-641. Landy, D., & Aronson, E. The influence of the character of the criminal and his victim on the decisions of simulated jurors. Journal of Experimental Social Psychology, 1969, 5, 141-152. Lerner, M. J. Evaluation of performance as a function of performer’s reward and attractiveness. Journal of Personality and Social Psychology, 1965, 1, 355-360. Levinger, G., & Moreland, J. Approach-avoidance as a function of imagined shock threat and self-other similarity. Journal of Personality and Social Psychology, 1969, 12, 245-251. Liddell, H. S. Animal origins of anxiety. In M. L. Reymert (Ed.), Feelings and emotion. New York: McGraw-Hill, 1950. Little, K. B. Personal space. Journal of Experimental Social Psychology, 1965, 1, 237-247. Lott, A. J., Aponte, J. F., Lott, B. E., & McGinley, W. H. The effect of delayed reward on the development of positive attitudes toward persons. Journal of Experimental Social Psychology, 1969, 5, 101-113( a). Lott, A. J., Bright, M. A., Weinstein, P., & Lott, B. E. Liking for persons as a function of incentive and drive during acquisition. Journal of Personality and Social Psychology, 1970, 14, 66-77. ( a ) Lott, A. J., & Lott, B. E. Group cohesiveness, communication level, and conformity. Journal of Abnormal and Social Psychology, 1961, 62, 408412. Lott, A. J., & Lott, 8. E. Group cohesiveness as interpersonal attraction: A review of relationships with antecedent and consequent variables. Psychologicul Bulletin, 1965, 64, 259-309. Lott, A. J., & Lott, B. E. Group cohesiveness and individual learning. Journal of Educational Psychology, 1966, 57, 61-73. Lott, A. J., & Lott, B. E. A learning theory approach to interpersonal attitudes. In A. G. Greenwald, T. C. Brock, & T. M. Ostrom (Eds.), Psychological foundations of attitudes. New York: Academic Press, 1968. Lott, A. J., & Lott, B. E. Liked and disliked persons as reinforcing stimuli. Journal of Personality and Social Psychology, 1969, 11, 129-137. Lott, A. J., & Lott, B. E. Some indirect measures of interpersonal attraction among children. Journal of Educational PsychoZogy, 1970, 61, 124-135. Lott, A. J., Lott, B. E., & Matthews, G. Interpersonal attraction among children as a function of vicarious reward. Journal of Educational Psychology, 1969, 60, 274-282. ( b ) Lott, A. J., Lott, B. E., Reed, T., & Crow, T. Personality-trait descriptions of differentially liked persons. Journal of Personality and Social Psychology, 1970, 16, 284-290. ( b ) Lott, A. J., Lott, B. E., & Walsh, M. L. The learning of paired-associates relevant to differentially liked persons. Journal of Personality and Social Psychology, 1970, 16, 274-283. ( c ) Lott, B. E. Group cohesiveness: A learning phenomenon. Journal of Social Psychology, 1961, 55, 275-286. Lott, B. E., & Lott, A. J. The formation of positive attitudes toward group members. Journal of Abnormal and Social Psychology, 1960, 61, 297-300.

146

ALBERT J. L O T AND BERNICE E. LO'IT

Lubow, R. E., & Marcuse, F. L. The effect of the presence of mother on abnormal behavior in the rat. Journal of Comparative and Physiological Psychology, 1956, 49, 462464. Maccoby, E. E., & Wilson, W. C. Identification and observational learning from films. Journal of Abnormul and Social Psychology, 1957, 55, 76-87. Mason, W. A. Socially mediated reduction in emotional responses of young Rhesus monkeys. Journal of Abnormal and Social Psychology, 1960, 60, 100-104. Matthews, G. Interpersonal attraction and stage of learning as determinants of social facilitation, Unpublished doctoral dissertation, University of Kentucky, 1968. McCoy, N., & Zigler, E. Social reinforcer effectiveness as a function of the relationship between child and adult. Journal of Personality and Social Psychology, 1965, 1, 604-612. McKeachie, W. J., Yi-Guang Lin, Milholland, J., & Isaacson, R. Student affiliation motives, teacher warmth, and academic achievement. Journal of Personality and Social Psychology, 1966, 4, 457461. McPartland, J. The relative influence of school and of classroom desegregation on the academic achievement of ninth grade Negro students. Journal of Social Issues, 1969, 15, 93-102. Mehrabian, A. Relationship of attitude to seated posture, orientation, and distance. Journal of Personality and Social Psychology, 1968, 10, 26-30. Miller, N. E. Experimental studies of conflict. In J. McV. Hunt ( E d . ) , Personality and the behauior disorders. New York: Ronald Press, 1944. Miller, N. E. Some reflections on the law of effect produce a new alternative to drive reduction. In M. R. Jones (Ed.), Nebraska symposium on motivation. Lincoln: University of Nebraska Press, 1963. Miller, N. E., & Dollard, J. Social learning and imitation. New Haven: Yale University Press, 1941. Mischel, W., & Grusec, J. Determinants of the rehearsal and transmission of neutral and aversive behaviors. Journal of Personality and Social Psychology, 1966, 3, 197205. Moore, M. Aggression themes in a binocular rivalry situation. Journal of Personality and Social Psychology, 1966, 3, 685-688. Moran, G . Dyadic attraction and orientation consensus. Journal of Personality and Social Psychology, 1966, 4, 94-99. Mowrer, 0. H. Learning theory and behauior. New York: Wiley, 1960. Murray, L. B. The influence of interpersonal attraction on binocular resolution. Unpublished master's thesis, University of Kentucky, 1968. Mussen, P. H., Conger, J. J., & Kagan, J. Child deoelopment and personality. (3rd ed.) New York: Harper, 1969. Mussen, P. H., & Parker, A. L. Mother nurturance and girl's incidental imitative learning. Journal of Personality and Social Psychology, 1965, 2, 94-97. Natsoulas, T. Converging operations for perceptual defense. Psychological Bulletin, 65, 393-401. Newcomb, T. M. An approach to the study of communicative acts. Psychological Review, 1953, 60, 393404. Newcomb, T. M. The prediction of interpersonal attraction. American Psychologist, 1956, 11, 575-586. Nunnally, J. C., Duchnowski, A. J., & Parker, R. K. Association of neutral objects with rewards: Effect on verbal evaluation, reward expectancy, and selective attention. Journal of Personality and Social Psychology, 1965, 1, 270-274.

THE POWER OF LIKING

147

Park, J. N., & Smith, A. J. Recognition thresholds for value-related words: Differences between inner-directed and other-directed subjects. J o u m l of Personality and Social Psychology, 1966, 3, 248-252. Patterson, G. R., & Anderson, D. Peers as social reinforcers. Child Deuelopment, 1964, 35, 951-960. Pettigrew, T. F., Allport, G. W., & Barnett, E. 0. Binocular resolution and perception of race in South Africa. British Journal of Psychology, 1958, 49, 265-278. Rittle, R. H., & Cottrell, N. B. Cognitive bias in the perception of interpersonal relations. Psychonomic Science, 1967, 9, 551-552. Rosekrans, M. A. Imitation in children as a function of perceived similarity to a social model and vicarious reinforcement. Journal of Personality and Social Psychology, 1967, 7, 307315. Sampson, E. E., & Insko, C. A. Cognitive consistency and performance in the autokinetic situation. Journal of Abnormal and Social Psychology, 1964, 68, 192-194. Sapolsky, A. Effect of interpersonal relationships upon verbal conditioning. Journal of Abnormal and Social Psychology, 1960, 60, 241-246. Sawyer, J. The altruism scale: A measure of cooperative, individualistic, and competitive interpersonal orientation. American Journal of Sociology, 1966, 71, 407416. Schoeninger, D. W., & Wood, W. D. Comparison of married and ad hoc mixed-sex dyads negotiating the division of a reward. Journal of Experimental Social Psychology, 1969, 5, 483499. Sechrest, L., & Wallace, J. Figure drawing and naturally occurring events: Elimination of the expansive euphoria hypothesis. Journal of Educational Psychology, 1964, 55, 42-44. Smith, E. W. L., & Dixon, T. R. Verbal conditioning as a function of race of the experimenter and prejudice of the subject. Journal of Experimental Social PsychoZogy, 1968, 4, 285-301. Solley, C. M., & Haigh, G. A. A note to Santa Claus. Topeka Research Papers, The Menninger Foundation, 1957, 18, 4-5. Spence, K. W. The role of secondary reinforcement in delayed reward learning. Psychological Review, 1947, 54, 1-8. Spence, K. W. Behavior theory, and conditioning. New Haven: Yale University Press, 1956. Staats, A. W. Social behaviorisni and human motivation: Principles of the attitudereinforcer-discriminative system. In A. G . Greenwald, T. C. Brock, & T. M. Ostrom ( Eds. ), Psychological foundations of attitudes. New York: Academic Press, 1968. Staub, E., & Sherk, L. Need for approval, sharing behavior, and reciprocity in sharing. Child Deoelopment, 1970, 41, 243-252. Stevenson, H. W., Keen, R., & Knights, R. M. Parents and strangers as reinforcing agents for children’s performance. Journal of Abnormal and Social Psychology, 1963, 67, 183-186. Stotland, E., Zander, A., & Natsoulas, T. Generalization of interpersonal similarity. Journal of Abnormal and Social Psychology, 1961, 62, 250-256. Tajfel, H. Social and cultural factors in perception. In G. Lindzey & E. Aronson (Eds.), Handbook of social psychology. Vol. 3. (2nd ed.) Reading, Mass,: Addison-Wesley, 1969. Tannenbaum, P. H., & Gaer, E. P. Mood change as a function of stress of protagonist and degree of identification in a film-viewing situation. Journal of Personality and Social Psychology, 1965, 2, 612-616.

148

ALBERT J. LOlT AND BERNICE E. LOT"

Taylor, D., Altman, I., & Sorrentino, R. Interpersonal exchange as a function of rewards and costs and situational factors: Expectancy confirmation-disconfrmation. Journal of Experimental Soda1 Psychology, 1969, 5, 324-339. Tomtzky, L., & Geiwitz, P. J. The effects of threat and attraction on interpersonal bargaining. Psychonomic Science, 1968, 13, 125-126. Vernon, M. D. Perception, attention, and consciousness. In P. Bakan (Ed.), Attention. Princeton, N. J.: Van Nostrand, 1966. Wahler, R. G. Infant social attachments: A reinforcement theory interpretation and investigation. Child Deuelopment, 1967, 38, 1079-1088. Walker, E. L. Experimental psychology and social responsibility. American P&IoZOgist, 1969,24, 862-868. Ward, C. D. Own height, sex, and liking in the judgment of the heights of others. Journal of Personality, 1967, 35, 381-401. Wike, E. L. Secondary reinforcement: Selected experiments. New York: Harper, 1966. Willis, F. N., Jr. Initial speaking distance as a function of the speakers' relationship, Pqdwtwmic Science, 1966, 5, 221-222. Woodmansee, J. J. The pupil response as a measure of social attitudes. In G. F. Summers (Ed.), Attitude measurement. Chicago: Rand McNally, 1970. Wright, P. H. Attitude change under direct and indirect interpersonal influence. Human Relations, 1966, 19, 199-211. Y m w , M. R., Campbell, J. D., & Yarrow, L. J. Acquisition of new norms: A study of racial desegregation. Joumal of Social Issues, 1958, 14, 8-28. Zajonc, R. B. Social facilitation. Science, 1965, 149, 269-274.

SOCIAL INFLUENCE, CONFORMITY BIAS, AND THE STUDY OF ACTIVE MINORITIES

Serge Moscovici Claude Faucheux2 ECOLE PRATIQUE DES HAUTES ETUDES PARIS, FRANCE

.

I. The Conformity Bias and the Reasons to Challenge It . . . A. Conformity as a Form of Social Influence . . . . . . B. Control and Change of Social Behavior . . . . . . 11. The Source of Social Influence: Dependency or Style of Behavior . . A. Coercive, Normative Relationships and Power . . . . . B. Dependency v e r w Style of Behavior . . . . . . C. Consistency and Attribution Processes: Some Consequences . . 111. Conformity, Normalization, Innovation, and Forms of Social Conflict . A. The Influence of the Group in the Absence of an Alternative to Consensus . . . . . . . . . . . B. The Norm as a Result of Reciprocal Concessions . . . . . C. Changing Social Norms or Judgments . . . . . . . IV. Six Experiments in Which Consistency Is Considered Directly as a Source of Influence . . . . . . . . . A. Preliminary Remarks . . . . . . . . . B. Approach to and Avoidance of the Answer as a Consistent Minority . . . . in a Situation Where There Is No Group Norm . C. The Influence of the Minority Response on an Implicit Norm . . D. Change of Response and Change of Code in a Perceptual Task . . V. Summary and Conclusion . . . . , . . . . References . . . . . . . . . . . . .

.

.

.

.

150 150 150 156 156 157 160 165 166 170 174 183 183 183 189 191 198 199

'Fellow, 1988-1989, at the Center for Advanced Study in the Behavioral Sciences, Stanford, California. I wish also to acknowledge the assistance extended to me by the James Marshall Fund, and I am very grateful to L. Berkowitz, M. Deutsch, and C. Nemeth for their helpful remarks and discussions. ' Visiting Professor at the Graduate School of Business, Stanford, California ( 1968-1969). 149

150

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

I. The Conformity Bias and the Reasons to Challenge It A. CONFORMITY AS A FORM OF SOCIALINFLUENCE The study of social influence constitutes one of the most important branches of social psychology. To date, research on this phenomenon has focused on such themes as the structure of a stimulus, the uncertainty of the individual, patterns of compliance, etc, and has followed mainly the traditions established by Durkheim, Cooley, and the behaviorists. These themes and traditions have largely shaped the research on social influence. Once embodied in specific paradigms, they have tended to select from among a large number of common sense propositions and theoretical intuitions those which are most plausible and best suited to empirical verification. However, by emphasizing too heavily certain aspects of the phenomenon at the expense of others, these very themes and traditions have acted as a brake to further advances. For example, there is a tendency to assume that any type of influence leads to conformity, and that conformity is the sole result of social interaction achieved by means of influence. This has as a consequence on the one hand that the individual is perceived only insofar as he tries to answer such questions as: ‘3hould I follow the group or the minority?” “How can I avoid disagreeing with the group?” “What rewards are the group or the leader likely to give me in exchange for my submission?” At the same time, most researchers seek only to describe ( a ) the nature of the individual or social factors which make a deviant give in to group pressure, ( b ) the role played by pressures toward uniformity in the maintenance of the general equilibrium of the individual or the group, and ( c ) the manner in which social norms or stimuli determine the movement of the members of the group toward the majority or toward the source of authority. A few dissenting voices have been raised against this implicit or explicit emphasis on conformity. Some years ago, Kelley and Shapiro (1954) emphasized that conformity may constitute an obstacle to a group’s adaptation to a changing reality, and that nonconformists can be popular individuals who have approval by their peers rather than by marginals or deviants. They deplored the neglect of nonconformity in current research which they felt disregarded its importance in real life. Asch, who was more influenced by ethical ideals and the wish to defend the individual against group suppression of his free choice, stressed the possibility open to each of us to resist social pressures by relying on the evidence of our senses. In line with these views, he asserted that “. . . We ought to treat with reserve the widespread assumption that there is a single form of social influence which is a prototype for all

SOCIAL INFLUENCE, BIAS, AND ACTIVE MINORITIES

151

others . . . In particular, it is not justifiable to assume in advance that a theory of social influence should be a theory of submission to social pressure [Asch, 1956, p. 21.” These and other observations have hardly encouraged the socialpsychologists to attempt to draw more subtle distinctions in this area and to determine under what conditions the individual who finds himself controlled by a group tries to influence the group itself. The most that has been granted the individual is his urge or his capacity to stick to his position, to be less compliant, to resist. It is seldom even considered that the individual might ask himself: “What should I do in order to make the majority adopt my point of view? How can I change the conceptions of the members of the group? How can we accept proposals that are so out of line from what we believe to be true or correct?” While individuals have been shown to be incapable of living together in peace without consensus (Newcomb, 1961), no effort has been made to prove that the individual can indefinitely enjoy this kind of peace. Conformity can generate boredom, stereotyping, and rigidity. The majority norm is sometimes experienced as violence by the individual who does not conform to it; it can be a source of permanent frustration to the individual who subscribes to it too mechanically. Society, the group, our peers, tend to reward less and less the repetition of conforming acts and in the end consider as an obligation what originally was seen as a valuable act, a gift, or an index of good will. The very evolution of conforming interactions forces the individual who wishes to reestablish his initial level of satisfaction to abandon the usual norms and to seek new relations and contacts. Conformity, far from being only an element of solidarity and psychological equilibrium, may become, in the long run, a source of instability and conflict. In this light, change and variety turn out to be an imperative of individual and collective survival (Coser, 1967). In some situations, it is the normal occurrence. A more thorough examination of the manifestations of social influence would define both the legitimate role of conformity and that of the processes by which the social body protects itself against its harmful effects. Such an examination cannot take place as long as the phenomenon is viewed only from a narrow frame of reference biased in favor of the “conformity or deviance” dichotomy.

B. CONTROL AND CHANGE OF SOCIALBEHAVIOR As we have just seen, the study of social influence has so far been nearly tantamount to a study of conformity. The reason for this can no doubt be found in the basic assumptions about social interactions. In the first place, the relationship of the individual to the group is viewed as a

152

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

unilateral one, the majority or the authority source always being considered as norm senders and the minority as norm receivers (Rommetveit, 1954). The former are identified with the group and the latter with the would-be deviants. By imposing this condition, one inevitably confines the individual to a closed social system in which he has no alternatives of choice. Asch (1962) wrote “There is no alternative to the language of one’s group, to the kinship relations it practices, to the diet that nourishes it, the arts it supports.” Within such limitations, the person or the minority subgroup seems to have no choice but to accept or to reject what is imposed by the group. From the viewpoint of the group, passivity or the seeking of social approval takes on a positive coloration, and dissenting activity a negative one. This selection from among the wider gamut of possible behavior, which leads to a definition of the minority (individual or subgroup) as one who has no system of answers or norms of his own, has had, in turn, deep consequences for the analysis of conformity and for the interpretation of experimental resuIts. We will show this definition to be upheld by the Asch experiments. These experiments involve placing seven to nine subjects in front of a screen on which are shown, successively, 12 sets of cards. The card on the left shows a standard vertical line, and the card on the right shows three vertical lines of unequal lengths, numbered from one to three, one of which is equal to the standard line, The subjects have to indicate orally which of the three lines is equal to the standard line. In the presence of the nai‘ve subject, seven to nine confederates of the experimenter declare that certain lines are equal to the standard line, although they obviously are not. We see here two forces contending with each other, the one representing the “norm” of the group, which is in the majority in the laboratory, and the other the “physical” or “sensory” evidence of the isolated minority individual. The latter seems to feel a conflict between what he sees and his need not to be considered aberrant by the others. The divergence between the consensus of the confederates and the “perception” of the naive subject is frequently resolved to the advantage of the group: the naive subject, in many instances, will yield. From this fact one draws an obvious conclusion: the majority can influence the minority eoen if it expresses opinions contrary to objective physical evidence. Would we reach the same conclusion if we considered that this minority individual has another answer besides the data provided by the senses on “physical reality”? Let us reexamine the phenomena that take place in the laboratory within a wider social context. The subject entering this laboratory is certainly an isolated individual; he is also the carrier of a “general norm,” the unconscious representative of a more

SOCIAL INFLUENCE, BIAS, AND ACTIVE MINORITIES

153

extensive collectivity, within which a certain consensus has been established as to what constitutes the equality of two straight lines. The perceptual judgment here reinforces an established norm; it is part of the cultural “evidence” shared by millions of people and one which has been strengthened by the action of parents and educators over the years. In the experimental room, the so-called nai’ve subject meets a group of individuals who are consistently proposing a different rule, a different way to judge length, a rule which inverts the rule that is usually applied. Do they in his eyes constitute a representative sample of the collectivity to which he belongs or of the majority of the people, as Asch (1962) suggests? We have no evidence for this conclusion. If this were really the case, if this “majority” really symbolized humanity as a whole, how could one explain that the naive subjects do not yield loof%? One is forced to assume that in terms of the original group relations of the subject, the group of confederates in the laboratory constitutes a minority. In facing this group, he is not, as one might think, in the position of an ordinary person faced with pressures from the collective system, forced to choose between submission and independance, between adopting other people’s judgments or relying on the evidence of his senses. It would be more accurate to compare his situation with that of a captive audience-prisoners of war, people submitted to brainwashing, political propagandizing, religious conversion, or professional initiation-people who are suddently cut off from their group of origin and systematically confronted with entirely new visions and norms (Schein, 1960). Deprived of all contact with the social context where norms are regularly reinforced, naive subjects, like all socially isolated minorities, are overwhelmed by uncertainty and submit. All the more so since they are confronted with a number of individuals who firmly abide by their own point of view. Thus, conflict takes place between the response current within the casual social context and the response adopted within the particular group with which one is confronted in the laboratory and not only between sensory data and the attraction toward the majority. If these remarks are plausible, and if the effect is due on one hand to the isolation of the naive subject and on the other hand to the unanimity of the confederates confronting that subject, then the Asch experiments show that a consistent minority can modify, under certain circumstances, the predominant norm. This conclusion, which we will verify experimentally, contradicts what was said earlier. It brings to light the ambiguity involved in the interpretation of a phenomenon which permits the confusion of anoniic, in-group deviance with outgroup deviance which is defined by the presence of a different norm.

154

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

Asch always thought that the subjects giving a “wrong” answer were conformists, while those giving the “correct” answer were independent. However, it is obvious that the latter simply obey a different norm, which happens to be the same as the experimenter’s. Cohen (1963) is aware of this ambiguity and of these repercussions. He writes From one point of view, correct responses can also be regarded as conformity; the individual, when he responds correctly, conforms not only to an internalized standard but also to an external norm. It can be argued that the experimenter acts as an external enforcing agent for the correct response. Indeed, some subjects regard the experimenter as a sanctioning rather than a neutral individual, as is illustrated by the comment: “I didn’t go along with them because I was afraid I’d ruin your experiment.

. . .”

Having noted this possible interpretation of correct responses, we shall continue to speak of “correct” and “conforming” or “incorrect” responses. But it is obvious now that this categorization of answers is not compelling, and that the meaning attributed to them does not permit a decision regarding who is really conformist or nonconformist in the situation described above. This example clearly shows how the interpretation of conformity is linked with a concept according to which the individual is subjected to the group in a position of passive dependence. It takes into account only minorities deprived of a point of view of their own, always eager for acceptance and belonging, but fails to consider those active minorities who are sure of their opinions and are determined to propagate them. Political, artistic, and scientific reforms or revolutions would be inconceivable without such minorities, without such deviants who are not content with mere dissent or resistence but who wish to attain their goals and to propose alternatives to the existing systems. In the second place, social psychologists have limited themselves to a relatively narrow region of the whole social milieu. The social interactions and the groups which they normally study are influenced by mass society (anonymous and temporary) or the family, industrial, or school environments. In these groups, consensus and the need for approval would be preconditions for a conflict-free existence. No one would dream of denying the importance of this aspect of social life, the purposes derived from it, or the motivations which underlie it. Nevertheless, the limited nature of such an approach is striking, as Kelley and Shapiro have pointed out: “. . . this type of research has been highly selective in the kind of situation studied. It has dealt primarily with settings where, in some sense, conformity is ‘good’ for the group [Kelley & Shapiro, 1954, p. 6681.’’ To appreciate the meaning of this statement, one need only look at scientific or artistic circles or to glance at social

SOCIAL INFLUENCE, BIAS, AND ACTIVE MINORITIES

155

movements or important political events. There originality and the clash of ideas are highly valued. In these spheres, all influence attempts have as their object the introduction of something new and the establishment of differences between individuals. The maintenance of the status quo is frowned upon. The principal motive is not affiliation with others or sharing their opinions or codes at any price. It is the modification of norms, of knowledge, and of social relations which constitutes the accepted dynamic force. Assuredly, such instances of transformation are not as common as cases of stability and uniformity, yet they are no less decisive for our society. As a result, one can hardly grasp the fundamental dimensions of behavior or groups without taking some cognizance of these facts. In the third place, the prevailing view of the function and goal of social influence must be mentioned. Unquestionably, what is most significant in understanding the process of social influence is to understand how it permits and reinforces social control, since “only through one sort of social control or another can individuals accomplish a concerted action or become a group [Hare, 1965, p. 231.” As a result, interest is focused on how groups can preserve their cohesion and how society, through the means of persuasion at its disposal, can keep intact the chains of command between its different parts and assure the convergence of these parts toward legitimate norms or intentions. Individuals figure in this process only to the extent that they are driven by the desire to be like others and to have their approval. The channeling of attitudes and behaviors toward these goals defined by the majority, thus becomes the crux of any influence-related action. It is evident, however, that while this is generally an accepted approach, it is not the only possible one, nor is it adequate for understanding the nature of relations between individuals and groups. The analysis of the process of influence which finds its expression in social change is just as legitimate a concern. Daily, political parties, organizations, and individuals attempt to modify the conditions in which we live, think, and act. In all these cases, the renewal of norms and social institutions leads to a recasting of interindividual relations and, at the same time, to the appearance of new modalities of action, communication, or even new leaders. We need not belabor this point to demonstrate that social change is just as much a function of influence as is social control. We also know that the majority of the theories developed by social psychology, to the extent that they deal above all with the pressure of the majority on a passive minority, are inadequate when applied to interactions where the minority plays a dynamic role, or to a reality in which positive value is assigned to originality, the renewal

156

SERGE MOS<XIVICI AND CLAUDE FAUCHEUX

of norms, and to actions promoting change. One must therefore refer to other notions and conceptual models. If we want to grasp the process of social influence in its totality, our attention must be directed toward another class of phenomena.

II. The Source of Social Influence: Dependency or Style of Behavior A. COERCIVE, NORMATIVE RELATIONSHIPSAND POWER Because of the bias toward conformity, there is some confusion in this field between power and influence. In fact, in this case the majority is simultaneously the source of normative relations, which presuppose the sharing of a code, and aIso the source of coercive relations, which deal in threats as well as material or psychological rewards. We intend to distinguish between those two types of relations as well as the social pressures they represent. The power pressure describes the phenomenon of opinion convergence by A's dominance over B by the means at A's disposal to compel B to do his bidding without any reference to the inherent properties of this opinion. What makes B change his stand is not a genuine modification of his preferences after what A says, but the gratifications A can grant him and the punishments A is authorized to inflict on him. Hence, it seems that B, the target person or subgroup, does not have entire responsibility for his changes of opinion, and that his conformity is a function of the sheer magnitude of power difference. The influence pressure springs from the need felt by the individuals to seek approval from the group or the leader of the group, from the need felt by the leader to seek a consensus from the individuals or the subgroups, and from the need felt by both to think about each other in terms of reducing the variability of their responses and the instability of their social and material relations, Group or individual judgments and opinions fulfill this role and are acceptable to the extent that they are organized and predictable. What matters, thus, is the existence of a norm and its coherence and the requirements it poses. It exerts an attraction because each of the partners must recognize it and recognize each other if they wish to reach a stabilization of their relationship and of their evaluations of a physical or social object. The way in which they behave toward each other is a decisive factor in the reciprocal influence they exert. If one accepts the distinction we make between power and influence, terms often used synonymously, as well as the distinction between their underlying processes, it becomes necessary to challenge what generally has been considered, up to now, as a source of influence.

SOCIAL

INFLUENCE,

BIAS, AND ACTIVE MINORITIES

157

B. DEPENDENCY VERSUS STYLEOF BEHAVIOR In almost all research undertaken up to now, only one source of influence has been recognized both theoretically and experimentally: dependency. The impact of expertise, in particular, has been studied in terms of the acceptance of a judgment or an opinion by the members of a group. As soon as a socially sanctioned differential competence comes to the fore, one notes conformity shifting to individuals reputed to have a greater knowledge or greater experience in a given field. The presence of a majority, a difference in status, as much as the need for approval and affiliation, create the right to direct the behavior of individuals or subgroups, determine the extent of this right, and the possibility to control each one’s opinions and judgments. It follows that the lower an individual’s status, the more he feels in a minority position and dependent upon the evaluation of others, the more he is threatened with exclusion or mistreatment by the majority or those in position of power, and thus the more likely he is to conform (Berkowitz, 1957). Of course, since dependence, whatever its guise, is a concrete expression of the gap between majority and minority, of the role hierarchy and/or the distribution of skills, and of the possibilities for reward or penalty within the group, it easily lends itself to experimental and theoretical investigation of conformity. On casual inspection, dependency seems to relate particularly to the power dimension of social relations. It does not appear to be the key concept, however, when all social influences, including innovation, are viewed. One might even say that dependency, with respect to an innovating individual or subgroup, is a consequence rather than a cause of the influence process. For instance, the need to follow the advice of experts in connection with television or modern mathematics is contingent upon the adoption of television or modern mathematics. Then too, a truly innovating minority, one which sets in motion a social or scientific revolution, does not have initially any advantages from the point of view of power and resources, and thus of status and competence. When, at the end of the nineteenth century, a small group actively promoted experimental methods in psychology, nobody immediately attributed to its members any special expertise as psychologists. There is no need to mention the many scholars who never received recognition from universities or from the great artists who were regarded by their contemporaries as devoid of merit. Thus, dependency is not, from the point of view under consideration, either an independent variable or a differentiating factor which might account for the influence in question. We believe that we can find the source of influence in the style of

158

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

behavior of the one who proposes a solution to a problem or a norm for a group, and who, by modifying the psychological field, can have some impact on the receiver of the solution or norm. There remains even today a certain incongruity between the study of influence and the study of communications as a form of social interaction, although the two phenomena are closely interrelated. While the former is concerned only with the variables related to power and resources (in short, with instrumental variables), the latter takes into account the impact of the self-presentation of the sender, the sequence of arguments, and other oratorical devices which are symbolic variables or behaviors. There are good reasons to believe that symbolic variables intervene in the influence process. The organization of behavior, its “rhetoric,” might by itself explain the acceptance or rejection of a proposed judgment or model, Its meaning would play a decisive role for the partners in a social exchange. It particular, behavioral consistency which is felt to be a sign of certainty, of commitment to a coherent choice, must be a powerful source of influence which cannot be ascribed either to differences in capacities or to an explicit dependency. Now, it is true that behavioral consistency can range from persevering repetition of a given response, through the avoidance of any contradictory behavior, to the elaboration of a logical system of demonstration, For our present purpose this is not the main issue; intuition and experiments will have to elucidate the matter progressively. Since we are trying to isolate the power process from the influence process by opposing dependency, which is rather a source of the former process, to style of behavior, which is a source of the latter, we must be able to prove that: ( a ) variations in amount of dependency do not lead to a corresponding variation in yielding; ( b ) the degree of yielding is a function of the consistent or inconsistent character of social pressure; and especially ( c ) the number of people involved in the interaction is not a decisive factor. In other words, using data referring to conformity, we will try to show that this conformity is an effect of the consistency of the majority and not of the dependency toward this majority. To substantiate these proposals we will limit ourselves to Asch‘s experiments or those derived from them, since they have served as a paradigm for the study of conformity. We know that in these experiments a majority of confederates can induce a single naive subject to give answers opposed to the evidence. The essential conditions in all these experiments are the introduction of an unambiguous stimulus, the need to respond publicly, and the presence of a unanimous majority. This majority, according to Asch, institutes a propensity in the minority to adopt its “erroneous”

SOCIAL INFLUENCE, BIAS, AND ACTIVE MINORITIES

159

responses. He writes: ‘We may conclude that the experimental effect described earlier was a function of the presence of a majority [Asch, 1962, p. 4771.” Of course, our own conclusion will be different, but first, let us look at the data and their meaning. We can consider that unanimity in a group corresponds to interindividual consistency, to consistency which results from the coincidence of response of several subjects to a given stimulus. At the same time, identical, consistently “erroneous responses” from each subject through a series of stimuli express internal, intraindividual consistency. What do we see when we examine Aschs experiments? We see that a unanimous majority of 3-16 subjects leads to the acceptance of its “erroneous” responses by about a third. The increase of the number of confederates beyond three has no repercussion on the frequency of these responses. There is thus no direct relation between the magnitude of social pressure and the degree of conformity. But let us assume that this is a matter of threshold, and let us imagine two sets of groups exerting the same amount of pressure, one in which the members of the group are unanimous, the other in which they are not. It has been observed that approximately 32%of responses are conformist when subjects are confronted with the unanimous group, but only 10.4%or even 5.5%are conformist when, in the second group, a single confederate breaks the unanimity by giving a “correct” answer or one that differs from that of the others. One can thus say that the consistency of social pressure is a more effective factor in conformity than the amount of dependency. Indeed, Graham ( 1962) also found unanimous agreement among all the members of the majority to be the largest contributing single factor to the conformity of the subjects. This conclusion, however, is not warranted unless we know Erst why the minority subject gives fewer erroneous answers in the presence of the second group. In terms of power, this decrease should result from a coalition, from the fact that someone gives him social support and diminishes his dependence. In terms of influence, it should arise either from the lack of consensus within the group reducing the subject’s need to acknowledge the majority view, or else from the fact that the majority opinion does not constitute a valid alternative to the norm concerning the equality of line length which does have unanimous approbation outside the group. It is therefore extremely important to be able to decide whether the observed variation is due to: ( a ) the presence of a partner agreeing with the subject, who reduces his dependence upon the group by forming a coalition with him; or ( b ) the lack of group unanimity which weakens the norm itself and the value of the sought-after consensus.

160

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

Asch (1962) and Allen and Levine (1968) suggested that if social support were the important factor, the dissenter would have to give the answer that the naive subject privately believed to be correct in order to reduce the conformist pressure. Conversely, if group consistency were the crucial variable, then a dissenter’s disagreement with the group would be sufficient to decrease conformity, whether or not his answer would be correct or would agree with the subject’s private judgment. The conclusion of the two experiments was that it was the lack of unanimous consensus which was the decisive factor. This means that in the absence of the group’s unanimous acceptance of a rule, group judgment is unreliable, and other responses are conceivable; consequently, either subject’s own answer is also acceptable, or he must look elsewhere for a partner with whom to agree. Therefore, it seems that social pressure elicited by the style of behavior is undoubtedly the true source of conformity, especially when the individual is isolated,

C. CONSISTENCY AND ATTRIBUTION PROCESSES: SOMECONSEQUENCES We will now try to analyze the impact of consistency on the way one can contemplate social influence. We may assume that each individual or group seeks to cope with his material and social environment, to organize this environment, and to confirm his knowledge of it. For these purposes both the individual and the group come to terms with reality, make predictions about it, and control its future course by separating ephemeral, variable behaviors or events from their invariable, permanent foundations; they introduce a causal, temporal sequence into a situation where everything might seem to be fortuitous or arbitrary. When we meet a person, we can only communicate with him and behave in a successful way toward him if we manage to sift out from the innumerable impressions a few predominating features, such as intentions, feelings, ability, etc. Objects are submitted to the same sifting operations of comparison, classification, and selection of transmitted stimuli; these operations lead to a characterization of the dimensions of the objects in terms of size, color, speed, or applicability for some useful purpose. Whether we are dealing with persons or with objects, we set in motion a process of inference or of perception of dispositional properties. Heider ( 1958) defined the conditions for this differentiation of the physical and social spheres : he also described the way in which a person endows the two areas with stable features when faced with their varying components: ( a ) one discriminates the presence of an effect when its cause is presumed to be present and its absence when the cause is absent; ( b ) the reaction to this effect, pro-

SOCIAL INFLUENCE, BIAS, AND ACXIVE MINORITIES

161

duced by a person or an object, remains the same, irrespective of the form it may take; ( c ) the answer given agrees with the answer of other persons in the same circumstances and with reference to the same effect. To put it more succinctly, distinctiveness, consistency over time, consistency over modality, and consensus are the four criteria which permit one to separate phenotypical properties from genotypical properties, as well as to validate the information acquired in respect to them (Kelley, 1967). Furthermore, one can see that consistency has a decisive role in the process of discovery and organization of information as provided by our environment. It may play this role either as internal, intraindividual consistency ( consistency over time and modality, in Heider’s terminology ) or as external, interindividual, social consistency (consensus). There is no question that these are just two ways of looking at the same thing: consistency over time and modality is merely a sort of consensus each person establishes within himself among his various acts of cognition, while consensus is a form of consistency imposed upon thd judgments and opinions of different individuals. Either kind of consistency results in a reduction in the variability of responses. This reduction is the ordinary, visible indicator that a pattern of acting has been acquired through which desired properties are revealed and by which the norms which determine our behavior toward our environment are validated. Or, at least, this is the feeling generated by the reduction of variability and the meaning it has for us. Assuming that this analysis of the attribution process and the relevance of the consistency factor in this context are correct, one must note two very important consequences in order to understand influence phenomena. The first consequence is that consensus, preferably unanimous, is a necessary ingredient of the sequence of operations involved in sorting out and organizing the stable properties of the material and social environment. According to Kelley, in the pursuit of the satisfaction of the consensus criterion, persons in interaction with one another will often exercise iduence in an attempt to produce unanimity of attribution. Attribution theory carries the important implication that mere compliance by a person initially expressing deviant opinions may not be satisfactory to his colleagues. Therefore, if a consensus is serving attributional purposes, genuine agreement will be sought [Kelley, 1967, p. 2061.

This is hardly surprising if one thinks of scientific methods, where this analogy is obvious. Scientific methods imply that one can only know whether a measurement and the information it provides are correct if there is a theory for the measuring instrument. This theory implies the acceptance by the whole scientific community of a common paradigm

162

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

or protocol which establishes the conditions for making observations, the position of the observers, the acceptable degree of inaccuracy, etc. As a result, a judgment can only be validated when the judges themselves are validated or when the judges try to validate each other. Reproducibility of an experiment is crucial because it assures the convergence of procedures and results. Social consistency ( or consensus) in relation to the individual who infers and perceives and individual consistency in respect to identical acts of inference and perception have complementary functions. We stress this point because it has been asserted that social and individual consistency are completely separate and separable phenomena. Regarding the circumstances which permit an individual to seek the support or agreement of others, it has been said that the individual would do so only when he himself is unable to establish a stable relation with his environment (Festinger, 1950). It was assumed that a judgment or estimation could be verified without relying on another person, since the nonpersonal evidences were adequate. Thus we can determine by ourselves, without asking anyone else, whether two lengths are equal, whether substances are crumbly, etc. Should such nonpersonal evidences be lacking, we would be uncertain about the accuracy of our information. The uncertainty would naturally impel1 us to rely on the standard of the group if we were anxious to form an opinion or to confirm or reject a judgment. Pressure toward uniformity and conformity results exclusively from the absence of technical, nonpersonal means for verifying an assertion. On closer inspection, one sees that the arguments underlying these assumptions are based on the idea that consensus exists only where it is explicit. Of course, we often make physical estimates without consulting anyone else, either because we feel that there is no need to do so or because we have some sort of instrument. But the reason we do not rely on anyone else is that we assume that any other person in his right mind would see or feel the same way that we do. This is often true for ordinary events in life, when we try to determine whether it is day or night, whether a line is long or short, etc. If we have a socalled objective means of verification available, we are even less likely to worry about the judgment of others. When we measure a length with a yardstick or test the quality of an object with a tool, we accept the result we have obtained, considering it valid without anyone else’s approval. Our certainty is not due to any particular property of our senses or of reality. It is due to the fact that the yardstick or the tool represents in some way a crystallized consensus, a “canned” consensus, so that if we individually reach consistent conclusions, social consistency is automatically assured. The independence of the judge and of the

SOCIAL INFLUENCE, BIAS, AND ACI‘IVE MINORITIES

163

observer is thus a consequence of consensus rather than an absence of one, In all these circumstances involving physical, objective reality or judgments, social consistency is not nonexistent. On the contrary, it is taken for granted on the strength of past experience, the uselessness of making a special effort to reconfirm what we have already confirmed. It is reasonable that we should not seek out testimony when we have instruments or nonpersonal evidences, as it is not practical to establish a new consensus time and again. What is important to ascertain, instead, is the nature of the already implied consensus; one should determine the reciprocal orientation of the observers as they seek or adhere t o a universal or particular validation. The requisite degree of consistency depends on these two options. A universal validation (objective, true) of a judgment or opinion demands a high degree of consistency, and a specific validation (conventional, likely agreement) calls for a lower degree. The organization of the segment of reality and the influence one has are closely related to the kind of validation one seeks and to the objectives of the consensus. At least this should be so if attributional analysis is carried through to its logical conclusion and if it refers to observable phenomena. The second consequence of the analysis of attribution, which is intimately related to the first, consists in the impossibility of considering other persons as identical with physical stimuli or as mere substitutes for physical stimuli. An assumption generally accepted in the study of social influence is the following: the use or reliance on persons is a form of reliance on physical objects, and social relations are instruments to perfect, where necessary, one’s relations with one’s environment. If, as we have seen, the individual becomes immersed in a reality which he is incapable of understanding on his own, he will turn to other individuals; they will become points of reference, information carriers permitting him to make up for a gap in his own knowledge and to estimate the probable degree of accuracy in his own answer. Interaction of individuals and groups is thus entirely conditioned by the environment, and the partners in this interaction view each other as parts or substitutes of parts of the environment. In short, social relations are seen as one facet of relations between objects. One might wonder why it was believed necessary to single out consensus (social consistency in our terms) in the process of stabilization of the properties of the external world, if personal or social agents are nothing more than agents and represent nothing specific. In fact, it is because they do indeed contribute something more, something specific, that it was thought desirable to single out this criterion and to draw it to our attention when we try to understand how the material and

164

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

social environment is conceptually ordered. In such attempts, we must, for instance, take into account individual capacities, efforts, feelings, or intentions, and this is unnecessary when we deal with nonpersonal, nonsocial entities. It is true that we may endow such entities with malevolent or benevolent intentions ( Faucheux & Moscovici, 19fl8), but that only happens when we personalize objects, not when we objectify persons. If relations between people are envisaged as an independent form of relations and not simply as one mode of relations between objects, then the divergences in opinion that these relations attempt to overcome or emphasize imply a form of either exchange or conflict. Insofar as the relations lead to consensus, they constitute further proof, if proof is needed, that the influence and the negotiation processes are closely rehted and that the properties of the former are derived from the same conditions that determine the latter. It is a striking feature of social psychology that no one has shown the connection between these two processes nor examined its theoretical implications. This is not surprising inasmuch as consensus has been considered a secondary phenomenon, and the individual’s link with other persons has been assimilated into his link with a physical entity. And yet, everyday life makes us aware of the fact that a norm, an institution, is a kind of contract, and influence contributes to the elaboration of norms. In this respect, influence also contributes to what has been described as a passage from the state of nature or violence to the state of society; the role of influence is not merely to reabsorb the uncertainties we experience regarding the accuracy of our evaluations of reality. In other words, influence tends to substitute a rule or a relation where there was incompatibility and difference. It does not serve only as a social prosthesis for our psychophysical handicaps. What is more, the operation of judging is never purely a treatment of information, an internal organization of stimuli from the outside. It is also a dialogue or controversy. As Churchman (1961) puts it, essentially a judgment is a group opinion. The “group” may consist of the same individual at different points in his reflective life, but for all practical purposes, we can talk as though the group in question had several different members. We shall want to argue that judgment is a group belief which occurs when there are differences of opinion among the group members because in the spirit of the earlier discussion, we want to say that sound judgment occurs when the judgment is subject to strong opposition. Thus, the essence of judgment is the establishment of “agreement” in the context of disagreements. Judgment is a type of negotiation [p. 2931.

These remarks apply even to individual estimations of stimuli that we customarily use in the laboratory: “Averages of a set of readings are judgments and have the same structure of social agreement that other

SOCIAL INFLUENCE, BIAS, AND ACTTVE MINORITZES

165

judgments have. Despite the precision of formalism embodied in statistical methods, basically these methods are ways of reaching a satisfactory negotiated ‘peace’ among disputants [Churchman, 1961, p. 2961 .” Thus, intra- and interindividual consistency are both products of the reduction of conflict achieved by eliminating, through hierarchical ordering, different and possibly incompatible responses. To summarize, consistency serves the function of criterion both for the judgment-validation mechanism and for the mechanism of differentiating inferred and perceived properties of our environment. As such, it is an essential component of both mechanisms. Consistency manifests itself in its intraindividual form when it persists over time and modality in the presentation of an entity, it manifests itself in its interindividual form when there is agreement among several people regarding the judgment of a given entity. Intraindividual consistency strives for stability, while interindividual consistency strives for unanimity. The fact, that they coexist and act in harmony, each retaining a specific nature, has definite implications for the influence process. Thus, the influence process reflects the dual (objective or subjective) character of the norms determining relationships between observers more strongly than it reflects the existence of one reality generating or failing to generate pressure toward uniformity. In this connection, attributions concerning quasi-physical entities and their interactions are distinguished from attributions concerning persons, social agents, and their interactions. When differences of opinion or judgment focus on quasi-physical entities, they result in uncertainty or ambiguity; when they deal with persons, they result in exchanges or conflicts. In the latter case, influence must be seen as a negotiation process between persons; consensus results either in a contract that permits regular transactions, or in a hierarchy which makes certain choices preferable to others. We can imagine that in most experiments on conformity a tacit negotiation is taking place, during which each of the participants either tries to make his own point of view prevail or attempts to find out what the consequences of a concession on his part would be. If this is so, then each form of influence corresponds to a type of negotiation, a way of dealing with social conflict. 111. Conformity, Normalization, Innovation, and Forms of Social Conflict

The influence process, as is now obvious, extends beyond the limits usually applied to it. A theory of this process cannot be exclusively a theory of conformity; it must also be a theory of the two other

166

SERGE hlOSCOVICI AND CLAUDE FAUCHEUX

modalities of social influence, normalization and innovation. The need to study several modalities of influence has been accepted for some time. In his logical analysis of research on pressure to uniformity, Simon and Guetzkow (1955) showed that normalization and conformity must first be distinguished clearly and then integrated into a more general implicit model. In order to make this model truly general, one must allow for innovation mechanisms for a minority which is not only the target but also the source of social pressures. Above all, it should be recognized that each of the modalities of social influence represents: ( a ) a behavioral type in the course of the evolution of codicts; and (21) a mode for establishing norms and social codes. The existence of three modalities can be attributed to the particular forms assumed by social conflicts and to the norms channeling or resulting from these conflicts. A. THEINFLUENCE OF THE GROUP IN ALTERNATIVETO CONSENSUS

THE

ABSENCEOF AN

The behavior of an individual or subgroup is defined as conformist when this behavior is determined by legitimate rules and expectations of the group, and when the individual or the subgroup subscribes to judgments and opinions of the real or ideal group, irrespective of initial differences, The function of conformity is the resorption of deviance which constitutes a permanent threat to the equilibrium and integrity of the social system. This function operates successfully when the normative pressure is exerted on an individual or subgroup who is “confined in the “culture” of the group or lacks (theoretically) other articulate alternatives. In the conformity phenomenology, it is usually useful to differentiate between behavioral or external compliance and belief-oriented or inner compliance. Despite its practical importance, however, this distinction does not deeply alter the basic features of social influence. Let us have a closer look at the conformist game. By the nature of things, one of the partners, the majority, has its very own code, its very own set of responses to the material and social environment. The attraction or cohesion of the group commits each one to respect previously learned rules or norms; unanimity plays a reassuring role in confirming the choices previously made and excluding the possibility of adhering to other rules or norms. The other partner is the minority subgroup or individual. If he moves away from the group and defies it, or if he is singled out by the group as being different, he has no possibility, in principle, of receiving support. He will not find another source of normative and cognitive support, a reward for his behavior, or reach a consensus in favor of his opinions or judgments. Festinger

SOCIAL

INFLUENCE,

BIAS, AND ACXIVE MINORITIES

167

(1950) rightly stressed that if an individual can confirm his opinions or judgments by using an instrument, he will not need to consult the opinion of other individuals. However, when such “physical” or other alternatives to the norms of the group either do not exist, or when a minority has not succeeded in securing them, then it suffers more heavily under the weight of internal uncertainty and the divergence between it and the majority. It has been shown repeatedly that differences in regard to the evaluation of an objective stimulus create a tension which finds its expression in measurable psychophysiological changes ( Steiner, 1966). A divergence between the majority and the minority may become too pronounced when, in terms of a code previously accepted by all, “normal” individuals have become separated from “deviant” individuals. -4t such time, the majority will be forced to channel or to eliminate the conflict if it wishes to survive. The minority will be forced to do the same thing, for its uncertainty concerning the majority’s norm creates difficult psychological situations. Neither side is capable of avoiding the conflict because the group must restore stability in the relationships between its members and the outside world, and the individual or subgroup lacking internal consistency is unable to get its bearings and survive socially. As one can see, the majority has a very narrow margin of freedom in moving away from its position and has no reason for making many concessions, while the minority can do nothing but concede. The more firm the consensus of the majority, the stronger its determination will be to exclude any compromise or dialogue and seek to impose its own perspective. The minority is well aware that it can be forced into isolation or that it can suffer internal conflict which may eventually break down its resistance, leading it to join the majority. It has submitted when it abandons its stand and either wholly assumes the stand of the group or undergoes an inward splitting of its social personality. Of course, this description best fits the portrait of a “closed group” facing a “life and d e a t h issue; actually, there are less dramatic situations. This does not invalidate what has just been said, however. First of all, closed groups and life and death issues are not lacking in the real world today. Second, the theoretical description of a phenomenon is meant to portray an “ideal” type in the Max Weber sense of the word. It might otherwise become a mere reflection of the infinite variety of things. Moreover, Aschs experiment which has served as a prototype for the greater part of the research done on conformity does succeed in creating such an ideal type. The data which have been obtained by applying his experiment favor our “negotiation-of-a-conflict” hypothesis and emphasize the importance of the behavioral style. The basic paradigm goes something like this (Allen, 1965): ( a )

168

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

there is a ”correct” answer involved in the task (but it might be merely model or average); ( b ) there is a significant divergence between the group answer and that of the isolated individual; ( c ) the communication of judgment is the only permissible interaction between the “orthodox,” “correct” majority and the potentially deviant individual; and ( d ) the social constraint does not appear to be intentional. This constraint demonstrates the circumstances under which the tacit negotiation for resolution of the main conflict is engaged. To start with, the majority and the minority represent two mutually exclusive outlooks, and the divergence is disturbing because it destroys, for many minority individuals at least, a preestablished equilibrium with the outside world. They are compelled to look for and reach a consensus which is different from the one to which they have been accustomed. They will reckon with the consensus according to the nature of group pressure. When group pressure is consistent and the majority leaves no hope that concession on its part can be expected, then the individual must choose between submission and opposition; when, however, the group’s consistency is imperfect and the individual can expect a certain reciprocity, then he may surrender his position, but only partially. It is not so much that we theoretically expect a significant decrease in the numbers of “errors,” or more independence, but we do expect a change in the categories of errors, each of which has a specific psychological meaning. In Asch‘s experiments, one knows that: ( a ) there are two types of trials, the “critical” trials where the confederates give the “erroneous” answer, and the “neutral” trials where the confederates give the “correct” answer; and ( b ) the naive subject must compare three lines with a standard line and choose between one correct and two erroneous responses. The two errors are basically different, however. It is an “extreme” error if the line that the group judges to be equal to the standard line is either the largest or the smallest one; it is “moderate” if the line chosen is located between the line equal to the standard line and the one most unequal to it. In one experiment, Asch (1956) varied the proportion of “neutral” trials in relation to the “critical” trials in these proportions: 1/6,1/1, 4/ 1. Although the frequency of “erroneous” responses by the subjects does not lessen, one can observe that, as the majority becomes less consistent, the influence does decrease and the frequency of erroneous “compromise” responses does increase (see Table I). Another experiment is also informative in this respect. The subjects are asked to compare two rather than three lines with the standard line. Since the possibility of any compromise, i.e., “moderate” error, is eliminated, the minority subjects are forced either to accept what is proposed to them

SOCIAL

INFLUENCE, BIAS,

AND ACTIVE MINORITIES

169

TABLE I INFLUENCE OF INCREASING INCONSISTENCY IN AMOUNTAND TYPE OF CONFORMIST RESPONSES Proportion of neutral to critical trials

Proportion of compromise errom on extenuat,ion

Conformist responses

5.1 19.1 20.5 42.7

53.0 36.8 38.6 26.2

%

or to go into opposition and take the change of a schism. From the results it appears that the total number of “errors” is relatively large; confronted with the choice of capitulation or withdrawal, many subjects submit (Asch, 1956; Gerard, 1965). There is, however, a change in the distribution of their responses: the subjects cluster around the extremities, showing less variability in their responses. “This mode of operation,” writes Asch (1956), “has the effect, for reasons which are not clear, of increasing the proportion of independent and totally submissive reactions [p. 641.” For us, on the contrary, the reasons are clear. In the standard experiments, the individual thinks that he can come closer to the estimations of the other individuals by making a minimal concession, with the expectation that the consensus will be established somewhere between the two alternatives. When subjects make “compromise” errors, they are as untrue to the evidence of their senses as when they make extreme errors. They believe however, that they will avoid facing a strong conflict, at least to a degree. This proves that it is the evolution of the minority-majority relations which determines the nature of the responses and not the degree of dependence or the desire for independence as such. In the experiment just described, when the consistency of the social pressure is maximized, the possibility of avoiding the main conflict is nonexistent, and the subjects are logically forced to choose compliance with the group on every trial or to choose defiance. The difference between a form of negotiation where there are reciprocal offers and one where ultimata are merely exchanged is clear: the reduction of negotiable alternatives leads to extremism; conversely, the provision of loopholes implies that movement toward the majority norm is transformed into a movement toward compromise. Diversification of responses, by creating a certain degree of inconsistency and giving the impression that the majority is not too committed to its judgment, weakens the appeal of the majority. While the majority has less strength to draw on in its negotiations, the minority

170

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

feels less obliged to accept its “offers,” retaining confidence in its own capacity to influence the norm of the group. Tuddenham (1958) repeated Asch‘s experiments but increased the number of possible responses forming an ordered sequence, so that the subjects were to choose not between 3 but between 10 lines. The results show that, this time, the “compromise” responses were more frequent than the responses reflecting absolute “dependence” or “independence.” Thus the “deviant” subjects were less “deviant” because they tended more often to respond similarly to the group; and the conformist subjects were less conformist, choosing the least “unreasonable” error. Cohen ( 1963), having systematically analyzed influence processes where errors were “moderate” and where errors were “extreme,” attempted to integrate them into the same mathematical model. He failed because the model that suited the “moderate” conditions did not suit the “extreme” condition, and vice versa. Beyond that, he questioned whether it was justified to make no distinction between these two forms of “erroneous” responses: When the group made a moderate error, no subject ever made an extreme error. On the other hand, when the group made an extreme error, the subject often made a moderate error. In this case, the subject’s behavior represented a compromise between complete yielding to the group and complete independence. When this form of compromise does occur, the categoiy “incorrect,” which combines compromise and direct yielding, does distort the data somewhat [Cohen, 1963, p. 311.

It does more than distort the data; it distorts reality. For, as becomes clear from our analysis, two distinct phenomena are at work. On the one hand, consistent social pressure expresses a commitment, a clear-cut attraction, and a warning from the group to the minority individual to adopt the proposed norm, to give up his own judgment as truly erroneous, and to do so without quibbling. In these circumstances, each individual is aware of and must confront the conflict. On the other hand, as soon as the differences in the estimations are unclearly expressed, there is less conflict, tension, and less need to take a stand. The individual does not become more “independent”; he merely feels less uncertain, less threatened by isolation, and he changes in a less striking manner, The influence exerted in this case is in the direction of compromise, a third alternative. At the same time, the modality of conformity is transformed into that of normalization.

B. THENORMAS A RESULTOF RECIPROCAL CONCESSIONS What happens when the interaction of individuals or subgroups results in a compromise? Viewed from the previously suggested perspective, normalization assumes, on the one hand, that an individual who observes the fluctuations of values given a stimulus tends to stabilize it

SOCIAL INFLUENCE, BIAS, AND ACTIVE MINORITIES

171

around some modal values or features. The classic study of Sherif (1936), using the visual effect of the autokinetic phenomenon, has demonstrated this tendency. The person looking at a stationary point, the source of light in a darkened room, which seems to be moving, validates his perceptions by observing the uniformity of his reaction over time or over modality for each appearance of the stimulus. Since we are dealing here with an optical illusion-which is used as a model for most empirical research on normalization-responses differ considerably from one person to another. One person gives answers varying between 10 and 20 cm, another between 30 and 50 cm, etc. Regardless of these variations, however, individuals who are confronted with an ambiguous stimulus do succeed in making a selection from the information they are given, “classifying” it in the relatively invariant category which is their individual norm. On the other hand, once this individual norm-the symptom of an acquired internal consistency-is established, there is an expectation that the consensus will be easy. Since an individual normally considers himself to be a typical specimen of humanity, he supposes that everyone centers around the same value that he has chosen. If this is so, his individual norm would turn into a social norm, reinforced by coincidence. However, when confronting another person, he faces a divergence. If these divergences can be justified because one is told, for instance, that the stimulus induces subjective reactions (see Sperling, cited in Asch, 1962), then the person can adhere to his own scale and refrain from seeking any external reinforcemenk3 If the divergence cannot be attributed to the subjective nature of the reactions, however, it arouses anxiety and surprise, and the partners are impelled to eliminate the source of anxiety, to reassure themselves about their judgment capacity, and to preserve somewhat the already attained consistency ( Kelman, 1950). Normalization defines the pressure each exei-ts on the other during an interaction, with the aim of reaching either a judgment acceptable to all individuals or one which approaches complete acceptability, By nccessity the completely acceptable judgment does not coincide with any of the individual assessments. Why is this true? First of all, since equality of capacities and competence prevails among the interacting The above theoretical remarks are supported by the results of two experiments by Alexander, Zucker, and Brody ( 1970). ThLse experiments demonstrated that a convergence of judgments occurs only when there is an expectation of orderliness and stability generated by the norms and behavioral styles of the confederate. When these expectations are lacking, the Sherif findings do not apply. That is, without expectations of orderliness or stability, there is no influence and no normalization of responses.

172

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

partners, no one can pretend to impose his opinion legitimately upon the other (French, 1956). Second, the individual is not very committed to his response; neither does he have a way of imposing his point of view on others. Under these conditions, no one has valid reason to feel justified in adhering rigidly to his own judgment or in taking the chance of being different simply to obtain an advantage over his partners; he would not wish to grant others this possibility either. Hence, it is useful for him to avoid extreme responses and to take an interest only in responses approximating his own (de Montmollin, 1966). If this were not so, chances for an agreement would decrease, and chances for a perceptual state of tension would increase. Since intrapersonal consistency is dependent upon interpersonal consistency, the perpetuation of divergence is undesirable and possibly dangerous; the existence of a plurality of individual norms may destroy each one of them by maintaining uncertainty and anxiety. In such a case, a tacit negotiation must take place and the answers must be coordinated so as to avoid conflict, even though no one will be a winner. This calls for each individual to make concessions which first create the convergence of opinions or judgment and only afterward remove or level off differences. Thus, in the autokinetic phenomena, the estimations of the individuals, which were widely scattered at the beginning, converge toward a common value after a few trials. These successive mutual concessions ensure that everyone will have the validity of his own judgment partially confirmed, and that no one will impose a totally different point of view on anyone else. Consensus is bought at the expense of blurring the “extreme” edges of judgment; by adopting the least controversial value, one even avoids having to face a choice. This evolution is all the more marked (Sherif & Harvey, 1952) when the individuals display a greater range of perceptual judgments, when the distance between the judgments is smaller, and when external cues are more limited. Allport (1924) has observed during a weight-guessing experiment that, in the presence of other persons, an individual tends to underestimate the heaviest weights and overestimate the lightest ones. Similarly, under comparable conditions, a subject judges the most fragrant aroma as less pleasant and the most noxious stench as more pleasant they really are. Kogan and Wallach (1966) set out more recently to investigate whether similar effects would be obtained by comparing individual judgments with those obtained by genuinely interacting groups. Their experiment consisted of asking a group of subjects “to make a series of judgments individually and then constituted as a group, with instructions to achieve consensus in a group discussion for each of their prior judgments [Kogan & Wallach, 1966, p. 1681.’’ Their findings

SOCIAL INFLUENCE, BIAS, AND ACTIVE MINORITIES

173

showed that the group judgment represented the average of the prior individual judgments. Summarizing these findings, Kelley and Thibaut (1954) note that “while reacting with other persons, the person reacts to them . , . by tempering his judgments so as to avoid the possibility of being extremely different from others [p. 7491.” Thus, the interference of a group results in the elimination of those answers might provoke a polarization of positions; it simultaneously leads to the giving of answers which underestimate the properties of a stimulus. The answers can thus serve as a rallying point for all and can reduce the risk for each of seeing his judgment contradicted by others. In this respect, the norm is a ”supraindividual” (Sherif, 1936) social product, where each person is right and no one is wrong. This compromise, engendered by the desire to avoid conflict and the necessity to establish interindividual consistency for the group, reduces intraindividual variability and conflict, as is demonstrated by the fact that individuals adopt social norms for long periods of time. This, however, does not necessarily have to be so. Sometimes the discrepancy between the compromise norm and the previous norms of the individuals is not really reduced; the negotiation which takes place aims only to prevent an eventual clash of opinions, it does not settle the question as to which is the true or good response. As Riecken (1952) writes: “The consensus that is reached in many cases is nothing but an agreement not to disagree [p. 2521.” To summarize, the elaboration of a norm corresponds to the functional necessity of relating to the environment and to the other invariant by setting up a frame of reference for ordering stimuli and response. It has been shown with the autokinetic phenomenon and, later, with many other phenomena, that individuals in a group tend to reduce progressively the diversity of their judgments when confronted with an ambiguous or relatively unimportant social stimulus ( Moscovici & Zavalloni, 1969). Because, within a group, each individual uses the other group members as anchorage points and does not rely solely upon his personal assessment, persons in a group are said to influence each other. It has been stated that such a positive convergence of several norms permits the establishment of the most rational and valid possible frame of reference (Allport, 1924; Sherif, 1936). Our own view of the normalization process is somewhat different. We claim that convergence is motivated not by the inner qualities of the physical properties of the environment, such as ambiguity, but by the need for consensus. In this modality of influence, the essential dynamic is less the search for a common and assured frame of reference than the tendency to avoid conflict. We see in the influence process not only a positive movement toward cooperation, but also a flight from choosing between probably incom-

174

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

patible alternatives; this is accomplished through suppression of differences and a leveling off at the lowest common denominator.

C. CHANGING SOCIALNORMSOR JUDGMENTS Theoretical or experimental studies of innovation and the influence of minorities are extremely rare. This is probably due to preoccupation with social control and to the typically negative conception of individuality and deviance. In fact, individuals and subgroups can establish and maintain themselves in a positive manner through affirmation of their convictions, preferences, and capacities. The only truly independent response in one of Asch‘s experiments was to leave the room. Sometimes, when faced with intense conformist pressures in a group incapable of tolerating diversity, an individual can feel he has no alternative but to hold fast to a personal vision of things and events. William Blake highlighted this dilemma clearly: ‘‘I must create my own system or be enslaved by another man.” If someone builds up such a system, or, on a more modest scale, firmly states a logically coherent and consistently presented opinion, he can have an impact on the outlooks of other people. In such a case, innovation refers to the proposal and acceptance of a model which either makes the prevailing social rules and judgments obsolete or, minimally, influences the production of these social norms in the direction of the new model. One must not equate invention and innovation: invention is closely related to the process of problem solving, while innovation is linked to the influence process. From the standpoint of social consequences the fate of an innovation is as important as its conception . , . a study of cultural change takes us beyond the appearance of a new idea into a consideration of its acceptance and rejection [Barnett, 1963, p. 2911.

The few studies that have tried to come to grips with innovation phenomena have been based on the idea that innovators and innovating behavior belong to the category of deviance. When the problem is expressed in such terms, one may wonder what conditions will allow a minority to become “deviant,” or nonconformist, without being eliminated from the group or being subjected to counterinfluence attempts. Ziller and Behringer ( 1960) showed, for instance, that a minority individual who seems to have capacities superior to those of most of the group members is seen as a facilitating agent for changes when the group itself experiences failure. At such time, an individual who is a newcomer to the group may be able to impose his solutions, acquiring a leadership position. In a frequently cited experi-

SOCIAL INFLUENCE, BIAS, AND ACTIVE MINORITIES

175

ment, Kelley and Shapiro (1954) set out to analyze the repercussions of deviance in circumstances where the pressures toward uniformity are harmful to the group, hindering its efforts to adapt to reality. They assumed that, when the group’s performance is becoming less effective because the group members are hesitant to deviate from the norm, the most popular individuals will be the ones who take the risk of by-passing the norm and instigating change. Kelley’s and Shapiro’s experiment does not show unequivocally that a group’s adaptation to reality requires either nonconformity or innovation; neither does it prove that only high status individuals are capable of bringing about this adaptation because they can more easily divorce themselves from the norm. Nevertheless, the underlying hypothesis does seem plausible and in agreement with other observations ( Harvey & Consalvi, 1960). Are the individuals who are likely to exercise social control on others also the ones who are best able to free themselves from such control? This would mean that the person who increases his influence already has a certain amount to begin with. Hollander (1958) started this view in a systematic fashion. He proposed that each individual in a group has a certain “idiosyncrasy credit” which represents an accumulation of other members’ positive dispositions toward him. The greater this credit, the greater his partner’s confidence in him, and consequently, the more he is in a position to act without regard for the opinion of the majority. In two experiments, Hollander proved that the individual who has acquired a great deal of idiosyncrasy credit, through his competence or his earlier adhesion to group goals, can afford to behave in a nonconformist way. Both the hypothesis and the results of these experiments seem to conflict with the generally held opinion backed up by other experiments that leaders must respect the established consensus of the group more than other members. Hollander (1960) attempted to reconcile the two series of facts by showing empirically that a temporal sequence is involved here: An individual must first be a conformist, reach a high status, and establish a dominant position; thereafter, he can successfully introduce changes and deviate from the norm in proportion to the competence accredited to him or to the dependency he has imposed on others. Hollander’s investigations presuppose that (1) the minority can influence only when it has power or resources, and ( 2 ) the initiative for innovating is propagated from the top toward the bottom of the social psychological ladder. The paradoxical consequences of such a conception, when applied to history, are clear. Although social psychologists view each innovation as the result of a compromise, and the greatest innovators are described as being, in addition, the cleverest conformists, history proves the opposite: Stringent intransigence often typifies the attitude of individuals

176

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

who have had a great impact on our ideas and behavior. We have telling examples in Copernicus for astronomy, Galileo for mechanics, Freud for psychology, Marx for economics and sociology, Robespierre for politics, and Luther for religion, Typically, for social psychology, innovation or nonconformity is made to appear as an outgrowth of leadership, conducive to the preservation of the leader’s dominance; innovations are not seen demanding the replacement of old leaders with new ones. Yet, if Lenin had followed this scheme described by Hollander, he would first have had to become Czar of Russia before he could begin to transform it into a socialist country. As everyone knows, he followed exactly the opposite path: he first built up a party, and carried out a revolution, and then became the leader of the Soviet Union. If it is true that an individual is in fact obliged to use his credit to deviate, and that any innovation brings with it a social “loss” rather than a “gain,” two questions arise: ( 1 ) How did the individual acquire this credit? and ( 2 ) what is his motivation for using it up? Unless one assumes that the individual was more conformist than the other members of the group-which would be contradictory and certainly has not been proved-r that he did not need to be influential in order to become a leader, one must attribute the credit to a prior innovation. Of course, there are leaders who are impelled forward by their “followers,” but this is not the typical situation. If the theory of idiosyncrasy credit (taken here as a prototype of similar theories) merely explains the reinforcement of credit and not its origin, then it does not solve the problem. At this point, it should be clear that the prevailing conception of innovation takes off from the idea of social control. The innovating agent, the nonconformist or “deviant” member of the group, is consistently seen as one who enjoys a high social status; he is the one on whom others depend. There is no point in denying that this is a genuine aspect of innovation-one can observe it in highly centralized or rigid societiesbut it is only a partial view of reality and does not explain the total phenomenon. One cannot even claim that it is characteristic of the most interesting situations that may arise. On the other hand, when one views innovation in the context of social change and originality, things are altogether different. Under those conditions, there is no need to consider innovation as a special case of deviance; it is either a normal behavior which fits expected or well-known patterns of interaction ( Coser, 1967), or it is the expression of a state of conflict which fulfills the needs of a subgroup to modify the habits or the organization of the group (Menzel, 1960). The influence of such a minority is consciously directed toward the modification of instituted social uniformities. For this reason, we need to analyze both the actions of the minority and the conditions under

SOCIAL INFLUENCE, BIAS, AND ACTIVE MINORITIES

177

which it can find acceptance for its opinions and judgments; this will constitute a preliminary step to gaining theoretical and empirical insight into the innovation phenomenon now viewed as an autonomous process. Let us look at the real social interaction. The presence of a norm is revealed by the spontaneous conformity of those who share it. Each member of a group or culture recognizes it in his group and readily accepts a certain way of dressing or talking, etc. As long as there is a table of categories showing what is allowed and what is forbidden, the “ g o o d citizen, scientist, etc., can separate promptly what he must do and what he must not do, what is good and what is bad, and so on. When a single individual or a subgroup strays from this generally accepted classification, the confluence of individual responses toward the same norm is interrupted, and the constant conformist pressure shows its limitations. At the same time, this minority casts doubt on the value of the majority’s judgment by envisioning a different model to attain the same goals and explain the same phenomena. With the substitution of diversity for uniformity, the group’s uncertainty is increased in two ways. On the one hand, unanimity is threatened. Where only one “culture” prevailed, there are now two “cultures” facing each other, the new offering an alternative solution to the old. On the other hand, since new dimensions have been introduced or old dimensions have been combined, and since differences or overlappings that previously went unnoticed have been brought to light, intraindividual consistency is disturbed. Thus, at the time of the student revolt in Paris, many people saw their scale of values thrown into confusion when students identified themselves with workers. When two groups which previously had symbolized the two mutually exclusive classes of the bourgeoisie and the proletariat became embodied in a single group, tension was produced. Previous responses were now incompatible. The bewilderment and the reaction of the police and political circles was caused by this new combination of social dimensions. In general, the majority, upon seeing the number of alternatives increased, is compelled to view what was once an absolute, almost natural option, as being now merely an arbitrary convention. Before Copernicus’ work on the heliocentric system was published, the learned astronomers and theologians saw the world as Ptolemy had described it, with the earth at the center and the planets revolving around it. After the publication of the book, they tried to maintain that there were two equally likely hypotheses. The solidity of the world had dissolved, to be replaced by human “arbitrariness.” Using Lewinian terminology, one could say that in such a situation, norms had “thawed.” What is more, they had come into conflict with each other. For a genuine thaw to take place, the minority must show

178

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

coherence and decisiveness; it must exert a continuous pressure, thereby making its own vision equivalent to the prevailing vision. The group is then faced with a choice between two alternatives that are nearly equal in value or intellectual attraction. When the need to recognize what is objective is insisted upon, the conflictual relations acquire a special character, since it is more difficult to dismiss the minority. If the reality in question is submitted to the test of truth and objectivity, the result cannot be peculiar to a part of humanity; it must be, in principle, universal. Confronted with such a reality, anyone should have the same reaction. In other words, a group or an individual cannot have his own objective reality; his reality is objective only to the extent that he is sure of shaiing it with everyone else. Minority judgments thus acquire a great weight, for agreement is an indispensable, logical condition of the coherence of the whole. If a response must be objective, disagreement is more serious, and conflict between two contradictory evaluations cannot be resolved by compromise. The reason minorities frequently appeal to supraordinated entities, values or categories-truth, beauty, history, rationality-is probably at least in part due to their desire to impose their agreement as an indispensable element and to exacerbate that tension which leads to choice, Their aim, at least at the beginning, is not to establish uniformity, but to introduce and gain recognition for individuality, an original, different truth: to get recognition for difference. The sharpening of these divergences, of these individualities, is the conditio .sine qua non for moving from one order to a different order. In this sense, one might say that the innovator is a creator of conflicts, that the negotiation set in motion between majority and minority revolves around the creation of conflict where none had existed and the establishment of new relations with respect to the issues at stake. Here we have the specific quality of this innovative modality of influence: it revolves around the creation of conflicts, just as normulixation revolves around the avoidance, and conformity around the control or resolution of conflicts. In this context, consistent social pressure by the minority is one of the preconditions of success. First of all, through such pressure the minority defines the solutions in its own terms and channels the majority response in a certain direction. Second, because the minority is resolute, its behavior is generally more predictable and visible than that of other group members; establishing a consensus with the minority individual or subgroup appears more secure than establishing a consensus with any other member of the group. In addition, the consistency of the minority implies confidence which, in turn, may mean that the minority is right.

SOCIAL INFLUENCE, BIAS, AND ACTIVE MINORITIES

179

Simple as they may seem, these remarks deserve to be studied more thoroughly. We have some empirical evidence which tends to agree with them, In 1937, Gurnee had already noted the following: In a group, when there is a problem to be solved, there is occasionally one subject who reacts faster and proposes an hypothesis at the beginning; he then takes the lead in the whole sequence of decision processes. Shaw (1963) attempted to study how the quantity of information at the disposal of a group member affects his behavior and the group’s performance. His experiment confirms Gurnee’s (1937) clinical observations. In it, each group of three subjects discusses two cases and, depending upon conditions, two, four, or six solutions. The two other members are not informed about these solutions. It may be observed that the fewer solutions the informed member has: ( a ) the sooner he enters the discussion in relation to the uninformed members; ( b ) the more readily his solutions are accepted; ( c ) the greater his contribution to the solution of the problems is considered to be; and ( d ) the more likely he is to be chosen as leader. The implication is quite clear: the subject who presents fewer solutions, who is thus both systematic and a source of clear choices, has a greater influence on the group. To carry this argument just a step further, a single individual is also more consistent than, say, three individuals, unless he has reached a prior agreement with them, which permits him to focus the group’s attention more successfully on his proposals. In other words, a minority of one can be more influential than a minority of three. Of course, this inference can only be proved indirectly. Torrance (1959) attempted to observe the impact of an individual’s testimony on a group when the individual has made a prior test of an object and the rest of the group is seeing the object for the first time. The groups are made up of 6 1 2 individuals who are asked to eat a food product called “pemmican.” The number of witnesses vanes; some have had a favorable, others an unfavorable experience with the product. The influence of the witnesses on the novices is measured following their interaction. A side result of this study is of particular interest to us. Torrance observes that groups in which a single subject has had an unfavorable experience react more unfavorably than groups in which two or more subjects have shared this prior unfavorable experience. One can sce that the testimony of a single subject who is perforce more consistent with himself is more influential than the testimony of a theoretically less consistent subgroup. Torrance ( 1959) writes: Apparently, if an individual is the only mernber of the group who has experienced the attitude object, he alone is the star. Everyone looks to him for this judgment. He can speak without fear of contradiction. If there are others who have experienced the attitude-object, he is reduced at least to a co-starring role. Even though he

180

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

may be in agreement in his judgment with the other experienced member or members, he probably does not feel as free from fear of contradiction as he does when he alone has experienced the object. Of course, should the experienced people choose to form a coalition and consolidate their “stories,” their influence might be sharpened thereby [p. 2551.

Thus, experimental evidence supports the idea that the expression of a coherent point of view by a minority induces others to go along with it; the minority’s coherence influences the perception of an object or the inferences made with respect to it. The next hypothesis is that a minority which deviates conspicuously from the common point of view, simultaneously adhering systematically to its own positions, will create a persuasion pole by redirecting the field of forces and communications in the group. Research by Festinger and his colleagues has amply demonstrated that a deviant who asserts his position attracts most of the communications aimed at reestablishing social cohesion. Instead of remaining passive, however, the minority individual or subgroup can become active, utilize this same communication flow, and direct a counterpressure against the group, compelling it to choose between the “deviant” and the group response. Since the minority thus becomes the center of attention, its arguments and actions attract more notice than those of the conformist group members. Thus, whereas the group relies on this focal pressure to resorb deviance, resolve the conflict, and checkmate divergence, an active minority stresses this divergence in order to channel messages of the group toward it and to gain attention. Such a tactic is often used as a first step in the influence process. What is more, this tactic, while disruptive for the majority with its need for uniformity, may be momentarily reinforcing for the minority which needs reassurance that it is different. In any case, the opening up of communication channels which focus on the deviant minority increases the minority’s chances of changing the majority. This is clearly shown by Emerson’s (1954) replication of Schachter’s (1957) experiment. Using the general framework of Festinger’s theory on informal communication, Schachter predicted that, with an increase in the relevance of the task and of group cohesion, there would be a corresponding increase in efforts and, therefore, in communications directed toward the deviant individual, to bring him back to the right path of the group. If the deviant remained unresponsive to these pressures, then a tendency would arise for the group to reject the deviant and to lower the volume of communications addressed to him; this tendency would be all the greater, the more coherent the group and the more relevant the task. These predictions were verified on the whole. Emerson does observe the rejection of the deviant in this

SOCIAL INFLUENCE, BIAS, AND ACTIVE MINORITIES

181

experiment. He also notes however, contrary to Schachter’s predictions that the number of communications directed toward the deviant increases to the very end, and that the shift of opinion in his favor is noticeable. As the author states, “pressures towards uniformity took the direction of changing self . . . [Emerson, 1954, p. 6931.” The majority subjects seem to have no way to escape the conflict imposed by the minority than by yielding to the minority. It must be remembered that the bargaining power of the minority is, in these circumstances, very high, since it confronts a group which generally hesitates to proceed with a genuine process of separation or rejection. A minority of this sort, by resisting or questioning standards, gives an example of liberation from social constraints. In social life, most people usually think that there is only one way of acting or of doing something. Even if they are dissatisfied or unhappy with the established ways of life, they do not believe in the possibility of bringing about change; they do not believe there are other ways of living which can be created and adopted without any danger. At the same time they reside in pluralistic ignorance, not knowing that the other members of the community have similar desires and thoughts. When a minority has the courage to reject established practices or to propose what has been considered “forbidden,” “unlawful,” or “impossible,” then it demonstrates that other ways of acting or living are valid and that the consequences are not necessarily dangerous. Consequently, it undermines the authority of norms and institutions. With respect to this exemplary value of the minority, de Monchaux and Shimmin (1955) related what happened in England during a crosscultural study led by Schachter (1957). In this study, different groups were asked to choose an airplane model that they were supposed to build. The confederate was instructed to choose systematically the model most frequently rejected by the group; the group in turn was instructed to reach a unanimous decision which would necessarily be different from the choice of the deviant. Out of 32 groups, 12 reached an agreement against the choice of the confederate, 6 agreed on his choice, and 14 failed to reach any agreement. The English psychologists write: Our assumption has been that, as discussion progressed the need to reach a group decision would master individual needs. But it is possible that the behavior of the deviate in persisting quietly and unaggressively with his own choice may have reinforced the strength of individual goal preference. (If he won’t give up his choice, why would I?). We have therefore to consider the possibility that the presence of the deviate may have worked as a pressure towards individuality of choice rather than towards group unanimity [de Monchaux & Shimmin, 1955, p. 591.

182

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

Up until now, we have reasoned as though the only possible outcome of this situation is a move toward the model proposed by the individual or subgroup who innovates or seeks social change. Actually, in this conflict created by the need, on one side, to obtain agreement and, on the other, to retain individual integrity, there are three possible outcomes: ( a ) approach toward the minority; ( b ) polarization; and ( c ) avoidance of the minority expressed in terms of mistrust. The first possible outcome has been described. The second, polarization, results from the complete rejection of the proposed norm and leads to a reinforcement of existing standards. In general, the studies on the effect of divergence between the group response and the individual's response have stressed that if the differences are too large, there will be no influence. Actually, what people have observed is only that there is no approach toward the proposed judgment; they have failed to observe that there is also a polarization of responses. This polarization must be described as a result of influence; indeed, sometimes polarization itself is the goal, increasing the distance between minority and majority so that the price of later negotiations will have risen, compelling individuals to commit themselves and making impossible any compromise. The third possible outcome, avoidance, can only happen when the difference can be explained away as a function of a perceptual of intellectual bias. There is then no reason to seek consensus with the minority or to be troubled by the power of its arguments. Conflict of responses is transformed into conflict of attribution. For example, socialism is seen by its advocates as a political movement aiming to change society and to materialize the profound needs of the mass of the people, but it is presented by those who reject it as a doctrine used by foreign groups to seize power and to satisfy their appetites. Conservative critics assert that the style of modem painters merely allows them to look original and to get more money for their paintings; none of the modem works expresses a genuine creative urge. Such strategy by avoidance results from the need to raise a barrier between what is the accepted norm, upheld by the political, professional group, and what is new, upheld by an individual or a subgroup, so that no interchange can ensue. These three possible reactions of approach, polarization, and avoidance must always be kept in mind. They are especially conspicuous during the process of innovation. Thus, in summary, an individual or subgroup who has already gained the group's confidence is quite able to influence his followers when the environment changes and there is some need for adjustment. There is no doubt, however, that an unpopular minority can also impose itself as a demanding partner in an ongoing collective bargaining, using

SOCIAL INFLUENCE, BIAS, AND ACXIVE MINORITIES

183

consistency and resolution to its benefit. Such a minority can provide an example of differentiation in the social system, and, in the limiting case, can constitute a new center of attraction. These are the characteristics of innovation.

IV. Six Experiments in Which Consistency Is Considered Directly as a Source of InfIuence A. PRELIMINARY REMARKS We have shown that behavioral consistency is a source of influence and that its effects conform with certain general psychological mechanisms. We have relied exclusively upon indirect evidence, however, and we shall now complement it with more direct proof. We shall verify the following hypotheses (or conjectures) : ( 1) a “minority” subject expressing a consistent preference in an informal judgment situation will induce the “majority” subjects to adopt his answer; ( 2 ) a subject consistently expressing a preference in an informal judgment situation will interfere with the leveling of answers and will encourage the other subjects to retain their own preferences; ( 3 ) a minority subject consistently making a certain choice in an informal judgment situation will induce the majority subjects to adopt his answer, even at the cost of a modification of an implicit group norm; ( 4 ) a minority subgroup offering an alternative answer to a well-established physical norm is capable of bringing about at least a partial adoption of this alternative answer; and ( 5 ) if the consistently expressed preferences of a minority subject can be attributed to personal idiosyncrasies, then the majority will avoid adopting them.

B. APPROACH TO AND AVOIDANCEOF THE ANSWER OF A CONSISTENT MINORITYIN A SITUATION WHERETHEREIS No GROUPNORM 1. Description of the Experimental Paradigm To validate these hypotheses, we required an experimental procedure with the following characteristics: a. The task must involve a choice between several dimensions or values that are equally acceptable. If the minority answer can be judged true or false, one might have to attribute its influence to its competence rather than to its behavior style. Only when other options are removed can the effect be attributed solely to the subject’s consistency. The answer must not be related to a stimulus whose correctness can be tested (Asch effect) or which results from an illusion (Sherif effect). As well, where two answers are possible, neither can be considered

184

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

better. The minority, in short, must not command or represent a choice which would give it some advantage over the majority; otherwise, any influence it might have would be the effect of dependency. It is true that Crutchfield's (1955) investigations have postulated that no influence action can exist in the realm of preference judgments. We believe that this theory must be reexamined in light of the role which consistency plays, as a source of influence. b. The subjects should give a series of judgments on the stimuli presented to them. c. The minority individual must make a fixed type of judgment through a whole series. d. Differences between individuals should be noticeable, and they may refer both to the content and to the style of behavior. e. Social constraint must be unintentional, resulting only from the differences displayed during the emission of judgments. f. No communication unrelated to the task is to be permitted. This paradigm was used in two experiments.

2. First Experiment a. Experimental procedure. The experiment is presented to the subjects as an investigation of a problem in applied psychology. They are told that in connection with recent developments in information transmission, notably in aerial and space navigation, operators have to read information projected on cathode screens. For this reason it is important to determine people's preferences for different codes in order to find a way of increasing the legibility and discriminability of the transmitted information. A concrete example is offered: men in the control towers of airports must keep track of and simultaneously guide a large number of planes getting ready either to land or to take off. From the point of view of airport safety, the work of these operators must be simplified as much as possible; the relevant information (altitude, position, speed, priority, etc.) must be presented to them in the clearest possible way, with the most appropriate signals, so that they can make rapid and accurate decisions. The subjects are then told that they will take part in an experiment which is a simplified version of such a situation of perceptive selection and decision making. For this purpose, they will be shown a series of drawings varying in four respects: size (either large or small); color (either red or green) ; shape (either round or angular); and Zine (either dotted or continuous). The subjects are shown a sample of each of these drawings. They are then informed that they will be shown a long series of drawings, and that for each drawing there will always be four possible correct answers. However, they are asked only to give one

SOCIAL

INFLUENCE,

185

BIAS, AND ACTIVE MINORITIES

answer-the one which, whatever the reason, seems to them the most appropriate for a given drawing under the circumstances. Each subject must give his answer orally and write it down on a piece of paper with which he is supplied. The order of answers is vaned systematically: the subject who gives his answer first on the initial trial will speak last on the following trial, etc. The series of stimuli consists of 64 drawings which are shown in a sequence such that only onc dimension remains constant from one drawing to the next, the other three being switched around (example: large-green-rounded-dotted, then large-red-angular-continuous, then small-red-rounded-dotted, etc. ) . The subjects in groups of four or five fill three sides of a rectangular table, while the fourth side is occupied by the experimenter who holds up the drawings one after the other. In the experimental groups, a confederate of the experimenter consistently chooses the color answer from the first to the last trial. The control groups consist only of naive subjects. The subjects are American students in Paris, 19-22 years old, taking a special language program to improve their knowledge of French. li. Results. Table I1 shows that the number of “color” responses increased significantly in the experimental groups compared to the control groups. It is noticeable that this increase in the number of “color” responses accompanies a significant decrease in the number of “shape” responses. However, this latter variation cannot be attributed to any definite factor. The “shape” dimension is not the one most frequently chosen by the control groups and does not seem to be related in any special way to “color.” A second measure of this influence of the minority subject is the following: normally, when expressing a preferential choice, subjects do not respond in an isolated manner but rather in a series of two or more successive responses. It can be observed that in the experimental groups, the “color” judgments are more frequently given in series of TABLE I1 FIRSTEXPERIMENT COMPARISON O F T H E EXPERIMENTAL AND CONTROL GROUPS FOR AVERAGE NUMBER OF CHOICES I N EACHDIMKNSION Dimension

Color

Line

Size

Average of the experimental groups (A: = 8 )

20.86

16.18

16.09

10.88

Average of the control groups (k = 6)

15.28

18.93

14.20

15.59

t of student Significance level

p

2.46 .2


.lo

p

>

.75 .10

Shape

2.74

.2 > p

> .05

186

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

two or more responses ( x 2 = 17.84, p < .001). For the other dimensions, however, there is either no significant difference (size) or else the number of “isolated answers increases (shape: x2 = 5.45, .05 > p > $02; x2 = 22.397, p < .001). Thus, the consistent behavior of a minority determines not only the frequency of responses of the majority but also its organization.

3. Second Experiment a. Experimental procedure. The material, arrangement, instructions, and number of groups are the same as in the preceding experiment. However, instead of having the subjects register their own answers on a piece of paper, the answers are taken down by a research assistant on a paper board located in the middle of the room., Now each one can see clearly who gave what answer as well as the answer pattern for each individual and for the group as a whole. In this way, the relative situation of the subjects becomes visible to all by simple inspection, so that ( a ) the 100% consistent character of the confederate is highlighted, (27) the “coalition” movement of converging or diverging choices can be seen (one can tell “who answers the same as whom”), ( c ) everyone can look at a table of his past answers-what he says assumes a public and, in a way, a formal aspect. This manipulation seems to us to result, on the one hand, in committing the subjects to their answers and, on the other, in allowing them to notice that the confederate is never once untrue to his own choice or influenced by the decisions of others. Thus, any consensus implies that the change must come from them. At the same time, this consistency of the confederate does not seem to be motivated by reality and is in contrast with the answer pattern of the others. b. Results. If one compares the number of “color” responses given in the first half of the 64 tries with the number of “color” responses in the second half, it can be seen that there is a significant reduction ( t = 2.65, significance level .02 > p > .01). Nothing of this sort can be noted either in the control groups of this experiment or in the control or experimental groups in the previous experiment. Moreover, the deviation between the frequency of the “color” choices and of the “size” and “line” choices becomes significant only in the second part of the trials ( t = 2.52, significance level .05 > p > .02). We find no difference, on the whole, between the experimental and the control groups in terms of the number of answers with different dimensions. There is thus avoidance of the consistent answer of the confederate. Moreover, if one considers the average of the subjects’ answers polarized around the “color” answer-and therefore who was influenced-one can see that this aver-

SOCIAL INFLUENCE, BIAS, AND ACTIVE MINORITIES

187

age is significantly lower in this experiment ( N = 21.44) than in the first experiment ( N = 28.76). The differences between these two averages is significant at .02 > p > .01 for t = 2.75. We did not get a similar deviation (either positive or negative) for the other dimensions. It seems that the confederate, on the whole, exerts less influence. These results are parallel to those obtained by Cohen (1963). Cohen, using Asch‘s paradigm, notes that when the majority answers are at one extreme, the amount of conformity is greatest at the beginning, after which subjects become increasingly “correct.” 4. Salience, Avoidance, and Leveling of Answers

In view of the materials we have used, we can expect that in any given situation subjects will develop preferences for a particular dimension. We can expect a phenomenon similar to the one observed by Sherif when, during a series of judgments, isolated individuals reduce the variation in their subjective estimations of a point source of light. However, because the individuals in these control groups must choose between the dimensions of the same figures, a compromise effect is to be expected. Each one resigns himself, in this case, not to express a preference. Sometimes he answers like his neighbor on the left, sometimes he answers like his neighbor on the right; in return, he expects the person on his left or, alternately, his right-hand neighbor to reciprocate by giving an answer identical to his own. Such rotating coalitions lead to a much more even distribution of individual judgments and to an avoidance of differential stands. Consensus is reached at the price of no one’s being different, of no one’s coming into conflict with the others. When a consistent minority intervenes, two aspects are particularly striking. On the one hand, the minority interrupts the negotiation process leading to Compromise; on the other hand, it makes the stimulus dimensions more conspicuous and forces the other individuals to be more consistent. It sets an example by making a choice, acting in opposition, and at the same time it seems to refuse to make any sort of compromise. This is true when there is an approach phenomenon. Conversely, when there is avoidance, when one can attribute the difference between the individual and the rest of the group to a sort of arbitrary idiosyncrasy, then there is no way to take it into account and try to make a compromise. More than that, avoidance of conflict in responses which might arise between individuals becomes all the more urgent since the whole majority must resist the influence attempts of the confederates. In our second experiment we would expect to see such a tendency develop. But let us see what happens before presenting the results. The

188

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

subjects of the control groups of this second experiment have their series of answers written out in front of them on a board. In view of the greater “visibility,” they tend to apply the instructions more strictly and to make personal choices. The experimenter authorizes them to differentiate themselves from each other, and after the first trial, perhaps, they feel committed to their initial answer (Gerard, 1965). The salience of the dimensions is imposed by objective conditions. In the experimental group, on the other hand, a tension develops which prevents both total intraindividual consistency-which would amount to acting like the c o n f e d e r a t e and divergence by increasing the tension already existing between them and the confederate. This tension, which was admitted in postexperimental interviews, leads to a neglect of both the instructions and relations with the experimenter. Group relations become more important than relations with the task. We can follow this process by taking intraindividual variance to be a measure of the degree of preference for internal consistency, as is customary. The greater this variance, the more frequently the individual chooses one dimension at the expense of another. We can make three predictions: (a)intraindividual variance is greater when individuals are alone in making their choice than when they are in a group; ( b ) in the first experiment, intraindividual variance will be smaller in the control group, where a normalizing influence is felt, than in the experimental group, where a minority tends to attract the majority to its point of view; and ( c ) in the second experiment, intraindividual variance will be greater in the control groups than in the experimental groups. The results (Table 111) confirm our predictions. The average of intraindividual variances is higher when the drawings are sorted in an individual setting than in all the group situations. The quest for external consistency thus clearly has the effect of reducing internal consistency. Comparing these averages, one also notes that in experiment No. 1 they are lower for the control groups than for the experimental groups. Thus, two types of influence modalities are represented: normalization and innovation, Experiment No. 2 shows us that this normalization, this attempt of the majority individuals to lessen differences between and to oppose their compromise to the rather arbitrary commitment of the confederate, is more pronounced in the experimental groups. In the face of a very clear-cut norm and the very rigid behavior of the minority, the majority opts for a sort of common sense compromise consensus, where the different “points of view”-the different dimensions-are fairly represented. In the control groups, intraindividual variance is increased, for reasons we have mentioned. (If one wanted to exploit the results to the limit, one would have to say that there is a polarization phenomenon demonstrated in this experiment.) This is just a tentative interpretation. The study of intra-

SOCIAL

INFLUENCE,

189

BIAS, AND ACTIVE MINORITIES

TABLE I11

AVERAQEOF INTRAINDIVIDUAL VARIANCES ~~

Experimental groups Control groups Individuals

Experiment No. 1

Experiment No. 2

85.62 58.04

55.39 92.24 188.90

individual variances makes the mechanisms we have described more plausible; it is not a substitute, in our opinion, for a more direct experimental approach to these mechanisms.

C. THEINFLUENCE OF THE MINORITY RESPONSE ON

AN

IMPLICIT NORM

1. Experimental Procedure An implicit norm is a norm which regulates our behavior without our being clearly conscious of its general nature or of the fact that it determines most of our decisions. Such norms are ubiquitous in the realms of fashion, food, taste, language, etc. For convenience, we chose linguistic material in this experiment; one might well say that most verbal habits are normative regularities in a collectivity which shares these habits. We selected 89 verbal associations from a list established by Nunnally and Hussek (1958) who had measured the frequency of associative choices in an American student population.' For each stimulus word ( e.g., orange) there are two corresponding word associations, one being qualitative ( e.g., round) the other being supraordinated ( e.g., fruit). The subject is given a sheaf of five pages on which the 89 associations are printed. The experimenter reads the stimulus word and the subjects must say aloud which of the two words located on the same line seems to correspond most closely to the stimulus word. Then each subject takes down his answer on the sheet of paper in front of him. The order in which the replies are given is varied systematically, the subject answering first during the previous trial, speaking last on the following trial, etc. The groups, each consisting of four individuals, are seated in the same way as in experiment Nos. 1 and 2. In the experimental groups one confederate always chooses the supraordinated answer; in the control groups there are only nai've subjects. We organized the word associations into two different lists by increasing or decreasing probability of the choice of the supraordinated word among the general population, ' W e are thankful to the Graduate School of Industrial Administration of the Carnegie Institute of Technology (Pittsburgh) for the help extended to us at this occasion.

190

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

In the first list (list A ) the probability of having a supraordinated association for the stimulus word is greater in the beginning; thus, the association chosen by the confederate happens to correspond to the norm. As this probability decreases, his behavior appears more “conservative,” preventing adaptation to the shift in verbal habits. In the second list (list B ) where the probability of association of the supraordinated words is smaller at the beginning, the confederate’s answer is “deviant.” We used both these lists to show that: ( a ) the true source of influence is the consistency of the behavior of the minority, not its degree of deviance in relation to the norm; ( b ) the initial conformity of a minority, contrary to what seems to be the findings of Kelley and Shapiro (1954) and Hollander (1960), may amplify the tendency to adopt its answer, but is not itself the cause ( Doise & Moscovici, 1970). 2. Results Regardless of the order in which the list of associations is introduced, the increase in supraordinated answers in the experimental group, compared to the control groups, is significant. The confederate’s influence on the majority’s answer is undeniable, whether he acts as a “deviant” or as a “conservative,” as shown in Table IV. One might wonder if the minority individual‘s choice affected the majority individuals’ choice for the whole list or only for that part of the list where he was nearer the norm. If all associations were modified, there would be a modification of the majority norm. Such a modification can indeed be observed, since the proportion of “supraordinated associations for the subjects in the experimental groups is significantly higher than for the control groups in the second half of the list A ( t = 3.41, v = 34, .01 > p > .OOl), and in the first half of list B ( t = 2.38, v = 34, .01 > p > .001). What is the confederate’s relative influence when he is a “deviant” and when he is a “conservative”? Let us start with our evidence: we have found that the order of presentation itself does not affect the direction of the associations. The proportion of “supraordinated answers in the control groups is identical for list A and list B. Whatever difference TABLE IV COMPARISON OF THE EXPERIMENTAL GROUPSAND CONTROL GROUPSFOR THEIR AVERAGES OF “SUPRAORDINATED” ANSWERS List A

List B ~~

Experimental group averages ( R = 6 ) Control group averages ( k = 6) t of student Level of significance

74.01 57.61 2.24

p

< .05

63.67 53.89 1.91 .10 > p > .05

SOCIAL INFLUENCE, BIAS, AND ACTIVE MINORITIES

191

appears in the experimental groups must be due to the position of the confederate. For list A, where the confederate is a “conservative,” the frequency of “supraordinated associations is higher ( t = 1.91, 0 = 10, .10 > p > .05) than for list B, where the confederate acted as deviant. Thus, the initial conformity of the minority allows it to exert a greater influence, but such conformity is in no way a precondition for this influence (Doise & Moscovici, 1970).

D. CHANGE OF RESPONSEAND CHANGE OF CODE IN A PERCEPTUAL TASK In the preceding experiments, we highlighted the impact of a consistent minority on a majority, with respect both to preference judgments on equally probable stimuli and to modifications of an implicit norm. We also wanted to obtain some evidence that the same impact would be noticed where the norm to be changed was explicit or quasi-physical. Taking as our starting point the analysis of minority influence when there is a conflict of norms, we devised two experiments where: ( a ) this conflict is sharpened by the consistency of the minority and the consensus among its members; ( b ) objectivity is an implicit requirement for the judgment; ( c ) minority and majority responses are mutually exclusive without being negations of each other (as when one says that two equal lines are unequal); ( d ) the difference in judgment cannot be attributed to differences in individual qualities; and ( e ) the judgment of the majority in the laboratory is identical with that of anyone outside the laboratory, while the minority judgment is exactly the opposite of normal expectations. The two experiments differ in the area in which influence is exerted. In the first experiment we try to bring about a change solely in the response-the way in which the subject names the stimulus. In the second experiment, our goal is to produce a change of code-a change in the way in which the subject discriminates the stimulus from the other stimuli, when he is alone after having been under the influence of the minority ( Moscovici, Lage, & Naffrechoux, 1969). The difference between these two experimental conditions is not identical to the difference between “public” and “private” situations. In those situations, two subjects normally complete the same task with the same stimulus, and the only difference is the presence or absence of other group members. The kind of experiment we have done corresponds to other theoretical concerns.

1 . Experimental Procedure

a. First experiment. The subjects were liberal arts, law, and social science students. These subjects were preferred because of their greater involvement in evaluating the color of an object. The stimuli used con-

192

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

sisted of slides with two different types of filters mounted in them: (1) Photo filters permitting the passage of a beam of light of the dominant wavelength ( A = 483.5) in the blue scale; ( 2 ) neutral filters which reduced light intensity in certain proportion. In a set of six slides, three slides were more luminous than three others. These variations in light intensity were studied in order to make the task more realistic and less boring. Their effect in this experiment was controlled. Each experimental group consisted of four naive subjects and two confederates. Once the subjects were seated in a row before the screen on which the slides were to be projected, they were told that this would be an experiment on color perception; they were informed that they would be asked to judge the color and variation in light intensity of a series of slides ( a brief explanation of the meaning of light intensity was given). Before passing any judgments, the whole group was asked to take a Palack test collectively in order to check the participant’s “chromatic sense.” This test had a twofold objective: (1) to eliminate those subjects who perchance might have had visual abnormalities; ( 2 ) to emphasize the fact that everyone in the group had normal vision, so that the confederate’s response would not be attributed to a difference in vision, i.e., to a personal factor external to the experimental situation. After the collective correction of the test result and the ascertainment that everyone had normal vision, the subjects were instructed about what responses might be given and how the experiment would be conducted: they would reply aloud, naming a simple color as well as estimating the light intensity in numerical terms (ranging from 0 for the dimmest to 5 for the brightest). Subjects were also told that the preliminary trial would be for practice only; each subject would make only a light intensity judgment. The real purpose of these preliminaries was to enable the subjects to get acquainted with the color of the stimulus and “immunize” them, in McGuire’s (1965) sense of the word, against the future onslaught of the instructed minority which would not share the norm. (During these preliminary trials, the confederates answered at random. ) Following this, the series of six different slides was presented six times, the order of the slides varying systematically from one series to the next. Thus, there were 36 in all; each lasted 15 seconds and was separated from the next by approximately 5 seconds of darkness. In each trial the two confederates exerted influence by calling the color “green.” The confederates were thus both internally consistent and consistent with each other from one trial to the next. At the time of the experiment, the subjects filled out a questionnaire

SOCIAL

INFLUENCE,

BIAS, AND ACTIVE MINORITIES

193

concerning the stimuli and the other members of the group. As usual, the real objectives of the experiment were explained before the subjects left. 6. Second experiment. We wondered whether the subjects had experienced an influence which-even if not resulting in a change in verbal response during the experinient-did have a lasting effect on their perception. This would entail, we expected, a shift in the blue-green designation threshold, revealing a reaction that had been repressed during the social interaction. Certain subjects had refused to adopt the minority response openly, feeling compelled to remain loyal to the general norm even when they had begun to doubt its validity. In this case, one might expect x latent attraction to manifest itself by an extension of the designation “green” to stimuli in what a control group would call a “blue” zone. The opposite reaction (extension of the notion “blue” to stimuli into the “green” zone) would be the result of polarization. The first stage of this experiment was identical to the preceding experiment, in that the minority exerted its influence on the majority. At the end of this phase the experimenter thanked the subjects, telling them telling them that another researcher who was also interested in vision phenomena would like to solicit their participation in another research project, independent of the one in which they had just participated. He left the room and the second experimenter entered immediately. He repeated his request and, having obtained agreement from the subjects seated around a table, told them that this experiment was related to the effect of training upon vision. He then described the material, isolated the subjects by means of cardboard screens, and instructed them to write down their responses individually on a sheet of paper. The material constituted 16 disks in the blue-green zone of the Farnsworth 100-hue set perception test. Three disks from each end of the “blue” and “green” scale were absolutely unambiguous, but the 10 other stimuli might appear ambiguous. After having made sure that the subjects understood the instructions well, the experimenter announced the beginning of the test. Each disk was presented on a neutral background for a period lasting approximately 5 seconds; it was placed in the center of the table so that it would be visible to everyone. The series of 16 disks was presented 10 times in the continuous method. The order of presentation was randomized. After the discrimination test, the first experimenter returned, the subjects filled in the postexperimental questionnaire, and the experiment ended in the same manner as the previous one. Ten groups participated in this experiment. c. Third experiment. In this experiment, identical to the first one, we merely diversified the consistency degree of the confederates: They answered “green” 24 times and “blue” 12 times, the dispersion of “blue”

194

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

answers being randomized. Eleven groups participated in this experiment. The control group was the same for the three experiments. In the control group, the presentation of the stimulus was continuous; the subjects also took the discrimination test after the initial experimental phase. In all we had 22 subjects in these groups-four groups of 6 subjectswith the elimination of two subjects who failed to give the discrimination response according to the construction.

2. Results and Discussion “Green” responses (responses which express the influence of the minority in the experimental groups) constituted 8.42%of the answers of the 128 naive subjects in both experiments. As we can see (Tables V and VI), there were no significant differences between the two series of groups. In the third experiment, however, the percentage of “green” responses was only 1.25. Among the 22 subjects of the control group, only one gave two “green” responses, totaling .25% of the responses of the uninfluenced subjects. This means that the latter perceived the stimulus unquestionably as blue; this norm was therefore firmly established socially. The difference between control and experimental subjects on the basis of Mann-Whitney’s U-test ( z = 2.10) was significant ( p = ,019, one-tailed). When one compares the percentage of “green” reTABLE V AMOUNTOF INFLUENCE PRODUCED IN EXPERIMENT 1

Frequency of “green” responses Groups with influenced subjectS,a N = 9 (40.9%) Groups without influenced subjects, N = 13 (59.1y0) Total For influenced subjects,b N = 24 (27.3%) For noninfluenced subjects, N = 64 (72.7%)

Mean num- Mean number of in- ber of nonPercentage fluenced influenced of “green” subjects per subjects per responses group group

239

18.44

2.22

1.78

4

0.20

.31

3.69

243 243

7.67 28.13

0

0%

By “group with influenced subjects” we mean the groups where a t least one subject gave more than two “green” responses. By “influenced subjects” we mean subjects who gave a t least one “green” response.

195

SOCIAL INFLUENIX, BIAS, AND ACTIVE MINORITIES

sponses with the percentage of “erroneous” responses given in the Asch experiments, in which a majority emitted a judgment contrary to the perceptual evidence, one can see that they are of the same order of magnitude. As a matter of fact, in the Asch experiments where the minority was not isolated, and where one of the members broke the unanimity (precisely the situation of our subjects), between 5.5%and 10.4%of “erroneous” responses were obtained. Other data show the influence we expected. Subjects changed their responses (providing four or more “green” responses) in 43.75%of the group. This means that at least one individual gave more than two green responses (we actually observed between 14 and 48 green responses in these groups). The percentage of subjects who yielded was 32%.Thus, we have two categories of groups, those in which there were no subjects who had been influenced and those in which there were subjects who had been influenced. In the latter, it can be seen that 57%of the subjects, or two subjects per group on the average, gave the same response as the confederates; 18.75%green responses were obtained in these groups. Thus, the quantity of green responses which we obtained was not so much the result of isolated individuals yielding to the confederate as it was the result of a modification of judgment within the group. We shall now analyze the results of the discrimination test. We want to find out whether the subjects who changed their social response under TABLE V I 2 AMOUNT OF INFLUENCE PRODUCED IN EXPERIMENT Mean number of inFrequency Percentage fluenced of “green” of “green” subjects per responses responses group

Mean number of noninfluenced subjects per group

~

Groups with influenced subjects, N = 5 (50%) Groups without influenced subjects, N = 5 (50%) Total For influenced subjects, N = 17 (42.5%) For noninfluenced subjects, N = 23 (57.5%) For all control groups, N = 6 For influenced Subjects in control groups, N = 1 (4.5%)

139

19.31

2.4

1.0

6

.83

1.6

3.0

145 145

10.07 23.69

0

0

2 2

.25 .25

196

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

the influence of the consistent minority also changed their perceptual code. In addition, we would also like to verify the hypothesis that the subjects who did not change their social response at least changed their perceptual code, even when they had been mmbers of groups where the majority was not at all influenced at that time by the minority. Regarding response frequencies (Fig. l ) , one can see immediately that the four disks at the extremity of the blue scale (53, 54, 55, 56) and the three disks at the extremity of the green scale (41, 42, 43) did not present any ambiguity for the control subjects; they judged these disks blue or green 9% and 100% of the time, respectively. For the experimental subjects, however, the entire series of disks fell into the zone of ambiguity. Not a single disk was judged 100%blue or 100%green. The generalization of the notion of green in the blue zone can be observed in the experimental groups. Conversely, the notion of blue has a tendency to extend towards the extreme green pole of the green scale slightly more in the experimental groups than in the control group (from disk 4145) . Thus, two modes of reaction to a minority response can be differentiated: attraction and polarization. The first mode of response is much more frequent than the second. These results provide a general view of the phenomenon. The measurement of the threshold makes it possible to verify our hypothesis with greater precision. The calculations bear on the threshold values which are obtained by a graphic method on the smoothed out curve of individual responses. We obtained three values: ( 1 ) the 50% threshold indicating the point in the ordered

lo90o----% 807060-

z 50-

$ 4030-

! l

20 -

41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 Color chip numbers

Fig. 1. Results of the discrimination test. Experimental group “blue” responses (-); experimental group “green” responses (-x-x-); control group control group “green” responses .).-.-.( “blue” responses (000);

SOCIAL

INFLUENCE,

197

BIAS, AND ACTIVE MINORITIES

sequence of stimuli where the subject gives as many “blue” as “green” judgments; ( 2 ) the lower threshold value indicating the point where the subject gives 75%green and 25%blue judgments; and ( 3 ) the upper threshold value indicating the point where the subject gives 25%green and 75%blue judgments. To study the influence of the consistent minority, we shall eliminate subsequently the results of three subjects in the experimental groups who polarized. Their 50%threshold is lower than that of all the control thresholds, as is their lower threshold value, which indicates a generalization of the notion of blue in the green zone. By comparing the 50%, 75%,and 25%thresholds of the experimental groups (37 subjects) and the control groups (22 subjects), we should obtain the expected shift (Table VII) . All of the data reflect the effect of the interaction between minority and majority in the modification of the perceptual code. We propose that this modification affects more subjects than does the change of overt responses, and we will use two series of data for support. On the one hand, although a distinction may be apparent in the experimental groups between subjects who sometimes adopted the minority response and subjects who never adopted it, no such difference emerges from the discrimination test for the three thresholds under consideration. The shift is, in fact, more pronounced for groups where the majority did not change than it is for those where it did change, and the t of student of 1.50 is close to the 1.68 value, while it would be significant at .lo. We had assumed that in the groups where there was no change in social response, or where the “green” response had been in some way “repressed,” we would observe a greater number of “green” judgments in the discrimination test. This was indeed the case. The difference between the groups where the majority did not change and where the majority did change is significant ( x ’ ) = 14.94, significant at p < .002). We can conclude that the consistent minority has an even greater influence on the perceptual code of the subjects than it does

SHIFTIN

THE

TABLE VII THRESHOLD FOR PJCRCEPTION OF THE COLOR GREEN

Control group

Experimental group p (One-tailed

Threshold

Mean

S.D.

Mean

S.D.

50 75 25

47.39 46.16 48.41

1.21 1.42 1.14

48.03 46.85 49.19

1.38 1.54 1.28

t

1.78 1.68 2.33

level) ,038 ,047 . 01

198

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

on their overt response to the slides: Subjects who respond spontaneously in accordance with a color designation norm change their perceptual code after having been subjected to minority pressure, whether or not they change their verbal responses. Thus subjects who, in theory, initially shared a universal norm [seeing (naming) a color and responding to the color] but who were subsequently pressured by a minority norm either changed their response (seeing as “green” what most people see as “blue”) or continued to answer “blue” to something that they later defined perceptually as being “green.” In both cases, the change seems to be the result of intense cognitive activity on the part of the experimental subjects. Confronted with a minority that vigorously supports a different viewpoint, the majority subjects, having no reason to reject the minority position, tried both to understand it and to see what the other subjects see. At no time did they remain passive, nor were they content to blindly accept or reject a norm opposed to their own. This cognitive work, induced by individuals (our confederates) who seemed convinced of the validity of their responses (see also Riecken, 1958), was probably the source of the modification of their own perception which we observed. Our interpretation is supported by the findings of Zimbardo (1965) who showed that the cause of attitude changes in role-playing resides more in the subject’s effort than in an intellectual weighing of the arguments pro and con. In our experiments, as well, the minority forces the majority to change by creating an unusual conflict, compelling the majority to role-play by taking into account the minority’s norm. The present analysis may be incorrect in whole or in part. Additional research obviously is needed, especially studies exploring the more subtle mechanisms of influence than those which are at work in direct, visible, and yielding responses. V. Summary and Conclusion The basic aim of this chapter has been to: ( a ) criticize the postulates underlying the current studies on social influence; ( b ) discard the one-sided association of social influence with questions and facts related to conformity and deviance; ( c ) indicate several new directions, both conceptual and experimental, which could lead to an understanding of innovation; and ( d ) demonstrate the importance of behavior style as a general source of social influence. In order to achieve these goals, we have carefully differentiated between power effects and influence effects and have shown that the latter play a specific role in assigning attributes, i e . , in inferences based on information that we

SOCIAL INFLUENCE, ETAS, AND ACTIVE MINORITIES

199

have about others and about the environment. Intraindividual consistency (across time and across modality) and interindividual consistency (tradition or consensus) have a vital function in these inferential phenomena. They also have a vital function in processes related to influence. Hence, we have shown that different ways of influencing correspond to different ways of establishing interindividual consistency; these, in turn, have an impact on intraindividual consistency since all are different moments of the same chain of inferences. We have indicated the different ways in which a consensus is arrived at: when the majority partner has a norm and the minority partner is deprived of any norm; when partners are equals and each is in possession of his own norm; and when the minority partner has a norm which he opposes to the norm of the majority partner. Since a consensus is a result and a form of treating social conflict, by reducing the variation of interand intraindividual responses, we have also dealt with the modalities of social influence ( conformity, normalization, innovation )-the dynamics of social conflict. Thus, we have met another trend of thought that distinguishes questions of status (power) from institutional questions (influence), sees the influence process as a process of transformation from a natural state (violence) to a social one, rather than simply a method of establishing dependency and, finally, considers that the goal of social life is not merely to adapt oneself to a reality that exists, but rather to establish norms for a man-made reality that one masters. In this general framework, we have developed a set of hypotheses and an experimental paradigm that will allow us to begin a study of innovation and of the influence of minorities on majorities. We have shown that this is possible regardless of the nature of the norm (implicit or explicit), and that it can occur at the response level as well as the coding level. Special attention has been given to the direction of this influence, distinguishing between approach, avoidance, and polarization in relation to the minority. Certainly, if one wants to follow the theoretical and experimental orientation that we have outlined, one will have to look differently at the evolution of relations in a group, and at leadership. We will want to change carefully and with more feeling those parts of reality that we have been inclined to neglect unless, of course, reality itself forces us to do so, which is already the case. REFERENCES Alexander, C. N., Jr., Zucker, L. G., & Brody, C. L. Experimental expectations and autokinetic experiences: consistency theories and judgments convergence. Sodome t y , 1970, 33, 108-121. Allen, V. L. Situational factors in conformity. In L. Berkowitz (Ed.), Advances in

m

SERGE MOSCOVICI AND CLAUDE FAUCHEW

experimental social psychology. Vol. 2. New York: Academic Press, 1965. Pp. 133175. Allen, V. L., & Levine, J. M. Social support, dissent and conformity. Sociometry, 1968, 31, 138-149. Allport, L. H. Social psychology. Boston: Houghton Mifflin, 1924. Ash, S. E. Studies in independence and conformity: a minority of one against an unanimous majority. Psychological Monographs, 1956, 70 (Whole No. 416). Asch, S. E. Social psychology. Englewood Cliffs, N. J.: Prentice-Hall, 1962. Barnett, H. G. Innovation. New York: McGraw-Hill, 1963. Berkowik, L. Effects of perceived dependency relationship upon conformity to group expectations. Journal of Abnormal and Social Psychology, 1957, 55, 350354. Churchman, C. W. Predictions and optimal decisions. Englewood Cliffs, N. J.: Prentice-Hall, 1961. Cohen, B. P. Conflict and conformity. Cambridge, Mass.: MIT Press, 1963. Coser, L. Continuities in the study of social conflict. New York: Free Press, 1967. Crutchfield, R. S. Conformity and character. American Psychologist, 1955, 10, 191198. de Monchaux, C., & Shimmin, S. Some problems of method in experimental group psychology. Human Relations, 1955, 8, 53-60. de Montmollin, G. Influence sociale et jugements perceptifs; effets de la dispersion des informations sur les changements individuels. Ann& Psychologique, 1966, T, 111-128. Doise, W., & Moscovici, S . Approche et kvitement du dbviant dans des groupes de cohbsion diffkrente. Bulletin de Psychologie, 1970, 23, 522526. Emerson, R. Deviation and rejection: an experimental replication. American Sociological Review, 1954, 19, 688-693. Faucheux, C., & Moscovici, S. Le style de comportement d'une minoritk et son influence sur les rbponses d'une majoritb. Bulktin du C.E.R.P., 1967, 16, 337460. Faucheux, C., & Moscovici, S. Self-esteem and exploitative behavior in a game against chance and nature. Journal of Personality and Social Psychology, 1968, 8, 83-88. Festinger, L. Informal social communication. Psychological Review, 1950, 57, 217282. French, J. R. P. A formal theory of social power. Psychological Reuiew, 1956, 83, 181-194. Gerard, H. B. Deviation, conformity and commitment. In J. D. Steiner & M. Fischbein (Eds. ), Current studies in social psychology. New York: Holt, 1965. Pp. 263277. Graham, D. Experimental studies of social influence in simple judgment situations. Journal of Social Psychology, 1962, 56, 245-269. Gurnee, E. A comparison of collective and individual judgment of fact. Journal of Experimental Psychology, 1937, 21, 106-112. Hare, A. P. Handbook of small group research. New York: Free Press, 1965. Harvey, 0. J., & Consalvi, C. Status and conformity to persuasion in informal groups, Journal of Abnormal and Social Psychology, 1960, 60, 182187. Heider, F. The psychology of interpersonal relations. New York: Wiley, 1958. Hollander, E. P. Conformity, status and idiosyncrasy credit. Psychological Reuiew, 1958, 65, 117-127.

SOCIAL INFLUENCE, BIAS, AND ACTIVE MINORITIES

201

Hollander, E. P. Competence and conformity in the acceptance of influence. Journal of Abnormal and Social Psychology, 1960, 61, 360-365. Kelley, H. H. Attribution theory in social psychology. In L. Levine (Ed.), Nebraska symposium on motivation. Lincoln: University of Nebraska Press, 1967. Pp. 1 9 2 241. Kelley, H. H., & Shapiro, M. M. An experiment on conformity to group norms where conformity is detrimental to group achievement. American Sociological Reuiew, 1954, 19, 667-677. Kelley, H. H., & Thibaut, J. W. Experimental studies in group solving and process. In G. Lindzey (Ed.), Handbook of social psychology. Cambridge, Mass.: AddisonWesley, 1954. Kelman, H. C. Effects of success and failure on suggestibility in the autokinetic experiment. Journal of Abnormal and Social Psychology, 1950, 45, 267-285. Kogan, N., & Wallach, M. Modification of a judgmental style through group interaction. Journal of Personality and Social Psychology, 1966, 4, 165-174. Menzel, H. Innovation, integration and marginality: a survey of physicians. American Sociological Review, 1960, 25, 704-713. McGuire, W. J. Inducing Resistance to Persuasion. In L. Berkowitz (Ed.), Adoances in Experimental Social Psychology. Vol. 1. New York: Academic Press, 1965. Pp. 192-232. Moscovici, S. La psychanalyse, son image et son public. Paris: Presses Universitaires de France, 1961. Moscovici, S., Lage, E., & Naffrechoux, M. Influence of a consistent minority on the responses of a majority in a color perception task. Sociometry, 1969, 32, 365-379. Moscovici, S., & Zavalloni, M. The group as a polarizer of attitudes. Journal of Personality and Social Psychology, 1969, 12, 125-135. Newcomb, T. M. A study of consensus. In R. K. Merton, L. Broom, L. S. Cottvell, Jr. (Eds. ) Sociology Today. New York: Harper, 1969. Nunnally, J., & Hussek, T. R. The phony language examination: an approach to the measurement of response bias. Educational and Psychological Measurement, 1958, 18, 275-282. Riecken, H. W. Some problems of consensus and development. Rural Sociology, 1952, 17,245-252. Riecken, H. W. The effect of talkativeness on ability to influence group solutions of problems. Sociometry, 1958, 21, 309-321. Rommetveit, R. Social norms and roles. Oslo: Olso University Press, 1954. Schachter, S. Duration, rejection and communication. Journal of Abnormal and Social Psychology, 1957, 46, 190-207. Schein, E. H. Interpersonal communication group solidarity and social influence. Sociometry, 1960, 23, 148-161. Shaw, M. E. Some effects of varying amounts of information exclusively possessed by a group member upon his behavior to the group. Journal of Genetic Psychology, 1963, 68, 71-79. Sherif, M. The psychology of social norms. New York: Harper, 1936. Sherif, M., & Harvey, 0. J. A study in ego functioning: elimination of stable anchorages in individual and group situations. Sociometry, 1952, 15, 282405. Simon, H. A. Models of Man. New York: Wiley, 1957. Steiner, 1. D. Personality and the resolution of interpersonal disagreements. In B. A.

202

SERGE MOSCOVICI AND CLAUDE FAUCHEUX

Maher (Ed.), Progress in experimental personality research. Vol. 3. New York: Academic Press, 1966. Pp. 195-239. Torrance, E. P. The influence of experienced members of small groups on the behavior of the inexperienced. Journal of Social Psychology, 1959, 49, 249-257. Tuddenham, R. D. The influence of a distorted group norm upon individual judgment. Journal of Psychology, 1958, 46, 227-241. Ziller, R. C., & Behringer, R. D. Assimilation of the knowledgeable newcomer under conditions of group success and failure. Journal of A b n o m l and Social Psychology, 1960, 60, 288-291. Zimbardo, P. G. The effect of effort and improvisation in self-persuasion produced by role-playing. Journal of Experimental and Social Psychology, 1985, 1, 103-120.

A CRITICAL ANALYSIS OF RESEARCH UTILIZING THE PRISONER’S DILEMMA PARADIGM FOR THE STUDY OF BARGAINING Charlan Nemeth DEPARTMENT OF PSYCHOLOGY UNIVERSITY OF CHICAGO CHICAGO, ILLINOIS

I. Overview and Scope . . . . . . . . . . . . 11. Debitions . . . . . . . . . . . . . . A. The Prisoner’s Dilemma Paradigm . . . . . . . . B. A Bargaining Situation . . . . . . . . . . 111. Historical Perspectives . . . . . . . . . . . A. Origins of the Paradigm and Original Questions . . . . . B. Strategy Studies with the Prisoner’s Dilemma Game . . . . IV. A Comparison of Research on the Prisoner’s Dilemma and on Reciprocity . A. Communication and Ambiguity . . . . . . . . B. Ambiguity of Motives . . . . . . . . . . C. Other Miscellaneous Variables . . . . . . . . . V. General Conclusions . . . . . . . . . . . References . . . . . . . . . . . . .

203 206 206 207 208 208 211 215 215 219 224 228 231

I. Overview and Scope In this attempt to review and critically assess the significance of research utilizing a Prisoner’s Dilemma game, I find myself alternating between sympathy and irritation, respect and scepticism. In general, I consider this paradigm to be an elegant and simple derivation from game theory which at first seemed to be a concise and controlled vehicle for looking at the bargaining process, particularly the aspect of mixed 203

204

CHARLAN NEMETH

motives. This game also appeared to be an excellent framework for studying other problems of interest to social psychologists, e.g., person perception and decision making. Yet my general viewpoint is that research in bargaining utilizing the Prisoner’s Dilemma paradigm has become less concerned with questions concerning cooperation, competition, and the bargaining process, and more concerned with studying the Prisoner’s Dilemma paradigm itself. Thus, we now find researchers asking “how do we raise the level of ‘C‘ responding in the Prisoner’s Dilemma?” rather than studying “how do we raise the level of cooperation in a bargaining situation?” These are not the same questions; I will take the position that a “ C choice has little to do with cooperation as ordinarily defined, and that the Prisoner’s Dilemma situation is not a simulation of a bargaining situation and may have very few variables in common with one. We must be very precise in understanding exactly what is being studied when the Prisoner’s Dilemma is used in order to assess the relevance of such studies to any given phenomenon of concern. Here we will be concerned primarily with its relevance for the phenomenon of bargaining. It is my contention that, in contrast to the simplicity for the experimenter, the psychological situation for subjects in a Prisoner’s Dilemma game is one of great abstraction and ambiguity. This is primarily so because subjects have little possibility for communicating their goals, for making exchanges, or even for indicating agreement on a potential solution. Further, the subjects typically don’t have much opportunity to convey information about their personalities and motives. They are placed in a very isolated situation where interaction with one another is at an absolute minimum and where little, if any, information is communicated between the participating individuals. In such an exceedingly “minimal” social situation, subjects may not consider themselves as relating to another person but instead may be relating to an abstract matrix or to abstract feedback. If this is correct, much of the social psychology of bargaining is missing in the Prisoner’s Dilemma situation, and we might question whether the context that remains allows for empirical investigation of particular parameters with applicability to the bargaining process. For example, subjects may be in such an abstract situation that they are more concerned with what the experimenter thinks of them (Rosenberg, 1965) or with what the point of the game is that they no longer consider themsehes to be involved in anything akin to interpersonal bargaining. Even if it is theoretically derived, a particular variable must occur in some context; it is possible that the Prisoner’s Dilemma game, as ordinarily presented to subjects, is too

THE PRISONER’S DILEMMA AND THE STUDY OF BARGAINING

u)5

minimal a context. In such a minimal situation, ethical motivations, such as reciprocity and help giving, which rely on regard for another person for their manifestation, may not be found in studies utilizing the Prisoner’s Dilemma game but may in fact be operative in most bargaining encounters. Perhaps more importantly, attribution and impression formation processes appear less potent in this game than they undoubtedly are in actual bargaining situations. The problem of reciprocity is a case in point; experimenters studying bargaining in the Prisoner’s Dilemma setting have often at least implied the importance of a concept like reciprocity in their studies. Early research in this area assumed that increasing the cooperation (or, rather, the percentage of “C” choices) by a simulated partner should increase reciprocal cooperation on the part of the naive subject. Sometimes this assumption was made on the basis of the motive of maximization of gain; on other occasions, investigators assumed that there was a moral obligation norm prescribing reciprocity. Hence, there appears numerous studies on strategy simulation. Bixenstine, Potash, and Wilson (1963), for example, state: “We are primarily interested in testing the hypothesis that a predominantly cooperative strategy on the part of the other player will tend to elicit similar behavior in return (p. 308) .” However, these expectations have not been confirmed by the data. Studies have repeatedly failed to find simulated high degrees of cooperation eliciting a high degree of cooperation in return from a naive subject. My position is that a concept of reciprocity, while possibly not a potent consideration in most bargaining situations in the “real world,” is at least applicable to bargaining, but it may not be found in research utilizing the Prisoner’s Dilemma game. The reason that reciprocity is not usually found in Prisoner’s Dilemma research is probably because of variables intrinsic to the paradigm that render reciprocity inappropriate or at least decrease the probability of its occurrence. Such variables, as I mentioned previously, include ambiguity of motives and goals. A systematic review of the pertinent research literature can help reconcile the difference between the relative lack of reciprocity in the Prisoner’s Dilemma game and the frequent observation of reciprocity in naturalistic settings (Gouldner, 1960). Further, this review might help clarify the psychological situation for the subject in a Prisoner’s Dilemma situation. Through such clarification we hopefully will be better able to assess the contribution of this literature to date and, more importantly, can suggest directions for its use in the future, whether it be for the study of bargaining or for any other phenomenon in social psychology. ,

206

CHARLAN NEMETH

11. Definitions

A. THEPRISONER’SDILEMMA PARADIGM The original anecdote of the Prisoner’s Dilemma concerns two individuals accused of a crime but who are interviewed separately by the police without being able to communicate with one another. Each prisoner is faced with two alternatives: either to confess to the crime or not to confess to it. If both individuals do not confess, both will be acquitted. On the other hand, if one prisoner confesses and the other does not, the individual who confessed will not only go free but will also receive a reward for turning state’s evidence, while his partner who did not confess will be given a more severe sentence than if he had confessed. However, if both prisoners confess, both will be convicted. We can see that it is to each prisoner’s advantage to confess regardless of what the other individual does. If the other individual confesses also, both would be given the lesser sentence rather than the more severe sentence. If the other individual does not confess, then one ends up free with a reward instead of simply being freed. However, the dilemma is that if both prisoners operate in such a fashion, i.e. if both confess, they end up in a worse position than if both of them do not confess and they will each be convicted of the crime rather than acquitted (Rapoport & Chammah, 1965). The dilemma just summarized has been placed in a more abstract matrix form and given the name Prisoner’s Dilemma game. The game, which provides the basis for this paper, can perhaps best be illustrated by a typical format as depicted in Fig. 1. We can see that the outcomes for the players are shown as ordered pairs, one such pair for each cell entry. For each ordered pair, the outcomes for Player 1 are on the lefthand side; outcomes for Player 2 are on the right-hand side. Players are allowed a choice between two alternatives, ordinarily labeled “C and “D,” which allows for four possible combinations. In comparing this Player I1

C

D

C

Player I

D Fig. 1. Format of the Prisoner’s Dilemma game.

THE PRISONER’S DILEMMA AND THE STUDY OF BARGAINING

207

game with the original verbal anecdote, “ C is comparable to the choice “do not confess,” and “D” is comparable to “confess.” Following the matrix, it can be seen that if both players choose “C,” each receives 3 points; if one player chooses “ D and the other chooses “C,” the player choosing “ D receives 4 points and the player choosing “ C receives nothing; if both play “D,”each receives 1point. If one assumes, as do the game theorists, that each player wishes to maximize his own gain, and that such maximization is the primary motive in a situation like the Prisoner’s Dilemma game, then the predicted strategy would be as follows: If the game were for one trial only, then clearly the optimal strategy would be to play “D,” since the expected outcome is greater. Given that the partner plays “ C on that given trial, a player makes 4 points if he plays “D,”whereas he makes only 3 points if he plays “C.” If the partner plays “ D on that given trial, one gains 1 point by playing “D,” but no points by playing “C.” Clearly, his outcomes are greater in either case if he plays “D.” Yet, if both players assume this posture, each player receives 1 point, which is obviously less than the 3-point outcome that would have occurred if both had chosen “C.” There is, then, no ambiguity as to what is in their collective interest, since joint “C” plays would result in maximum payoffs collectively. Thus, individual interest and collective interest are at odds with each other. Game theorists have assumed that while an individual is likely to play “D” for a one-trial game, this strategy would be maladaptive in the long run and, over a period of trials, each individual will learn that by pursuing his own individual interest, i.e. by playing “D,” he is receiving less return than if he and the other player would jointly play “C.” Therefore, these theorists have assumed that, over the long run, subjects will come to make joint “ C responses.

B. A BARGAINING SITUATION Bargaining, as ordinarily defined, entails a process by which at least two individuals with initially different preferences engage in a series of exchanges and compromises in search of a solution which is mutually satisfactory. In isolating the components of such a process, it is obvious, first of all, that the initial preferences are at least not identical and may be antithetical. However, each individual finds it in his best interest to achieve some sort of solution, i.e. agreement, with the other person. Then there is the process by which this is achieved; each participant must at least be able to stipulate preferences and exchange information relative to acceptable solutions. Finally, the concept of compromise suggests that at least one but more probably both participants will move their preferences in the direction of the other

208

CHARLAN NEMETH

individual to a point where both individuals can agree on a solution. Considering a very common example of bargaining, a seller may like to sell an item for $10.00 while the buyer would like to buy it for $5.00. Their initial positions are clearly not identical and are, to some extent, antithetical, In the process of bargaining, the participants may engage in a series of communication exchanges and compromises. This may take the form of alternative bids. For example, the seller may suggest a price of $9.00 and the buyer may counter with $6.00; the seller may then suggest $8.00 and the buyer may say $7.00, etc. The individuals are clearly communicating to one another their preferences in the situation and are proposing possible solutions that they would consider acceptable. Possibly the individuals may come to an agreement of, say, $7.50 for the item, an outcome which we will assume is mutually satisfactory. At this point, one would say that a bargain had been made. It should be clear that in defining the bargaining situation, one need not have all the variables which were present in the above example. Individuals need not communicate to one another verbally, nor do they necessarily have to communicate via alternate suggestions. They must, however, be able to exchange minimal information regarding their preferences and the solutions they would find acceptable. Further, the individuals need not necessarily reach a bargain, but they must search for a solution if we are to consider it as a bargaining situation. The example of the buyer and the seller could still be considered a bargaining situation even if the people had not come to a solution; the seller, for example, may have offered to sell the product for $9.00, while the buyer may have offered to pay $6.00 for it, but if neither was willing to budge beyond his respective points, the sale might never have taken place. We could assume that they had been engaged in a bargaining process, but that a bargain had not been made. III. Historical Perspectives

A. ORIGINSOF THE PARADIGM AND ORIGINAL QUESTIONS Perhaps the most important element in the above example of a bargaining situation is the matter of mixed motives. It is clear that subjects in a bargaining situation need to find a mutually acceptable solution in order to realize their goals. However, the particular solution preferred by each is, to some extent, antithetical to the other’s desired solution. Thus, there is a motive for cooperation in order to reach a mutually agreeable solution and, simultaneously, a motive for competition in order to gain at the other person’s expense. In the previous example, such a mix of motives was reflected in the seller’s wish to sell

THE PRISONER’S DILEMMA AND THE STUDY OF BARGAINING

209

the product at the highest price possible and the buyer’s desire to buy it at the lowest price possible. Each, however, wants to come to some solution that both will find acceptable, since the seller wants to sell the product and the buyer wants to buy it. It is precisely this aspect of bargaining situations that has attracted the interest of researchers and led to the adoption of the Prisoner’s Dilemma paradigm as a way of studying them. In the Prisoner’s Dilemma game, each participant has two motives operating simultaneously. On the one hand, each individual would like to maximize his own interest at the expense of the other person, that is, play “D.” However, if both people play in this fashion, they will each accrue less gain than if both had made the “C” choice. As I noted above, game theorists assume that, over a period of trials, each individual will come to learn that continually playing “D” is maladaptive, and he will tend to play the “ C choice. We can see the simplicity and elegance of the Prisoner’s Dilemma game in allowing a researcher to study mixed motives in an experimental setting, and can also understand why social psychologists utilized this paradigm so heavily, at least in the beginning of this research endeavor. However, I personally believe that the Prisoner’s Dilemma is not a good analog of a bargaining situation, even a very minimal one. Subjects are usually told that it is a “game,” and they perceive it as such. They therefore are rarely set to bargain with each other. Further, the set for a game may imply considerably different motivations and definition of appropriate behavior from those in a bargaining situation. For example, players may be motivated to “beat the other” in a “game” such as the Prisoner’s Dilemma (Marwell, Ratcliff, & Schmitt, 1969; Messick & Thorngate, 1967), a motive less likely to be important in a bargaining situation. Further, from the viewpoint of the subjects, there is no clear way to communicate their preferences or to interpret them; nor can they actually agree on a strategy. Further, the series of trials in a Prisoner’s Dilemma game does not constitute negotiation toward a solution. It is difficult to know what a solution is in a Prisoner’s Dilemma game: is it a joint “C” lock, a joint “ D lock, or a sequence of plays that the researcher thinks he comprehends? In considering the pertinence of research with the Prisoner’s Dilemma game to the mixed-motive element in bargaining, we should bear in mind the history of these studies, In borrowing the Prisoner’s Dilemma paradigm from game theorists, bargaining researchers have also been strongly influenced by the writings and approach of game theorists (e.g., Luce 81 Raiffa, 1957), and such historical precursors as Pareto and Bernoulli (see Plon, 1970). One suspects that these re-

210

CHARLAN NEMETH

searchers may have adopted the game theorist’s way of looking at the problem as well as the game theorists’ paradigms. Game theorists have concerned themselves primarily with the problem of maximization of outcomes; they are concerned with optimal solutions, economically speaking. Rather than studying the process whereby people reach an agreement, whatever that agreement might be, they tend to concentrate on the ability of an individual to see the solution that will optimize his own gains. As one of the leaders in this area, Rapoport, has stated, the approach of game theorists is basically “de-psychologized in that it reflects “purely structural features of the game itself, not the psychological characteristics of the players [Rapoport, 1970, p. 31.’’ Thus, they inquire whether subjects will find the optimal solution in a mixedmotive situation such as the Prisoner’s Dilemma game. The game theorists, it will be remembered, hypothesize that, over a period of trials, subjects will learn that joint “C” play is the optimal strategy. They assume that short-term gain and self-interest considerations inherent in the dominant “ D strategy will be outweighed by the Iong-term gain and collective interest considerations dictated by the optimal solution of joint “ C choices. The question is whether subjects do, in fact, find the optimal solution, i.e. resolve the conflict of motives. More specifically, these game theorists are interested in conflict resolution, not in bargaining per se. Even where the game has no one rational solution, they are concerned with which of the possible solutions is realized by subjects. The psychological, sociological, or cultural factors influencing subject behavior are only considered if systematic departures from the rational solutions are found. Then, and only then, is there much interest in the “cognitive misconceptions, ethical considerations intruding into strategic calculations, or the personalities of the players ( Rapoport, 1970, p. 39) .” It is the position of this paper that social psychologists studying bargaining have, in fact, borrowed the concepts of optimization and maximization of gains along with their paradigms from game theorists and have approached the study of bargaining almost exclusively from a game theory point of view. However, in this wholesale borrowing, social psychologists have concentrated on optimization questions to the neglect of questions pertaining to the process of bargaining. They, too, consider ethical considerations, attribution of motives, and personality only as explanations for deviations from the rational solution as posited by the game theorists. They have asked, “Why do subjects not play the ‘C choice?” in the game of the Prisoner’s Dilemma rather than, W h y do or don’t subjects cooperate in a mixed motive or bargaining situation?” One would expect that the selection of a research paradigm would

THE PRISONER’S DILEMMA AND THE STUDY OF BARGAINING

211

follow theoretical concerns instead of determining the questions. It appears, however, that while the Prisoner’s Dilemma game originally seemed a proper vehicle for studying mixed motives with the hope of eventual applicability to a bargaining situation, it has become an autonomous, self-perpetuating research area with questions focused on itself rather than on a naturally occurring phenomenon. I suspect that as long as investigators attempt to understand patterns of behavior in a Prisoner’s Dilemma game, they will never be able to answer questions concerning cooperation, competition, and bargaining. The psychological situation for the subject may be so removed from anything remotely connected with bargaining-or even with mixed motives for that matter-that one is studying purely a laboratory phenomenon.

B. STRATEGY STUDIESWITH THE PRISONER’S DILEMMAGAME Examining the research employing the game theory paradigm, we see a considerable emphasis on the effectiveness of various strategies in inducing increased “C” responding or the achievement of the collectively rational solution. Many of these studies were stimulated by findings, contrary to the predictions of game theorists, that subjects do not in fact adopt the mutual “C” choice in a Prisoner’s Dilemma game. Findings indicate that subjects play the “C” choice on only approximately one-third of the trials. Psychologists investigating this phenomenon have concentrated on the reasons for increased or decreased “C” plays. They tend to assume that the lack of “ C choices, which they term the cooperative choice, is due to a lack of trust (Deutsch, 1958; Solomon, 1960) or to the previously selfish plays of their partner (Bixenstine et al., 1963). Thus, considering the sequence of plays, some experimenters have been concerned with the fact that if the subject or the partner starts out by playing “ D due to short-term interest, the other may reciprocate that choice. The participants therefore end up in a joint “D” lock rather than in the predicted joint “C” lock. This observation, it should be pointed out, implies that subjects are not governed only by a single motive to maximize their gain. If they were, even if the other individual started out by playing “D,” the two individuals would, in the long run, come to a joint “ C play. However, these initial “ D choices might induce mistrust, a desire to save face, and the urge for revenge, all of which would negate the possibility of finding the joint solution to maximize gain. In an attempt to raise the level of cooperative choices, numerous experimenters have programmed the strategy of the partner as basically cooperative. Some experimenters ( Oskamp & Perlman, 1965; Scodel,

212

CHARLAN NEMETH

1962; Swingle & Coady, 1967) have even: given their subjects partners who were cooperative. In other words, the simulated partner chose “Con every one of the trials. In other cases, experimenters have given subjects information that their partner played ”C” on 95% (B’ixenstine & Wilson, 1963) or 85%(McClintock, Harrison, Strand, & GaIlo, 1963) of the trials whereas on the remaining 5%and 15%of the trials, they were informed that he played “D.” The sequence of these choices was random. In yet other cases, experimenters established various percentages in between (e.g., Sermat, 1964), where the confederate was programmed to play “ C randomly on 20%, 40%,60%,or 80% of the trials. Clearly, these experimenters assumed that these stark differences in strategy would lead to considerably different “ C responding on the part of the subject. Contrary to such expectations, however, subjects still cooperated (or played “ C choices) on approximately one-third of the trials. For example, in the Swingle and Coady (1967) experiment, subjects playing against an unconditionally cooperative subject, i.e. one who played “ C on 100% of the trials, returned a “ C choice on only 32.9%of the trials. When playing against an individual who played unconditionally competitively, i.e. played “D” on looW of the trials, subjects still played ”C” on 28.3% of the trials. The difference between these percentages is not statistically significant. Similarly, in the McClintock et al. (1963) experiment, subjects given partners whose level of cooperative responding was either 85%, a,or 15%showed no significant differences in the amount of ”C” responding. Seeking to investigate further the effects of varying strategies, several experimenters have looked at the consequences of changes in strategy on “C” responding by the subject. For example, in the Swingle and Coady (1967) study, after 50 trials of unconditionally uncooperative plays by the partner, he altered his behavior and displayed either O%, 25%, 501%, or 100% cooperation on the last 50 trials. The findings indicate that “strategy change resulted in increased variability of subject’s strategy, although mean number of cooperative responses was not significantly affected [Swingle & Coady, 1967, p. 3571.” I . Questions Raised In one sense, the above mentioned experimenters have answered Rapoport’s (1970) question: “How do people behave in response to specific strategies of their co-players? [p. 391.” As mentioned earlier, Rapoport had originally assumed that where there were various solutions to a particular game, one might be concerned with which of the solu-

THE PRISONER’S DILEMMA AND THE STUDY OF BARGAINING

213

tions are realized and with what frequency. If there were systematic departures from rational solutions, he suggested, we could try to identify the psychological, sociological, or cultural factors that operated to bring these nonrational solutions about. The problem is, however, that subjects do not realize the optimal solution as predicted by game theorists and do not, in fact, respond differentially to different strategies of the co-player. In the search for the psychological reasons underlying these discrepancies from a rational solution, some theorists have suggested that an initial “ D choice may engender mistrust and therefore produce reciprocal “ D choices. Other researchers have attempted to look at the motivation of individual subjects and have assumed that subjects may conceivably be bored during the course of the game, particularly when they are playing for imaginary points rather than for real money. Some studies have thus looked at the relative effect of playing for imaginary points, or for imaginary money or real money, but the results are mixed. However, in trying to make sense out of the subjects’ %onoptimal” behavior, investigators have necessarily directed their attention toward variables (e.g., imaginary points) intrinsic to the normal Prisoner’s Dilemma paradigm. I would like to pursue this type of venture further in order to suggest how these variables intrinsic to the paradigm affect the psychological situation in which subjects in a Prisoner’s Dilemma game find themselves. My belief is that the seemingly irrational behavior of subjects in a Prisoner’s Dilemma game is due primarily to the essential incomprehensibility of the situation in which the subject is placed. While the paradigm may be quite simple and elegant from the point of view of the experimenter, who sees it basically as a relatively simple optimizing situation which happens to entail mixed motives on the part of individual subjects, closer analysis reveals that subjects do not find it either simple or elegant. They usually have so little information about the other individual-who he is, what his intentions are, what he will and will not accept, and the basis for his moves-that the subjects are in a quandary as to how to act in such a situation. Their irrationality may be a product of the essentially ambiguous and absurd situation confronting them. This ambiguity, I think, stems mainly from the subjects’ inability to communicate their goals or preferences with any degree of precision and their concomitant difficulty in interpreting the goals of the other individual or getting any feedback that would help them to know their partner. The situation is one of extreme social isolation and, as such,

214

CHARLAN NEMETH

may render otherwise useful concepts from other areas of social psychology inapplicable since these depend on awareness of another individual for their appearance.

2. The Case of Reciprocity The principle of reciprocity is a case in point. This concept assumes that there is a moral obligation prescribing that “one should help or at least not hurt one who has helped him.” This notion, which is postulated as a universal norm by Gouldner (1960), has been at least implied in much of the research utilizing the Prisoner’s Dilemma game, but the findings have not been in accord with it. Some of the strategy studies referred to previously, it will be remembered, suggested that raising the level of cooperation, i.e. “C” choices, for a simulated partner could result in increased cooperative play by the subject. Reciprocity in the sense of returned “C” choices can occur for several reasons, however. An individual may reciprocate “ C choices because he realizes it is the best way to maximize his own and the other’s outcomes. He could also reciprocate because of some moral obligation in the sense described above: he “owes” it to the person who played “C.” A third reason for reciprocity might be for the development and maintenance of a pleasant relationship with the other individual. Regarding this last point, numerous theorists (e.g., Berscheid & Walster, 1969; Blau, 1968) contend that giving and reciprocating benefits is the basis of attraction and of the development of a social relationship. One gives not out of altruism, but as a manifestation of one’s attraction to another or his desire to develop a relationship with that person. Implied in this idea is the expectation that the other will eventually provide some kind of benefits in return. While all of these findings suggest at least three possible bases for reciprocity in a bargaining situation, subjects usually do not reciprocate when playing a Prisoner’s Dilemma game. As I noted previously, percentages of “C” choices by the partner, whether great or small, have little if any effect on the subject. Even where the other plays “C” on 100%of the trials, subjects return such a choice on only a fraction (approximately one-third ) of the trials. One might ask, “Are these concepts totally inapplicable to a mixedmotives situation?” To repeat, my own contention is that the lack of reciprocity is due to the variables intrinsic to the paradigm, the same variables that may account for other apparently irrational behaviors. If subjects are to reciprocate in order to maximize their gains, they must trust that the partner will adhere to the joint “ C move solution. Trust is based on information regarding the other individual, information which is exceedingly minimal in the typical Prisoner’s Dilemma game. Simi-

THE PRISONER’S DILEMMA AND THE STUDY OF BARGAINING

215

larly, if the moral norm of reciprocity is to operate, the same variables become important in that the subjects must perceive the receipt of a “C” choice as a benefit and as stemming from a benevolent motive. Further, they must feel that reciprocity is appropriate and justified, The ambiguity of the situation and the lack of communication again are important factors in reducing the applicability of the norm. For one thing, subjects do not interpret high “ C responses as a benevolent action. Furthermore, subjects are not really interacting with another human being in the Prisoner’s Dilemma. With no information about the other individual, how can one decide to develop a relationship? A subject does not know and cannot see or hear the other individual; nor can he get to know him since information is limited to a game choice between two alternatives, information that we will later see is highly ambiguous. There is no past, no real present, and no future between the participants in a Prisoner’s Dilemma game. Thus, the concept of reciprocity in terms of development and maintenance of a relationship is essentially inapplicable under this paradigm. The variables mentioned above will be considered more fully in the next section in the context of their importance for the moral norm of reciprocity. Since this literature is fairly well developed, it helps us understand the psychological situation of the subject in a Prisoner’s Dilemma game. Reformulating the data obtained from research of this kind may enable us to make more appropriate use of this paradigm in future research.

IV. A Comparison of Research on the Prisoner’s Dilemma and on Reciprocity

A. COMMUNICATION AND AMBIGUITY We have seen that a subject placed in a typical Prisoner’s Dilemma game has very little information as to what is going on. As with many paradigms and experimental contexts, he undoubtedly feels some ambiguity about what the experimenter actually wants to study and the actual purpose of the experiment. Moreover, he also is in the relatively unique position of having to deal with another individual with almost no opportunities for communication. Communication in the sense of verbal transmission of words is probably not very important, but communication of preference, intent, and agreement on particular issues is undoubtedly essential to a bargaining situation. Not only is he deprived of information about his partner, but perhaps even more importantly, the individual cannot send information either, since his own choice in the

216

CHARLAN N E M m

Prisoner‘s Dilemma game carries minimal information about himself and his intentions, Subjects also typically have no prior acquaintance with the other, so they cannot estimate information about him from that source. All they have is the other’s choice between two alternatives, the meaning of the alternatives being highly ambiguous. The player‘s relationship with the other individual and his interpretation of the other’s motives are very important if reciprocity is to occur. The moral norm of reciprocity only makes sense in the context of perceiving dependence, being the recipient of a benefit which is defined as such, and having the opportunity to return a benefit. Further, one of the fundamental reasons for reciprocity is the development of a relationship with another individual. Let us look more closely at the absence of verbal communication in the Prisoner’s Dilemma situation. Available research suggests that such restraints on verbal communication tend to lessen the amount of “C” plays. For example, Bixenstine and Douglas (1967) compared six normal persons with six persons exhibiting psychopathic tendencies in a Prisoner’s Dilemma game. The subjects played the Prisoner’s Dilemma game for a period of time and were then allowed a session of communication. Upon resumption of play, Bixenstine and Douglas found that the normal individuals became more cooperative, whereas the psychopathic individuals did not increase the number of “ C choices. It was assumed that the normal persons took the opportunity to discuss their various strategies and eventually settled upon the mutually advantageous solution of joint “ C plays. Those exhibiting psychopathic tendencies, however, either did not utilize the session to arrive at a cooperative agreement or, if they did, tended to mistrust that agreement and, as a result, refused to adhere to it. In another experiment, this one by Rapoport, Chammah, Dwyer, and Gyr (1962), subjects played the Prisoner’s Dilemma game for 2-3 hours, playing 300500 trials. During the rest session following this game, they ate and rested for 1% 2- hours, after which they resumed the game for a total of 1200 trials. Rapoport et al. found that the percentage of “ C choices in the second session was significantly higher than the percentage of “C” choices in the first session. Rapoport suggests that subjects took advantage of the rest period to decide on a mutually cooperative strategy. Thus, again, one sees that where communication is possible, subjects tend to utilize it to decide upon a mutually advantageous solution, namely joint “C” choices, and to adhere to such an agreement in subsequent trials. It should be mentioned, however, that there is some evidence which

THE PRISONER’S DILEMMA AND THE STUDY OF BARGAINING

217

suggests that the opportunity for communication in a Prisoner’s Dilemma game tends to increase cooperation only where the subjects are disposed to consider only their own welfare. In an experiment by Deutsch (1958), findings indicate that the opportunity for communication makes little difference when the subjects are under either competitive or cooperative instructions. In other words, when subjects are explicitly told to try to beat the other individual or when they are told to consider the welfare of the other individual as well as themselves, communication is relatively ineffective. This is probably because there is nothing to discuss. The appropriate play has been made clear: Under competitive instructions, subjects should play “ D ; under cooperative instructions, they clearly should play “C.” It is in the condition where subjects are instructed to think of themselves that the mix of motives and possible conflict come into play. They can best serve themselves in the long run by playing “C,” but on2y if the other plays “C,” and they might be able to achieve this arrangement through verbal communication. In Deutsch‘s study, when subjects were given individualistic instructions, the opportunity for communication resulted in “ C choices on 70.6%of the trials, whereas a lack of communication showed the usual, and comparatively infrequent, 35.9%“ C plays. Thus, in the Prisoner’s Dilemma game as it is ordinarily played, i.e. where subjects are in fact given individualistic instructions, it appears that the lack of communication results in decreased cooperation. This paper is not suggesting that verbal communication necessarily leads to increased cooperation. As Terhune (1968) summarizes: Finally, it should be realized that it is not concluded that communication necessarily increases cooperation. Communication provides greater opportunity for

cooperation, but that opportunity may either not be used, ineptly used, or used for deceit and vituperation [p. 221.

However, when the subjects are motivated to achieve joint maximization of gain, communication allows them to convey this intent rather clearly to the other and may then result in an agreement between subjects to choose “C.” Clearly, subjects could go back on such a manifest agreement, but they could well expect retaliation and an eventual sequence of “ D plays by the other. In fact, subjects entering into an agreement to cooperate appear to be quite honest. Tedeschi, Powell, Lindskold, and Gahagan ( 196913) and Tedeschi, Lindskold, Horai, and Gahagan ( 1969), for example, report that subjects sending a message of intent to cooperate tend to be truthful a high percentage of the time, females somewhat more than males. In a study broadening the concept of communication to include the

218

CHARLAN NEMETH

nonverbal sources, Wichman (1970) conducted an experiment in which subjects could either see or hear their partner, both see and hear their partner, or neither see nor hear their partner. Results indicate that subjects are most cooperative in the see and hear condition, less cooperative in the hear only condition, still less cooperative in the see only condition, and least cooperative in the isolated condition, a typical Prisoner’s Dilemma situation where subjects neither see nor hear one another. It should be pointed out, however, that it was only in the see-and-hear condition that subjects were able to attain a high and very stable percentage of cooperation over the series of 70 trials. An interesting point in this experiment stems from the low level of “C” responding in the see-only condition. Rather than this being an inefficient method of communication, Wichman suggests that subjects elected not to utilize the channel:

...

in several instances subjects in the see-only condition avoided using the opportunity to conmunicate. When one subject wanted to communicate, the other one would sometimes prevent it simply by not looking up [Wichman, 1970, p. 1181.

Wichman indicates that where subjects wish to utilize a communication channel, the opportunity to view another individual, even if one cannot speak to him, is sufficient to induce increased cooperation. It is also possible that such nonverbal cues are important as “reactive” information. A subject in visual interaction with a partner who was playing “D”could convey his disappointment and, by doing so, tell the individual who is playing very competitively that this is something that he ought not to do. Clearly, much work on the norm of reciprocity suggests that raising the salience of the norm raises the probability that it will be invoked. If one accepts the dictum that one ought to reciprocate in a given situation, the more it is brought to mind, the more it is likely to be followed. Such thinking is also compatible with some research on aggression. Buss (lW),for example, has demonstrated that feedback in the form of moans or groans in reaction to an aggressive act (which in this experiment was the delivery of an electric shock) was effective in reducing aggression intensity. Baron (1971) similarly reports that “signs of pain and suffering on the part of the victim produce significant decrements in the strength of the attacks against this person under all experimental conditions [p. 481.’’ Thus, the Wichman study points out the importance of both verbal and nonverbal information in a situation such as the Prisoner’s Dilemma game. It is precisely when both communication channels are open, namely the see-and-hear condition, that subjects in the Wichman investigation achieve the highest and most stable set of “C” responses.

THE PRISONER’S DILEMMA AND THE STUDY OF BARGAINING

219

B. AMBIGUITY OF MOTIVES Ambiguity of motives is clearly a subset of communication in the way it is defined in this paper. It is precisely because there are no channels for communication, either verbal or nonverbal, that motives and intentions are particularly ambiguous. There are numerous studies utilizing the Prisoner’s Dilemma game which suggest that the other person’s motives are extremely unclear to the subjects. Even where the other individual plays “C” on 100%of the trials, they are uncertain as to why he acted as he did. Solomon (1960) conducted one of the few studies in which subjects were actually asked their interpretations, When the players had equal power, which is the case in a typical Prisoner’s Dilemma game, subjects were asked to check the motive for the other person’s acting cooperatively on 1 W of the trials. Rather than checking the category of “cooperation” or “individualistic” as the motive, subjects tended to rate this behavior as “out of field,” indicating that subjects did not comprehend why the other individual played as he did. Clearly, the subjects could have interpreted such unconditional cooperation on the part of the other individual in many ways. He may have been attempting to maximize his own gains as well as the subject’s own gains; he may have been playing in such a fashion out of regard for the subject; or he may have been playing the game incorrectly. It is also possible, however, that a person would behave in such a manner to deceive the other; in attempting to convey good intentions, he may actually be setting up the subject in order to exploit him. This latter possibility tends to be rather implausible over a number of trials unless the other eventually does attempt exploitation. The difficulty is that the subjects are not interpreting such behavior as stemming from either benevolent or malevolent motives. Unable to attribute any motive whatsoever, they register “I don’t know.” This uncertainty is important in understanding the lack of reciprocal “ C choices in a Prisoner’s Dilemma game, regardless of which motive we posit as the basis for reciprocity. If an individual wants to maximize his gains in a Prisoner’s Dilemma situation, the tendency to play “ C a high percentage of the time and to reciprocate the other’s “C” choices make sense only when the individual feels the other person also wishes to maximize his gains and is willing to play in this fashion for a maximum joint return. From the point of view of reciprocity as a moral norm, the subjects clearly need to feel that the donor intended to benefit the recipient before they feel obligated to reciprocate. Two experiments (Goranson and Berkowitz, 1966; Nemeth, 1970b) found that subjects worked harder on

220

CHARLAN NEMETH

behalf of another person when that person had previously helped them voluntarily than when that help had been made compulsory by the experimenter. The implication is that where the benefactor is seen to behave under constraints, the recipient does not feel motivated to reciprocate, However, when it can be reasonably concluded that the donor had really wanted to benefit the recipient, then the recipient is motivated to return such benefits. In fact, there is some suggestion that where an ulterior motive is even conceiuable, some subjects tend not to reciprocate benefits. Dickoff ( 1964) illustrated this in an experiment where his subjects were given positive evaluations by an interviewer. For one condition it was explained that the purpose of the experiment was to determine how accurately people form impressions of one another. In another group, subjects were told the same thing but it was added that the interviewer wanted them to participate in an additional experiment of her own. Findings indicated that the interviewer was rated much more favorably in the former situation than in the latter. In the “accurate impression” condition, subjects could reasonably assume that the evaluations were an honest re3ection of the other’s views. Where it was even possible that the donor of the favorable evaluation wanted the subject indebted to her for purposes of her own, the subjects rated her much more negatively. The latter finding is important because it suggests that if there is even a possibility of an ulterior motive for a positive action, one may not only refuse to reciprocate benefits but may actually become antagonistic. In a situation such as the Prisoner’s Dilemma game where the motives are so ambiguous and where an ulterior motive is certainly possible (the subject may be attempting to set up the other subject in order to exploit him), the other subject may be motivated to not reciprocate a “ C choice and may instead turn to a competitive sequence of plays. This reasoning can probably account for findings such as that of Komorita, Sheposh, and Braver (1968). In their experiment with the Prisoner’s Dilemma game, a benevolent person who had power over the subject was considerably more effective in inducing cooperation by the subject than an individual with equal power. Komorita et al. suggest that a benevolent, powerful person conveys the impression that he had benevolent intentions, presumably because he has more choice in the situation. Since he clearly has the opportunity to take more than an equal share, his good will appears to be intentional. Under such conditions, Komorita et al. (1968) suggest that “trust develops and cooperative behavior is facilitated [p. 1411.” There are games structurally equivalent to the Prisoner’s Dilemma game which have had more success in inducing reciprocal cooperation, but they appear to be less ambiguous in terms of motives for making the

THE PRISONER’S DILEMMA AND THE STUDY OF BARGAINING

221

“cooperative” choice. One such game was investigated by Evans and Crumbaugh (1966). In their experiment, both matrix and nonmatrix forms of the Prisoner’s Dilemma game were used. In this experiment, illustrated in Fig. 2, the subjects were asked to choose between the two alternatives of “give him three” or “give me one.” As can be seen from Fig. 2, such alternatives are “equivalent” to the alternatives of “ C or “D,” respectively, in the typical Prisoner’s Dilemma game. One can see that if both individuals choose to “give him three,” each subject receives 3 points, which is equivalent to a joint “C” choice. If one chooses to “give me one” while the other individual chooses to “give him three,” the individual who chooses “give me one” receives 4 points, while the individual who chooses “give him three” receives nothing. This is an identical outcome to the situation where one player chooses “C” while the other chooses “D.” If both individuals choose to give themselves 1 point, then each subject receives 1 point, an outcome equivalent to the joint “D” response. However, even though such a game may be quantitatively equivalent to the ordinary Prisoner’s Dilemma game, Evans and Crumbaugh report significantly greater cooperation in the nonmatrix form of the game than in the matrix form. This finding has since been corroborated by Pruitt ( 1967). A number of explanations could account for the difference in cooperation between these two quantitatively equivalent games. On the one hand, subjects may perceive greater control over the distribution of outcomes, and consequently, may feel more morally responsible toward their partner. However, it seems to me that the choices of “give him three” and “give me one” more clearly represent benevolent and malevolent intentions than do their counterparts of “C” and “D.”A subject who has found that his partner has decided to “give him three” points may reason-

(A)

Player P

C

D

C Player S

D

0,4

“Give me 1” versus “Give him 3” Fig. 2. Paradigm used by Evans and Crumbaugh (1966). ( A ) matrix presentation. ( B ) nonmatrix presentation. Copyright 1966 by the American Psychological Association.

222

CHARLAN NEMETH

ably assume that the individual in fact wanted to give him 3 points through a benevolent gesture, Assuming that the motives for the “C choice are benevolent, one might expect that subjects would reciprocate such cooperation, which they in fact do. Another point to be made in this context is that labeling a choice in terms of a motive to “give” may well heighten the salience of a norm of altruism or reciprocity for the subject. In addition to considering the other subject’s motive as one of benevolence or malevolence based on his choice, the subject may also interpret his own behavior in the same light. When giving the other person points, the subject may clearly see himself as generous, whereas when he decides to keep points for himself, he may consider himself to be selfish, Such labels may then serve to indicate for him the appropriateness or the inappropriateness of his action, and this may allow for a norm of social responsibility or reciprocity to function. I should note, however, that clarifying the other individual’s motives does not necessarily lead to reciprocity of cooperation. For one thing, this clarification may make it obvious that a person’s motives are malevolent. And even if these motives are interpreted as benevolent, subjects may not feel that reciprocity is appropriate, or they may consider other motives which are incompatible with reciprocity more important at the time. For example, they might want to win the game. However, where the individual sees reciprocity as a desirable mode of action for himself and appropriate in a given situation, clarification of motives at least allows the possibility of reciprocity, provided the other’s intentions are reasonably good. With ambiguity, however, reciprocity is, by definition, inappropriate. A case in point is a study by Shure, Meeker, and Hansford (1965) which indicates that even where benevolent motives are clear, subjects still will not reciprocate generosity or cooperation. In their study, an individual acting as a pacifist made initial cooperative gestures and, although provided with a means of retaliation (shock) did not use it but simply blocked the other individual’s responses. In this situation, the pacifist’s intentions were explained to the subject as a product of his Quaker background. While one might assume that clarifying his intentions to be a pacifist might increase the probability of cooperation, the study indicates that this strategy led to enhanced exploitation. Thus, the other player’s unambiguous intentions and even the disarmament that he exhibited during the game both were ineffective in inducing cooperation and reciprocity. Perhaps one of the most important aspects of the Shure et al. experiment is that subjects were repeatedly pressured into maintaining the dominating strategy. In their experiment, in order to maintain the situation as an encounter between a pacifist and an aggres-

THF, PRISONER’S DILEMMA AND THE STUDY OF BARGAINING

223

sor, it was necessary to coerce ordinary subjects into becoming very aggressive. Subjects were sent messages from two other individuals (who were programmed confederates) urging the subject to adopt the dominating strategy. While the subject had final say on the choice, he clearly was being pressured into choosing the exploitative strategy. It is possible that even if the subject was inclined to reciprocate the generosity of the pacifist partner, he may have yielded to the continued pressure from peers not to do so. On the other hand, peer pressure may simply have justified his own inclination to consider his personal welfare and to disregard the benevolent gestures of another. I am not maintaining that reciprocity as a moral norm is particularly potent in the bargaining situation, nor am I suggesting that it cannot be outweighed by more powerful considerations. However, I do believe that its general lack of appearance in a Prisoner’s Dilemma game is due, to a large extent, to the ambiguity of the situation for the subject, particularly the ambiguity regarding the other’s motives. Other bargaining situations where motives are made clearer indicate that reciprocity may be a significant factor. In the Deutsch, Epstein, Canavan, and Gumpert (1967) study, for example, six choices were available: 4$ to the subject, regardless of what his partner chose; 9~ to each, if matched, that is if both plarers chose blue; 1+ to the subject if his partner played anything but blue; White: 7t to the partner, regardless of how the partner played; Attack move, that is 66 taken from the partner, unless the partner played Red: green, i.e. the defensive choice; in the latter case, no points are given or taken from either player; Green: Defense move, explained above; Beige: Play of disarming, that is the subject destroys all the red pegs he has accumulated; Orange: Attack indicator.

Black: Blue:

Utilizing this game, Deutsch et al. found that in a situation roughly comparable to the unconditional cooperation employed by the previous experimenters with the Prisoner’s Dilemma game, subjects returned cooperation on a high percentage of trials. Their condition, termed “turn the other cheek,” consisted of a partner playing blue, then white and then blue, after which he consistently chose blue. Such strategy led to approximately 60%cooperation by subjects, a percentage which is notably higher than that previously found by experimenters using the Prisoner’s Dilemma game, It is probable that the greater effectiveness of simulated cooperation in this study is due primarily to the game situation, not to the effectiveness of cooperative strategy per se. In the Deutsch et al. game,

224

CHARLAN NEMETH

the partner’s moves were very clearly interpreted as a communicative gesture reflecting a motive of cooperation. From a list of adjectives, subjects considered a play of blue as “cooperative,” “mutual,” and “COworking.” Similarly, other colors were associated with other particular intentions. For example, black was seen as “cautious,” “independent,” and “selfish.” A white play was seen as “benevolent,” “philanthropic,” and “foolish.” In sum, we can see that where a subject communicates to his partner that he wishes to be cooperative, the subject will reciprocate with cooperation on his part, namely playing a blue choice, which in many ways is comparable to a “ C choice in the Prisoner’s Dilemma situation. To some extent, the concept of ambiguity may explain why a tit-fortat strategy is more effective in inducing “cooperation” than is unconditional cooperation. Solomon ( 1960), for example, found that subjects in a Prisoner’s Dilemma game play “ D significantly more often when the opponent is unconditionally cooperative than when his cooperation is conditional upon the subject’s cooperation. In the latter condition, conditional or tit-for-tat strategy, the partner plays the response which the subject played on the previous trial. It may well be that tit-for-tat strategy provides information regarding the motives of the other. There is at least a clear pattern to this strategy, and subjects are likely to assume that the other is communicating “I’ll match what you play.” Thus the subject may feel that he has the choice of eventually getting into a “ C lock or a “ D lock. He clearly will prefer the “ C lock to the “ D lock, since the outcomes are definitely greater. Unconditional cooperation, however, carries no clear message. The partner playing in this fashion appears immovable and unsusceptible to influence. He plays “C” regardless of what the subject plays. The subject is then in the position of interpreting this strategy, which he may attribute to benevolence or stupidity, or which he may find totally incomprehensible. We can see, then, that the tit-for-tat strategy may be effective simply because it reduces the ambiguity for the subject. The other’s actions need not indicate benevolence. If they do indicate benevolence, reciprocity in the sense of a moral norm may come into play. Even if they simply convey what the other wants, at least the subject can consider this possibility in determining a solution. If the other prefers a “C-“C lock of trials but is willing to play “ C only if the subject plays in a similar fashion, the subject may well consider this a reasonable solution and adhere to it.

C. OTHERMISCELLANEOUS VARIABLES Other variables which affect the psychological situation for the subject in a Prisoner’s Dilemma game, and render concepts such as

THE PRISONER’S DILEMMA AND THE STUDY OF BARGAINING

225

reciprocity relatively inapplicable, are those of instructional set, magnitude of incentive, and prior relationship of the participants. It will be remembered that game theorists originally made their predictions on the assumption that subjects would want to maximize their gains. Borrowing this orientation, most experiments with the Prisoner’s Dilemma game contain instructions which make it quite clear to the subject that he should consider his own welfare only. Such instructions are important in determining the subjects’ choices. It is possible that by giving individualistic instructions, the experimenter may be implicitly fostering competition with the other person and therefore defining the rules of the game. Becker and McClintock (1967) have pointed out that instructions define what is rational in providing rules of the game. Thus, even though subjects may not appear to be acting rationally on the basis of pay-offs, they may be acting rationally on the basis of what the experimenter defines as the goal of the situation, or at least on the basis of their own interpretation of his instructions. For example, the instructions may suggest to the subject that he should “beat the other individual.” Deutsch (1958) studied the effect of the instructional set on “ C responding in a Prisoner’s Dilemma game. He gave individualistic instructions to some subjects, as indicated above. To other subjects he gave cooperative instructions, that is, the subject was to consider his own welfare as we22 as the welfare of the other individual. A third set of subjects was told to consider their own welfare and to defeat their opponent. The percentage of cooperation was significantly different for these three conditions. Cooperatively-set subjects played “ C on a great number of trials, while individualistically-set and competitively-set subjects showed a very low percentage of cooperation and did not differ significantly from each other. One might conclude that subjects given individualistic instruction tend to interpret them in the light of competitive instructions and therefore define “beating the other individual” as a goal of the game. Magnitude of incentive for cooperation is another variable conceivably linked with the lack of reciprocity. Some experimenters have suggested that subjects playing for imaginary points, which is normally the case in these experimental situations, may be bored and may play in particular ways out of curiosity about how their partner will respond or may play simply as a way of alleviating their boredom. They suggest that subjects playing for imaginary money or points which are of little importance to them may not be motivated to maximize their outcomes. Furthermore, they might not be sufficiently motivated to give benefits or to reciprocate them. Some research has attempted to investigate the relative effects of

226

CHARLAN NEMETH

real us. imaginary money, but the findings are somewhat mixed. Gallo (1966), in replicating the Deutsch and Krauss (1960) bargaining experiment, drew considerable attention to this variable. In his experiment, there is evidence that subjects are more cooperative when real money is used than when the money is imaginary. Research on this variable with the Prisoner’s Dilemma game offers rather equivocal findings. Evans ( 1964) , Wrightsman (1966), and Knox and Douglas (1968) show little or no effect of monetary awards; whether the reward is large or small, subjects tend to cooperate approximately the same percentage of time. On the other hand, McClintock and McNeel (1967) found that Flemish subjects were more cooperative when there was a high reward than when there was a low reward in a maximizing difference game. Oskamp and Perlman (1965) show a small but significant tendency for a high reward to lead to greater cooperation. Radlow, Weidner, and Hurst (1968) found that subjects playing for real money were significantly more cooperative than those playing for imaginary money in a Prisoner’s Dilemma game. Oskamp and Kleinke (1970) show no differences in cooperation as a result of five different reward levels. In a second study, however, there is a reversal; the trend, though not significant, is for subjects in a no-reward condition to be more cooperative than those playing for either low or high rewards. Gumpert, Deutsch, and Epstein (1969) found that subjects playing for imaginary dollars were more cooperative than those playing for real dollars. Thus, the results on the effects of the magnitude of incentive for cooperation are quite inconclusive. It is highly likely, however, that the existing differences are due to slight modifications in experimental procedure. Gallo, Funk, and Levine (1969) offer some evidence for the importance of procedural changes in exhibiting the effects of incentive for cooperation. In this experiment, subjects either played the regular Prisoner’s Dilemma game or played a comparable pay-off matrix which was decomposed. Using decomposed matrices, Gallo et al. allowed subjects a differential number of response options. Some subjects were given one chip on which “ten cents for me” was written on one side and “twenty cents for him” was written on the other side. Other subjects were given either five or ten chips. In the five-chip condition, each chip had “two cents for me” on one side and “four cents for him” on the other side. In the ten-chip condition, each chip was lettered “one cent for me” on one side and “two cents for him” on the other side. While the pay-offs were identical in all four conditions, Gallo d al. report a differential effect of the variable of real money as opposed to imaginary money. In the regular Prisoner’s Dilemma game and in the one-chip condition, the money variable showed little effect. However, in both the five- and ten-chip condition, subjects were significantly more cooperative in the reward condition than in the

THE PRISONER’S DILEMMA AND THE S

~ OFY BARGAINING

227

imaginary money condition. Gallo and his co-investigators (1969) suggest that “the motivational effects of large rewards may show up with expanded response options, even under conditions which prevent their appearance in two choice games [p, 2431.” It is possible that increasing the number of alternatives may lead to the possibility of communication and classification of preferences and motives. Subjects who assume that they are likely to reach a bargain through effective communication with their partner might well be expected to do so under large rewards. However, if the game is basically incomprehensible to the subject, increasing the desirability of reaching a bargain is not particularly motivating. Certainly, it is not necessarily true that a greater number of alternatives may lessen ambiguity for the subject. This clearly depends on the logical relationships of the alternatives. If the alternatives are systematically varied, increasing the number of alternatives may tend to increase the uniqueness of a motive for any given choice. Thus, provided that an increase in the number of alternatives increases the specificity of intention, increase in the reward level will have a significant effect on the level of cooperation. Regarding reciprocity and social responsibility, the importance of motives and clarity of attribution has already been discussed in the previous section. However, the lack of real incentive in a Prisoner’s Dilemma game has another effect on reciprocity. Generally, one does not consider the obligatory part of this concept unless one is talking about reasonably important outcomes. Even where the concept refers to development of a relationship, one gives and returns favors that are of some importance to the recipient. Otherwise these favors are unlikely to result in the development of a relationship. The lack of a relationship between the participants in the Prisoner’s Dilemma game is another important variable. Subjects have no prior relationship when they come into the situation and, as I have pointed out previously, have little possibility of developing one. Research by Swingle and Gillis (1968) suggest the importance of this variable for cooperation in the Prisoner’s Dilemma game. Their study shows greater initial cooperation by subjects playing against a partner whom they liked, as compared to a partner whom they disliked or with whom they had no specified relationship. McClintock and McNeel ( 1967) report similar findings with the maximizing differences game. When we consider reciprocity as a vehicle for developing and maintaining a relationship, we can see that subjects in the Prisoner’s Dilemma siutation cannot know if they even want to develop a relationship. Since there is no history, they have no relationship to maintain. Since there is no future, they have none to develop. Regarding reciprocity as a moral norm of obligation, there is some evidence from the altruism literature that

228

CHARLAN NEMETH

subjects are more willing to help a person who is highly liked than one who is less liked (Daniels 81 Berkowitz, 1963). However, there is also the suggestion that subjects in the Prisoner’s Dilemma and in the maximizing differences game tend to treat persons with whom they have no prior acquaintance in a manner similar to those toward whom they are hostile. The variable of no prior relationship thus emerges as an important one in understanding the lack of cooperative responding in a Prisoner’s Dilemma game.

V. General Conclusions In the preceding section, studies utiIizing the Prisoner’s Dilemma game and the literature of reciprocity were reviewed and compared with reference to variables intrinsic to the Prisoner’s Dilemma paradigm. Uncertainty about the meaning of the responses, the difficulty of assessing the motives behind particular strategies, the barriers against conveying or interpreting preferences or acceptable solutions, the lack of incentive for maximizing gains or perhaps even for taking the task seriously, and the absence of interaction with another human being all tend to make this a situation of possible irrelevance, certain ambiguity, and perhaps even incomprehensibility for the subject. Not only do these variables generally make concepts like reciprocity irrelevant, they tend to draw into question what is actually being studied in the Prisoner’s Dilemma game. The Prisoner’s Dilemma game is certainly not an example of bargaining. There appears to be no real solution from the point of view of the subjects. Even if there were a solution, subjects cannot convey or interpret their preferences to one another in order to arrive at it, and it is almost impossible to communicate agreement to any given solution. It is the theorist and the experimenter who define joint “ C choices as a solution, who assume that the subject is motivated to maximize his point return according to the given matrix, and who define the situation as one of simply finding the optimal solution. There is a real question whether subjects even are motivated to maximize their point return; the stakes are so low that they may be more interested in “beating the other,” testing how another will respond, making a good impression on the experimenter, or trying to figure out what the point of the game or the experiment is. Further, even if they were concerned with gain return, they might not know how to go about achieving it. How does one convey the impression that he prefers and is willing to adhere to a sequence of joint “ C responses? Does he continuaIIy play “C” himself? Available research suggests that the other won’t comprehend his

THE PRISONER’S D~LEMMA AND THE STUDY OF BARGAINING

229

motives for doing this. The only strategy that appears to work, but which any given subject may not think of, is a tit-for-tat strategy, i.e. playing what the other person played on the previous trial. Assuming he picks up the strategy, he may decide to adopt the “ C position as well. However, results from Solomon’s (19600)study indicate that subjects reciprocate the “ C choice only 524: of the time under such a strategy. Admittedly, this is better than the 39% return from an unconditionally cooperative strategy, but it falls considerably short of an “optimal” solution. The subject playing this way receives no points on nearly half of the trials. Thus, a given subject will probably try various other plays, assuming he even wants to arrive at a “C-“Clock. It is unlikely that the other will comprehend his moves, however, and the feedback regarding choice is so minimal that the subject remains confused, It would perhaps take on more meaning if the subject could see, hear, or speak to the other, or if he had a greater set of alternatives from which to choose, alternatives that would more clearly specify intention. However, in the typical Prisoner’s Dilemma paradigm, none of these paths to information is allowed. In addition to the ambiguity of the situation in terms of interacting with and comprehending another person, one must consider how the subject defines the situation in which he finds himself. Is it a “game” or is it a problem-solving task, or is it a bargaining situation? I would contend that he does not perceive it as a bargaining situation primarily because there are no real outcomes, no clear solution, and no way to communicate preferences and agreements. The concept of optimal solution originally posited by the game theorists appears to be closer to a problem-solving task. A game, at least as ordinarily defined, entails the concept of competition and “beating the other.” In the usual instructions defining the situation, experimenters refer to the Prisoner’s Dilemma as a “game” and suggest that they are investigating “gameplaying behavior.” Even if subjects are instructed to consider only their own welfare, which is the usual case, this does not necessarily translate itself into a search for maximum self-return or the optimal solution for most subjects. Placed in the context of a game, subjects may translate self-concern into winning and beating the other. There was some suggestion in the preceding review that subjects tend to operate under such individualistic instructions much the same way they operate under competitive instructions. Such a definition clearly changes the strategy. If you want to “win” in the sense of beating the other, why reciprocate “ C choices from the point of view of an optimal solution? The only winning strategy is to set up a situation where you play “D” while the other plays “C.”Otherwise, you lose or are equal.

230

CHARLAN NEMETH

There is also the question of social interaction. Subjects pIaying a Prisoner’s Dilemma game have not known each other previously and are not allowed to see, hear, or speak to each other. Information regarding the motives and personality dispositions of the other are highly ambiguous or nonexistent. There is no relationship to develop or to maintain. In many ways, this game is set up almost as if one were playing with a machine. Another person exists only in a very abstract sense, Even the feedback regarding his choice comes via the experimenter. Concepts like regard for the other, attraction, aggression, and revenge take on very little meaning in this context. Where is the interaction in a Prisoner’s Dilemma game? It appears that research on the Prisoner’s Dilemma game may have very little transferability to a naturally occurring bargaining situation. This is not because it is too simple or too minimal a situation to be relevant to the complex situations that one might wish to comprehend. Rather, the Prisoner’s Dilemma game is not really a simulation of even a very simple bargaining situation. Then, does it add information relevant to optimal solutions? It probably has little to say here since subjects do not appear to consider the game in this light. However, it is being increasingly used as a vehicle to comprehend other psychological phenomena. Kelley and Stahelski (1970), for example, used the Prisoner’s Dilemma game to focus on individual differences in terms of perceptions and attributions of one’s partner based on his play. Here the authors took advantage of the ambiguity inherent in the Prisoner’s Dilemma game, since this ambiguity allows individual differences in attribution to emerge. Horai and Tedeschi (1969) utilized the paradigm in conjunction with verbal threats and responses which allowed them to investigate concepts such as credibility of threat. Used in this way, the game becomes better defined, and information regarding the other becomes available. Benton, Gelber, Kelley, and Liebling ( 1969) staged a modified version of the Prisoner’s Dilemma utilizing messages which a subject could either believe or distrust. This made it possible for them to look at reactions to degrees of deceit. None of these studies dealt at all with questions related to optimal solutions or to the effects of particular strategies unless in the context of attributions and beliefs or where increased information was available. Not only are different questions being posed in these studies, but the original paradigm is modified in a way that allows for investigation. Perhaps the realization that new questions can be more appropriately scrutinized in the Prisoner’s Dilemma situation has only become possible as a result of the previous research on the game. This may be the way that science normally develops and changes. Kuhn ( 1970), in

THE PRISONER’S DILEMMA AND THE STUDY OF BARGAINING

231

his book The Structure of Scientific Revolutions, suggests that “normal science” progresses, in part, by articulation of the paradigm itself. Although he feels that this renders the course of science relatively inflexible and prone to ignore new phenomena which do not fit the paradigm, he concludes: Perhaps these are defects. The areas investigated by normal science are, of course, miniscule; the enterprise now under discussion has drastically restricted vision. But those restrictions, born from confidence in a paradigm, turn out to be essential to the development of science. By focusing attention upon a small range of relatively esoteric problems, the paradigm forces scientists to investigate some part of nature in a detail and depth that would otherwise be unimaginable. And normal science possesses a built-in mechanism that ensures the relaxation of the restrictions that bound research whenever the paradigm from which they derive ceases to function effectively. At that point, scientists begin to behave differently, and the nature of their research problems change. In the interim, however, during the period when the paradigm is successful, the profession will have solved problems that its members could scarcely have imagined and would never have undertaken without commitment to the paradigm. And at least part of that achievement always proves to be permanent [pp. 24-25].

We may be in the process of relaxing these restrictions and opening up our questions. Since I believe that the Prisoner’s Dilemma paradigm has ceased to function effectively for the study of bargaining, I cannot help but view the recent studies utilizing this paradigm for other psychological phenomena, e.g., attribution processes, as a sign of progress in the field, a progress that undoubtedly owes a debt to the research that preceded it. REFERENCES Baron, R. A. Aggression as a function of magnitude of victim’s pain cues, level of prior anger arousal, and aggressor-victim similarity. Journal of Personality and Social Psychology, 1971, 18, 48-54. Becker, G. M., & McClintock, C. G. Value: Behavioral decision theory. Annual Review of Psychology, 1967, 18, 239-286. Benton, A. A. Productivity, distributive justice, and bargaining among children. Journal of Personality and Social Psychology, 1971, 18,67-78. Benton, A. A., Celber, E. R., Kelley, H. H., & Liebling, B. A. Reactions to various degrees of deceit in a mixed-motive relationship. ]ourno1 of Personality and Social Psychology, 1969, 12, 170-180. Berkowitz, L. Resistance to improper dependency relationships. Journal of Experimental Social Psychology, 1969, 5, 283-294. Berkowitz, L. Beyond exchange: Ideals and other factors affecting helping and altruism. In R. Willis & M. Greenberg (Eds.), Social exchange. New York: Wiley, 1970. Berscheid, E., & Walster, E. H. Interpersonal attraction. Reading, Mass.: AddisonWesley, 1969. Bixenstine, V. E., & Douglas, J. Effect of psychopathology on group consensus and

232

CHAlUAN NEMETH

cooperative choice in a six-person game, Journal of Personality and Social Psychology, 1967, 5, 32-37. Bixenstine, V. E., Potash, H., & Wilson, K. V. Effects of level of cooperative choice by the other player on choices in a prisoner’s dilemma game. Part I. Journal of Abnormal and Social Psychology, 1963, 66, 308-313. Bixenstine, V. E., & Wilson, K. V. Effects of level of cooperative choice by the other player on choices in a prisoner’s dilemma game. Part 11. Journal of Abnormal and Sodal Psychology, 1963, 67, 139-147. Blau, P. M. Interaction: social exchange. In D. L. Sills (Ed.), International encyclopedia of the so& sciences. Vol. 7. New York: Macmillan and Free Press, 1968. Pp. 452-457. Brehm, J. W., & Cole, A. H. Effect of a favor which reduces freedom. Journal of Personality and Social Psychology, 1966, 3, 420-426. Buss, A. H. Instrumentality of aggression, feedback, and frustration as determinants of physical aggression. Journal of Personality and Social Psychology, 1966, 3, 153-162. Daniels, L. R., & Berkowitz, L. Liking and response to dependency relations. Human Relations, 1963, 16, 141-148. Deutsch, M. Trust and suspicion. Journal of Conflict Resolution, 1958, 2, 265-279. Deutsch, M., Epstein, Y., Canavan, P., & Gumpert, P. Strategies of inducing cooperation. Journal of Conflict Resolution, 1967, 11, 345-360. Deutsch, M., & Krauss, R. The effect of threat upon interpersonal bargaining. Journal of Abnormal and Social Psychology, 1960, 61, 181-189. Dickoff, H. Reactions to evaluations by another person as a function of self-evaluation and the interaction context. Unpublished doctoral dissertation, Duke University, 1961. Cited by E. E. Jones, Ingratiation. New York: Appleton, 1964. Evans, G. Effects of unilateral promise and value of rewards upon cooperation and trust. Journal of Abnormal and Social Psychology, 1964, 69, 587-590. Evans, G. W., & Crumbaugh, C. M. Effects of prisoner’s dilemma format on cooperative behavior. Journal of Personality and Social Psychology, 1966, 3, 4 8 s 488. Gallo, P. S., Jr. Effects of increased incentives upon the use of threat in bargaining. Journal of Personality and Social Psychology, 1966, 4, 14-20. Gallo, P. S., Jr., Funk, S. G., and Levine, J. R. Reward size, method of presentation, and number of alternatives in a prisoner’s dilemma game. Journal of Personality and Social Psychology, 1969, 13. 239-244. Goranson, R. E., & Berkowitz, L. Reciprocity and responsibility reactions to prior help. Journal of Personality and Social Psychology, 1966, 3, 227-232. Gouldner, A. W. The norm of reciprocity: A preliminary statement. American Sociological Reuiew, 1960, 25, 161-178. Gumpert, P., Deutsch, M., & Epstein, Y. Effect of incentive magnitude on cooperation in the prisoner’s dilemma game. Journal of Personality and Social Psychology, 1969, i i , 6 e 6 9 . Horai, J., & Tedeschi, J. T. Effects of credibility and magnitude of punishment on compliance to threats. Journal of Personality and Social Psychology, 1969, 12, 164-169. Kelley, H. H., & Stahelski, A. J. Social interaction basis of cooperators’ and competitors’ beliefs about others. Journal of Personality and Sodal Psychology, 1970, 15, 66-91. b o x , R. E., & Douglas, R. Low payoffs and marginal comprehension: Two possible

THE PRISONER’S DILEMMA AND THE STUDY OF BARGAINING

233

constraints upon behavior in the prisoner’s dilemma. Paper presented at the meeting of the Western Psychological Association, San Diego, April 1968. Komorita, S. S., Sheposh, J. P., & Braver, S. L. Power, the use of power, and cooperative choice in a two-person game. Journal of Personality and Social Psychology, 1968, 8, 134-142. Kuhn, T. S. The structure of scientific reuolutions. Chicago: University of Chicago Press, 1970. Luce, R. D., & RaifFa, H. Game and decisions. New York: Wiley, 1957. Marwell, G., Ratcliff, K., & Schniitt, D. R. Minimizing differences in a maximizing difference game. Journal of Personality and Social Psychology, 1969, 12, 158-163. McClintock, C. G., Harrison, A. A., Strand, S., & Gallo, P. Internationalism-isolationism, strategy of the other player, and two-person game behavior. Journal of Abnormal and Social Psychology, 1963, 67, 631-636. McClintock, C. G., & McNeel, S. P. Prior dyadic experience and monetary reward as determinants of cooperative and competitive game behavior. Journal of Personality and Social Psychology, 1967, 5, 282-294. Messick, D. M., & Thorngate, W. Relative gain maximization in experimental games. Journal of Experimental Social Psychology, 1967, 3, 85-101. Nemeth, C. Similarity, altruism and reciprocity. Unpublished manuscript, University of Chicago, 1969. Nemeth, C. Bargaining and reciprocity. Psychological Bulletin, 1970, 74, 297308. ( a ) Nemeth, C. The effects of free vs. constrained behavior on attraction between people. Journal of Personality and Social Psychology, 1970, 15, 302-311. ( b ) Oskamp, S., & Kleinke, C. Amount of reward in a variable in the prisoner’s dilemma game. Journal af Personality and Social Psychology, 1970, 16, 133-140. Oskamp, S., (Jr Perlman, D. Factors affecting cooperation in prisoner’s dilemma games. Journal of Conflict Resolution, 1965, 9, 358-374. Plon, M. Rbflexions sur la rencontre entre la psychologie sociale et la thkorie des jeux. Unpublished manuscript, Laboratoire de Psychologie Sociale de I’Rcole Pratique des Hautes Etudes, Paris, 1970. Pruitt, D. Reward structure and cooperation: The decomposed prisoner’s dilemma game. Journal of Persona!ity and Social Psychology, 1967, 7, 21-27. Pruitt, D. G. Reciprocity and credit building in the laboratory dyad. Journal of Personality and Social Psychology, 1968, 8, 143-147. Radlow, R., Weidner, M. F., & Hurst, P. M. The effect of incentive magnitude and “motivational orientation” upon choice behavior in a two-person non-zero-sum game. Journal of Social Psychology, 1968, 74, 199-208. Rapoport, A. Conflict resolution in the light of game theory and beyond. In P. G. Swingle (Ed.), The structure of conflict. New York: Academic Press, 1970. Pp. 1-43. Rapoport, A., & Chammah, A. M. Prisoner’s Dilemma. Ann Arbor, Mich.: Univ. of Michigan Press, 1965. Rapoport, A., Chammah, A., Dwyer, J., & Gyr, J. Three person and non-zero-sum non-negotiable game. Behavioral Science, 1962, 7, 39-58. Rosenberg, M. J. When dissonance fails: On eliminating evaluation apprehension from attitude measurement. Journal of Personality and Social Psychology, 1965, 1, 28-42. Scodel, A. Induced collaboration in some non-zero sum games. Journal of Conflict Resolution, 1962, 33-40.

234

CHARLAN NEMETH

Sermat, V. Cooperative behavior in mixed motive games. Journal of Social Psychology, 1984, 62, 217-239. Shure, G. H . , Meeker, R. J., & Hansford, E. A. The effectiveness of pacifist strategies in bargaining games. Journal of Conflict Resolution, 1965, 9, 106-117. Solomon, L. The influence of some types of power relationships and game strategies upon the development of interpersonal trust. Journal of Abnormal and Social Psychology, 1960, 61, 223-230. Swingle, P. G. (Ed. ). The structure of conflict. New York: Academic Press, 1970. Swingle, P. G., & Coady, H. Effects of the partner’s abrupt strategy change upon the subject’s responding in the Prisoner’s Dilemma. Journal of Personality and Social Psychology, 1987, 5, 357-383. Swingle, P. G., & Gillis, J. S. Effects of the emotional relationship between protagonists in the prisoner’s dilemma. Journal of Personality and Social Psychology, 1968, 8, 180-185. Tedeschi, J. T. Threats and promises. In P. G. Swingle (Ed.), The stnrcture of conflict. New York: Academic Press, 1970. 4. 155-191. Tedeschi, J. T., Lindskold, S., Horai, J., & Gahagan, J. P. Social power and the credibility of promises. Journal of personality and Social Psychology, 1969, 13, 253-261. ( a ) Tedeschi, J. T., Powell, J., Lindskold, S., & Gahagan, J. P. The patterning of “honored” promises and sex differences in social conflicts. Journal of Social P ~ c h o l o g y ,1989, 78, 297-298. ( b ) Terhune, K. W. Motives, situation, and the interpersonal conflict within prisoner’s dilemma. Joumal of Personality and Social Psychology, 1988, 8(3, Part 2 ) . Wichman, J. Effects of isolation and communication on cooperation in a two-person game. Journal of Personality and Social Psychology, 1970, 16, 114-120. Wrightsman, L. S. Personality and attitudinal correlates of trusting and trustworthy behaviors in a two-person game. Journal of Personality and Social Psychology, 1966, 4, 328-332.

STRUCTURAL REPRESENTATIONS OF IMPLICIT PERSONALITY THEORY'

Seymour Rosenberg Andrea Sedlak DEPARTMENT OF PSYCHOLOGY RUTGERS UNIVERSITY NEW BRUNSWICK, NEW JERSEY

I. Introduction . . . . . . . . . . . . 11. Scaling Methods and the Interpretation of Scaling Solutions . . A. Scaling Methods . . . . . . . . . . B. Interpretation of Scaling Solutions . . . . . . . 111. Laboratory Analogs of Trait Inference and Personality Description A. Trait-Inference Approach . . . . . . . . B. Personality Descriptions with Trait Terms Supplied by . . . . . . . . . . the Experimenter C. Conclusions . . . . . . . . . . . IV. Naturalistic Personality Descriptions . . . . . . . A. Multidimensional Representations . . . . . . . B. Typological Representations . . . . . . . . V. Individual Differences . . . . . . . . . A. Multidimensional Models of Individual Differences . . . B. Study of Individuals: An Example . . . . . . . VI. Personality Perception in Children . . . . . . . VII. Some Future Directions . . . . . . . . . . References . . . . . . . . . . . .

.

. . . . . . .

235 238 239 245 247 247

.

250 258 259

. . . . .

.

.

. .

.

261

269 274 275 279 287 291 293

I. Introduction This chapter is concerned with (1) the traits that a person perceives as characteristic of himself and of others and the beliefs that 'The writing of this chapter was supported in part by NIH Research Scientist Award # l-K5-MH-29,329-04 and NSF Grant GS-2552 to Seymour Rosenberg. 235

236

SEYMOUR ROSENBERG AND ANDREA SEDLAK

he holds about which traits tend to go together and which do not, ( 2 ) parsimonious ways of representing this type of complex and loosely organized system of perceived traits and relations. The perception of stable physical and psychological characteristics and the beliefs about their co-occurrences are important objects of research for social psychologists, not only because of the ubiquity with which a person thinks and talks about himself and others but also because of the central role that personality perception plays in everyday social interaction and social decisions. The trait elements and relations that a person perceives are frequently referred to by psychologists as the person’s implicit personality theory. The “theory” is characterized as “implicit” because a person’s trait categories and beliefs are inferred from his descriptions and expectations about individuals and groups rather than being stated by him as a formal theory. It is unlikely that an individual could make many of his categories and beliefs explicit or, given these basic elements, could organize them into a formal and parsimoniously stated theory of personality. Implicit theories, and there are presumably as many such theories as there are individuals albeit with many overlapping elements and relations, have also been dubbed “common sense,” “lay,” and “naive” to distinguish them from scientific theories of personality. A person’s implicit theory reveals itself in at least two different ways in everyday behavior. One is in the way he describes himself and others in conversation and writing-“She is very sweet and gullible”-or in the general propositions he makes about people-“Physically attractive men and women are likely to be vain and empty-headed.” Certain trait terms occur frequentIy in such protocols, and among these trait terms, some tend to co-occur while others do not. For example, intelligent, friendly, selfcentered, ambitious, and lazy are among the terms most frequently used by college students in their ad-lib descriptions of people they know ( Rosenberg & Sedlak, 1972). Among these terms, intelligent and friendly tend to be ascribed to the same person, but intelligent and self-centered do not. A second and related way in which implicit personality theory is revealed is in the trait inferences that a listener (or reader) draws from a personality description given to him. If a person is said to be “intelligent,” for example, then a college student may also think of him as “friendly” whether or not the speaker says so explicitly. How can a person’s categories and beliefs be discovered from his everyday descriptions and inferences of people and then represented in at least a quasiformal manner? The answer is not simple in spite of the

IMPLICIT PERSONALITY THEORY

237

ubiquity with which people think and talk about themselves and others. To begin with, the task of codifying ad-lib descriptions of personality is a forbidding one in terms of the time and effort involved. Even more of an obstacle, however, has been the absence of techniques for extracting and representing the structure implicit in such data. Until recently, freeresponse protocols were simply summarized in terms of the general categories of description present in such protocols; for example, the degree to which terms referring to appearance, emotional adjustment, interests, etc., show up in the descriptions by children and adults (Beach & Wertheimer, 1961; Dornbusch, Hastorf, Richardson, Muzzy, & Vreeland, 1965; Yarrow & Campbell, 1963). However, the individual trait terms and their co-occurrences were not extracted and analyzed in these studies. Because of the difficulties involved in dealing with free-response data, and because such data are not appropriate in answering certain substantive questions, social psychologists have developed laboratory analogs of personality description and trait inference. The personality descriptions are obtained from trait sorting, check lists, and rating scales, and the trait inferences are obtained with tasks similar in format either to the semantic differential or to similarity judgments. In all instances, the trait vocabulary available to the respondent is limited to the terms supplied by the investigator. The first part of the chapter describes these laboratory analogs and summarizes a number of findings that have resulted from their use, particularly studies involving the application of multidimensional scaling and, to some extent, clustering and factor analysis. These multivariate studies are emphasized because the dimensional analysis of free-response data, which is the focus of the remaining part of the chapter, developed directly out of such work. Free-response descriptions, also called “naturalistic personality descriptions,” refer either to solicited descriptions or to descriptions as they are actually found in everyday life. In the solicited descriptions that have been analyzed with multidimensional scaling and clustering techniques, the person is free to choose both the people he will describe (except for the work with children) and the trait names and phrases he will use to describe these people. The guiding principle in the collection and analysis of solicited descriptions is to obtain protocols which approximate as closely as possible the descriptions found in everyday life. The application of scaling and clustering to free-response descriptions and inferences, and their laboratory analogs, is a recent development. Even in the brief decade that scaling and clustering have been applied to studies of implicit personality theory, however, there are

238

SEYMOUR ROSENBERG AND ANDREA SEDLAK

already extant studies representative of a wide range of diverse phenomena within this area including (1) representations of the culturally shared aspects of implicit personality theory, ( 2 ) individual differences in personality perception and their correlates in perceiver characteristics, (3) intensive analysis of an individual's view of people, and ( 4 ) personality perception in children. This chapter summarizes these various applications even where the number of representative studies is small since such studies often provide new research paradigms and substantive questions for future work. The next section contains a brief overview of multidimensional scaling techniques and the use of multiple regression to empirically interpret the configurations obtained from scaling. Readers familiar with these basic tools may wish to skip this section.

II. Scaling Methods and the Interpretation of Scaling Solutions The application of multidimensional scaling starts with the selection of a set of interrelated entities. The basic input datum required for multidimensional scaling is a number for each pair of entities in the set reflecting how closely the two entities are related. What multidimensional scaling then does in psychological research is to provide as output a geometric configuration of the entities in which the interpoint distances correspond to the empirical input values of psychological relatedness. In personality perception, the entities are usually personality traitsoccasionally, persons. The input measure of trait relatedness may reflect either how close two traits are (the closer the traits the larger the number) or how distant they are (the closer the traits the smaller the number). Serviceable measures of trait relatedness can be obtained from any of a variety of the data-gathering techniques described in the following sections of this chapter. The notion of psychological relatedness between two traits may be interpreted in at least two ways, each of which implies somewhat different substantive questions. One interpretation is that of trait similarity. For example, a rating of the similarity between two traits would probably reflect the synonymity or semantic substitutability of the two traits. A second interpretation is that of trait co-occurrence, that is, the degree to which two traits are perceived as going together in the same individual. Intelligent and friendly are not likely to be judged as synonyms, but they are often found together in a description of an individual; adept and skillful, on the other hand, are judged as highly similar in meaning (Haagen, 1949), and hence would also be more or less descriptive of the same individual. Since the focus of this chapter is on person-

IMPLICIT PERSONALlTY THEORY

239

ality perception rather than on meaning, per se, studies involving the notion of trait co-occurrence will be emphasized. The relations between meaning similarity, perceived co-occurrence, and actual co-occurrence are discussed more fully in Section 111, B, 2. The situation is different when a study in personality perception involves people as the entities to be scaled rather than traits (e.g., Hays, 1958; Jackson, Messick, & Solley, 1957). In such instances, a measure for each pair of persons reflecting their similarity in personality is usually more appropriate than one reflecting their co-occurrence-the latter would presumably refer to the degree of social contact between the two persons.

A. SCALING METHODS There are three basic methods of multidimensional scaling: fully metric, nonmetric, and fully rwnmetric (Green & Carmone, 1970). The choice of which method to use depends on the assumptions that the investigator makes about the input data and about the geometric model for representing the relations among the scaled entities. 1. Fully Metric Methods In fully metric (also termed linear or classical) methods, it is assumed that the input distances are on an interval scale, in the sense that these methods seek an output configuration with interpoint distances that match the input values up to a linear transformation. Consider, for example, the distances between the 10 cities shown in Table I. A fully metric method, applied to these distances, or any linear transformation of these distances, would yield the configuration shown in Fig. 1A. When the input measure does not contain any error, the relation between input and output distances form a perfect straight line, such as that shown in Fig, 2A. Fully metric methods are generally based on linear factoring methods and were the first multidimensional techniques to have been developed (Torgerson, 1952, 1958). 2. Nonmetric Methods In nonmetric methods, it is assumed that the input distances (or proximities) are on an ordinal scale but that output distances are on a ratio scale. (The dimensions of the output configuration are on an interval scale.) The implication of this assumption is that the correspondence which is sought between input and output is not restricted to a best-fitting linear function but can be any monotone relation. Shepard (1962a, 1962b), who was responsible for initiating the development of nonmetric methods, showed that the reconstruction of a

84

TABLE I

5

DISTANCE IN MILESBETWEEN TENU.S. CITIES

Sail N.Y. Chicago Miami L.A. San Francisco

New Orleans Washington Denver Dallas Seattle

N.Y.

Chicago

0 818 1325 2911 3082 1406 229 1852 1649 3025

0 1386 2189 2233 977 696 1043 955 2184

Miami

L.A.

Francisco

New Orleans

$ Washington

Denver

Dallas

!s

@

B

0

0

2885 3238 881 1096 2126 1394 3496

0

407 1947 2754 1189 1431 1193

0 2300 2897 1267 1791 866

0 1165 1314 509 2731

0 1707 1414

2880

5 5 0 805 1426

* 0 2222

8 k

IMPLICIT PERSONALITY THEORY

I A-

ACTUAL - -

---

241

MAP -

- --- --- - -

- - -

------

' SEATTLE

AN

FRANCISCO 'DENVER

. c H I c ~ ~ 0' N E W YORK 'WASHINGTON

'LOS ANGELES 'DALLAS 'NEW ORLEANS *MIAMI

B MAP RECONSTRUCTED BY NONMETRIC MDS ---- _ _ - _ _ - - -- - - - _ _ _ _ _ - _ ~

'SEATTLE 'NEW Y O R K 'DENVER

* C H ' C A G Q ~ ~ ~ ~ ~ ~ ~ ~ ~ ~

'S4N FRANCISCO LOS ANGELES * N E W ORLEANS *MIAMI

Fig. 1A. Actual configuration of 10 U.S. cities; B. Configuration of the 10 cities reconstructed by nonmetric scaling from the rank order of the distances,

known configuration from only the rank order of the interpoint distances is essentially perfect with 15 or more points, and is usually quite good with even fewer points (Shepard, 1366). The reason for this, roughly speaking, is that even the rank order of distances contain enough metric information under most conditions to recover the original configuration. Figure l B , for example, shows the mapping which was produced by one of Kruskal's (1967) nonnietric programs when only the rank order of the distances in Table I was used as input. Fig. 2B shows the relation between input ranks and output distances for the ten cities. Nonmetric scaling represents an important advance in multidiniensional scaling, because the assumption that input data may be only ordinal in nature appears to be more realistic in psychological research than the interval-scale assumption required by linear methods. That is, any of a large variety of monotonic transformations of the original measure of relatedness is as acceptable as the original measure itself. In practice, this arbitrariness means that a linear method may indicate the presence of more dimensions than are warranted. For example, the application of a linear method to the ranks of the distances in Table I would suggest that the cities lie in three or more dimensions. The source of this distortion is the assumption that the ranks are on an interval scale

242

SEYMOUR ROSENBERG AND ANDRE4 SEDLAK

A

ln

w

LINEAR

t 3500 1 z

W

w 3:

t-

2800 .

W

m ln W

2

z

z 0 W

z

2100 2100

1400

I -

-

/

700 -

a

t-

u, a

v)

w

5

B

MONOTONE

z W W

z

I-

W

m w 0

z

5B Y

0

U W P a 0 Y

z a U

DISTANCE

IN TWO-DIMENSIONAL

SOLUTION

Fig. 2A. Linear transformation of the distances between the cities as plotted in Fig. 1A; B, scattergram of distances between the cities as plotted in Fig. 1B us. the rank order of the actual distances. Stress = 1%.

(not from the fact that the cities actually lie on a sphere). Any monotonic (except linear) transformation of the true distances would have produced some distortion of this type. Nonmetric methods also have certain disadvantages which fully metric methods do not. For one thing, nonmetric methods involve a complex search procedure which cannot examine all possible configurations, so that the final solution may not be the best possible one. A second

IMPLICIT PERSONALITY THEORY

243

problem is that under certain special conditions, nonmetric methods produce “degenerate” solutions ( Torgerson, 1965). We turn now to the question of determining the appropriate number of dimensions for representing the input data. Do the entities fall on a line (one dimension), a plane (two dimensions), etc.? Two criteria are commonly considered in answering this question. One criterion is that the multidimensional solution adequately represents the input data with as few dimensions as possible. The application of this criterion is complicated by the presence of an unknown amount of measurement error in the data. Kruskal (1964a, 1964b) has, however, provided a measure of the goodness of fit of a solution and some heuristic rules based on this measure in order to estimate optimal dimensionality. The procedure is first to obtain a one-dimensional solution, a twodimensional solution, and so on up to some number of dimensions considered more than sufficient for the input data. The goodness-of-fit measure, termed stress, is also calculated for each solution. The stress measure is essentially a normalized sum of squared residuals in which O% ! stress means that a perfect monotone relation exists between input and output distances. Stress necessarily decreases or remains unchanged with increasing dimensionality since the goodness of fit must improve (or remain unchanged) when more dimensions are used to represent the input data. The rate at which stress decreases with increasing dimensionality is useful, however, in determining optimal dimensionality. For example, even though the program will arrange the traits on a line, a plane may be required to capture adequately the distance relations in the input data. The stress measure would usually reflect this fact by decreasing sharply in value between the one- and two-dimensional solutions. If additional dimensions are superfluous-for example, a three-dimensional solution when a two-dimensional solution is adequate-little drop occurs in the stress measure. For the sake of parsimony, the dimensionality after which there is little improvement in fit is chosen as the optimal dimensionality. The nature of the stress measure and its use are illustrated with data from a study by Rosenberg, Nelson, and Vivekananthan (1968) on personality perception. They obtained multidimensional configurations of 60 traits in one through five dimensions. The scattergrams of input (ordinate) us. output (abscissa) distances for the one-, two-, and threedimensional configurations, along with the stress values, are shown in Fig. 3. The scatter of the points decreases most sharply between one and two dimensions, and this decrease is reflected in the large decrease in stress value. The decrease in scatter, and hence stress, between two and

244

SEYMOUR ROSENBERG AND ANDREA SEDLAK

TWO DIMENSIONS

THREE DIMENSIONS

STRESS ~ 6 %

DISTANCE, d-

Fig. 3., Scattergram of output distances d vs. input distances 6, for solutions of one, two, and three dimensions. From Rosenberg et al. ( 1968).

three dimensions is considerably less dramatic, indicating that two dimensions may be sufficient to represent the data. Nevertheless, one may want to explore the nature of the three-dimensional configuration. This brings us to the second criterion for estimating the appropriate dimensionality-the interpretability of a configuration. The rule is that if an n-dimensional configuration can be satisfactorily interpreted but the ( n 1)"dimensional configuration reveals no additional meaningful structure, it may be appropriate to stop at the n-dimensional configuration. Empirical methods for interpreting a configuration are discussed in Section 11, B.

+

3. Fully Nonrnetric Methods Fully nonmetric methods are an outgrowth of unfolding methods ( Coombs, 1950, 1958,19f34). In these methods, both the input and output are assumed to be nonmetric; that is, the output is not a geometric configuration but only a rank order of the entities on each of a set of dimensions, in turn (Bennett & Hays, 1960; Hays & Bennett, 1961). Section 111, A describes an early and pioneering application by Hays (1958) of a fully nonmetric method to the scaling of traits. In general, however, fully nonmetric methods have not been used extensively, primarily because they do not take advantage of the metric information available in the input data, and they yield output structures that are of limited use. [Note: Multidimensional scaling techniques have also been developed for representing individual and group differences in structure. A description of these techniques is given in Section V as an introduction to the empirical work on individual differences in implicit personality theory. A description of clustering is also deferred until the relevant empirical work is discussed ( Section IV, B ) .] Readers interested in a more complete but nontechnical description of scaling and clustering than it was possible to give in this chapter should consult Green and Cannone (1970). This book also describes

IMPLICIT PERSONALITY THEORY

245

multidimensional scaling techniques for the analysis of preference data (in contrast to proximity or distance data). Analyses of preference data will not be covered in the present chapter because they have not as yet been systematically exploited in studies of personality perception, although some starts have been made in the area (Posavac, 1971; Wiggins & Wiggins, 1969).

B. INTERPRETATION OF SCALING SOLUTIONS Most of the empirical methods which provide input data for multidimensional scaling make it unnecessary to prejudge the underlying dimensions of a set of traits (or persons). When a person i s asked to judge the degree to which traits co-occur in the same individual, for example, neither the investigator nor the judge need be aware of the underlying dimensions for this judgment. The same is true for a measure based on the degree to which traits actually co-occur in descriptions of personality. In fact, one of the attractive features of these datagathering methods, in combination with multidimensional scaling, is that no assumptions are required about the nature of the trait dimensions. Thus, multidimensional scaling of co-occurrence-based data is not the direct and simple result of the investigator’s preselection of dimensions. Instead, meaningful psychological dimensions are sought in the multidimensional configuration after it has been obtained from the cooccurrence data. We shall refer to this identification of dimensions as the interpretation of a solution. It should be noted that the orthogonal axes in a multidimensional solution do not necessarily correspond to meaningful psychological dimensions; the location of these orthogonal axes is typically arbitrary and serves simply as a convenient way to describe the output configuration. The interpretation of a trait configuration is analogous to the interpretation of a multidimensional configuration of entities for which there are physical continua. For example, a two-dimensional solution was obtained from similarity judgments among 14 colors, and this solution resembled the configuration of the colors on the conventional color circle (e.g., Shepard, 1962b). Unlike the physical properties of colors, geometric shapes, etc., the important properties of entities such as trait names may not be so apparent by inspection. The first step in the interpretation of a trait space relies rather more heavily on the intuitions of the investigator about the trait properties that are likely to be present in the space. These properties are sometimes also suggested either by previous research findings, e.g., the semantic differential factors, or by an examination of the solution itself. Suppose, for example, the investigator hypothesizes that an im-

246

SEYMOUR ROSENBERG AND ANDREA SEDLAK

portant dimension in the perception of oneself and others is the amount of interpersonal dominance that a person possesses. The question then is: How can he empirically confirm or disconfirm the importance of dominance as one of the properties of the traits that a perceiver attributes to people? The first step in answering this question is to obtain ratings (or any other quantitative estimate) of each of the traits on this property: for example, the average rating on a dominant-submissive scale of each of the traits in the space. This property rating is analogous to the values of a physical attribute of entities for which meaningful physical continua exist ( e.g., wavelength). Note that the property values are data collected independently of the data used as input for scaling. Multiple regression techniques are then used to locate an axis in the trait space which corresponds to the rated property. A separate axis is located for each different property. These techniques also provide a quantitative estimate of the degree to which each rated property does, in fact, satisfactorily interpret a dimension in the trait space. Thus, a formulation of the dimensions of personality perception which may start with the intuitions of the investigator can be validated empirically. Linear and nonlinear multiple-regression methods are available to locate axes in the trait space which correspond to the rated properties of the traits. For each property, the linear method locates an axis in the configuration such that the linear correlation between the external property ratings and the values of the traits along this axis is maximized. Any rotation of the axis would decrease this correlation. The linear method is based on linear multiple regression so than an ordinary multiple R provides a quantitative estimate of the degree to which a property actually corresponds to a dimension in the trait space. The R is calculated by using the property value as the dependent variable and arbitrary orthogonal coordinates of the traits in the configuration as the independent variables. For a two-dimensional solution, for example, the independent variables are the arbitrary x and y coordinates by which the program specifies the trait configuration. Solutions of higher dimensionality are also each represented by the program with a set of arbitrary orthogonal coordinates, and these coordinates serve as the independent variables in the calculation of R's for each of these spaces. The axis for each property is plotted by using the regression weights obtained from the multiple regression as direction numbers of a line through the origin. Carroll and Chang (1966) have developed a general nonlinear method for the interpretation of multidimensional configurations. The

247

IMPLICIT PERSONALITY THEORY

Carroll-Chang method locates an axis in the trait space for each property such that some continuous function, not necessarily linear or even monotonic, relates the values of the traits along this axis to the external ratings. In effect, their method optimizes a nonlinear “correlation” between the external ratings and the axis values in the trait space.

III. Laboratory Analogs of Trait Inference and Personality Description

A. TRAIT-INFERENCE APPROACH In the typical trait-inference study, the investigator presents his subject with personality trait names as descriptive of some (hypothetical) individual and then asks him to make inferences about the presence of other traits in that individual. The subject usually makes these inferences on one or more rating scales. The scales may be similar in format to that used to obtain similarities judgments; for example, the subject might be instructed as follows: A person is warm. How likely would he be:

Very likely generous intelligent

Neutral

Very unlikely ? ?

Alternatively, the scales may resemble the semantic differential; for example, the subject might be instructed as follows: A person is warm. To what extent is he likely to be: generous etingy? intelligent unintelligat?

The choice of a measure of relatedness between any two given traits (or trait scales) depends to some extent on which of the two formats is used. The type of measure in turn influences the choice of an appropriate psychometric technique. Multidimensional scaling is typically applied to the direct judgments of similarity or to a profile distance measure, whereas factor analysis is usually chosen to extract structures from inference data when the measure is correlational (crossproducts, covariances, correlations). Three different studies, each illustrative of a different combination of measures and methods, will be summarized here. Also, the substantive results from these trait-inference studies, in combination with the results from the personality description approach, lead to several general conclusions which will be summarized in Section 111, C.

248

SEYMOUR ROSENBERG AND ANDREA SEDLAK

I, Hays’ Trait Similarity Study In a pioneering, albeit small-scale study in this area, Hays (1958) used the similarities format with the following eight traits: warm, cold, dominant, submissive, intelligent, stupid, generous, stingy. Each subject was asked to judge the likelihood that trait x is present in an individual given that he has trait y, and vice versa, for all possible pairs of the eight traits. The resulting 56 judgments were pooled across subjects and then analyzed with a fully nonmetric multidimensional technique. Hays found the following two rank order dimensions: Dim. I: warm generous submissive stupid intelligent dominant stingy cold; Dim, 11: intelligent dominant cold generous warm submissive stingy stupid. He did not name the two dimensions because of the small number of traits involved, but he did note from them the presence of evaluation in the subjects’ judgments of similarity. 2. Osgood-Ware Personality Diferential In a more extensive study, C. Osgood and E. Ware (Osgood, 1962) constructed two sets of 40 bipolar scales using personality trait names selected from the Allport and Odbert (1936) list of 18,000 trait names. Subjects rated each scale against every other in a format akin to the semantic differential. According to Osgood, factor analyses of the two samples of “scale-on-scale judgments yielded such large evaluative factors (58%and 69%of the total variance, respectively), that little else could be determined ( p. 25) .” There are certain drawbacks in scaling trait terms as linked pairs as Osgood and Ware apparently did rather than scaling single terms as Hays (1958) did. One problem is that the grammatical opposites which form each pair and which anchor the end points of each scale may not correspond to psychological opposites ( Green and Goldfried, 1965). This problem was encountered by Wisher (1960) in his reanalysis of the Asch (1946) studies. It would be preferable to determine empirically and not to decide a priori the extent to which two trait terms are at the extremes of some dimension. A second problem is that the psychological distance between polar words may differ for different pairs; these distances can and probably should be determined empirically, These and other problems associated with the use of the semantic differential format in personality perception research have been summarized by Jackson (1962).

IMPLICIT PERSONALITY THEORY

249

3. Peabody’s Anulysis of Individual T e r m - a n d a Reanalysis Peabody (1967) used the semantic differential format but, in contrast to Osgood and Ware, did a factor analysis of individual trait names, that is, the concepts (as well as of the bipolar inference scales). Peabody’s results differed from those of Osgood and Ware in that he found no evidence of an evaluative dimension in his factor analysis of the trait names. Results from the factor analysis suggested that general descriptive dimensions such as “ego-control vs. impulse-expression,” “decidedness vs. undecidedness,” and “self-assertiveness us. unassertiveness” could account for trait inferences that are seemingly based on evaluation. However, an artifact was discovered in the particuhr way Peabody used the correlation coefficient as a measure of trait relatedness (Rosenberg & Olshan, 1970). This artifact accounts for the absence of an evaluative dimension in his factor analysis. A close examination of Peabody’s factor analysis revealed the following. Each bipolar scale on which the trait concepts were rated contained an evaluative component, that is, a favorable trait at one end and an unfavorable trait at the other end ( for example, timid-bold, frivolous-serious, lethargic-alert, etc.). Since all the scales were oriented the same way for scoring (7step scales with the unfavorable end being scored as l ) , most if not all the information about evaluative similarity between any two traits is contained in the difference between their means over the scales. The product-moment correlation of the average ratings across the scales, which Peabody used in his factor analysis, discards the differences between the means and hence the information about the evaluative similarity between the trait concepts. Even if the scales were not oriented in the same way, the use of the correlation coefficient is questionable, since the size of the r can be fortuitously altered by changing the scoring orientation of some (but not all) of the scales. One way to have avoided this loss of information is by the me of cross-products instead of correlations in the factor analysis. Another way to avoid any loss of information about the evaluative (or any other) dimension is the use of a “profile distance” measure (Cronbach & Gleser, 1953; Osgood & Suci, 1952). A reanalysis by Rosenberg and Olshan (1970) of Peabody’s trait-inference data with a profile distance measure in combination with multidimensional scaling ( Kruskal, 1967) revealed the presence of a strong evaluative dimension in his data. Verification of the presence of an evaluative dimension was ’ done empirically using linear multiple regression. Ratings of each of the traits

250

SEYMOUR ROSENBERC

A N D ANDREA SEDLAK

TABLE I1 MULTIPLE R’s BETWEEN EVALUATIVE PROPERTIES AND THE TRAITSPACES OBTAINEDFROM PEABODY’S (1967) TRAIT-INFERENCE DATA

Property Gmd-btid Intellectual good-bad Social good-bad 0

b 0

=p =p =p

Multidimensional scaling of profile distance

Peabody’s factor analysia

.895a

.354

.887a

.381‘ .474c

.814“

5 .001. 5 .05. 5 .01.

From Rosenberg and Olshan (1970).

were first obtained on three evaluative scales: good-bad, intellectual good-bad, and social good-bad. The good-bad ratings were obtained by asking a number of persons to use a 7-point scale to rate each trait “according to whether a person who exhibited each of the traits would be a good or a bud person.” Intellectual good-bad ratings were obtained by asking persons to rate the traits “according to whether a person who exhibited each of the traits would be good or bad in his intellectual activities”; for social good-bad ratings, the word “social” was substituted for “intellectual.” The multiple correlations for each of the three evaluative properties us. the three-dimensional configuration and Peabody’s ( 1967) original factor analysis (also three factors) are shown in Table 11. The R values for the multidimensional configuration are very large and highly significant, whereas the corresponding R values for the factor analysis are small and, in the case of the general evaluative scale, nonsignificant. These results clearly support the notion that the absence of an evaluative dimension in Peabody’s factor analysis was simply the result of his having scored all the bipolar scales in the same direction with respect to evaluation and then correlating across the scales. It seems likely that any dimension along which a substantial group of traits can be reliably rated and balanced will be artifactually washed out in this way.

B. PERSONALITY DESCRIPTIONS WITH TRAITTERMS SUPPLIED BY THE EXPERIMENTER This section begins with a description of trait sorting-the most recent addition to the set of laboratory techniques in which the person describes himself and/ or other people with trait categories supplied by the investigator. The results of two trait-sorting studies, one involving

IMPLICIT PERSONALITY THEORY

251

the same traits used by Peabody (1967), will be described. Other more traditional methods such as the check list and the rating scale will also be described briefly.

1. Trait Sorting In the trait-sorting method each subject is asked to describe several persons of his own choice by sorting personality trait names into different groups, each group representing a different person. The subject is free to distribute the traits supplied to him among his referent persons in any way he wishes. A “‘miscellaneous”category may be provided for those traits which do not seem to the subject to go together with any others. The subject may be either restricted to assigning each trait to only one referent person or free to use any trait to describe as many different referent persons as he wishes? The trait-sorting method was first used by Rosenberg et al. (1968). In their study, each subject used each trait exactly once in his s ~ r t i n g . ~ They used a profile distance measure as input for multidimensional scaling, and this measure was extracted from the group data as follows: A “disagreement score” was first obtained for each pair of traits by counting the number of subjects who assigned the two traits to two different persons they had in mind. For example, 9 of the 69 subjects assigned warm and intelligent to the same person, so the disagreement score for this pair of traits is 60. Any trait assigned to the miscellaneous category was considered to be in its own category. The disagreement score for any two traits, i and i, is denoted sij. By definition, sij = sji, and sii = 0. The profile distance measure, denoted by 8ij, was defined as follows:

where T is the set of traits. A 8 i j value is essentially equal to twice Sij2 plus a series of other squared terms. That is, when k is i or i, ( s i k - S j k ) * = sij2. The other squared terms in Eq. (1) provide a measure of indirect trait co-occur*For applications of the sorting method to the analysis of semantic structures, see Burton ( 1968), Miller ( 1967, 1969), and Steinberg ( 1967). Interesting applications of sorting, in combination with either clustering or multidimensional scaling, have also been made in studies concerned with the growth of word meaning in children ( Anglin, 1970) and young adults (Friendly & Glucksberg, 1970). ‘In another study (Wing & Nelson, 1972) involving the same traits, subjects were permitted to use each trait as many times as they wished. The differences between this procedure and the one used by Rosenberg et aZ. (1968) in terms of the resulting multidimensional configurations were not substantial.

252

SEYMOUR ROSENBERG AND ANDREA SEDLAK

rence. An indirect trait co-occurrence for any two traits i and j refers to an instance in which i and k co-occur in one description, i and k co-occur in another description, and i and j do not co-occur in either description. Thus the %-measure “contains” the s-measure of direct trait co-occurrence plus a measure of indirect trait co-occurrence. A rough analogy might be helpful for an intuitive understanding of Sij. Consider, for example, a social distance measure between two persons which is inversely related to how often they interact directly with each other and how often they interact with the same other individuals, whether or not they do so on the same occasions. Some of the substantive results of the Rosenberg et 4Z. (1988) study are given in Fig. 4, which is a plot of the two-dimensional configuration (stress = @). It is interesting to note, among other things, that a trait and its antonym are generally located opposite each other in the space. A plot of the one-dimensional configuration is not given because it did not provide a good fit of the input data (stress = 26%). The five labeled axes in Fig. 4 were obtained empirically using

IF??)

GOOD-INTELLECTUAL ACTIVE

H

P I 4 1L . 4 2 7 )

I OSTERN

UNSOCIABLE*

*COLD HUMORLESS

~~~~~~

IRRESPONSIBLE0 OWAS E UNINTELLIGEN?~~

’

CmTICAL

NAIVE

/

IMAGINAJIM*

,

INTELLIGENT

*SHREWD DISCRIMINATING

1 I

P A S S ~ ~ BAD-INTELLECTUAL E

ClENTlFlC PERSISTENTO-. DETERMINED .SKILLFUL’ .&NMISTRIOUS

.HELPFUL

-... .-

aSINCFRF -

SENT MENTAL

tAppy

HUMOROUS 0 GOOD-NATURED. WRM.

BCCIALABLE

I

SOFT

Fig. 4. Two-dimensional configuration of the 80 traits scaled by Rosenberg et al. (1968) showing the best-fitting axis for five properties. Each number in parentheses indicates the multiple correlation between projections on the best-fitting axis and property values.

253

IMPLICIT PERSONALITY THEORY

linear multiple regression. That is, ratings of each of the trait names were obtained on each of five bipolar scales. These scales correspond to the labeled dimensions in Fig. 4. The R values associated with each scale are given in the figure in parentheses. Nonlinear multiple regression (Carroll & Chang, 1966) gave essentially the same results. Five properties are obviously not needed to interpret a twodimensional space. Moreover, there are alternative pairs of properties, all with high R values, which can be used to interpret this space. It is possible, for example, to interpret the two-dimensional space in Fig. 4 with the two general semantic differential factors, good-bad and hard-soft. Also, if we consider the three-dimensional solution, the R value for active-passiue increases to .585 ( p < .OOl), and the angles between the fitted axes for the three semantic differential factors are: hd-soft adwepassive

gmd-bad 83" 92"

hard-soft

76"

Thus, the results from the three-dimensional solution support the presence of the three semantic differential factors in personality perception with each factor more-or-less orthogonal (angles E 90") to the others. An alternative interpretation, at least for the two-dimensional space in Fig. 4, is the use of the two descriptive-evaluative properties, social good-bad and intellectual good-bad. It is interesting to relate this interpretation to Hays' (1958) findings that the extreme traits on one of his rank-order dimensions were warm and cold (social good-bad?), and on the second rank-order dimension they were intelligent and stupid (intellectual good-bad?). These two descriptive-evaluative properties are not orthogonal, however, either in the two-dimensional ( 65" ) or the three-dimensional ( 67" ) spaces. Nevertheless, while orthogonality is convenient, it is not a necessary feature of an acceptable interpretation of a space. Nor are the two sets of interpretations of the two-dimensional solution incompatible. The heuristic distinction between connotative and denotative meaning is relevant here. The interpretation based on hard-soft and good-bad might be thought of as concerned with connotative meaning, whereas the interpretation of the space in terms of social and intellectual desirability is concerned with denotative meaning. A multidimensional analysis of the basic denotative meanings of traits may not result in orthogonal coordinates. Indeed, intellectual traits are likely to be perceived by college students as relevant for social activities. (Unfortunately, in examining the three-dimensional configuration, the authors were not able to specify a denotative property in addition to social and intellectual desirability. )

254

SEYMOUR ROSENBERG AND ANDREA SEDLAK

In another trait-sorting study, Rosenberg and Olshan (1970) used the 60 traits from Peabody’s (1967) “balanced sets” shown in Table 111. Only 6 of these 60 traits overlapped with the 60 used by Rosenberg et al. TABLE I11 PEABODY’S (1967) BALANCED TRAIT SETSUSEDIN ROSENBERG AND OLSHAN (1970) STUDY Set no.

4

Traitsa

1

Cautious Timid

Undiscriminating Rash

2

Self-controlled Inhibited

Uninhibited Impulsive

3

Serious Grim

Gay Frivolous

4

Alert Tense

Relaxed Lethargic

5

Committed Fanatical

Open-minded Noncommital

6

Steady Inflexible

Flexible Vacillating

7

Modest Self-disparaging

Confident Conceited

8

Thrifty St,ingy

Generous Extravagant

9

Skeptical Distrustful

Trusting Gullible

10

Selective Choosy

Tolerant Undiscriminating

11

Firm Severe

Lenient Lax

12

Discreet Secretive

Frank Indiscreet

13

Individualistic Uncooperative

Cooperative Conforming

14

Pragmatic Opportunistic

Idealistic Unrealistic

15

Cultivated Artificial

Natural Naive

In groups of four.

IMPLICIT PERSONALITY THEORY

255

( l W ) ,thus providing some test of the generality of the dimensions found by Rosenberg et al. (1968). In addition, the use of Peabody’s trait list in a sorting task made it possible to compare directly the spaces obtained from the two data-gathering procedures : trait sorting and trait inference. A three-dimensional solution (stress = 108) of the sorting data was obtained by Rosenberg and Olshan ( 1970), using Kruskal’s (1967) program. To answer the question of whether this space and the one obtained from the reanalysis of Peabody’s ( 1967) trait-inference data are substantially the same, a canonical correlation was computed between the two spaces.‘ The resulting canonical r’s were .96, .93, and .90.The magnitude of all three r values indicates that the two datagathering procedures yield approximately the same multidimensional structure. Nine properties including the five from Rosenberg et al. (19sS) were also fitted to each of the two spaces with linear multiple regression. The four additional properties, introverted-extraverted, impulsiueinhibited, dominant-submissive, and decided-undecided, were suggested in part by Peabody’s (1967) descriptive dimensions of perception. The results from the 18 multiple regressions are summarized in Table IV. All the R values are highly significant, but nine properties are obviously not needed to interpret a three-dimensional space. However, a factor analysis of the nine properties yielded only three factors. The properties in Table IV are grouped according to their factor loading, and an examination of this grouping suggests that the three factors correspond to the semantic differential factors, evaluation, potency, and activity. The three properties in a given subset are generally loaded highly on the same (and only that) factor with each subset containing one of the connotative properties. The descriptive properties in each subset serve to amplify the meaning of a connotative property in the context of personality perception.

2. Check Lists and Rating Scales The adjective check list-a variant of trait sorting-has been used in personality research and clinical diagnosis for at least 40 years (e.g., Gough, 1955, 1960; Hartshorne, May, & Maller, 1929; Hartshorne, May, ‘Given two configurations of the same entities, each with n dimensions, the method of canonical correlation first locates an axis in each configuration such that the projectioiis of the entities on the two axes are maximally correlated. Additional pairs of axes are then located in sequence according to the same criterion until n pairs are determined, with the restriction that each successive axis in a configuration be orthogonal to the previous one( s ) (Cooley & Lohnes, 1962).

256

SEYMOUR ROSENBERC AND ANDREA SEDLAK

TABLE IV MULTIPLE R’s BETWEEN NINEPROPERTIES AND THE THREE-DIMENSIONAL CONFIQURATIONS OF PEABODY’S (1967) BALANCED TRAITS Multidimensional scaling of traitsorting data

Multidimensional scaling of traitinference data

Good-bad Intellectual good-bad Social good-bad

.903 .877 .766

.895 .887 .814

Hard-soft Dominant-submissive Decided-undecided

.881 .868 .858

.872 .889 .819

Active-passive Introverted-extraverted Impulsive-inhibited

.798 ,753 .774

.790 .667 .796

Property

Note: All R values in this table are significant beyond the .001 level. From Rosenberg and Olshan (1970).

& Shuttleworth, 1930; Hathaway & Meehl, 1951; LaForge & Suczek,

1955; Leary, 1957; Zuckerman, 1960). In a typical application of the chqck-list method, the respondent is given a list of trait adjectives and asked to check those adjectives descriptive of himself or some person( s ) known to him. Thurstone (1934) was apparently the first investigator to give a structural representation to the co-occurrences of items checked on a list of traits. In his study, each subject was given a list of 60 adjectives and asked to describe one person whom he knew well by underlining every adjective descriptive of that person. The tetrachoric r was used as a measure of co-occurrence of each pair of adjectives by pooling the data across subjects. A factor analysis yielded five factors. The factors were not interpreted, making it difficult to compare Thurstone’s factors with those obtained in the more recent multidimensional studies summarized in this paper. He did, however, report the following five relatively distinct clusters: 1. friendly, congenial, broad-minded, generous, cheerful; 2. patient, calm, faithful, earnest; 3. perseuering, hard-working, systematic; 4. capable, frank, self-reliant, courageous; 5. self-important, sarcastic, haughty, grasping, cynical, quicktempered.

IMPLICIT PERSONALITY THEORY

257

Evaluation appears to be one basis for the clustering, along with some finer denotative distinctions among the positive traits. Aside from Thurstone’s original study, there is little indication in the literature that the check list has been used as a data source for structural studies of personality perception. This paucity is somewhat surprising since the check list is frequently used as a clinical and research tool, is easily administered, is not subject to the criticisms made of bipolar rating scales, and the resulting data readily lend themselves to analyses by a variety of scaling methods. Empirical comparisons of the structures obtained from check-list data with those obtained from other, more laborious, techniques are obviously of methodological interest and are likely to be of substantive interest as well. By contrast with the small literature on the scaling of check-list data, the literature on factor analyses of rating-scale data is voluminous, too voluminous to summarize here. Moreover, the work is aimed for the most part at the development of a scientific theory of personality rather than at the description of the dimensions of personality perception. The use of the rating scale in the construction of a scientific theory of personality has been brought into question, however, by studies which point up the presence, in these ratings, of strong linguistic factors, as well as stereotypic beliefs and other characteristics of the rater (DAndrade, 1965; Levy & Dugan, 1960; Loehlin, 1961, 1967; Mulaik, 1964; Passini & Norman, 1966; Peterson, 1965). That is, the “personality dimensions” obtained from trait ratings may represent not only the actual co-occurrences of certain characteristics in the person being rated but also the various beliefs of the rater and similarities in the perceived subjective meaning of the trait names used to label the scales. Recent evidence strongly suggests, for example, that the dimensions underlying personality trait ratings can be recast into the semantic differential framework ( Hallworth, 1965a, 1965b; Kuusinen, 1969; Peterson, 1965)a reinterpretation which is consistent with the results from the traitinference and trait-sorting studies described in this section. This is not to say that “lay personality theory” may not provide a useful start for the development of n scientific theory of personality. The two systems are likely to be related. The point is that lay judgments of personality contain contaminating effects not of interest to the personality theorist, effects not easily unconfounded. By contrast, when lay judgments are used to study personality perception, the confounding of these various effects is not a serious problem. That is, the lawfulness in perceived co-occurrences of traits is a legitimate and important object of psychological interest, whether such perceptions are valid or not.

258

SEYMOUR ROSENBERG AND ANDREA SEDLAK

Of course, if the degree of co-occurrences of trait names in trait-inference data and in personality descriptions reflects only the degree of synonymity of the terms, such data would be of little value in the study of personality perception (although they would be of linguistic interest). The possibility that synonymity alone accounts for the observed co-occurrences of trait names seems very unlikely, however, from a perusal of the co-occurrence patterns obtained in the various studies discussed in this chapter.

C. CONCLUSIONS The studies summarized in this section point to the existence of at least two, more likely three, dimensions in trait inference and personality description. These studies also suggest that the dimensions may be interpretable in terms of the semantic differential factors, eualuation, potency, and actiuity. Of the three factors, evaluation (good-bad) is typically the strongest and is present in virtually all the trait configurations described in this section. Nevertheless, the presence of a general evaluative dimension in a multidimensional trait space does not mean that the interpretation of the space must of necessity include this general dimension. The dimensions which are chosen, even if they fit the space well are, in a sense, arbitrary. Rosenberg et al. (1968) showed, for example, that a two-dimensional space interpreted in terms of good-bad and hard-soft could be interpreted equally well in terms of the evaluative-descriptive properties, intellectual desirability, and social desirability. The same general point obviously holds for the other two general factors of the semantic differential. The procedure for fitting trait properties is useful in eliminating completely redundant or poorly fitting (low R ' s ) trait properties but provides no basis for choosing from among alternative interpretations, that is, alternative sets of nonoverlapping properties with high R values. However, if it were assumed that one interpretation is more meaningful psychologically than the others, the problem of choosing such an interpretation might be resolved in terms of the perceptual and conceptual processes involved in personality impressions. For example, are a person's spontaneous evaluations of himself and others limited to responses on a global good-bad ( o r like-dislike) dimension, or do evaluative responses also contain descriptive content such as perceived intelligence, sociability, etc.? Unfortunately, very little is known about the underlying processes of personality inipressions. The data that are available do not reveal much about the processes themselves, but do indicate that any single

IMPLICIT PERSONALITY THEORY

259

trait inference is governed by both descriptive and evaluative consistency ( Felipe, 1970; Peabody, 1967, 1970), and that the intellectual-social distinction is a useful one in explicating Asch‘s (1946) original &dings on the role of “central traits” in impression formation (Rosenberg et d., 1968). There are at least two other considerations involved in assessing the utility of the semantic differential, or any other system of trait properties, for describing the basic dimensions of implicit personality theory. One consideration is the generality of the trait structures obtained from laboratory analogs of trait inference and personality description. A basic problem with the laboratory approach described in this section is that the results are based on trait categories fortuitously selected by the investigator and that, as a consequence, may not be representative of the trait categories used by people in everyday life. This problem and others that may be inherent in the laboratory approach are discussed more fully in the next section where studies involving free-response protocols are summarized. A second consideration is the problem of individual differences. The trait structures summarized in this section, and hence the interpretations of them, are all based on aggregated data and are, therefore, not necessarily the best description of the trait structures and properties of the individuals comprising the aggregate. Even if the semantic differential were the best interpretation of individual trait structures, it is unlikely that the same traits would be perceived as, say, good or bad for all individuals. The perceived trait of intelligence that college students value may even have a negative connotation in other subcultures. In fact, almost any seemingly pure descriptive dimension could conceivably be evaluative for some individuals. The same may be true for the so-called nonevaluative properties of the semantic differential: hard traits may be valued as good (in men but not in women?) by some and as bad by others. The individual-difference issue and the multidimensional methods that have been developed to describe individual differences are discussed in Section V.

IV. Naturalistic Personality Descriptions All the investigators cited so far have chosen their trait samples intuitively, fortuitously, or according to criteria which may have little to do with obtaining a representative trait sample. Although the trait samples have varied considerably from one study to another, the question of the generality of an interpreted structure remains unanswered.

260

SEYMOUR ROSENBERG AND ANDREA SEDLAK

Thus, the interpretation of a multidimensional configuration, even when the interpretation is based on such objective procedures as the multiple regression of independently rated properties, may reflect the traitsampling biases of the investigator as well as the psychological dimensions of personality perception, That is, if the traits in the various samples are relatively constant with respect to a particular dimension of personality perception, then that dimension cannot differentially affect subjects’ responses, and so will not be observable in the experimental results, In addition, certain dimensions, while perhaps of only secondary importance in personality perception, may be prominent in a multidimensional configuration simply because they were the only dimensions along which the items in the trait sample varied. The number of trait terms, even the commonly used ones, is too large to deal with all at once in an empirical study, and we still know too little about the structure of personality impressions to select or eliminate terms which are representative of the various regions of the structure. Thus, the task of empirically mapping perceived trait relationships turns into the task of mapping the structure of modest-sized samples of trait terms. One approach to the problem of obtaining a representative but modest-sized sample of traits can be found in Cattell’s (1943, 1945, 1948, 1957) classic work on personality. In an attempt to obtain a trait population which covered every area of personality, Cattell (1943) reduced the personality trait terms collated by Allport and Odbert (1936) from Webster’s New Unabridged International Dictionuy (192.5 edition) to a list of 160 basic terms by instructing two judges, a psychologist and a “student of literature,” to group all synonymous terms together, labeling each synonym group with a key term. An objection to this approach, one that was noted by Cattell, is that such judgments may “trespass from semantic into psychological judgments ( Cattell, 1943, p. 487).” It is difficult to know whether such an objection has been met in practice. Thus, the representativeness of the resultant 160 terms is questionable since they may have been strongly biased by the categories and beliefs in the implicit personality theories of his particular judges. A second approach to the problem of obtaining a representative trait sample, the one we shall describe in this section, is the use of traits and trait co-occurrences generated by subjects themselves in their ad-lib descriptions of the personalities of others. The notion of using a freeresponse approach to the study of person perception is not a new one. Hastorf, Richardson, and Dornbusch ( 1958) advocated this approach because free-response protocols might contain information about cate-

IMPLICIT PERSONALITY THEORY

261

gories and beliefs not captured by a fixed-response format. Beach and Wertheimer ( 1961 ) subsequently published an experiment which showed the feasibility of using a free-response approach. They demonstrated that the coded content of subjects’ descriptions varied systematically with variables such as the sex of the perceiver and perceived and the social status of the two persons with respect to each other. “Dimensions” were extracted from the protocols by a judge who went through them “to obtain a global impression of what dimensions were used (p. 369).” A similar intuitive approach was used in extracting categories and dimensions from children’s free-response descriptions (Dornbusch et al., 1965; Yarrow & Campbell, 1963). Section IV, A which follows describes the first two studies in which dimensions were extracted from free-response protocols by means of multidimensional scaling ( Rosenberg & Sedlak, 1972; Sedlak, 1971). The resulting trait configurations were then interpreted in terms of the nine trait properties introduced in Section 111. Section IV, B describes a clustering analysis of another set of free-response protocols collected by R. A. Jones and Rosenberg (1971). Clustering analysis provides what might be considered a typological (rather than dimensional ) representation of the implicit theory; that is, the nonoverlapping clusters of traits may each be considered to be a basic personality type in the theory. A. MULTIDIMENSIONAL REPRESENTATIONS Rosenberg and Sedlak (1972) instructed college students to describe ten different people of their choice, five of whom they knew well, and five of whom they knew by reputation. They were asked to include at least five adjectives in each description, but could use as many as they considered necessary. The only restriction imposed was the exclusion of any description of physical features. Two editing steps were performed in sequence on the set of descriptions. The first step involved the extraction of trait units. Each item in a subject’s list of traits was recorded verbatim. One-word trait names such as intelligent, perceptive, resourceful, and mature were readily identified. Where a subject used a trait phrase, however, a judgment was frequently necessary as to whether the particular phrase referred to a single trait or could, in fact, be divided into two or more units, each descriptive of a different trait. Phrases which were coded as single units were those in which the trait name was simply preceded or followed by Some modifier or modifying phrase, for example, v e y intelligent, rather cold. Phrases which were written by subjects as single phrases

262

SEYMOUR ROSENBERG AND ANDREA SEDLAK

but coded as more than one trait unit were those which were judged to contain two or more key words, for example, knowledgeable but doesn’t try to impress anyone, inexperienced but open-minded. The second step involved the grouping of trait units (tokens) into trait categories (types). In order to reduce the number of discrete trait tokens to a manageable number of trait types, certain variations in phrasing were collapsed into one trait category which was then identified with a key trait name. Trait units which were judged as referring to the same trait were grouped into one trait category, provided they contained the same basic morpheme. In addition, it was necessary that the meanings be judged as similar. For example, lies and liar were placed together in one category, under the key trait name liar, and snob and snobbish were grouped together in the category snobbish, whereas artful was not grouped with artistic. (Note that this criterion is much more restrictive than Cattell’s method of grouping synonyms regardless of their morphological dissimilarities.) A more extensive description of the guidelines on which the categorization decisions were based is given in Rosenberg and Sedlak ( 1972). Eighty of the most frequently occurring trait categories were selected for multidimensional scaling. The subjects’ data were aggregated to obtain a profile distance measure that reflected the direct and indirect co-occurrences of these traits in the free-response protocols. The measure of profile distance used by Rosenberg and Sedlak (1972) has since been modified and the 80 traits rescaled (Kruskal, 1967) using this new measure. The results of this reanalysis are reported here, and a complete description of the new distance measure is available in Sedlak ( 1971) , Figure 5 is a plot of the stress for each configuration obtained in this rescaling uersuS the number of dimensions in a configuration. The curve in Fig. 5 does not exhibit any “elbow” which might serve as a criterion for choosing a specific dimensionality for interpretation. Also, the stress in this scaling analysis, even at five dimensions, is noticeably above the value of 6%found by Rosenberg et al. (1968) and by Rosenberg and Olshan (1970) in their three-dimensional configurations of 60 traits. Although the stress is relatively high, it is substantially below that of a random input (Stenson & Knoll, 1969), so the possibility that the data lack structure is not plausible. Possible reasons for the relatively high stress in the naturalistic data are discussed later in this section. Ratings of the 80 traits were obtained on each of the nine properties used by Rosenberg and Olshan (1970). The purpose of these ratings, as in previous studies, was to interpret the multidimensional configurations. The five configurations were interpreted by fitting each of the nine

263

IMPLICIT PERSONALITY THEORY

1\

( 49.0)

45

50

I-

40

CL

2

I

3

5

4

DIMENSIONALITY

Fig. 5. Stress of the multidimensional configurations of 80 trait categories, based on a rescaling of the naturalistic co-occurrence data obtained by Rosenberg and Sedlak (1972).

TABLE V MULTIPLE R'S BETWEEN NINEPROPERTIES AND EACHOF TEE FIVEMULTIDIMENSIONAL CONFIGURATIONS OF 80 TRAIT CATEGORIES FOUND IN NATURALISTIC DESCRIPTIONS Dimensionality Property

1

2

3

4

5

Good-baxl Social good-bad Hard-sof t

.777a . 738a .726"

.816" .79@ ,7574

.84P .815O .785'

.84la .803a .798a

.858a .816" .7860

Active-passive Introverted-extraverted Impulsive-inhibited

.096 ,175 ,170

.219 ,3306 .%2c

.218 ,3590 .307

.216 .398c .379c

.233 .403c .432b

Intellectual good-bad Dominanbsubmisive Decided-undecided

.397@ .35w ,148

.439a .453a .244

,439 .506a

.436* .547a .290

.502a ,5360 .300

= b

p

=p

.280

5 .001. 5 .01.

= p 5 .05. Note: Properties are grouped in sets of three based on a factor analysis of the intercorrelations of the ratings (Rosenberg & Sedlak, 1972). c

264

SEYMOUR ROSENBERG AND ANDREA SEDLAK

properties to each configuration with linear multiple regression. The results are summarized in Table V. According to Table V, only good-bad, social good-bad, and hard-soft fit all five spaces, with p < ,001 and with R values comparable in magnitude to those found in previous studies ( R 2 .6). Several of the remaining R values are statistically significant, beyond the .001 level, however. Figure 6 is the two-dimensional configuration and shows the bestfitting axes for the five properties whose R values are significant beyond the .001 level at this dimensionality. Consistent with previous findings, evaluation emerged as a strong dimension, with the good-bad and social good-bad axes virtually coinciding in the space. Although hard-soft is also strongly represented in the naturalistic descriptions, the angle between it and good-bad is only 13"-far short of the near orthogonality obtained by Rosenberg et al. (1968) and usually found in semantic differential studies. Neither does hard-soft differentiate itself from eualzuztion in the spaces of higher dimensionality: the angle between it and good-bad is 21" in the five-dimensional solution. This finding is consistent with the results of the factor analysis of the ratings on the nine properties which showed hard-soft clustering with good-bad and social good-bad. While intellectual good-bad is significant beyond the .001 level in DOMINANT .INOEPENDENT

Fig. 6. Two-dimensional configuration of 80 trait categories showing the best-fitting axis for five properties. Each number in parentheses indicates the axis and multiple correlation between projections on the best-fitting property values. This figure is based on a rescaling of the naturalistic co-occurrence data obtained by Rosenberg and Sedlak (1972).

IMPLICIT PERSONALITY THEORY

265

the two-dimensional configuration, the R is low in an absolute sense ( .439), and the traits most obviously associated with intellectual functioning are not located consistently with respect to the best-fitting axis. Note that while very intelligent, smart, and intelligent are generally in the intellectual good or neutral region, both brilliant and unintelligent lie rather close to each other in the intellectual bad extreme. The activity set-active-passive, introverted-extraverted, and impulsive-inliibited-is very weak in the naturalistic descriptions. The R's are low across dimensionalities, and none of the properties ever reach the .001 level of significance. As a final consideration, the configuration in Fig. 6 may not be capable of being completely interpreted satisfactorily. Although there appear to be no traits which violate common sense in relation to the dominant-submissive axis, and this property seems to be as good an interpretation as any for the upper-left to lower-right dimension, the R is rather low ( .453). In addition, dominunt-submissive is not orthogonal to good-bad; the angle between these two properties is 55" in the twodimension solution and does not change appreciably with increasing dimensionality. The trait configurations based on free-response data are apparently more dScult to interpret than those based on the fixed-response formats described in Section 111. This difficulty may be related to the relatively high stress values obtained with naturalistic data. The following three factors were considered by Rosenberg and Sedlak (1972) to account for the high stress:

1. 80 traits were scaled instead of 60; 2. naturalistic descriptions yield a less stable measure of trait relatedness for a given sample size of traits than either trait sorting or trait inference; 3. the dimensionality of a sample of traits extracted from naturalistic descriptions is simply larger than that of a set typically selected by an investigator. The first possibility was easily checked by randomly selecting 60 of the 80 traits and scaling them in a separate analysis. This possibility was ruled out because the stress values obtained (40.9%,2 9 3 , 22.7%,18.4%, and 15.1X for dimensionalities 1, 2, 3, 4, and 5, respectively) were only slightly lower than those which appear in Fig. 5. If only the second possibility were true, then the dimensionality found with a sorting task using the traits extracted from naturalistic descriptions should be comparable to dimensionalities obtained in previous research with trait sorting and trait inference. If only the third

266

SEYMOUR ROSENBERG AND ANDREA SEDLAX

possibility were true, then the stress values found with a sorting task using these traits should be comparable to those found in the naturalistic description data. If both the second and third possibilities were true, then a scaling of these traits using sorting data should yield intermediate stress values: lower than those associated with the naturalistic scalings because the measure is more stable, but higher than the stress values found in previous research with trait sorting and trait inference because the dimensionality of the naturalistically obtained traits is greater. In order to check on the second and third possibilities, Sedlak (1971) asked subjects to sort 99 of the most frequently occurring traits in the naturalistic descriptions. In order to duplicate other features of the freeresponse situation, subjects were free to use any trait in the sample to describe as many different people as they wished and could add traits not provided in the sample. However, since every subject was required to use each of the traits at least once, the amount of usable co-occurrence data obtained from this sorting procedure was much greater than those obtained from the naturalistic descriptions. The profile distance measure based on the sorting co-occurrences was submitted to the multidimensional scaling program ( Kruskal, 1967). The resulting stress values, which are graphed in Fig. 7, are all comparable in magnitude to those obtained in previous studies using traitsorting data. As in these previous studies, two to three dimensions would appear to give an adequate description of the data. The stress values in 50 45 40 I-

$ 35 u

a

z

2

30 35

20

K

$ 15 10 5 I

2

3

4

5

DIMENSIONALITY

Fig. 7. Stress of the multidimensional configurations of 80 trait categories, based on trait-sorting data obtained by Sedlak ( 1971).

267

IMPLICIT PERSONALITY THEORY

TABLE VI BETWEEN THE CONFIGURATIONS DERIVEDFROM THE CANONICAL CORRELATIONS NATURALISTIC DESCRIPTIONS AND THE CONFIGURATIONS OBTAINED IN THE TRAIT-SORTING STUDY ~~

Dimensionality 1

838.

2

3

4

5

.a811 .378"

.897a .447" .148

,912" ,498.

.907" .526O

.230

,351 .138

.027

.023 4

p

5 .01, according to Bartlett's test (Cooley & Lobes,

1962).

Fig. 7 suggest that the high stress values obtained in the naturalistic scaling were due to an instability of the measure of trait relatedness (the second possibility) rather than to a higher dimensionality of the trait sample obtained from such data (the third possibility). Since the set of traits used in the trait-sorting study was identical to that extracted from the naturalistic protocols, canonical correlations between configurations of corresponding dimensionality were also calculated. The results, which are summarized in Table VI, indicate that there are at most two dimensions common to the naturalistic and trait-sorting data. The R values obtained from fitting the nine properties to the threedimensional configuration based on trait-sorting data are listed in Table VII. The only major increase in the size of R values for the four- and five-dimensional solutions is the R value for decided-undecided: R = .719 ( p < .OOl) in the four-dimensional solution and ,728 ( p < .oOl) in the five-dimensional solution. As in the naturalistic study, evaluation emerged as a strong dimension in trait sorting, and probably represents one of the two dimensions found by the canonical analysis to be common to the two data sets. Dominant-submissive is probably the second dimension more or less common to the two data sets. However, the two fitted properties, goodbad and dominunt-submissive, are more nearly orthogonal to each other in the trait-sorting configuration (84") than in the naturalistic configuration (55" ). (The same is true to a lesser extent for good-bad vs. hurdsoft: 47" in trait-sorting data us. 13" in naturalistic data.) By contrast with the naturalistic configurations, the activity properties-actiue-passiue, introverted-extraverted, and impulsiue-inhibite& are all highly significant in the trait-sorting configurations. However, they

268

SEYMOUR ROSENBERG AND ANDREA SEDLAK

TABLE VII MVLTIPLE R's BETWEEN NINEPROPERTIES AND THE THREE-DIMENSIONAL OBTAINEDFROM .A SORTING OF 80 TRAITCATEQORIES CONFIGURATION R value

Property

-7

b

Good-bad Social good-bad Hard-sof t

,943-7 .948-7 .839-7

Active-passive Introverted-extraverted Impulsive-inhibited

.701" * 812'7 .764'

Intellectual good-bad Dominant-submissive Decided-undecided

.6960 .703" .365b

= p 5.001. = p 5.05.

Note: Properties are grouped in sets of three based on a factor analysis of the intercorrelations of the ratings (Rosenberg & Sedlak, 1972). Table from Sedlak (1971).

do not represent a new orthogonal dimension-the angle between each of the three activity properties and dominant-submissive ranges from 27" to 32". Intellectual good-bad also emerges as a strong property in the trait-sorting data, although it also is not completely orthogonal to either good-bad ( 5 8 " ) or dominant-submissiue (46"). It is not surprising that intellectually desirable traits co-occur with generally desirable traits (see discussion of a similar result in Section 111, B, 1) and with traits associated with a fairly dominant personality. In summary, it appears that trait-sorting data yield results similar to those found with the naturalistic-description data, but provide more clearly interpretable solutions with lower stress values. In both the naturalistic data and the trait-sorting data, the three connotative factors from the semantic differential-good-bad, hard-soft and active-pasdueare not independent of each other according to either the factor analysis of property ratings or the orientation of the axes associated with these properties in the trait spaces. The angles among these three properties in the three-dimensional configuration based on trait sorting are as follows: good-bad

har&ofl active-passwe

47" 63"

hUTCh0ft

70"

This finding contrasts with the results of Rosenberg et al. (1!368),who found nearly orthogonal relations among the three properties in their

IMPLICIT PERSONALITY THEORY

269

three-dimensional space, and Rosenberg and Olshan ( 1970), whose factor analysis produced three distinct subsets of properties, each containing one of the three connotative properties. Taken together, the trait-sorting studies (Rosenberg et al., 1968; Rosenberg &I Olshan, 1970; Sedlak, 1971) indicate that the idealized semantic differential structure, with its three orthogonal dimensions, emerges more clearly when the trait sample is fortuitously selected by the investigator than when the sample consists of terms most frequently used by subjects in their free-response descriptions. The basic issue raised by these findings is the extent to which the semantic differential represents biases entering the investigator’s selection of traits rather than the structure underlying personality perception.

B. TYPOLOGICAL REPRESENTATTONS The use of multidimensional scaling to describe implicit personality theories involves the assumption that perceivers think of personality differences among individuals in terms of gradations in each of a set of basic trait properties. An alternative assumption is that perceivers think of these differences among individuals in terms of more or less distinct personality types. These types may be hierarchically organized as Kelly (1955) has suggested or relatively unrelated to one another. In any case, no empirical work has been reported in the literature in which trait cooccurrences were represented with a typological structure. Hierarchical clustering techniques such as those developed by Johnson (1967) are an excellent way to explore the usefulness of this kind of structure. R. A. Jones and Rosenberg (1971) used solicited free-response descriptions to compare multidimensional with typological structures, In their study, college students were asked to describe five people whom they knew well, using 25 adjectives or phrases for each of the persons described. The coding scheme used was similar to that employed in the naturalistic study by Rosenberg and Sedlak (1972). A profile distance measure, related to the one described by Rosenberg and Sedlak ( 1972), was obtained for the 99 most frequently occurring traits. This profile distance measure was used as input for the multidimensional scaling of the 99 traits. The results of this scaling will not be reported here since the stress values of the configurations were high, comparable in value to those reported in the previous section (see Fig. 5 ) , and since the R’s for the nine properties were not particularly impressive. The clustering procedure (diameter method of Johnson, 1967) used by Jones and Rosenberg accepts the same type of measure for input as does multidimensional scaling. The program partitions the traits into a

270

SEYMOUR ROSENBERG AND ANDREA SEDLAK

set of nonoverlapping clusters on the basis of the input measure. Each successive clustering is produced when an increasingly larger distance is taken as the criterion for putting traits into the same cluster. The clustering is hierarchical which means that a sequence of clusterings (partitionings) is produced, in which each new clustering consists of the union of two clusters from the preceding clustering (partitioning). Thus, the procedure starts with each trait in its own cluster and ends with all traits in the same cluster. Until recently, the interpretation of a hierarchical clustering of objects consisted largely of a subjective analysis by the investigator as to what properties differentiate the objects in the different clusters. This approach may be acceptable when the underlying bases for the partitioning are clear-cut and obvious; for example, when a clustering analysis based on semantic judgments of similarity partitions a set of nouns into two types: those which are names of things and those which are not (Miller, 1969). The interpretation of a clustering of trait names is, in general, not so simple. In a recent study (which will be described in detail in Section V, B ) , Rosenberg and Jones (1972) proposed an interpretative scheme based on a separate analysis of variance of the trait ratings on each of a set of properties. Their method for interpreting a clustering analysis with ANOVA parallels the use of multiple regression to interpret a multidimensional configuration. Consider, for example, the ratings of the 99 traits on the good-bad property. A clustering analysis is interpreted by grouping these ratings so as to correspond with the way the traits are grouped at any given level of clustering. Thus, at each level of clustering one can partition the total sum of squares of the 99 good-bad trait ratings into a sum between clusters and a sum within clusters. The F ratio based on this partitioning reveals the extent to which the property in question is an appropriate interpretation of cluster differences at that particular level of clustering. The optimal level of clustering for a property is the one for which the F ratio has the smallest p value. When the minimum p value occurs at two or more levels, the level with the smallest number of clusters is chosen as the optimal level. A plot of the exact p value associated with each F uersus the clustering level is given in Fig. 8 for good-bad and the other eight properties used in the studies discussed earlier. Since the minimum p for good-bad and social good-bad occurred at the same level of dissimilarity, they are .01 as a criterion, there are six properties plotted together. With p

(dominunt-submissive, actiue-passive, intellectual good-bad, hard-soft, irnpulsive-inhibited, good-bad, and social good-bad) which, at different levels of dissimilarity, are appropriate interpretations of the clusters.

271

IMPLICIT PERSONALITY THEORY

DOMINANT- SUBMISSIVE

INTELLECTUAL GOOD-BAD

--

"/o -:.

-

.OoOI.oOOoOI-

ACTIVE

W

-

-

- PASSIVE

HARD- SOFT

:

: .01I

:\

J

2

.OoOl-

;.000001d

INTROVERrrD-EXTRAVERTED

m m

IMPULSIVE- INHIBITED

1+J----y-d4---+

5

-01

0

.OOOl-

.000001-

- GOOD - BAD

-+

-

DECIDED UNDECIDED

-

: .01 I

-

.0001.OOOoOl-, .80

,

,

.as .so

, 95

1

1

1.00 1.05 1.10

- I

SOCIAL GOOD-BAD I

I

.eo .es .so .a 1.00

,

I

1.05 1.10

Figure 9 is a tree diagram of the levels at which these different properties achieved their minimum p , together with the cluster memberships at each of these six levels of dissimilarity. The optimal clustering level for dominant-submissive is .835. The left-most part of the tree shows the traits which merge when the maximum profile distance between traits in a cluster is .835. For example, the distances among honest, sincere, selzsitive, and understanding are all less than or equal to 335. At this level, a number of clusters contain only one trait, indicating that the distance between such traits and all other traits is greater than .835.The second level in the tree structure indicates the clusters which are produced when the maximum profile distance between traits in a cluster is ,853, the level at which the active-passive

loving narvoua conaldarotr h8Iptul

I

I I

I

C

I 1

1

DE F

A

fun

Domlnonl- Submiaaivr

B Activr-Poaaivr

qUl8t concrmrd amotlonoi honrat sincrra aanaltlvr undrrsiondlng wnrrous

Intallactuol Good-Bod

kind

friandly Intel1W n t warm indrprndant hard workrr

rary-going happy

!:G:!ga

truatworthy Critical dominant

raw atlo

omblka IntdlrctuaI aanual rltrowrt mot8~loliatlo idraliatlc very intaltlgrnt proud drinkar

1

!$%a-mindrd wit-conaciour loqlcal !%c

lovar

IOYOI

:$It

confident lnarcurr atrong 9-11

clron

wolbdrraaad

mu.Ica1 ... -. ... .

wall mannarad talantad s8ntimantol conaarvativa atubborn proctlcal naot thoughtful g8nllr

-

,

op8n m Indrd liatrn8r aorcoslic baouliful lihaoblr atudioua OttroCtIva rroda odvanturoua octivr truating

k”.%n.

olhlalic quick-tam rrrd good-lookf& Intrrratly raap nalbl8 miiJoua tolbotlvr rrlloblr potlrnt

c

witty

young

DISTANCE (8-MEASURE)

Fig. 9. Tree diagram of six levels of clustering of 99 trait categories. Each level is optimal with regard to a trait property in distinguishing the trait clusters at that level. From R. A. Jones and Rosenberg (1971).

IMPLICIT PERSONALITY THEORY

273

analyses yielded p 5 .01. For example, two of the existing clusters, formed at an earlier level of clustering, merge at .853: generous and kind unite with friendly, intelligent, and warm to form a single cluster. A number of clusters remain unchanged in membership at this level. Thus, at each level in the tree diagram the various clusters can be considered different personality “types,” maximally differentiated by the property at that level. A question of interest is how to conceptualize the fact that the different content properties achieve their minimum p values at different levels in the hierarchy. In order to answer this question let us consider the way clusters at one level are differentiated at another level. For example, the hard-soft property achieves its minimum p at level 1.034, and at that level there is a “hard cluster of five traits: critical, dominant, loud, moody, selfish. However, at the next clustering level in Fig. 9 where intelbctual good-bad achieves its minimum p value, these same five traits are differentiated into two clusters: critical, dominant which is “intellectually g o o d and loud, moody, seljislz which is “intellectually bad.” Thus, the one “hard cluster has become differentiated into two clusters-one “hard-intellectual g o o d and the second “hard-intellectual bad.” This suggests that the properties at the higher levels of clustering (i-e., where there are few clusters) are more basic than those at the lower levels (i.e., where there are many clusters). For example, good-bad is a more elementary discrimination in our subjects’ perceptions of others than is intellectual good-bad, which in turn is more elementary to their perceptions of others than is dominant-submissive.Thus, the order of the six significant properties from most basic to least basic is: good-bad, impulsive-inhibited, hard-soft, intellectual good-bad, active-passive, and dominunt-submissive. This hierarchical arrangement of the properties appears to be evidence for Kelly’s (1955) idea that ordinal relationships exist among the constructs used in our perceptions of others: “Thus, man systematizes his constructs by concretely arranging them in hierarchies . . . he builds a system embracing ordinal relationships between constructs for his personal convenience in anticipating events ( Kelly, 1955, p. 58).” While clustering yielded more interpretable solutions than multidimensional scaling in the Jones and Rosenberg study, one cannot assume that implicit personality theories are more hierarchical than they are multidimensional on the basis of this finding alone. In comparing these two structural models of personality perception, interpretability is undoubtedly a criterion of some significance, but any determination of relative interpretability is likely to be a function of the sample of properties used in the analysis. The nine properties used by Jones and Rosen-

274

SEYMOUR ROSENBERG AND ANDREA SEDLAK

berg were those used in studies described earlier. It is questionable whether such a set of properties, derived as they were from the intuitive processes of several investigators, adequately represents all properties of possible relevance to personality perception. A better basis for determining whether the underlying structure is multidimensional or hierarchical can be found in the pattern of the input distance scores themselves. If the objects are in fact best represented by a hierarchical structure, then at each and every level of organization the distances among objects within a cluster are all smaller than those between any two objects, each from a different cluster. Unfortunately, no statistical tests have been developed to estimate the degree to which input distances which contain measurement error deviate significantly from this ideal pattern. Until such significance tests are developed, a choice between a dimensional and a hierarchical structure, based on interpretability alone, is a highly tentative one. Nevertheless, a hierarchical representation of the data is obviously worth obtaining, and except for the two studies (R. A. Jones & Rosenberg, 1971; Rosenberg & Jones, 1972) described in this chapter, the hierarchical model has been strangely neglected in investigations of personality perception.

V. Individual Differences A trait structure based on group data is not necessarily the most accurate description of the structure of the individuals comprising the group. The individuals may differ from the composite in any number of ways. Even when all the individuals share the same dimensions, the dimensions may have different relative weights for the different individuals; for example, some individuals may emphasize dominance in their perceptions of people while others may emphasize extraversion. A more extreme, but not unlikely, possibility is that some individuals use one or more dimensions that other individuals do not, in which case the composite structure contains dimensions not present at all in some of the individual structures. One way of representing individual differences in structure is to run a separate scaling (or clustering) for each individual. Such an approach still leaves the formidable problem of describing in a parsimonious way the similarities and differences among the individual configurations. In response to this problem, psychometricians have developed a number of individual-difference models to represent these similarities and differences (Carroll, 1972; Carroll & Chang, 1970; Horan, 1969; Kruskal, 1968; McGee, 1968; Tucker & Messick, 1963). The Carroll-Chang and Tucker-Messick models will be outlined in Section V, A in some

IMPLICIT PERSONALITY THEORY

275

detail since they each represent a different approach to the problem and are fairly representative of the models developed in this area. Section V, A also describes a number of empirical applications of these two models to personality perception. As we shall see, these applications have revealed certain reliable differences among individuals in the way they perceive people. However, investigators have generally not been as successful in finding personality or biographical correlates of the perceiver that account for these differences in perception. The difEculty in finding perceiver correlates may be due, in part, to a reliance on classical personality variables. The problem with these personality measures may be somewhat analogous to that of the fixedformat methods described in Section 111. That is, standard personality tests may not tap those features of a perceiver’s life and personality that are most relevant to his particular view of people. Nor have any of the applications of the individual-difference models involved the use of free-response descriptions. These models are, of course, applicable to the analysis of free-response descriptions. Such analyses were not attempted on the college-student samples (i.e,, R. A. Jones & Rosenberg, 1971; Rosenberg & Sedlak, 1972), however, because meaningful solutions require a much larger corpus of free-response descriptions than an individual can give in an hour or two of experimental time. Adequate-sized corpora of free-response descriptions are available, however, from a variety of writers, politicians, therapy patients, etc. Section V, B describes an application of clustering and scaling to Theodore Dreiser’s descriptions of a large number of persons he knew (Rosenberg & Jones, 1972). The resulting trait structures were then accounted for informally in terms of certain biographical facts about Dreiser’s life and preoccupations. This type of account is admittedly more subjective than one based on conventional methods of correlating trait structures with perceiver variables. Nevertheless, the striking connections between Dreiser‘s life and the trait structures obtained from his descriptions of people demonstrate the need to analyze particular individuals in this way-at least given the present state of the field. The study of Dreiser has also provided leads for a number of studies which can make use of the individual-difference models to be discussed next.

A. MULTIDIMENSIONAL MODELSOF INDIVIDUAL DIFFERENCES One of the first investigators to propose a multidimensional structure for representing individual differences in perception was Cronbach ( 1955, 1958) in his analytic treatment of social perception scores. His proposal, which is not actually based on multidimensional scaling, is that an indi-

276

SEYMOUR ROSENBERC AND ANDREA SEDLAK

vidual’s descriptions of others on a set of rating scales (preferably uncorrelated) be thought of as defining a distribution of points in a multivariate space. An individual is then represented in terms of the means, variances, and covariances of his ratings in this space. Cronbach‘s analysis has had an important impact on the study of perceptual accuracy because it pointed up a number of artifacts that had been introduced by researchers in their measurement of individual differences in accuracy. Nevertheless, Cronbach‘s approach has not been used extensively to represent individual differences in personality perception. One of the basic limitations of his approach is that it is applicable primarily to dimensions selected in advance by the investigator. Coincident in time with Cronbach‘s work were the first applications of multidimensional scaling techniques to personality perception (Hays, 1958; Jackson et al., 1957). These applications demonstrated that the investigator need not specify any perceptual dimensions in advance. Moreover, the individual-difference models of multidimensional scaling were also being developed in response to the need in a number of diverse areas, to systematically represent differences among individual structures. All multidimensional scaling methods for individual differences start with a set of data similar to that illustrated by the hypothetical example in Table VIII. In general, for n stimuli and N subjects the matrix will have n(n - 1)/2 rows for the stimulus pairs and N columns for the individuals. (The columns could also represent subject types or subgroups.) The entries in the matrix may be based on any of the measures of trait relatedness described in this chapter. The ideas underlying the Tucker and Messick (1983) model are quite simple. Cross-products, covariances, or correlations are first obtained for each pair of individuals across the set of stimulus pairs; for example, a correlation is computed across the rows in Table VIII for each pair of columns. The resulting values are then factored to obtain a spatial configuration of the individuals. The individuals are then grouped, either intuitively or by some clustering scheme and an “idealized” individual is selected to represent each group. Each cluster is called a “point of view.” A multidimensional scaling is performed separately for each point of view so selected, and the resulting stimulus configurations are systematically rotated in order to gauge the similarity among the configurations. As an adjunct to the Tucker-Messick method, characteristics ( demographic, personality, etc. ) of the individuals may be related to the spatial configuration of the individuals by means of canonical correlations between these characteristics and the spatial configuration. Pedersen (1982,1965) selected a sample of 50 traits and applied

277

IMPLICIT PERSONALITY THEORY

TABLE VIII AN EXAMPLE OF A DATA MATRIXFOR INDIVIDUA~DIFFERENCES SCALING: PAIRWISE DISSIMILARITIES FOR SEVENSTIMULI FROM EACHOF SIX JUDGES Judges ~~~

Stimulus pairs

1

2

3

4

5

6

6 13 12 13 6 14 13 13 14 2 14 3 14 13 15 13 12 15 14 3 15

1 2 2 3 1 4 3 2 4 1 4 1 4 3 4 3 2 3 3 1 4

6 11 9 10 4 10 8 9 12 2 10 3 12 9 10 10 8 11 11 2 11

4 12 10 13 6 13 12 11 14 5 12 5 13 11 14 12 9 14 12 5 13

3 9 8 8 5 9 8 8 9 4 8 2 9 8 9 9 8 9 8 3 9

4 9

7 6 3 9 6 8 9 4 6 4 8 8 7 5 5 8 6 5 9

the Tucker-Messick model to judgments of their similarity. He found three points of view, one of which was identified with authoritarian judges (California F Scale). A variety of other personality measures and biographical variables were also explored, but with little or no success in accounting for the different points of view. Walters and Jackson (1966) selected 30 traits from Asch's (1946) study and applied the TuckerMessick model to judgments of their co-occurrence. They found two points of view but could find no relation between these points of view and measures of perceiver intelligence and cognitive complexity. A study by Messick and Kogan (1966), yielded a number of marginally significant correlations between 13 different points of view and a large battery of cognitive and personality tests.5 The Carroll and Chang (1970) model, also called INDSCAL, assumes that all individuals in the data matrix share a common space, 'The Tucker-Messick model has also been applied to the analysis of individual differences in connotative meaning (Wiggins & Fishbein, 1969), in the social desirability of personality test items (Wiggins, 1966), and in the perception of political figures (Tucker & Messick, 1963).

278

SEYMOUR ROSENBERG AND ANDREA SEDLAK

called the “group space,” but that they may weight the dimensions of this space idiosyncratically. The model allows for the possibility that some individuals have zero weights for one or more dimensions. In effect, the space for an individual is obtained by stretching and shrinking the dimensions in the group space in proportion to the relative salience for the particular individual of the various dimensions. Figure 10 shows a hypothetical group space of two dimensions and two individual spaces obtained by differential stretching and shrinking of the two dimensions. A major advantage of INDSCAL over the Tucker-Messick approach is that INDSCAL provides a direct and parsimonious comparison among individual stimulus spaces. Another feature of INDSCAL is that the dimensions of the group space are uniquely oriented with respect to the scaled objects so that if the underlying model is correct, the dimensions found by this method should lend themselves directly to psychological interpretation without further rotation. Sherman (1972) selected a sample of 20 traits from the Rosenberg and Sedlak (1972) study and applied the INDSCAL model to judgments of their similarity. The analysis yielded a group space of four dimensions which he labeled openness, commitment, efectability, and dominance. Sherman distinguished five types of judges according to the patterns of weights they gave to these four dimensions. No personality variables were reported to account for the differences among the five types of judges. L. E. Jones and Young (1971), however, describe an application of INDSCAL in which the relative salience of certain perceptual dimensions was found to be closely related to independent information about the perceivers. They asked the faculty and graduate-student members of the Thurstone Psychometric Laboratory to rate all possible pairs of the Laboratory members according to their similarity, taking into account

GROUP STIMULUS SPACE

STIMULUS SPACE FOR SUBJECT I STIMULUS SPACE FOR SUBJECT 2

Fig. 10. Hypothetical example of a group stimulus space and two individual spaces obtained by the relative stretching and shrinking of the two dimensions in the group space.

IMPLICIT PERSONALITY THEORY

279

“whatever characteristics of the individuals that are relevant for making similarity judgments.” Jones and Young found that members of the Laboratory were judged to vary on three dimensions identified as status, professional interests, and political persuasion. For faculty members status was more salient than professional interests, while for graduate . students the reverse was true. Wish, Deutsch, and Biener (1970) also describe an application of INDSCAL to nation perception in which an interesting perceiver correlate was found to account for individual differences in the weighting of certain dimensions. They found the following four dimensions from the similarity ratings of 21 countries: political alignment and ideology, economic development, geography and population, and culture and race. Judges identified as “hawks” gave greater weight to political alignment and ideology than to economic development, while the reverse was true of “doves.” The studies by L. E. Jones and Young ( 1971) and Wish et al. (1970) point up the value of a rationale, when one is available, for selecting perceiver correlates. B. STUDYOF INDIVIDUALS: AN EXAMPLE The analysis of Theodore Dreiser is based on the physical and psychological trait categories that he used most frequently in A Galkry of Women (1929). This book is a collection of “fictionalized” verbal sketches of 15 different women known by Dreiser, and the focus is on character description rather than story-telling-an ideal emphasis from our standpoint. In all, Dreiser described 241 different characters in Gallery with at least one stable trait. The large number of characters made it possible to calculate relatively stable measures of co-occurrence among the most frequently occurring trait categories, and the diversity among the characters provided some assurance that these categories and their co-occurrences were representative of the elements and relations in Dreiser’s implicit personality theory. The guidelines for extracting trait terms (tokens) from Galley and for forming trait categories (types) are similar to those summarized in Section IV, A for coding the free-response descriptions solicited from college students (Rosenberg & Sedlak, 1972). Table IX is a list of the 99 most frequently occurring trait categories, identified by their key trait names, together with their frequency in all 15 stories. The s-measure in this study is somewhat simpler than those previously used for naturalistic descriptions, primarily because all the descriptions were attributed to one perceiver ( Dreiser) . The exact defini-

280

SEYMOUR ROSENBERG AND ANDREA SEDLAK

TABLE I X

THE99 MOSTFREQUENTLY OCCURRINO TRAITCATEGORIES IN A Gallmu of Wonen Trait category

Trait category

Frequency

Frequency

~~~~~~

Young Beautiful Attractive Charming, dreamer Poetic Interesting, worker Artistic Gay

Practical, romantic, writer Conventional, girl Free, strong Woman Unhappy Intellectual, radical, sensitive, sensual Kind, cold Vigorous Able, drinker, generous, troubled Colorful, graceful, intelligent, poor, r e d , religious, studiea, tall Fool, good-looking, literary, Pagan

100 67 44 41 39 32 29 26

25 24 23 22 20 19 18 17

Lspirant, determined, good, had means, sad, society person, not strong, tasteful unbitious, careful, defiant, different, erratic, genial, happy, lonely, man, varietistic :lever, genius, indifferent, Irish, manager, nice, old, pale, quiet, restless, serious, shrewd, sincere, successful, suffering, sympathetic, thin, understanding unerican, communist, crazy, critical, emotional, enthusiastic, fearful, fighter, forceful, great, handsome, hard, lovely, bad money, painter, physically alluring, playful, repressed, reserved, skill, sophisticated

13

12

11

10

16

15 14

tion of s is available in Rosenberg and Jones (1972). The &measure is a profile distance measure derived from the s-measure in the same way as in all the other work described in this chapter. The two measures (s and 8 ) of perceived trait relatedness were analyzed with both clustering and multidimensional scaling. In order to interpret the clustering and scaling results, ratings by college students of the 99 traits in Table IX were obtained on 12 properties, nine of which the reader is familiar with from previous sections (Table IV) . The three additional rated properties are interesting-dull, conform-does not conform, and young-old.A thirteenth property of the traits termed male-female, was obtained by measuring the proportion of times Dreiser actually used a trait to describe a female character. The tree diagram in Fig. 11 shows three of the clusterings generated by the analysis of the s-measurea using Johnson's (1967) diameter 'Since the clustering results from the s- and 8-measures were quite similar, only the results from the s-measure are described here.

IMPLICIT PERSONALITY THEORY

281

method. The clustering program actually produced a larger number of clusterings than the three diagrammed in Fig. 11. The three clusterings in Fig. 11 were selected on the basis of statistical analyses designed to discover which trait clusterings in the hierarchical sequence have the greatest psychological interpretability in terms of the various trait properties measured in this study. The analyses consist of a separate series of ANOVA calculations for each trait property, as described in Section IV, B. According to these statistical analyses, hurd-soft, male-female, and conforms-does not conform are the only measured properties useable for interpreting the clustering analysis. Also according to these analyses, the maximum differentiation among the traits in terms of these three properties occurs at the .756, .810, and 826 levels respectively, and these are the clustering levels diagrammed in Fig. 11. The positive results with three of the properties do not necessarily mean that the trait clusters consist exclusively of either hard or soft traits, either male or female traits, and either conforming or nonconforming traits. The average values of the traits for the different clusters are distributed over a wide range. The clusters with the smallest and largest values on each of the three properties are listed in Table X. These clusters may represent extreme “personality types” in Dreiser’s perception of people. A discussion of their content, particularly in connection with certain biographical ‘information about Dreiser, will be deferred until after the scaling results are presented. Kruskal’s (1967) program was used for the multidimensional scaling of the Dreiser data. The results from the stress measure suggested that the &measure behaves more like a distance measure than the s-measure does, in that the &measure yields configurations with less stress for all dimensionalities-a finding which corroborates a previously reported comparison between the two measures ( Rosenberg et al., 1968).Also, the spaces based on 6 were found to be more interpretable than those based on s-interpretable in terms of the statistical fitting of the various trait properties. For these reasons and to simplify the presentation, the discussion which follows is based exclusively on scaling results obtained from the S-measure. Figure 12 shows the two-dimensional configuration. This solution, with its relatively high stress (+28%),is likely to juxtapose a number of traits which are further apart in the input data-and which undoubtedly do move away from one another in spaces of higher dimensionality. Nevertheless, the general configuration in Fig. 12 is quite interpretable. In fact, the same properties that proved significant in distinguishing the trait clusters are the most significant properties in the two-dimensional space.

282

SEYMOUR ROSENBERG AND ANDREA SEDLAK

HARDSOFT

-

MALE CONFORMFEMALE NOT CONFORM

par

RELIGIOUS

PNTYfE GOOD STRONG

6'8

BSPt PANL'

-

I

T

I

HAD MEANS

'AM SNUEEt "4f8 l E R IC AN NICE MAN TALL ABLE A Y 81T IOU S

YANAGER GENIAL PRACTICAL SHREWD SKILL

$i%iff

I

PERSON 2 T ST FUL PALE THIN OE TERYINED QUIET

DEFIANT LOVELY RESERVE rlN DE RS TA NDlNG INDlFF E R E N 1 HARD RESTLESS

'

I

Fig. 11. Tree diagram of three levels of clustering of the 99 traits used most frequently in A Gollery of Women. Each level is optimal with regard to a trait property in distinguishing the trait clusters at that level. From Rosenberg and Jones ( 1972).

Interpretations of the three-, four-, and five-dimensional configurations can be summarized as follows: 1. Sex and conformity are highly significant and more or less orthogonal in the various spaces. 2. The angle between hard-soft and mab-femab increases from 10" in the two-dimensional solution to 67" in the four-dimensional solution. Dreiser apparently thought hardness to be related to sex but not redun-

IMPLICIT PERSONALITY THEORY

283

DIFFERENT FOOL READS SENSITIVE HA0 MONEY EMOTIONAL PLAYFUL

SYMPATHETIC VAR IE T IST I C

i[K#p

NG

CCUVENTIONAL STRONG

ASIIlilFJtVE

ElRL C L V?G& OUS

IRISH WOMAN

-1

IN1 R T l N G RO&lffIC PO€ TIC INTELLECTUAL WRITER DRINKER P IN1 R %NS&L

SERIO

SOP~~I#?ICATEO

"ldA GTIFUL

I)

756

DISTANCE

810 826

(s - M E A S U R E )

Fig. 11 (continued).

dant with it. Thus, the impression from the two-dimensional solution (Fig. 12) is somewhat misleading. 3. The idealized semantic differential structure ( evaluation, potency, and activity as orthogonal dimensions) does not emerge in any solution. However, an evaluative cluster does emerge in the four-dimensional solution: good-bad, interesting-dull, social good-bad. The acute angles between any of these properties and sex or conformity vary considerably: 55"-90". Nevertheless, the evaluative cluster does add interpretability to the four-dimensional space. 4. Interpretability drops sharply in the five-dimensional solution, suggesting that error is being "fitted at this point by the scaling program, and that the four-dimensional configuration is the most adequate repre-

284

SEYMOUR ROSENBERG AND ANDREA SEDLAK

TABLE X TRAIT CLUSTERS WITH THE SMALLEST AND LARQEST AVERAQEVALUESON THE THREE PROPERTIES: Hard-Soft ; Male-Female; Confoms-Does Not Conform Hard Cold, vigorous Fighter, forceful, pagan Indifferent, hard, restless Aspirant, critical

Soft (1.5) Emotional, playful, sympathetic, (1.7) varietistic (2.0) Good, kind, not strong (2.5) Painter, sensual

Male Man, tall Good-looking, careful, handsome, manager Poetic, intellectual, writer Crazy, great, erratic, eincere

Conforms Shrewd, skill Aspirant, critical Able, ambitious, good-looking, careful, handsome, manager American, nice Serious, sophisticated

(5.3) (5.7)

(6.0)

Female (18)

(24) (30) (32)

Attractive, charming, girl, cold, vigorous Defiant, lovely, reserved, understanding Pale, thin Painter, sensual

(80)

(81) (81)

(86)

Does not conform (2.0) Crazy, great, erratic, sincere (2.0) Genius, old, poor Communist, radical, troubled, (2.1) worker (2.5) Dreamer, free, lonely, sad (2.5) Fighter, forceful, pagan

(5.0) (5.0) (5.0) (5.0)

(5.7)

Note: Number in parentheses is the average value of the trait cluster on the property. Hard-soft and wnfoms-doea not wnjorm can range from 1 through 7, and malefemale can range from 0 through 100.

sentation of the input data. The stress for the four-dimensional configuration is ~ 1 4 % . The presence of sex as a highly significant property both in the clustering and scaling solutions appears to be explicable in terms of one of the most salient features of Dreiser’s life-that of his deep involvement with women. The traits typically ascribed to men are distinctly different from those ascribed to women. There are at least two distinguishable aspects in his relationships with women, one having to do with women as sexual partners and one having to do with women as intellectual companions. His sexual involvement with women is clearly reflected in certain of the traits that are associated primarily with women, including attractive, beautiful, graceful, lovely, physically alluring, and sensuul. ( A number of parallel terms were associated with men: good-looking,handsome.) But Dreiser also deviated noticeably from the contemporary stereotype of the “male chauvinist.” He did not characterize women as more con-

IMPLICIT PERSONALITY THEORY

285

DOES NOT CONFORM \UTW

Fig. 12. Two-dimensional configuration of the 99 traits used most frequently in A Gallery of Women. Also shown is the best-fitting axis for each of three trait properties. Each number in parentheses is the multiple correlation between projections on the axis and property values. From Rosenberg and Jones (1972).

forming than men, for example: sex and conformity are almost orthogonal in the scaling solutions. Dreiser himself was drawn to “attractive, unconventional women” (his own words). He also liked his attractive, unconventional women to be “clever” and depended on women with business acumen and intellectual strength. He employed women as his literary agents, and he often entrusted the extensive editing of the first drafts of his verbose manuscripts to his female companions. In our data, this aspect of his relation to women shows up in such female traits as reads, defiant, intelligent, cold, and clever. But he reserved for men the traits great, poetic, sincere, genius. The significance of conformity in the clustering and scaling appears to be related to a second major theme and also a major conflict in Dreiser’s life. From the very beginning of his literary career, Dreiser fought with the forces of convention in society. His first novel, Sister Carrie, was accepted for publication by Frank Doubleday who then tried to cancel the agreement because Mrs. Doubleday found the manuscript offensive. The troubles with prudery and convention continued throughout Dreiser’s life. He came to regard himself as the force for reaIism in

286

SEYMOUR ROSENBERG AND ANDREA SEDLAK

American letters and sought out those who shared his nonconformist views. Dreiser also longed for fame and fortune-the fruits of conformity-which he finally achieved with An American Tragedy. It is interesting that the scaling also captured what Dreiser probably saw as frequent accompaniments to nonconformity : sad, sufering, lonely, troubled. There is little doubt that he often experienced these negative feelings as a result of his own nonconformity. In summary, the study of Dreiser has yielded a number of methodological and substantive gains. On the methodological side, the study demonstrates that feasible techniques now exist for obtaining a parsimonious description of an individual’s view of people from his free-response protocols. Developments in the area of computer-assisted content analysis should facilitate the study of other individuals by reducing the labor now involved in extracting character descriptions. On the substantive side, the study of Dreiser indicates that perceived trait dimensions which are highly related to an individual’s life and personality can be uncovered from his free-response descriptions. In particular, it may be the case that basic conflicts-for Dreiser, conformity-in the individual’s personality are projected as preoccupations with traits that represent both sides of the conflict, and hence appear as trait dimensions. The question of whether an individual’s trait space is interpretable in terms of the idealized semantic differential structure is answered with a qualified %o” in the case of Dreiser. Even if an individual’s perceived dimensions can be cast into such a structure, many important details about his categories and beliefs are probably best explicated in terms of the individual’s particular life and personality. General dimensions of person perception are nevertheless necessary in the long-run development of theory, and an affective dimension such as eoalmtion is likely to be part of such a theory. Even with Dreiser, the relative weakness of the evaluative dimension may reflect more about Dreiser’s novelistic style than about his private views of people. Dreiser was certainly evaluative and extremely opinionated in his personal life and conformity, may have been his basic good-bad dimension. The importance of sex in Dreiser’s perception of people also invites speculation about the pervasiveness of this “dimension” in person perception, even if the sexual involvements of most people do not match those of Dreiser. Although the sex of the perceiver has been extensively studied as a variable influencing perception (see reviews by Shrauger & Altrocchi, 1964; Tagiuri, 1969; Warr & Knapper, lW), comparatively little work has been concerned with the influence on perception of the

IMPLICJT PERSONALITY THEORY

287

sex of the person perceived. However, the evidence that is available indicates that the sex of the object person is strongly represented in trait structures, for example, the emergence of sex as a property in adolescents’ perceptions (see Olshan, 1970 in Section VI) and the results of an unpublished replication of the Rosenberg et al. (1968) study with college students, in which sex was fitted as a property of the space. With both adolescents and college students, the intellectually desirable and socially undesirable traits tend to be ascribed to men, while intellectually undesirable and socially desirable traits tend to be ascribed to women. A variable of possibly greater cogency is the sex combinations involved in considering both perceiver and perceived (Beach & Wertheimer, 1961). However, the Dreiser study also underscores the need for research designs and analyses to be sensitive to what constitutes good and bad traits, male and female traits, etc. for different individuals or groups. Finally, it is interesting to compare the two types of structures, the typological us. the multidimensional, in terms of representing Dreiser’s view of people. Both structures were interpretable with the sex and conformity properties, the importance of which has just been discussed. However, a number of properties which were found to be relevant to the interpretation of the four-dimensional solution were not picked up at all as significant for the clustering solutions. The reader is referred back to the end of Section IV, B for a discussion of the issues involved in choosing between a typological and a multidimensional structure. The general conclusion from this discussion is that the choice of one structure over another cannot be definitely made on the basis of interpretability alone.

VI. Personality Perception in Children The early studies in children’s perception of personality were conducted within the framework of Werner’s (1948) organismic theory. This theory views all development as a process of transition from global, undifferentiated states to states of greater differentiation and hierarchical organization. The first study in this framework used subjects between the ages of 10 and 18 (Gollin, 1958). The subjects were shown a silent film depicting a boy in four different social interactions, two involving socially approved behavior, and two involving socially disapproved behavior. They were asked to give their opinions of the boy in writing. Two judges scored the statements according to the degree to which they contained “inferences” and “concepts.” Inferences were defined as

288

SEYMOUR ROSENBERG AND ANDREA SEDLAK

statements which went beyond the observable behavior in the movie by attributing a motive or unobserved reason for the behavior. “Concepts’’ were d e h e d as terms used to account for the diversity, including apparent inconsistency in the boy’s behavior. Gollin found that subjects generated a significantly greater number of both inferences and concepts with increasing age, a result that supported the notion of developmental differentiation and organization. In another study in this framework, Signell (1966) differentiated the notion of cognitive complexity into two distinct types: (1) the degree to which a person discriminates along a given trait dimension (Cronbach‘s variance measure), and (2) the number of different content categories present in the person’s trait vocabulary. In line with organismic theory, she hypothesized greater cognitive complexity of both types with increasing age, as well as a larger number of trait dimensions. Fourth-, sixth-, and ninth-grade subjects were given the Kelly Rep Test in order to obtain a set of object persons and traits. Each subject then used a 7-point scale to rate the object persons on each of the traits he had generated in the Rep Test. She found that older children did have higher scores on the first type of cognitive complexity, but in apparent contradiction to organismic theory, did not have a greater number of either content categories or trait dimensions. It is difficult to account for the discrepancy between Gollin’s (1958) and Signell’s (1968) findings, but one obvious possibility is the difference in the way each investigator operationalized the ambiguous concepts of complexity and differentiation. Instead of focussing on agerelated changes in complexity and differentiation, other investigators have attempted to identrfy the specific trait categories used by children. Two studies, both conducted at summer camps, examined children’s free-response descriptions of each other. In the first of these studies, Yarrow and Campbell (1963) interviewed each of a group of children between the ages of 8 and 13 years and asked them to choose the cabin mate about whom he thought he knew most and to “tell all about him.” Coding of the interviews revealed two strong tendencies in the children’s descriptions: (1) the presence of broad positive and negative judgments; and (2) frequent reference to “sociable” behaviors. The experimenters also grouped the trait categories on an intuitive basis into four major classes of interaction ( which they termed “dimensions”)-aggressive, afiliatiue, msettioe, and submissive. Almost half of the children used only one category in their descriptions, and nearly none used all four categories. In the second study at a summer camp, Dornbusch et al. (1965) also investigated the categories of behaviors used by children in their free-response

IMPLICIT PERSONALITY THEORY

289

descriptions of the behavior of others, but they were primarily concerned with determining the relative influence of the perceiver and the perceived on category use. They found greater overlap in descriptions from the same perceiver of different stimulus persons than in descriptions from different perceivers of the same stimulus persons. The latter descriptions were only slightly more specific than descriptions composed simply of terms that most perceivers used to characterize most stimulus persons. While both Yarrow and Campbell (1963) and Dornbusch et al. (1905) classified the categories they obtained into several major content areas, neither attempted to systematically determine either the dimensionality of the categories obtained or the nature of the underlying dimensions. Signell ( 1966), the reader will recall, investigated the dimensionality of the object persons and the traits she obtained, but did not identify the nature of the dimensions involved. Nor was she interested in the particular traits that were used by children of different ages. A recent doctoral dissertation by Olshan (1970) provides a synthesis of the aims of these previous studies. She determined both the particular traits used by children at different ages and the dimensionality of the trait co-occurrences. Olshan alsd used independently rated properties of the traits to interpret the multidimensional solutions obtained for the different ages. In the first part of Olshan’s (1970) study, third, sixth, and ninth graders were asked to describe a person from each of ten different categories or roles, such as “a boy friend,” “a girl friend,” “my mother,” “my teacher,” “a fireman.” The most frequently occurring traits at each grade level comprised the trait vocabulary of that grade level. In the second part of the study, a new sample of subjects at each grade level sorted the trait vocabulary of that grade level. The data from these sortings were grouped within each grade level, and the aggregated data for each group was scaled, yielding three trait spaces. In all cases the two- and three-dimensional configurations were found to adequately represent the sorting data. Ratings of the traits in each grade’s vocabulary list were obtained on the five properties: good-bad, smart-stupid, many friends-few friends, male-female, and adult-child. The first three properties correspond to the evaluative properties good-bad, intellectual goohbad, and social good-bad, respectively, and the latter two were suggested by an examination of the trait spaces themselves. Each set of traits was rated by children of the corresponding grade level. The relative location in the two-dimensional spaces of the significant

290

SEYMOUR ROSENBERG AND ANDREA SEDLAK

properties ( p 5 .Ol) are shown in Fig. 13. The R values are also given in the figure, but the traits themselves are omitted for simplicity. The R’s in Fig. 13 show that all three evaluative properties correlate highly with each of the three trait spaces. The increase in angle size among these properties with increasing age indicates an interesting developmental trend, suggestive of the organismic notion of differentiation. Figure 13 also shows that the angle between good-bad and many friends-few friends is smaller for all grade levels than the angle between good-bad and smart-stupid, a pattern which is reminiscent of that obtained with college students (see Figs. 4 and 6 ) . Another interesting developmental trend can be found in the interpretation of a second dimension in these spaces. While adult-child is a highly significant dimension underlying third graders’ personality perceptions, this dimension is no longer salient with the beginning of adolescence. On the other hand, preadolescents do not perceive sex as a dimension, whereas adolescents do. The relations of adult-child and MALE

\ SMART 1.79 I

FEW FRIENDS

SMART

FEW FRIENDS

MANY FRIENDS I CHILD

FEMALE f.78)

THIRD GRADERS

SIX GRADERS

SMART (.671

FEW FRIENDS BAD

STUPID

1.87)

’

MANY FRIENDS \FEMALE L561

NINTH GRADERS

Fig. 13. Relations among significant properties ( p I -01) in the twodimensional spaces of third, sixth, and ninth graders. From Olshan (1970).

IMPLICIT PERSONALITY THEORY

291

of sex to the evaluative properties also show the nature of certain stereotypes that children have at various ages. Preadolescents identify adults as “good and children as “bad.” Adolescents identify females as more “sociable,” and males as more “intellectual,” and this cultural stereotype is even stronger in the older adolescents (ninth graders). Evidence from a pilot study indicates the presence of these same sex stereotypes in an undifferentiated sample of college students. In summary, Olshan’s (1970) results do not support organismic theory in that the number of dimensions required to represent the personality description data shows no developmental change. However, the theory is supported by the finding that evaluation becomes differentiated with increasing age. Finally, the nature of the dimensions at different ages changes in psychologically meaningful ways, a finding that is apparently unreIated to the organismic concepts of differentiation and organization.

VII. Some Future Directions The studies described in this chapter pose a variety of questions, a number of which have already been discussed in previous sections. On the methodological side, for example, we have mentioned the need for ( 1) systematically comparing the different data-gathering methods such as similarity judgments, trait inferences, and personality descriptions, ( 2 ) tests that can compare the relative adequacy of multidimensional and hierarchical structures for a given corpus of data, and ( 3) more streamlined methods of coding naturalistic descriptions. There is also the need to (1) provide an adequate conceptual basis for the various measures of trait relatedness, and (2) develop psychometric models that can deal adequately with asymmetric matrices of trait relatedness ( e.g. : Does intelligent imply friendly more than the reverse?). Many of these and other methodological problems obviously fall outside the province of social psychology per se and depend for their solution on developments in psychometrics, linguistics, and computer technology. On the substantive side, there are even more interesting (for social psychologists at least) and diverse questions. To begin with, the work described in this chapter has raised a fundamental question about the validity and usefulness of the semantic differential for personality perception, This three-dimensional system has not provided an adequate interpretation of the trait spaces obtained from naturalistic descriptions. These findings, together with those from laboratory analogs of trait description and trait inference, have led to the conjecture that the semantic differential structure emerges primarily when the trait samples

292

SEYMOUR ROSENBERG AND ANDREA SEDLAK

are fortuitously selected by the investigator (Section IV, B ) . Yet it is difficult to substantiate or refute this conjecture with any particular set of studies. This type of issue is more likely to be resolved from experience with a body of literature as it accumulates. Even if the semantic differential system were adequate, the study of Dreiser has suggested that psychologically more meaningful trait properties can probably be found in the life and personality of the individual. For example, one of the fascinating possibilities is that basic conflicts in an individual’s personality are reflected in one or more of the trait dimensions underlying his perceptions of himself and others. Application of the technique to therapy protocols is likely to throw light on this question as well as others concerning the therapeutic process itself. The analysis of Dreiser also raises the question of the extent to which Dreiser is himself of a type: the talented artist or writer struggling between creative, unconventional expression and the achievement of fame and fortune-perhaps with a sex life that is also unconventional? This question suggests the potential usefulness of analyzing other writers, both men and women. Finally, will another sample of Dreiser’s writings such as his voluminous collection of personal letters yield the same structure as Gallery did? The findings concerning sex as a dimension suggest that the method may be useful for identifying the presence and nature of sex stereotypes in different groups of individuals. How do “feminists” characterize men and women? . . . “male chauvinists”? . . . homosexuals? Under what social conditions (family, peers, social class) do culturally-based sex stereotypes emerge in children? How do adolescent boys and girls differ in the stereotypes they acquire about each other? Finally, there is the need for theory development in this area. The answer to the question about the usefulness of the semantic differential or other properties may eventually emerge as a by-product of the important but as yet apparently intractable questions about the processes underlying personality perceptions. Structural representations and their interpretation with external properties provide very useful summaries of sizable chunks of data, but their psychological utility, as well as that of other research in personality perception, will be best realized in combination with a formal and testable theory. One possible framework for such R theory has been suggested by Warr and Knapper (1968). Short of a well-articulated theory, there are research directions that can make the structures more useful psychologically, as well as contribute to theory. One strategy is to relate these structures to other independently observable phenomena in personality perception. Rosenberg d al. (1%8), for example, were able to rationalize some of Asch‘s

IMPLICIT PERSONALITY THEORY

293

(1946) classic findings in impression formation with the help of the structure they obtained. In a related study, Hamilton and Huffman (1971) tested the generality of certain additive models of impression formation for nonevaluative, as well as for evaluative, dimensions. Cliff (1972) investigated the relation between favorableness ratings in an impression-formation paradigm with the positions of the component adjectives in a multidimensional space. These empirical connections between “structural representations of implicit personality theory” and “impression formation” are likely to be useful to both areas of research. REFERENCES Allport, G. W., 81 Odbert, H. S. Trait names: A psycholexical study. PsychologicaE Monographs, 1936, 47( 1, Whole No. 211). Anglin, J. M. The growth of word meaning. Cambridge, Mass.: MIT Press, 1970. Asch, S. E. Forming impressions of personality. Journal of Abnormal and Social Psychology, 1946, 41, 258-290. Beach, L., & Wertheimer, M. A free-response approach to the study of person cognition. Journal of Abnormal and Social Psychology, 1961, 62, 367-374. Bennett, J. F., & Hays, W. L. Multidimensional unfolding: Determining the dimensionality of ranked preference data. Psychometrika, 1960, 25, 27A3. Burton, M. L. Multidimensional scaling of role terms. Unpublished doctoral dissertation, Stanford University, 1968. Carroll, J. D. Individual differences and multidimensional scaling. In R. N. Shepard, A. K. Romney, & S. Nerlove (Eds.), Multidimensional scaling: Theory and applications in the behavioral sciences. Vol. I. Theory. New York: Seminar Press, 1972. Carroll, J. D., & Chang, J.-J. A general index of nonlinear correlation and its application to the problem of relating physical and psychological dimensions. Paper presented at the meeting of the Psychometric Society, New York, September 1986. Carroll, J. D., & Chang, 1.-J. Analysis of individual differences in multidimensional scaling via an n-way generalization of “Eckart-Young” decomposition. Psychometrika, 1970, 35, 283419. Cattell, R. B. The description of personality: Basic traits resolved into clusters. Journal of Abnormal and Social Psychology, 1943, 38, 476506. Cattell, R. B. The description of personality. Principles and findings in a factor analysis. American Journal of Psychology, 1945, 58, 69-90. Cattell, R. B. Description and measurement of personality. London: Harrap, 1946. Cattell, R. B. Personality and motioation: Structure and measurement. Yonkers, N. Y.: World Book, 1957. Cliff, N. Consistencies among judgments of adjective combination. In A. K. Romney, R. N. Shepard, & S. Nerlove (Eds.), Multidimemional scaling: Theoq and applications in the behavioral sciences. Vol. 11. Applications. New York: Seminar Press, 1972. Cooley, W. W., & Lohnes, P. R. Multivariate procedures for the behavioral sciences. New York: Wiley, 1962. Coombs, C. H. Psychological scaling without a unit of measurement. Psychological Review, 1950, 57, 145-158.

294

SEYMOUR ROSENBERG AND ANDREA SEDLAK

Coombs, C. H. An application of a nonmetric model for multidimensional analysis of similarities, Psychological Reports, 1958, 4, 511418. Coombs, C. H. A theory of data. New York: Wiley, 1964. Cronbach, L. J. Processes affecting scores on “understanding of others” and “assumed similarity.” Psychological Bulletin, 1955, 52, 177-193. Cronbach, L. J. Proposals leading to analytic treatment of social perception scores. In R. Tagiuri & L. Petrullo (Eds.), Person perception and interpersonal behaoior. Stanford: Stanford University Press, 1958. Pp. 353-379. Cronbach, L. J., & Gleser, G. C. Assessing similarity between profiles. Psychological Bulletin, 1953, 50, 456-473. DAndrade, R. G. Trait psychology and componential analysis. American Anthropologist, 1965, 67, 215-228. Dornbusch, S. M., Hastorf, A. H., Richardson, S. A., Muzzy, R. E., & Vreeland, R. S. The perceiver and the perceived: Their relative influence on the categories of interpersonal cognition. Jou~nalof PersonaZity and Social Psychology, 1965, 1, 434-440. Dreiser, T. A Gauery of Women. New York: Boni & Liveright, 1929. Felipe, A. I. Evaluative versus descriptive consistency in trait inferences. Journd of Personality and Social Psychology, 1970, 16, 627-838. Friendly, M. L., & Glucksberg, S. On the description of subcultural lexicons: A multidimensional approach, Journal .of Personality and Social Psychology, 1970, 14, 55-65. Collin, E. S. Organizational characteristics of social judgment: a developmental investigation. Journal of Personality, 1958,52, 39-42. Gough, H. G. Reference handbook for the Cough Adjectiue Check-List. Berkeley: University of California Institute for Personality Assessment and Research, 1955. Gough, H. G. The adjective check list as a personality assessment research technique. Psychological Reports, 1960,6, 107-122. Green, R. F.,& Goldfried, M. R. On the bipolarity of semantic space. Psychologicd Monographs: General and Applied, 1965,79, 1-31. Green, P. E., & Carmone, F. J. M u l t i d i m e n s i d scaling and related techniques in marketing analysis. Boston, Mass.: Allyn & Bacon, 1970. Haagen, C. H. Synonymity, vividness, familiarity, and association value ratings of 400 pairs of common adjectives. Journal of Psychology, 1949, 27, 453-463. Hallworth, H. J. The dimensions of personality among children of school age. British Journal of Mathematical and Statistical Psychology, 1965, 18, 45-56. ( a ) Hallworth, H. J. Dimensions of personality and meaning. British Journal of So& and Clinic01 Psychology, 1965,4, 161-168. ( b ) Hamilton, D. L., & Huffman, L. J. Generality of impression-formation processes for evaluative and nonevaluative judgments. Journal of Personality and Social Psychology, 1971, 20, 200-207. Hartshorne, H., May, M. A,, & Maller, J. B. Studies in the nature of character. Vol. 11. Studies in service and self-control. New York: Macmillan, 1929. Hartshorne, H., May, M. A., & Shuttleworth, F. K. Studies in the nature of character. Vol. 111. Studies in the organization of character. New York: Macmillan,

1930. Hastorf, A, H., Richardson, S. A,, & Dornbusch, S. M. The problem of relevance in the study of person perception. In R. Tagiuri & L. Petrullo (Eds.), Person perception and interpersonal behavior. Stanford: Stanford University Press, 1958. Pp. 54-62.

IMPLICIT PERSONALITY THEORY

295

Hathaway, S. R., & Meehl, P. E. The Minnesota Multiphasic Personality Inventory. Military clinical psychology. Washington, D.C.: U.S. Government Printing Office, 1951. Pp. 71-111. Hays, W. L. An approach to the study of trait implication and trait similarity. In R. Tagiuri & L. Petrullo (Eds.), Person perception and interpersonal behavior. Stanford: Stanford University Press, 1958. Hays, W. L., & Bennett, J. F. Multidimensional unfolding: determining configuration from complete rank ordering of preference data. Psychometrika, 1961, 26, 221-238. Horan, C. B. Multidimensional scaling: Combining observations when individuals have different perceptual structures. Psychometrika, 1969, 34, 139-165. Jackson, D. N. The measurement of perceived personality trait relationships. In N. F. Washburne (Ed.), Decisions, Values and Groups. Vol. 2. New York: Pergamon, 1962. Jackson, D. N., Messick, S. J., & Solley, C. M. A multidimensional scaling approach to the perception of personality. Journal of Psychology, 1957, 44, 311-318. Johnson, S. C. Hierarchical clustering schemes. Psychometrika, 1967, 32, 241-254. Jones, L. E., & Young, F. W. The structure of a social environment: A longitudinal individual differences scaling of the L. L. Thurstone Psychometric Laboratory. Unpublished manuscript, University of North Carolina, 1971. Jones, R. A., & Rosenberg, S. Structural representations of naturalistic descriptions of personality. Unpublished manuscript, Rutgers University, 1971. Kelly, G. A. A theory of personality: The psychology of personal constructs. New York: Norton, 1955. Kruskal, J. B. Multidimensional scaling by optimizing goodness of fit to a nonmetric hypothesis. Psychometrika, 1964, 29, 1-27. ( a ) Kruskal, J. B. Nonmetric multidimensional scaling: A numerical method. Psychometrika, 1964,.29, 115-129. ( b ) Kruskal, J. B. How to use MDSCAL, a multidimensional scaling program. Bell Telephone Laboratories, Murray Hill, New Jersey (Version 3 ) , May 1967. Kruskal, J. B. How to use MDSCAL, a program to do multidimensional scaling and multidimensional unfolding. Bell Telephone Laboratories, Murray Hill, New Jersey, 1968. Kuusinen, J. Affective and denotative structures of personality ratings. Journal of Personality and Social Psychology, 1969, 12, 181-188. LaForge, R., & Suczek, R. F. The interpersonal dimension of personality: 111. An interpersonal check list. Journal of Personality, 1955, 24, 94-112. Leary, T. F. Interpersonal diagnosis of personality. New York: Ronald Press, 1957. Levy, L. H., & Dugan, R. D. A constant error approach to the study of dimensions of social perception. Journal of Abnormal and Social Psychology, 1960, 61, 21-24. Loehlin, J. C. Word meanings and self-descriptions. journal of Abnormal and Social Psychology, 1961, 62, 28-34. Loehlin, J. C. Word meanings and self-descriptions: A replication and extension. Journal of Personality and Social Psychology, 1967, 5, 107-110. McGee, V. E. Multidimensional scaling of N sets of similarity measures: A nonmetric individual differences approach. Multivaride Behavioral Research, 1968, 3, 233-248. Messick, S., & Kogan, N. Personality consistencies in judgment: Dimensions of role constructs. Multivariate Behavioral Research, 1966, 1, 165-175. Miller, G. A. Psycholinguistic approaches to the study of communication. In D.

296

SEYMOUR ROSENBERG A N D ANDREA SEDLAK

L. Arm (Ed.), Journeys in science: Small steps-great

strides. Albuquerque: University of New Mexico Press, 1967. Miller, G. A. A psychological method to investigate verbal concepts. Journal of Mathematical Psychology, 1969, 6, 169-191. Mulaik, S. A. Are personality factors raters’ conceptual factors? Journal of Consulting Psychology, 1964, 28, 508-511. Olshan, K. The multidimensional structure of person perception in children. Unpublished doctoral dissertation, Rutgers University, 1970. Osgood, C. E. Studies on the generality of the affective meaning systems. American Psychologist, 1962, 17, 10-28. Osgood, C. E., & Suci, G. J. A measure of relation determined by both mean differences and profile information. Psychological Bulletin, 1952, 49, 251-262. Passini, F. T., & Norman, W. T. A universal conception of personality structure? Journal of Personulity and Social Psychology, 1966, 4 , 4 4 4 9 . Peabody, D. Trait inferences: Evaluative and descriptive aspects. Journal of Personality and Social Psychology, 1967, 7(4, Whole No. 644). Peabody, D. Evaluative and descriptive aspects in personality perception: A reappraisal, Journal of Personality and Social Psychology, 1970, 6, 639-846. Pedersen, D. M. The measurement of individual dgerences in perceived personality trait relationships and their relation to certain determinants. Unpublished doctoral dissertation, University of Illinois, 1962. Pedersen, D. M. The measurement of individual differences in perceived personalitytrait relationships and their relation to certain determinants. Journal of Social Psychology, 1965, 65, 233-258. Peterson, D. R. Scope and generality of verbally defined personality factors. Psychological Review, 1965, 72, 48-59. Posavac, E. J. Dimensions of trait preferences and personality type. Journal of Personality and Social Psychology, 1971, 19, 274-281. Rosenberg, S., & Jones, R. A. A method for investigating and representing a person’s implicit theory of personality: Theodore Dreiser’s view of People. Journal of Personality and Social Psychology, 1972, in press. Rosenberg, S., Nelson, C., and Vivekananthan, P. S. A multidimensional approach to the structure of personality impressions. J o u m l of Personality and Social Psychology, 1968, 9, 283-294. Rosenberg, S., & Olshan, K. Evaluative and descriptive aspects in personality perception. Journal of Personaltiy and Social Psychology, 1970, 16, 619426. Rosenberg, S., & Sedlak, A. Structural representations of perceived personality trait relationships. In A. K. Romney, R. N. Shepard, & S. Nerlove (Eds.), Multidimensional scaling: Theory and applications in the behavioral sciences. Vol. 11. Applications. New York: Seminar Press, 1972. Sedlak, A. A multidimensional study of the structure of personality descriptions. Unpublished master’s thesis, Rutgers University, 1971. Shepard, R. N. The analysis of proximities: Multidimensional scaling with an unknown distance function. I. Psychometrika, 1982, 27, 125-140. ( a ) Shepard, R. N. The analysis of proximities: Multidimensional scaling with an unknown distance function. 11. Psychometrika, 1962, 27, 219-246. ( b ) Shepard, R. N. Metric structures in ordinal data. Journal of Mathemutical Psychology, 1966, 3, 287315. Sherman, R. Individual differences in perceived trait relationships as a function of dimensional salience. Multiuariate Behavioral Research, 1972, 7, 109-129.

IMPLICIT PERSONALITY THEORY

297

Shrauger, S., & Altrocchi, J. The personality of the perceiver as a factor in person perception. Psychological Bulletin, 1964, 62, 289-308. Signell, K. A. Cognitive complexity in person perception and nation perception: A developmental approach. Journal of Personality, 1966, 34, 517537. Steinberg, D. D. The word sort: An instrument for semantic analysis. Psychonomic Science, 1967, 12, 541542. Stenson, H. H., & Knoll, R. L. Goodness of fit for random rankings in Kruskal’s nonmetric scaling procedure. Psychological Bulletin, 1969, 71, 122-126. Tagiuri, R. Person perception. In G. Lindzey & E. Aronson (Eds.), The handbook of social psychology. (2nd ed.) Vol. 111. Reading, Mass.: Addison Wesley, 1969. Pp. 395-449. Thurstone, L. L. The vectors of mind. Psychological Reuiew, 1934, 41, 1-32. Torgerson, W. S. Multidimensional scaling: I. Theory and method. Psychometrika, 1952, 17, 401419. Torgerson, W. S . Theory and methods of scaling. New York: Wiley, 1958. Torgerson, W. S. Multidimensional scaling of similarity. Psychometrika, 1965, 30, 379-393. Tucker, L. R., & Messick, S. An individual differences model for multidimensional scaling. Psychometrika, 1963, 28, 333-367. Walters, H. A., & Jackson, D. N. Group and individual regularities in trait inference: A multidimensional scaling analysis. Multivariate Behavioral Research, 1966, 1, 145-163. Warr, P. B., & Knapper, C. The perception of people and euents. New York: Wiley, 1968. Werner, H. Comparative psychology of mental development. Chicago: Follett Press, 1948. Wiggins, N. Individual viewpoints of social desirability. Psychological Bulletin, 1966, 66, 68-77. Wiggins, N., & Fishbein, M. Dimensions of semantic space: A problem of individual differences. In J. G. Snider & C. E. Osgood (Eds.), Semantic differential technique. Chicago: Aldine, 1969. Wiggins, N., & Wiggins, J. S. A typological analysis of male preferences for female body types. Multiuariute Behavioral Research, 1969, 4, 89-102. Wing, H., & Nelson, C. E. The perception of personality through trait sorting: A comparison of trait sampling techniques. Multivariate Behavioral Research, 1972, in press. Wish, M., Deutsch, M., & Biener, L. Differences in conceptual structures of nations: An exploratory study. Journal of Personality and Social Psychology, 1970, 3, 361-373. Wishner, J. Reanalysis of “Impressions of personality.” Psychological Reuiew, 1960, 67, 96-112. Yarrow, M. R., & Campbell, J. D. Person perception in children. Merrill-Palmer Quarterly, 1963, 9, 57-72. Zuckerman, M. The development of an affect adjective check list for the measurement of anxiety. j o u m l of Consulting Psychology, 1980, 24, 457462.

This Page Intentionally Left Blank

AUTHOR INDEX Numbers in italics refer to the pages on which the complete references are listed. Becker, G. M., 225, 231 Beecher, H. K., 12, 58 Abelson, R. P., 9, 20, 21, 26, 29, 30, 43, Behringer, R. D., 174, 202 57, 59, 61 Bern, D. J., 4, 6, 9, 10, 12, 13, 15, 18, Aderman, D., 87, 88, 93, 106 23, 24, 25, 27, 30, 31, 32, 33, 52, Alexander, C. N., Jr., 171, 199 55, 57, 58 Alexander, N. C., 25, 30, 57 Bennett, J. F., 244, 293, 295 Alexander, N. L., 137, 142 Benton, A. A,, 230, 231 Allen, H., 132, 144 Berger, S. M., 129, 130, 141 Allen, S. S., 131, 141 Berkowitz, H., 131, 141 Allen, V . L., 160, 167, 199, 200 Berkowitz, L., 34, 48, 49, 52, 56, 58, Allport, G. W., 121, 147, 248, 260, 293 69, 70, 71, 77, 80, 81, 87, 88, 89, Allport, L. H., 172, 173, 200 93, 94, 95, 96,98, 99, 100, 101, 103, Altman, I., l i 9 , 148 104, 106, 107, 124, 125, 128, 141, Altrocchi, J., 286, 297 142, 157, 200, 219, 228, 231, 232 Amsel, A,, 112, 141 Berscheid, E., 110, 141, 214, 231 Anderson, D., 131, 147 Biener, L., 279, 297 Anderson, N. H., 122, 125, 141 Bierbrauer, G. A., 41, 61 Andres, D., 125, 142 Bieri, J., 122, 144 Anglin, J. M., 251, 293 Bixenstine, V. E., 205, 211, 212, 216, Angulo, E., 34, 58 231 Aponte, J. F., 112, 145 Black, R., 112, 141 Appley, M. H., 110, 114, 142 Blake, R. R., 78, 107, 108 Aronfreed, J., 86, 106 Blau, P. M., 63, 64, 65, 107, 214, 232 Aronson, E., 9, 20, 26, 43, 57, 79, 80, Bless, E., 83, 107 106, 124, 132, 141, 145 Boffa, J., 14, 59 Arsenian, J. M., 135, 141 Bogdonoff, M. D., 135, 141 Asch, S. E., 5, 57, 151, 153, 159, 160, Bootyin, R. R., 120, 141 168, 169, 171, 200, 248, 259, 277, Bowers, K. S., 35, 58 292, 293 Braver, S. L., 220, 233 Asthana, H. S., 122, 141 Brehrn, J. W., 18, 22, 24, 25, 26, 27, 28, 29, 30, 31, 47, 58, 59, So, 72, 90, B 103, 105, 107, 232 Back, K. W., 135, 141 Bright, M. A., 11%145 Bagby, J. A., 121, 141 Brock, T. C., 52, 59, 126, 142 Bander, R. J., 12, 13, 57 Brody, C . L., 171, 199 Banfield, E. C., 103, 106 Broughton, A., 117, 143 Baratz, S. S., 135, 141 Brown, J. S., 12, 59 Barber, T. X., 12, 57 Bruner, J. S., 123, 142 Barnett, E. O., 121, 147 Bryan, J. H., 78, 79, 104, 107, 128, 142 Bamett, H. G., 174, 200 Bumstein, E., 126, 142 Baron, R. A., 126, 141, 218, 231 Burton, M. L., 251, 293 Beach, L., 237, 261, 287, 293 299

A

300

AUTHOR INDEX

Buss, A. H., 218, 232 Byrne, D., 110, 118, 125, 128, 142, 143

Crurnbaugh, C. M., '221, 232 Crutchfield, R. S., 124, 145, 184, 200

C

D

Calvin, H. D., 133, 142 Campbell, D. T., 113, 142 Campbell, E. Q., 137, 142 Campbell, J. D., 138, 139, 148, 237, 261, 288, 289, 297 Canavan, P., 22.3, 232 Carlsmith, J. M., 6, 15, 20, 24, 25, 57, 59, 83, 107 Carmone, F. J., 239, 244, 294 Carroll, J . D., 246, 253, 274, 277, 293 Cascio, W. F., 39, 59 Cattell, R. B., 260, 293 Chamniah, A. M., 206, 216, 233 Chance, J., 12Q, 142 Chang, J.-J., 246, 253, 274, 277, 293 Chapanis, A., 9, 59 Chapanis, N. P., 9, 59 Chappell, V. C., 2, 59 Cheder, P., 127, 142 Churchman, C. W., 164, 165, 200 Cialdini, R. B., 127, 144 Clausen, G . T., 73, 108 Cliff, N., 293 Clore, G. L., 110, 117, 142 Coady, H., 212, 234 Cofer, C. N., 110, 114, 142 Cohen, A,, 12, 13, 51, 53, 62 Cohen, A. R., 18, 22, 24, 25, 47, 58, 59 Cohen, B. P., 154, 170, 187, 200 Cole, A. H., 90, 107, 232 Collins, B. E., 25, 59, 117, 143 Conger, J. J., 118, 146 Connor, W. H., 80, 81, 89, 106 Consalvi, C., 175, 200 Cooley, W. W., 255, 267, 293 Coombs, C. H., 244, 293, 294 Corah, N . L., 14, 59 Coser, L., 151, 176, 200 Cottrell, N. B., 126, 132, 142, 147 Craddick, R. A., 123, 142 Crockett, W. H., 122, 142 Cronbach, L. J., 249, 275, 294 Crow, T., 122, 125, 145

D'Andrade, R. G., 257, 294 Daniels, L. R., 69, 70, 77, 94, 103, 106, 107, 128, 142, 228, 232 Darley, J. M., 66, 73, 74, 103, 107 Darlington, R. B., 83, 107 Davis, K. E., 44, 60 Davison, G. C., 34, 46, 51, 59 Davitz, J. R., 124, 135, 142 decharms, R., 39, 40, 54, 59 Deci, E. L., 39, 46, 59 de Monchaux, C., 181, 200 de Montmollin, G., 172, 200 Denner, B., 27, 62 Deutsch, M., 68, 107, 211, 217, 223, 22.5, 226, 232, 279, 297 Dickoff, H., 220, 232 Dinitz, S., 79, 108 Dison, T. R., 132, 147 Doise, W., 190, 191, 200 Dollard, J., 5, 61, 114, 146 Doob, L. W., 111, 114, 143 Dornbusch, S. M., 237, 260, 261, 288, 289, 294 Douglas, J., 216, 231 Douglas, R., 226, 232 Dreiser, T., 279, 294 Duchnowski, A. J., 117, 146 Dugan, R. D., 257, 295 Dworkin, L., 12, 13, 51, 53, 62 Dwyer, J,, 216, 233

E Eagly, A. H., 124, 144 Ebbesen, E. B., 20, 31, 32, 59, 62 Efran, J . S., 117, 143 Eisman(Lott), B., 111, 119, 143 Elms, A. C., 27, 59 Emerson, R., 180, 181, 200 Engel, E., 121, 143 Epstein, Y., 223, 232 Ervin, C. R., 118, 142 Evans, G., 226, 232 Evans, G. W., 221, 232 Exline, R., 117, 143

301

AUTHOR INDEX

F Faucheux, C., 164, 200 Feldman, R. E., 66, 107 Felipe, A. I., 259, 294 Festinger, L., 6, 15, 22, 24, 43, 59, 118, 143, 162, 166, 200 Firestone, I., 12, 13, 51, 53, 62 Fishbein, M., 277, 297 Frank, J., 46, 59 Fraser, S. C., 37, 38, 59 Freedman, J. L., 20, 37, 38, 59, 83, 107 French, J. R. P., 172, 200 Friedman, P., 98, 100, 106 Friendly, M. L., 251, 294 Frincke, G., 120, 144 Frisch, D. M., 95, 107 Funk, S. G., 227, 232

G Gaer, E. P., 129, 147 Gahagan, J. P., 217, 234 Gallo, P., 212, 233 Gallo, P. S., Jr., 227, 232 Geiwitz, P. J., 125, 148 Gelber, E. R., 230, 231 Gerard, H. B., 169, 188, 200 Gillis, J. S., 227, 234 Gleser, G. C., 249, 294 Glucksberg, S., 251, 294 Goldberg, G. N., 117, 143, 144 Goldfried, M . R., 248, 294 Goldman, R., 50, 59 Goldstein, A. G., 120, 142 Golightly, C., 128, 143 Collin, E. S., 287, 288, 294 Goodman, C. C., 123, 142 Goranson, R. E., 64,107, 219, 232 Cough, H. G., 255, 294 Gouldner, A. W., 64, 107, 205, 214, 232 Graham, D., 159, 200 Grant, G., 139, 144 Grant, L. D., 52, 59 Gray, D., 117, 143 Green, P. E., 239, 244, 294 Green, R. F., 248, 294 Greenbaum, C., 135, 144 Greenberg, M. S., 95, 107 Greene, D., 39, 46, 60 Creenwald, A. G., 115, 143

Griffitt, W., 124, 125, 143 Grinker, J., 50, 51, 60 Gross, A. E., 83, 107 Grusec, J., 127, 143, 146 Gumpert, P., 223, 232 Gumee, E., 179, 200 Gyr, J., 216, 233

H Haagen, C. H., 238, 294 Haigh, G. A,, 123, 147 Hain, J. D., 78, 107 Hallworth, H. J., 257, 294 Hamilton, D. L.,293, 294 Hammond, P., 20, 26, 27, 28, 29, 30, 61 Hansford, E. A., 222, 234 Harden, E. L., 133, 142 Hare, A. P., 155, 200 Harlow, H. F., 135, 143 Harris, R., 71, 107 Harris, V. A., 26, 29, 30, 44, 60 Harrison, A. A., 212, 233 Hartshorne, H., 255, 294 Hartup, W. W., 132, 143 Harvey, 0. J., 18, 60, 172, 175, 200, 201 Hastorf, A. H., 237, 260, 261, 288, 289, 294 Hathaway, S. R., 256, 295 Havinghurst, R. L., 140, 143 Hays, W. L., 239, 244, 248, 253, 276, 293, 295 Heider, F., 43, 44, 60, 115, 143, 160, 200 Heilman, M., 128, 143 Helmreich, R. L., 25, 59 Helson, H., 78, 107 Henchy, T., 132, 144 Henriksen, K., 34, 46, 48, 49, 60 Hess, E. H., 116, 143 Hoffman, D., 128, 144 Hoffman, F. K., 133, 142 Hollander, E. P., 175, 190, 200, 201 Homans, G. C., 64, 107 Horai, J., 230, 232, 234 Horan, C. B., 274, 295 Hornstein, H. A., 87, 107, 128, 143 Hornstein, M. G., 119, 144 Horowitz, I. A., 105, 107

302

AUTHOR INDEX

Hraba, J., 139, 144 Huffman, D. M., 128, 143 Huffman, L. J., 293,294 Hull, C. L., 111, 112, 115, 144 Hurst, P. M., 226, 233 Husek, T. R., 189, 201

I Insko, C. A,, 127, 144, 147 Irwin, M., 122, 144 Isaacson, R., 132, 146 Isen, A. M., 82, 107

J Jackson, D. N., 239, 248, 276, 277, 295, 297 Jackson, T., 124, 143 JaEa, M.,50, 59 James, G., 112, 144 Janis, J. L., 128, 144 Johnson, J., 131, 141 Johnson, R. C., 120, 144 Johnson, S. C., 269, 280, 295 Joncas, A. J., 20, 59 Jones, E. E., 42, 44, 60 Jones, L. E., 278, 279, 295 Jones, R. A., 19, 26, 27, 28, 29, 30, 31, 60, 61, 281, 269, 270, 271, 272, 274, 275, 280, 282, 285, 295 296 Jones, R. G., 27, 60

Kleinke, C., 226, 233 Klemp, G. O., 14, 60 Knapper, C.,286, 292, 297 Knight, G. W., 25, 30, 57 Knights, R. M., 132, 147 Knoll, R. L., 262, 297 Knowles, J. B., 132, 144 Knox, R. E.,226, 232 Knurek, D. A,, 125, 141 Koenig, K. P., 34, 46, 48, 49, 60 Kogan;. N., 172, 201,. 277, 295 Komorita, S. S., 220, 233 Krasner, L., 132, 144 Krauss, R. M., 125, 144, 226, 232 Krebs, D.L., 73, 78, 107, 128, 144 Krech, D.,124, 145 Kruskal, J. B., 241, 243, 249, 255, 262, 266, 274, 281, 295 Kuhn, T. S., 230, 233 Kuusinen, J., 257, 295

L

LaForge, R., 256, 295 Lage, E.,191,201 Lambert, W. W., 123, 145 Lamberth, J., 118, 142 Landy, D.,124, 145 LatanB, B., 68, 73, 74, 103, 107 Leary, T. F., 258, 295 Lepinski, J., 34, 58 Lepper, M. R., 20, 21, 38, 46, 59, 60, 61 Lerner, M. J., 124, 145 K Leventhal, H., 14, 56, 60, 61 Levine, J: M., 160,200 Kagan, J., 118, 146 Levine, J. R., 227, 232 Katz, I., 132, 135, 144 Levinger, G., 118, 145 Katz, P. A., 20, 61 Levy, B. I., 69, 107 Kay, B., 79, 108 Levy, J., 119, 143 Keen, R., 132, 147 Kelley, H. H., 21, 36, 41, 44, 45, 60, Levy, L. H., 257, 295 150, 154, 161, 173, 175, 190, 201, Liddell, H. S., 135, 145 Liebling, B. A., 230, 231 230, 231, 232 Linder, D. E., 19, 26, 27, 28, 29, 30, Kelly, G. A., 289, 273, 295 31, 60, 61 Kelman, H. C., 124, 136, 144, 171, 201 Linder, O., 132, 141 Kennedy, W. A., 132, 144 Kiesler, C. A,, 17, 18, 26, 27, 28, 29, Lindskold, S., 217, 234 Little, K. B., 117, 145 30, 31, 60, 117, 143, 144 Loehlin, J. C., 257, 295 King, D. L., 135, 144 Loewenton, E. P., 28, 27, 28, 29, 30. 61 Kissel, S., 135, 144 Loftis, J., 34, 46, 61 Klanderman, S. B., 71, 107 Lohnes, P. R., 255, 267, 293 Heck, R., 118, 144

303

AUTHOR INDEX

London, H., 56, 61 Lott, A. J., 110, 111, 112, 117, 119, 120, 122, 125, 130, 132, 133, 134, 144, 145

Lott, B. E., 110, 111, 112, 117, 119, 120, 122, 125, 130, 132, 133, 134, 145

Lubow, R. E., 135, 146 Luce, R. D., 209, 233

M Macaday, J. R.,79, 108 McCauley, C., 20, 26, 27, 28, 29, 30, 61 McClintock, C. G., 212, 225, 226, 227, 231, 233

Maccoby, E. E., 122, 146 McConnell, H. K., 23, 30, 31, 32, 33, 52, 58 McCoy, N., 131, 146 McGee, V. E., 274, 295 McGinley, W. H., 112, 145 McGuire, W. J., 9, 26, 43, 57, 61, 192, 201

McKeachie, W. J., 132, 146 Macker, C. F., 83, 107 McNeel, S. P., 226, 227, 233 McPartland, J., 137, 146 Madaras, G. R., 19, 13, 57 Maller, J. B., 255, 294 Marcuse, F. L., 135, 146 Marwell, G., 209, 233 Maslach, C., 12, 61 Mason, D. J., 135, 143 Mason, W. A,, 135, 146 Masor, H. N., 128, 143 Matthews, G., 112, 130, 133, 145, 146 Matthews, M. W., 104, 108 May, M. A,, 2-55, 294 Meehl, P. E., 256, 295 Meeker, R. J., 222, 234 Mehrabian, A., 117, 118, 146 Melzack, R., 12, 61 Menzel, H., 176, 201 Messick, D. M., 209, 233 Messick, S. J., 239, 274, 276, 277, 295, 297

Mettee, D. R., 79, 80, 106 Midlarsky, E., 67, 73, 108 Milholland, J., 132, 146 Miller, D. R., 98, 108

Miller, G. A,, 251, 270, 295, 296 Miller, N. E., 5, 61, 112, 113, 114, 146 Mills, J., 18, 27, 60, 61 Mischel, W., 57, 61, 127, 146 Monello, L., 56, 61 Moore, M., 121, 146 Moran, G., 119, 146 Moreland, J., 118, 145 Moscovici, S., 164, 173, 190, 191, 200, 201

Mouton, J. S., 78, 107 Mowrer, 0. H., 112, 146 Mulaik, S. A., 257, 296 Murray, L. B., 121, 146 Mussen, P. H., 118, 127, 146 Muzzy, R. E., 237, 261, 288, 289, 294

N Naffrechoux, M., 191, 201 Natsoulas, T., 120, 126, 141, 146, 147 Nelson, C. E., 243, 244, 251, 252, 254, 255, 258, 259, 262, 264, 268, 269, 281, 287, 292, 296, 297 Nemeth, C., 233 Newcomb, T. M., 9, 26, 43, 57, 115, 118, 146, 151, 201 Nisbett, R. E., 12, 15, 17, 18, 34, 39, 42, 46, 50, 53, 56, 60, 61, 62 Norman, W. T., 257, 296 Nunnally, J. C., 117, 146, 189, 201

0 Odbert, H. S., 248, 260, 293 Olshan, K., 249, 250, 254, 255, 256, 262, 269, 287, 289, 290, 291, 296 Osgood, C. E., 248, 249, 296 Oskamp, S., 211, 226, 233 Ostfeld, B., 20, 61

P Park, J. N., 120, 147 Parker, A. L., 127, 146 Parker, R. K., 117, 146 Passini, F. T., 257, 296 Patterson, G. R., 131, 147 Peabody, D., 249, 250, 251, 254, 255, 256, 259, 296

304

AUTHOR INDFX

Pedersen, D. hl., 276, 296 Pepitone, A,, A., 20, 61 Perlman, D., 211, 226, 233 Peterson, D. R., 2.57, 296 Pettigrew, T. F., 121, 147 Piliavin, I. M., 26, 27, 28, 29, 30, 61, 66, 108

Piliavin, J. A., 66, 108 Plon, M., 209, 233 Polly, S., 41, 61 Posavac, E. J., 145, 296 Postman, L., 123, 142 Potash, H., 205, 211, 231 Powell, J., 217, 234 Pruitt, D., 221, 233 Pruitt, D. G . , 95, 108, 233

R Radlow, R., 226, 233 Raiffa, H., 209, 233 Rapoport, A,, 206, 210, 212, 216, 233 Ratcliff, K., 209, 233 Rawlings, E. E., 85, 86, 108 Ray, A. A., 34, 46, 51, 62 Reckless, W. C., 79, 108 Reed, T., 122, 125, 145 Richardson, S. A., 237, 260, 261, 288, 280, 294 Riecken, H. W., 173, 198, 201 Rittle, R. H., 126, 147 Rodin, J., 34, 46, 61, 66, 108 Rommetveit, R., 152, 201 Rosekrans, M. A., 126, 147 Rosenbaum, M., 78, 108 Rosenberg, M. J., 9, 26, 43, 57, 204, 233 Rosenberg, S., 236, 243, 244, 249, 250, 251, 252, 254, 255, 261, 262, 283, 264, 270, 271, 272, 274, 280, 281, 282, 285, 296 Ross, L.,34, 41, 46, 61 Ryle, G., 2, 61

256, 285, 275, 287,

S Sampson, E. E., 127, 147 Sapolsky, A., 132, 147

258, 259, 2E8, 269, 278, 279, 292, 295,

Sawyer, J., 128, 147 Schachter, S., 4, 12, 15, 33, 34, 43, 46, 49, 50, 51, 52, 53, 56, 59, 61, 62, 180, 181, 201 Schein, E. H., 153, 201 Schicht, W., 120, 142 Schniitt, D. R., 209, 233 Schoeninger, D. W., 125, 147 Schopler, J., 97, 104, 108 Schuette, D., 117, 143 Schwartz, S. H., 67, 72, 73, 77, 108 Scodel, A., 211, 233 Sechrest, L., 123, 147 Sedlak, A., 236, 261, 262, 263, 264, 265, 266, 2E8, 269, 275, 278, 279, 296 Seltzer, A. L., 116, 143 Sermat, V., 212, 234 Shapiro, M. M., 150, 154, 175, 190, 201 Shaw, M. E., 179, 201 Shepard, R. N., 239, 241, 245, 296 Sheposh, J. P., 220, 233 Sherif, M., 172, 173, 201 Sherk, L., 128, 147 Sherman, R., 278, 296 Shimmin, S., 181, 200 Shlein, J. M., 116, 143 Shrauger, S., 286, 297 Shure, G . H., 222, 234 Shuttleworth, F. K., 255, 294 Signell, K. A,, 288, 289, 297 Simon, H. A., 166, 201 Singer, J. E., 4, 33, 43, 46, 50, 51, 52, 53, 62 Skinner, B. F., 2, 4, 6, 62 Skubiski, S . L., 127, 143 Smith, A. J., 120, 147 Smith, E. W. L., 132, 147 Snyder, M., 31, 32, 62 Sole, K., 128, 143 Solley, C. M., 123, 147, 239, 276, 295 Solomon, L., 211, 219, 224, 229, 234 Solomon, R. L., 123, 145 Sorrentino, R., 119, 148 Spear, P. S., 131, 141 Spence, K. W., 114, 147 Staats, A. W., 115, 147 Stahelski, A. J., 230, 232 Staub, E., 128, 147 Steinberg, D. D., 251, 297

305

AUTHOR INDEX

Steiner, I. D., 36, 62, 167, 201 Stenson, H. H., 262, 297 Stephens, M. W., 120, 142 Stevenson, H. W., 132, 147 Storms, M. D., 34, 42, 50, 56, 61, 62 Stotland, E., 126, 142, 147 Strand, S., 212, 233 Suci, G. I., 249, 296 Suczek, R. F., 256, 295 Swanson, G. E., 98, 108 Swingle, P. G., 212, 227, 234

T Tagiuri, R., 286, 297 Tajfel, H., 120, 147 Tamler, H., 26, 29, 30, 60 Tannenbaum, P. H., 9, 26, 43, 52, 57, 62, 129, 147 Taylor, D., 119, 148 Tedeschi, J. T., 217, 230, 232, 234 Terhune, K. W., 217, 234 Test, M. A., 78, 79, 107, 128, 142 Thibaut, J. W., 173, 201 Thomas, E., 68, 108 Thomson, C. W., 120, 144 Thorngate, W., 209, 233 Thurstone, L. L., 256, 297 Torgerson, W. S., 239, 243, 297 Tornatzky, L., 125, 148 Torrance, E. P., 179, 202 Tripodi, T., 122, 144 Tucker, L. R., 274, 276, 277, 297 Tuddenham, R. D., 170, 202 Turner, C., 34, 46, 49, 52, 58 Turner, E. A,, 20, 62

U Ullmann, L. P., 132, 144

V Valins, S., 4, 33, 34, 39, 46, 51, 53, 59, 61, 62 Vega, M., 132, 144 Vernon, M. D., 120, 148 Vivekananthan, P. S., 243, 244, 251, 252, 254, 255, 258, 259, 262, 264, 2E8, 269, 281, 287, 292, 296 Vreeland, R. S., 237, 261, 288, 289, 294

w Wagner, C. M., 67, 78, 108 Wahler, R. G., 131, 148 Walker, E. L., 136, 148 Wallace, J., 123, 147 Wallach, M., 172, 201 Wallington, S. A., 83, 107 Walsh, M. L., 133, 145 Walster, E. H., 110, 121, 214, 231 Walters, H. A., 277, 297 Ward, C. D., 123, 148 Warr, P. B., 286, 292, 297 Watson, P. D., 123, 145 Weick, K. E., 46, 51, 53, 62 Weidner, M. F., 226, 233 Weinstein, P., 112, 145 Weisenberg, M., 12, 13, 51, 53, 62 Werner, H., 287, 297 Wertheimer, M., 237, 261, 287, 293 Wheeler, L., 46, 50, 62, 67, 78, 108 Wichman, J., 218, 234 Wiggins, J. S., 245, 297 Wiggins, N., 247, 277, 297 Wike, E. L., 115, 148 Willis, F. N., Jr., 118, 148 Wilson, K. V., 205, 211, 212, 231 Wilson, W. C., 122, 146 Wing, H., 251, 297 Wish, M., 279, 297 Wishner, J., 248, 297 Wolosin, R. J., 27, 62 Wood, W. D., 125, 147 Woodmansee, J. J., 117, 148 Wright, J., 20, 62 Wright, P. H., 126, 148 Wrightsman, L. S., 226, 234

Y Yarrow, L. J., 138, 139, 148 Yarrow, M. R., 138, 139, 148, 237, 261, 288, 289, 297 Yi-Guang Lin, 132, 146 Young, F. W., 278, 279, 295

Z Zajonc, R. B., 132, 148 Zander, A., 126, 142, 147 Zanna, M., 26, 27, 28, 29, 30, 31, 60

306

AUTHOR INDEX

Zanna, M. P., 17, 18, 20, 21, 26, 60, Zimbardo, P. G., 12, 13, 34, 43, 46, 47, 48, 51, 52, 53, 54, 61, 62, 198, 202 61, 62 Zimmerman, R. R., 135, 143 Zavalloni, M., 173, 201 Zucker, L. G., 171, 199 Zigler, E., 131, 141, 146 Zuckerman, M., 258. 297 ZilIer, R. C., 174, 202

SUBJECT INDEX A Acquiescent responses, interpersonal attitudes and, 126-129 Altruism, see Helping Ambiguity, communication and, 215-218 of motives, 219-224 Approach, consistency and, 183-189 interpersonal attitudes and, 116-119 Attitudes, see Interpersonal attitudes Attractiveness, of other, helpfulness and, 94-95 Attribution, see also under Self-perception theory social influence and, 180-165 Avoidance, consistency and, 183-189 interpersonal attitudes and, 116-119

B Bargaining, 203-234 definition of 207-208 prisoner’s dilemma and reciprocity and, 214-215, 224-228 ambiguity of motives and, 21S224 communication and ambiguity and, 215-218 prisoner’s dilemma paradigm and, definition of, 206-207 origins of, 209-211 strategy studies with, 211-215 Behavior, mediation of by attributions, 5 0 5 4 social, change of, 151-156 style of, as source of social influence, 157-160

C Cartoons, self-perception and, 9-10 Children, personality perception in, 287-291

Choice, free, self-perception and, 22-23 Coercion, power and, 156 Cognitive dissonance, self-perception theory and, 15-16, 31-33 forced compliance studies, 16-22 free choice studies, 22-23 interpersonal simulations, 23-31 Communication, ambiguity in prisoner’s dilemma and, 215-218 Compliance, forced, self-perception and, 16-22 Confession, false, self-perception and, 10-12 Conflict, social influence and, 165-166 Conformity, see akio Social influence interpersonal attitudes and, 126-129 Consistency, as source of social influence, 160-165, 183 change of response and code and, 191-198 lack of a group norm and, 183-189 minority response and an implicit norm and, 189-191 Control, conformity bias and, 151-156 Cues, consequences of interpersonal attitudes and, 112-113

D

Dependency, helpfulness and, attractiveness of other and, 94-95 moral norms and, 68-77 previous helpfulness of other and, 95-98 reactance and legitimacy of other’s dependency and, 103-105 salience of helping ideal and, 77-80 social class and, 98-103 as source of social influence, 157-160 Descriptive responses, interpersonal attitudes and, 119-123 Dispositional properties, self-attribution of, 3739 Dissonance, see Cognitive dissonance 307

308

SUBJECT INDEX

E Empathy, helpfulness and, 86-89 Evaluative responses, interpersonal attitudes and, 123-126

F Feelings, helpfulness and, empathy, 86-89 good and bad moods, 80-83 guilt, 83-86 .self-concern, 89-94 Forced compliance, self-perception and, 16-22 Free choice, self-perception and, 22-23

G Generalization, consequences of interpersonal attitudes and, 115-116 Group, influence of, 166-170 norms and, 183-198 Guilt, helpfulness and, 83-86

H Helping, 63-108 dependency and, attractiveness of other and, 94-95 previous helpfulness of other and, 95-98 reactance and legitimacy of, 103-105 social class and, 98-103 moods and feelings and, empathy, 86-89 good and bad moods, 80-83 guilt, 83-86 self-concern, 89-94 reactions to dependency and, moral norms and, 68-77 salience of helping ideal and, 77-80 social exchange and, 63-67

I Individual differences, 274-275 multidimensional models of, 275-273 study of individuals and, 279-287 Influence, see Social influence Innovation, social influence and, 165-166 Insufficient justification, self-perception and, 16-22 Interpersonal attitudes, 103-148

approach and avoidance responses and, 116-119 conformity, modeling, and acquiescent responses and, 126-129 cue properties of persons and, 112-113 descriptive responses and, 119-123 evaluative responses and, 123-128 generalization and, 115-116 learning and, 130-134 motivational properties of persons and, 113 responses to stress and, 134-135 reward properties of persons and, 114-115 theoretical application of, 136-141 vicarious responses and, 129-130 Interpersonal simulations, self-perception and, 23-31

J Judgments, changing, 174-183 Justification, self-perception and, 16-22, 3940

L Learning, interpersonal attitudes and, 130-134 Liking, see Interpersonal attitudes

M Mediation, by attributions, 50-54 Minorities, see Social influence Misattribution, self-perception and, 33-37 Modeling, interpersonal attitudes and, 126-129 Moods, see Feelings Moral norms, helpfulness and, 68-73 direct test of, 74-77 Motivation, ambiguity of, 219-224 consequences of interpersonal attitudes and, 113

N Noncognitive response classes, self-perception theory and, 4 5 5 0 Norm( s ) , changing, 174-183 implicit, influence of a minority re-

309

SUBJECT INDEX

sponse on, 189-191 lack of, consistency and, 183-189 moral, helpfulness and, 68-77 as result of reciprocal concessions, 170-174

0 Overjustification, self-perception and, 39-40

P Pain, self-perception and, 12-15 Perception, see also Self-perception theory of personality in children, 287-291 Perceptual task, social influence and, 191-198 Personality theory, 235-297 individual differences and, 274-275 multidimensional models of, 275-279 study of individuals and, 279-287 laboratory analogs and, 258-259 of personality descriptions, 250-258 of trait inference, 247-250 naturalistic personality descriptions and, 259-261 multidimensional, 261-269 typological, 269-274 personality perception in children and, 287-291 scaling methods and, 238-245 interpretation of solutions and, 245-247

Power, coercive, normative relationships and, 156 Prisoner’s dilemma, see under Bargaining

R Reactance, helpfulness and, 103-105 Reciprocity, prisoner’s dilemma and, 214-215, 224-228 ambiguity of motives and, 219-224 communication and ambiguity and, 215-218 Reinforcement, secondary, interpersonal attitudes and, see Interpersonal attitudes Relationships, coercive, power and, 156 Response change, consistency and, 191-198

Response classes, noncognitive, self-perception theory and, 45-50 Reward, consequences of interpersonal attitudes and, 114-115

S Self-concern, helpfulness and, 89-94 Self-perception theory, 1-62 cognitive dissonance phenomena and, 15-16, 31-33 forced compliance studies, 16-22 free choice studies, 22-23 interpersonal simulations, 23-31 dispositional properties and, 37-39 experimental evidence for, 8-9 cartoon experiment, 9-10 false confession experiment, 10-12 pain perception experiment, 12-15 functional analysis and, 54-57 interpersonal perception and, 40-42 mediation of behavior by attributions and, 5 0 5 4 misattribution effects and, 33-37 noncognitive response classes and, conceptual status of, 45-50 ontogeny of self-attributions and, 2 4 overjustification effects and, 39-40 paradigm shift in social psychology and, 42-45 postulates of, 4-8 Social class, helpfulness and, 9tL-103 Social exchange, helpfulness and, 63-67 Social influence, 149-202 conformity as a form of social psychology and, 150-151 consistency as source of, 183 change of response and code and, 191-198 lack of a group norm and, 183-189 minority response and an implicit norm and, 189-191 control and change of social behavior and, 151-156 of group in the absence of an alternative to consensus, 166-170 norms and, changing, 174-183 as result of reciprocal concessions, 170-174

310

SUBJECT INDEX

minority response and an implicit norm, 189-191 source of, coercive, normative relationships and power and, 156 consistency and attribution processes and, 180-165 dependency versus style of behavior as, 157-160 Social psychology, conformity as a form of, 150-151

shift of paradigm in, 4 2 4 5 Stress, interpersonal attitudes and, 134-135

T Trait inference, 247-250, 258-259

V Vicarious responses, interpersonal attitudes and, 129-130