Journal of Personality and Social Psychology1980, Vol. 38, No. 1, 23-35

Social Anxiety, Self-Presentation, and the Self-Serving Biasin Causal Attribution

Robert M. Arkin, Alan J. Appelman, and Jerry M. BurgerUniversity of Missouri—Columbia

Two experiments were conducted to provide evidence concerning the contribu-tion of self-presentation concerns to the self-serving bias in causal attribution(individuals' tendency to assume more personal responsibility for a success thanfor a failure outcome) and its occasional, but systematic, reversal. In Experiment1 high- but not low-social-anxiety participants presented themselves in a far moremodest light when a committee of high prestige others was to join the experi-menter in evaluating their behavior than when the committee evaluation wascanceled. In Experiment 2 this reversal of the self-serving bias among high-social-anxiety subjects (in the evaluative context) was replicated, and it was also foundthat both high- and low-social-anxiety participants portrayed the causes of theirbehavior in a more modest fashion when they responded via the "bogus pipeline,"a measurement technique designed to reduce distortion and dissimulation inverbal responses, than when they responded in the traditional paper-and-pencilformat (although the influence of the bogus pipeline above and beyond the com-mittee evaluation in eliciting "honest" responses from subjects only reached sig-nificance for low-social-anxiety subjects). These findings are discussed in termsof the varying self-presentation strategies and differing self-concepts of individ-uals high and low in social anxiety, as well as the self-presentation component toapparently self-enhancing and self-effacing causal attributions for performance.

The term selj-presentation refers to themanner in which individuals plan, adopt, andcarry out strategies for managing the impres-sions they make on others. There is a growinginterest in this approach to the analysis ofinterpersonal relations. Self-presentation in-terpretations of such traditional social psycho-logical topics as group decision making (e.g.,Janis & Mann, 1977, pp. 129-133; Jellison &Arkin, 1977), attitude change (e.g., Hass &Mann, 1976; Tedeschi, Schlenker, & Bonoma,1971), obesity (Krantz, 1978), and equityresearch (Reis & Gruzen, 1976) have ap-peared recently. The basic premise of the self-presentation conceptual perspective is that

This research was supported by Grant 1 R0831910 from the National Institute of Mental Healthto the first author. The data from Experiment 1were presented at the annual meeting of the Mid-western Psychological Association, Chicago, 1979.

Requests for reprints should be sent to RobertM. Arkin, Department of Psychology, Universityof Missouri, Columbia, Missouri 65211.

individuals are highly concerned about win-ning the approval and avoiding the disap-proval of others. To achieve these ends, indi-viduals are said to spend considerable timeand effort discerning what factors will influ-ence the impression they make on others andthen behaving in the most positive mannerpossible so as to create a favorable impres-sion.

Research has shown repeatedly that indi-viduals tend to attribute their successes totheir own efforts, abilities, or other disposi-tions, whereas they attribute their failures tobad luck, task difficulty, or a variety of otherextenuating circumstances (cf. Bradley, 1978;Miller & Ross, 1975; Snyder, Stephan, &Rosenfield, 1978). Because this attributionalpattern stems so plausibly from a motive tomaintain or enhance esteem (e.g., Heider,1958), it has been termed the "self-servingbias" in causal attribution. Bradley (1978)has argued recently that the so-called self-serving bias in causal attribution may reflect

Copyright 1980 by the American Psychological Association, Inc. 0022-3514/80/3801-0023S00.75

23

24 R. ARKIN, A. APPELMAN, AND J. BURGER

self-presentational concerns. By taking per-sonal responsibility for praiseworthy acts anddenying personal responsibility for blame-worthy acts, individuals can maximize theesteem in which they are held by others and,as a consequence, maximize their own self-esteem (Bradley, 1978).

There are some notable exceptions to thegeneral finding that individuals assume greaterresponsibility for successful than for unsuc-cessful outcomes, however. Some researchershave taken these exceptions as evidenceagainst the notion of self-serving biases inthe attribution process (e.g., Ross, 1977).However, Gifford Weary Bradley (1978) hasincorporated these exceptions within her gen-eral self-presentation framework. As she hasnoted, "the potential for present or futureinvalidation of individuals' self-presentationtends to make them more modest about theirown abilities and attributes" (p. 66), some-times even reversing the typical self-servingbias effect. Seemingly, the individual at-tributor is aware that others may make judg-ments about how causally related the indi-vidual and his or her performance outcomereally are; if others are likely to be unper-suaded by—or even challenge—the individu-al's flattering interpretation of his or her ownbehavior, modesty becomes the best self-presentation strategy (Jones & Wortman,1973). Thus, individuals may often, for self-presentational reasons, assume more responsi-bility for failure than for success.

The two experiments reported here weredesigned to provide further evidence concern-ing the contribution of self-presentation con-cerns to individuals' attributions for successand failure. Experiment 1 was designed toprovide evidence concerning the hypothesisthat individuals become "more modest abouttheir abilities and attributes" when others areperceived as likely to be unpersuaded by, orperhaps even challenge publicly, an overlyflattering interpretation of events. In addi-tion, Experiment 1 was designed to providesome initial evidence concerning individualdifferences in self-presentation style, orstrategy.

Not surprisingly, some research and theoryconcerning individual differences in self-presentation has already been reported. How-

ever, these approaches have primarily empha-sized general individual differences in thepredisposition or tendency to manage one'simpression (Krantz, 1978; Snyder, in press).Although there is a growing interest in attri-butional styles generally (Ickes & Layden,1978), there is to date no solid evidence con-cerning individual difference factors thatmight distinguish among styles of strategiesused by individuals who engage in self-presentation.

However, Lepper (1970) and Carlsmith,Lepper, and Landauer (1974) found thatunder conditions of situationally inducedanxiety, children were far more responsive(compliant) to a threatening than to a warmand positive adult, whereas nonanxious chil-dren were more responsive to the warm andpositive than to the threatening adult. Lepper(1970) took these data concerning compli-ance to mean that "when the child is anxious. . . , he will be motivated largely to avoidaversive events; but when he is not anxious,he will be motivated to seek positive events"(p. 70S). Similarly, several authors havespeculated that this "cost orientation" amonganxious children exists among adults who are(situationally or chronically) high in socialanxiety; individuals high in social anxietyoften prefer to leave potentially embarras-sing social situations as quickly as possible andmay frequently fulfill this need instead ofresponding to other, perhaps more positive,social motives (Freedman, Wallington, &Bless, 1967; McPeek & Cialdini, 1977). Fen-igstein, Scheier, and Buss (1975) have simplydefined social anxiety as "discomfort in thepresence of others" (p. 523) and have ar-gued that individuals high in social anxiety(who describe themselves as shy, easily em-barrassed, and nervous in group situations)may have a more intense reaction to situa-tions that are potentially embarrassing thando individuals low in social anxiety. Thus,social anxiety seems to be a likely candidatefor distinguishing among individuals in termsof self-presentational style.

Specifically, individuals high in social anxi-ety should be particularly sensitive to publicscrutiny of their behavior and might there-fore be particularly likely to report modestcausal interpretations of their behavior if

SELF-SERVING BIAS 25

their behavior and attributions are subjectedto close public scrutiny (thus reversing thetypical self-serving attribution bias, as Brad-ley, 1978, would anticipate). Being less sensi-tive to the embarrassment that could ensuefrom a public invalidation of a positive self-appraisal, individuals low in social anxietyshould be much less likely to interpret intensescrutiny of their behavior as threatening.Thus individuals low in social anxiety wouldnot necessarily be expected to adopt a "costorientation" and portray themselves modestlywhen their behavior and attribution are sub-jected to close public scrutiny. (Indeed, thebehavior of individuals low in social anxietymight even be expected to reflect a "rewardorientation," with more personally flatteringattributions about the causes of their behaviorreported when their behavior is subjected toevaluation than when it is not.)

To test the prediction that high-social anxi-ety individuals would become "more modestabout their abilities and attributes" when anembarrassing public challenge to a flatteringinterpretation of events was likely, individualsscoring high and low on a measure of socialanxiety (Fenigstein, Scheier, & Buss, 197S)performed a task for which (false) feedbackabout successful versus unsuccessful perform-ance has reliably produced the "self-serving"pattern of performance attributions (e.g.,Arkin, Gleason, & Johnston, 1976; Federoff& Harvey, 1976; Harvey, Arkin, Gleason, &Johnston, 1974; Weary, in press). Subjectswere informed that their performance andquestionnaire responses would either be evalu-ated (with the subject present) during theremaining half of the experimental hour bya committee of prestigious individuals (im-mediate evaluation) or at some time in thedistant future with the subject absent (de-layed evaluation). It was predicted that bothhigh- and low-social-anxiety subjects wouldassume somewhat more responsibility for suc-cess than for failure in the delayed evaluationconditions but that in the immediate evalua-tion conditions high-social-anxiety subjectswould assume greater responsibility for fail-ure than for success, whereas low-social-anxi-ety individuals would assume greater respon-sibility for success than for failure, resulting

in a significant interaction of outcome, socialanxiety, and level of evaluation.

MethodExperiment 1

Subjects. Subjects were recruited from the pre-tested introductory psychology pool if they hadscores on the Social Anxiety subscale of the Fenig-stein et al. (1975) Self-Consciousness Scale in thehighest or lowest third of the distribution. A totalof 40 high- and 40 low-social-anxiety individualswere randomly assigned to the experimental condi-tions. Thirty-five female and 45 male students par-ticipated; men and women were assigned to condi-tions in a manner that maintained an equivalentproportion across conditions.1 Subjects were con-tacted by phone and were scheduled to participateby an experimental assistant, assuring that theexperimenter was kept blind to the subject's level ofsocial anxiety.

Procedure. Each subject reported to the "thera-peutic testing laboratory," where he or she was metby the experimenter. The confederate arrived soonthereafter, and the experimenter consulted hisrecords, apparently to identify the participants andtheir roles in the study. He then escorted both indi-viduals into the laboratory and conducted a tourof the facility.

While conducting his tour, the experimenter de-scribed the supposed purpose of the study and howthe equipment would be used. He explained thatthe study was designed to investigate the effects ofa new, shortened form of desensitization instruc-tions and that the present session was a follow-upto an earlier study in which a number of individu-als suffering from test anxiety had been identified.He then casually indicated to the actor that thepatient had been identified as test anxious and hadagreed to return for the therapeutic session. Theexperimenter then pointed out two electrodes drapedover a comfortable lounge chair and "reminded"the patient that an instrument designed to measuremuscular relaxation would be recording his or herresponses throughout the session. The experimenterthen applied electrode cream and the two electrodesto the left hand and arm of the patient, and thepatient was seated in the lounge chair.

The subject was then escorted to the other sideof a partition, was seated in a chair at a table, andwas shown that the effect of the therapy could beviewed on an oscilloscope, the endpoints of whichwere labeled "relaxed" and "not relaxed." The ex-perimenter then supposedly activated and calibratedthe equipment while the subject reviewed a one-pagesummary of instructions. Of course, the electrodes

1 Since inclusion of sex of subjects as a factor inthe analyses of variance did not reveal any sig-nificant main effects or interactions involving thisfactor, there will be no further discussion of thisvariable in this article.

26 R. ARKIN, A. APPELMAN, AND J. BURGER

Table 1Means for Subjects' Ratings of theSuccess of Therapy, Experiment 1

Success outcome Failure outcome

Social anxiety Social anxietyi.evei 01

evaluation

ImmediateDelayed

High

7.55,v8.44.

Low

8.00,,8.40n

High

1.44,,2.08b

Low

2.18b1.37b

Note. The higher the score, the more successfulsubjects perceived the therapy to be. Means withdifferent subscripts arc significantly different(p < .05) by Newman-Keuls procedure.

attached to the patient were not operative. Thevertical position of a signal on the oscilloscope wascontrolled by a potentiometer operated surrep-titiously by the patient.

After the actor finished reading the summary, theexperimenter explained that the therapist's perform-ance was always videotaped during the session sothat a campus-wide committee of psychologists andcounselors could evaluate the session more effectively.Subjects were told that a group of three profession-als were available to evaluate the session and wouldarrive shortly after it was completed. It was ex-plained that the researchers had found it most effi-cient to evaluate the session with the therapistpresent to answer questions or assist in any waybut that it had proved too distracting to conductthe evaluation during the therapeutic session.Following these instructions, a Sony videotaperecording apparatus was activated and adjusted bythe experimenter.

The actor was then given a therapeutic outlineto peruse and deliver. The therapeutic outlineconsisted of six short paragraphs that asked thepatient to imagine a series of events involving study-ing for and eventually taking a college exam.2

Outcome manipulation. During the therapeuticsession, the confederate manipulated the signal onthe oscilloscope so that it wavered slightly aroundthe midpoint of the scale and then crept progres-sively toward either the "relaxed" end (successoutcome) or the "not relaxed" end (failure out-come) as the therapy continued. The experimenterwas kept blind to this outcome manipulation untilit was actually delivered. Following the therapeuticsession, the experimenter stopped the videotape re-corder, removed the electrodes from the patient'shand and arm, and escorted the patient to an ad-joining room, supposedly to complete a question-naire.

Personal evaluation manipulation. When the ex-perimenter returned, he gave the subject either one(delayed evaluation) or three (immediate evalua-tion) copies of a questionnaire that appeared veryformal. Subjects in the immediate evaluation condi-tions were asked to complete one form of the ques-

tionnaire and then transfer their responses to theother copies so that each committee member wouldhave his own form. In the delayed evaluation con-ditions, the experimenter explained that he hadfound a note left for him in the adjoining roomindicating that due to a scheduling problem, thecommittee could not convene during the remaininghalf of the experimental hour. After apologizing forthis unexpected change in plans, the experimenterindicated that the evaluation session would berescheduled for a later date and that the subjectwould not be required to attend the session. Toensure that subjects would not be overly interestedin attending the session, a date during an extendeduniversity holiday season was scheduled. The experi-menter was kept blind to this personal evaluationmanipulation until just prior to its delivery.

The questionnaire was composed of several items,each anchored on 9-point scales (1 — not very much,9 = very much) . The dependent measures were theextent to which the outcome of the therapy wasperceived as attributable to the manner in whichthe therapist delivered it, to the therapist's ability(lack of ability), and to his or her effort (lack ofe f fo r t ) .

Upon completion of the questionnaire, subjectswere thoroughly debriefed, thanked, and dismissed.The comments of one subject indicated some aware-ness of the experimental purposes, and his data wereexcluded from the final analysis.

Results

Manipulation checks. The outcome ma-nipulation clearly produced the desired effects.Subjects in the success outcome conditionrated the therapeutic session much more suc-cessful (M — 8.09) than did subjects in thefailure outcome conditions (M — 1.77), F(l,71) = 782.86, p< .0001. A triple interac-tion, interesting in light of the results to bereported below, also reached significance, F(l,71) =4.36, p < .05. The means for thismeasure are presented in Table 1. Seemingly,high-social-anxiety subjects in the immediateevaluation conditions who were successfulwere somewhat less willing than other suc-cessful subjects to acknowledge the magni-tude of the positive outcome. Among unsuc-cessful subjects, low-social-anxiety subjects inthe immediate evaluation condition and high-social-anxiety subjects in the delayed evalua-tion condition were somewhat more reluctant

- A more complete description of the procedurescan be found in Harvey, Arkin, Gleason, and John-ston (1974). The therapeutic outline used in thepresent study is available from the authors.

SELF-SERVING BIAS 27

than the other unsuccessful subjects to ac-knowledge how unsuccessful the therapy hadbeen. Nevertheless, the manipulation wasclearly effective. No other main effects orinteractions emerged.

To check on the effectiveness of the evalu-ation manipulation, subjects were asked toindicate the "session evaluation date" as-signed to them. All subjects in the immediateevaluation conditions correctly filled in thedate of their participation in the therapeuticsession as the session evaluation date. Allbut two subjects in the delayed evaluationconditions correctly identified the future datesupplied them. Thus, at least in terms ofsubjects' understanding of the type of evalu-ation to which they would be subjected, theevaluation manipulation was highly success-ful.

Attribution of responsibility. For ease ofpresentation and because the three dependentmeasures were highly correlated, the threeitems were summed to form a single index ofdispositional attribution. Analysis of varianceof subjects' scores on this dispositional attri-bution index revealed the predicted interactionof outcome, social anxiety, and level of evalu-ation, f (1 ,70 ) =5.76, p < .02. No othermain effects or interactions in this analysiseven approached significance (p > .15).

To provide more specific tests of the pre-dictions, separate analyses of variance wereconducted within the immediate and thedelayed evaluation conditions. Within the im-mediate evaluation conditions, analysis ofvariance revealed only the predicted SocialAnxiety X Outcome interaction, F(l ,36) =4.60, p < .04. As can be seen from the meansdisplayed in Table 2, high-social-anxiety sub-jects attributed more responsibility to them-selves when they were unsuccessful than whenthey were successful (p = .OS),3 whereas low-social-anxiety subjects attributed somewhatmore responsibility to themselves when theywere successful than when they were unsuc-cessful (ns). High-social-anxiety subjects as-sumed significantly more personal responsi-bility for the failure outcome than did low-social-anxiety subjects ( / > < . 0 4 ) , but therewere no significant differences between theattributions of high- and low-social-anxiety

Table 2Means for the Composite Index of Subjects'Attribution to Self, Experiment 1

SuccessFailure

Immediateevaluation

Social anxiety

High Low

5.33 6.156.83 5.24

Delayedevaluation

Social anxiety

High Low

5.44 5.754.75 5.96

Note. The higher the score, the greater subjects'self-attribution for the outcome of the therapy.

subjects within the success outcome condi-tions.

There was no hint of the predicted outcomemain effect within the delayed evaluation con-ditions (p > .30). Although high-social-anxi-ety subjects assumed slightly more personalresponsibility for success than for failureacross the three measures of personal responsi-bility, low-social-anxiety subjects did not.However, this interaction of social anxietyand outcome within the delayed evaluationconditions did not approach significance (p> .30). The social anxiety main effect wasalso nonsignificant (p > .10).

Not surprisingly, given the analyses re-ported above, separate analyses of variancewithin levels of social anxiety revealed theexpected Outcome X Level of Evaluationinteraction among high-social-anxiety subjects,F( l ,36) = 6.29, p < .02, but no significanteffects whatsoever among low-social-anxietysubjects (ps> .20). As a result of the pat-tern of the interaction among high-social-anxiety subjects, a main effect of level ofevaluation also emerged in the analysis ofthese subjects' responses, F(l, 36) = 4.22,p < .05.

Discussion

The results of this study offer strong sup-port for Bradley's (1978) self-presentationframework for explaining some reversals ofthe self-serving attribution bias. As Bradley

3 For individual planned comparisons, two-tailedt tests were used.

28 R. ARKIN, A. APPELMAN, AND J. BURGER

noted, although individuals typically assumegreater personal responsibility for successfulthan for failing outcomes at tasks, there existseveral reversals of this trend. In the presentstudy, high-social-anxiety subjects presentedthemselves in a favorable light, assumingslightly more personal responsibility for suc-cess than for failure, when only the experi-menter was to evaluate their behavior, butpresented themselves in a more modest man-ner, assuming far more personal responsibilityfor failure than for success, when severalhigh prestige others were to join the experi-menter in evaluating their behavior (in theirpresence). This finding supports Bradley's(1978) contention that intense social evalua-tion could induce a reversal in the typicalself-serving attribution bias. As anticipated,low-social-anxiety subjects, who were not ex-pected to view the immediate evaluation as sothreatening, were much less affected by themanipulation of intensity of evaluation.

Although the present reversal of the self-serving attribution bias supports Bradley's(1978) general self-presentation framework,this finding taken alone provides little to aidin an understanding of the processes under-lying such reversals in attribution of responsi-bility for success and failure. In particular, itis not clear whether high-social-anxiety sub-jects reported their privately accepted percep-tions of causality in the immediate evalua-tion conditions, despite their unflattering na-ture, but reported a more flattering portrayalof their personal responsibility under condi-tions of less intense evaluation, or vice versa.Both interpretations seem reasonable.

Jones and Wortman (1973) have arguedthat a modest presentation of self can be aneffective self-presentational strategy. High-social-anxiety subjects presented themselvesas far more responsible for the blameworthyevent in the immediate than in the delayedevaluation conditions, perhaps as a strategyfor fostering an impression of their ownmodesty and even for receiving vindicationfrom the prestigious others when the video-tape was analyzed later. This analysis wouldargue that high-social-anxiety subjects werepresenting themselves favorably, albeit mod-estly, in the immediate evaluation conditionsand that their attributions in the delayed

evaluation conditions could reflect anythingfrom self-presentation concerns to an unbiasedprocessing of the available information (cf.Miller & Ross, 1975).

Alternatively, the desire to avoid aversiveevents, such as embarrassment, among indi-viduals high in social anxiety (a "cost orien-tation") might lead these individuals to en-dorse attributions that matched as closely aspossible the true nature of events. By en-dorsing attributions that would match others'perception of the true nature of events, sub-jects could expedite the discomfiting interac-tion with the others by ensuring a conflict-free, issue-free interaction. Perhaps just astest-anxious persons attribute their worry andinability to focus attention on a task (andsubsequent failure) to themselves (e.g.,Weiner & Sierad, 197S), high-social-anxietyindividuals may actually view their uncon-trollable anxious reactions in social situations(and consequent ineptitude) as due to them-selves. If so, high-social-anxiety subjects mayhave been portraying themselves honestly inthe immediate evaluation conditions but pre-senting themselves favorably in the delayedevaluation conditions.

Experiment 2 was conducted to shed furtherlight on the exact nature of high- and low-social-anxiety individuals' attributional be-havior. A second purpose of Experiment 2was to determine more precisely the contri-bution of self-presentation to the self-servingattribution bias generally.

Experiment 2

The bogus-pipeline technique was proposedby Jones and Sigall (1971) as a strategy forreducing distortion and dissimulation inverbal responses. The technique involves con-vincing subjects that the experimenter iscapable of detecting whether they are tellingthe truth, thus leading subjects to be morefrank in revealing socially undesirable infor-mation. As Gaes, Quigley-Fernandez, andTedeschi (1978) argue, it is presumably bet-ter from the subject's point of view to beperceived as having some undesirable char-acteristic or as holding some undesirablecognition than to be perceived as a liar. Be-cause it appears that the bogus-pipeline tech-

SELF-SERVING BIAS 29

nique actually reduces the tendency of sub-jects to present themselves in a sociallydesirable way, the measurement techniqueprovides a straightforward way to differenti-ate between the two interpretations (dis-cussed above) of the attributions reported bysubjects in Experiment 1.

Specifically, assuming that subjects wouldprefer to appear honest than to present acontrived presentation of self when their judg-ments are taken via the bogus pipeline, itcould be determined whether high-social-anxiety subjects privately accepted the attri-butions they reported in the delayed evalua-tion context (greater personal attribution forsuccess than for failure) and were intention-ally portraying themselves in a "modest"way in the immediate evaluation conditions(assuming greater personal responsibility forfailure than for success) or privately acceptedthe attributions they reported in the immedi-ate evaluation context but were portrayingthemselves in a more flattering, positive lightin the delayed evaluation conditions. To pro-vide a test between these two interpretations,it would only be necessary to replicate eitherthe findings in the delayed evaluation or im-mediate evaluation conditions of Experiment1 and include another set of conditions identi-cal in every respect, save the techniques usedfor collecting the dependent measures (i.e., thebogus pipeline).

The decision to replicate the immediateevaluation conditions was recommended bythe fact that the significant effects in Experi-ment 1 derived largely from the immediateevaluation conditions4 and because the pro-cedures in these conditions were slightly lesscomplex to carry out. Assuming that the find-ings in the immediate evaluation conditionsof Experiment 1 were essentially replicatedin the no-bogus-pipeline conditions of Ex-periment 2, an interaction of outcome, socialanxiety, and measurement technique (withthe pattern of the means in the bogus-pipe-line condition of Experiment 2 appearing simi-lar to the pattern in the delayed evaluationconditions of Experiment 1) would suggestthat subjects privately accepted the moreflattering portrayal of their responsibility buthad been overly modest when their behaviorwas closely scrutinized, presumably for self-

presentational advantage. However, if theresponses of subjects in the bogus-pipelineconditions mirrored the responses of subjectsresponding conventionally, this finding wouldsuggest that high-social-anxiety subjects pri-vately accepted their unflattering, modestattributions but would present themselvesmore favorably if given the opportunity. Adifferent experimental context or paradigm,as successful in producing the typical "self-serving" pattern of performance attributionsas the paradigm employed in Study 1 (Wells,Petty, Harkins, Kagehiro, & Harvey, 1977;Arkin, Petty, Burger, Pummill, & Reaves,Note 1), was used in Experiment 2 in orderto provide some evidence for the generalityof the attributional findings across tasks andcontexts.

Method

Subjects and procedure. A total of 40 high-(uppermost third of the distribution) and 40 low-(lowest third) social-anxiety female undergraduateswere recruited by phone and were randomly as-signed to the experimental conditions by an experi-mental assistant. Thus, the experimenter was keptblind to the subject's level of social anxiety.

Immediately after the subject arrived at the lab-oratory, she was joined by a female confederateposing as another subject. The experimenter gavethem each a short questionnaire to complete, indi-cating that another researcher had asked him toobtain responses from as many individuals as pos-sible. The questionnaire consisted of several itemsanchored on 9-point scales concerning currentsocial and political issues and personal matters (e.g.,how nervous she felt). While completing the ques-tionnaire, the confederate surreptitiously observedand then memorized the subject's responses. Com-pleted questionnaires were collected and placed in afolder that was set aside later.

The participants were then escorted into theexperimental room and were seated at a table. Itwas explained that the study concerned "the inter-action processes involved with a cooperative visualperception and manual dexterity task" and that thespecific task involved placing map pins of six dif-ferent colors on a corkboard so as to correspond toa geometric design (see Wells et al., 1977). Theexperimenter then conducted a (rigged) drawingthat assigned the confederate the role of "worker"and the subject the role of "assistant." The workerwas told that she would have 2 minutes to replicatethe geometric design in terms of its size, shape, and

* If the delayed evaluation conditions were repli-cated and the "mirror image" hypothesis (describedbelow) were supported, the data would be exceed-ingly difficult to interpret.

30 R. ARKIN, A. APPELMAN, AND J. BURGER

color pattern. The assistant was told that her taskwas to aid the worker in any way possible, such ashanding the worker pins or providing suggestions,without actually placing pins on the corkboard worksurface herself.

Just prior to beginning the task, the experimenterexplained that each session was videotaped so thata group of PhD candidates in clinical psychologycould evaluate the session more effectively. It wasexplained that the researchers had found it mostefficient to evaluate the session with the subjectspresent to answer questions and assist in any waybut that it had proved too distracting to try toevaluate the session while subjects were actuallyperforming the task. Subjects were then told thata team of three graduate students was available toevaluate the session during the second half of theexperimental hour and that they would be arrivingsoon. Following these instructions, the Sony video-tape apparatus was activated and the camera wasfocused on the work surface.

After allowing the participants a few seconds toview the geometric design, the experimenter beganthe timed task. The worker performed the task at apredetermined pace and only partially completed thedesign.

Following the Wells et al. (1977) and Arkin et al.(Note 1) procedures, the participants were informedafter the 2-minute period that the primary interestof the study actually concerned the helping be-havior of the assistant, and the experimenter askedthe assistant whether she would be interested inknowing how she performed in comparison withearlier participants. Every assistant expressed adesire to view the data.5

Outcome manipulation. After consulting his rec-ords, the experimenter pointed out on a chart thatthe assistant's performance was actually quite poor(15th percentile) or quite good (85th percentile)relative to the performance of prior participants. Theexperimenter was kept blind to this outcome manip-ulation until it was actually delivered.

Measurement technique manipulation. In everycondition the confederate was then given a set ofquestionnaires and was escorted to a nearby room.During this time the experimenter checked his notesto determine whether the subject was assigned tothe bogus-pipeline or to the conventional responseformat (paper-and-pencil) conditions. In the bogus-pipeline conditions, the subject was seated in frontof an impressive array of electronic gadgetry, in-cluding dials, meters, and switches, which the ex-perimenter identified as an elcctromyograph (EMG).Following the procedures of Jones and Sigall (1971)and Quigley-Fernandez and Tedeschi (1978) closely,the use of the EMG was introduced as part of asecond research program concerning the degree towhich individuals possessed good or poor contactwith their inner feeling states. Subjects were toldthat the EMG was a superior form of lie detectorand that the equipment measured both the directionand intensity of attitudes, feelings, or judgments.The experimenter then applied electrode paste andattached electrodes to the hand and wrist of the

subject. It was explained that the information ana-lyzed by the EMG was recorded on tape output bya computer located next door and would also appearin a somewhat more raw form on the EMG OUTPUTscreen (an oscilloscope adapted for use in thepresent study).

It was pointed out that the EMG sometimesrequired adjustments due to individual differencesin baseline responses. "In order to calibrate theequipment," subjects were asked to respond againto some of the same items they had been admin-istered earlier in the attitude survey. The subjectwas given a blank copy of the "attitude survey"and was asked to select any three of the items forthe test that she wished. After the subject had indi-cated three items, the experimenter read the firstaloud and waited until the subject had dialed in herresponse before flipping a prominently labeledswitch that supposedly "activated the analyzingprocess" of the EMG. The confederate, who waslistening over a concealed intercom to hear thequestion selected and the click of the "activator"switch, manipulated the oscilloscope reading to beabout one-half point from the subject's actual earlierresponse. The experimenter then retrieved the sub-ject's attitude questionnaire from its folder, both todemonstrate to the subject that her earlier responseshad not been tampered with and to demonstratethe accuracy of the EMG. After the experimenterfeigned adjusting the equipment, the EMG readingsfor the next two questions were near-exact copiesof the subject's earlier responses.

After establishing the supposed validity of theEMG, the experimenter turned the output screenaway from the subject's view, gave the subject thedependent measures questionnaire, and instructedher to record her response on the questionnaire first,then dial in her response (using the 9-point or 100-point register, as appropriate) for analysis by theEMG, and finally throw the activator switch. Theexperimenter watched the output screen and wrotesome comments while the subject responded to theitems.

In the no-bogus-pipeline conditions, the experi-menter arranged for two brief delays in order toequate the time elapsed from completing the taskto answering the questionnaire in the no-pipelineand bogus-pipeline conditions. Subjects then com-pleted their questionnaires in the conventionalpaper-and-pencil format.

Dependent measures. The first measure askedsubjects to divide 100 points between the followingtwo explanations of their behavior:

I chose to help as much or as little as I didbecause of SOME CHARACTERISTIC THAT I POSSESS.

I chose to help as much or as little as I did be-cause of SOME ASPECT OF THE SITUATION IN WHICHI POUND MYSELF.

5 A more complete description of the procedurescan be found in Wells et al. (1977) or in Arkinet al. (Note 1).

SELF-SERVING BIAS 31

Subjects were also asked to report (on 9-pointscales) the extent to which they felt three specificdispositional factors (her friendliness, her apathy, hersense of responsibility) affected her level of helpful-ness. The difference between the dispositional andsituational score on the 100-point index and thesum of each subject's scores on the three specificattribution items constituted the two dependentmeasures. Finally, subjects were asked to respond toseveral manipulation check items (as described inthe results), each anchored on a 9-point scale. Aftercompleting the questionnaire, subjects were thor-oughly debriefed, thanked, and dismissed.

Results

Manipulation checks. The outcome ma-nipulation was highly successful. Subjects inthe success outcome conditions indicated thatthey were more helpful than most previoussubjects (M = 7.60), whereas subjects in thefailure outcome conditions indicated that theywere less helpful (M = 1.78), F ( l , 7 2 ) =807.60, p < .0001.

Successful subjects also rated their per-formance on the pin-placing task much higher(M=6.88) than did unsuccessful subjects(M = 2.8S), /?(!, 72) = 243.26, p < .0001.Both high- and low-social-anxiety subjectsalso rated their performance better in theno-bogus-pipeline conditions (M = S.23) thanin the bogus-pipeline conditions (M = 4.90),F(l, 72) = 7.89, p < .007, a finding that isentirely consistent with the attribution datato be reported below. This tendency wasstronger for high- (Ms = 5.55, 4.25) than forlow-social-anxiety subjects (Ms = 4.90, 4.75),leading to a significant interaction of socialanxiety and measurement technique, F(l, 72)= 4.96, p < .03.

Subjects in the bogus-pipeline conditionsrated the EMG as highly accurate in evaluat-ing their judgments (M = 7.55) and rela-tively error free (M = 2.83), and there wereno significant differences between the ratingsof high- and low-social-anxiety respondents orsuccessful and unsuccessful subjects on theseitems. Subjects in the bogus-pipeline condi-tions reported feeling less comfortable in theirassigned roles (M — 5.78) than did subjectsin the no-pipeline conditions (M = 6.83),F ( l , 7 2 ) = 6.38, p < .02, as expected. Notsurprisingly, subjects in the success outcomeconditions (M = 7.28) reported feeling more

Table 3Means for the Standardized Index ofAttribution to Self, Experiment 2

Bogus pipeline No bogus pipeline

Social anxiety Social anxiety

Outcome High Low High Low

SuccessFailure

.08 .97

.32 -.62.29 1.40.14 -2.33

Nate, The greater the score, the greater the subjects'self-attribution for their performance at the task.

comfortable than did subjects in the failureoutcome conditions (M = 5.33), .F(l, 72) =21.99, p < .001. And fittingly, given theattribution data to be reported below, therealso emerged an interaction of outcome andmeasurement technique, F(l , 72) = 8.33, p <.006. Subjects who failed and then madejudgments on the bogus pipeline were far lesscomfortable (M — 4.20) than either subjectswho succeeded and reported judgments on thebogus pipeline (M = 7.35) or subjects in theno-bogus-pipeline conditions who succeeded(M ~ 7.20) or failed (M - 6.45).

Finally, high-social-anxiety subjects ratedthemselves on the initial questionnaire asfeeling more nervous (1 = very nervous; 9 =not at all nervous) as the experiment began(M — 5.63) than did low-social-anxiety sub-jects (M = 6.57), F(l, 72) = 5.33, p < .03.

Attribution of responsibility. For ease ofpresentation and because the two dependentmeasures were highly correlated, the two mea-sures were converted to standard scores andwere summed, creating a single index ofdispositional attribution. Analysis of varianceof subjects' scores on this attribution indexrevealed a significant interaction of socialanxiety and outcome, F(l, 72) = 16.63, p <.001, and no hint of a Social Anxiety X Out-come X Measurement Technique interaction,F( l , 72) = 1.42, p > .20. As can be seen fromthe means displayed in Table 3 and Figure 1,high-social-anxiety subjects assumed slightlymore personal responsibility for failure thanfor success ( n s ) , whereas low-social-anxietysubjects assumed far more responsibility forsuccess than for failure (p < .002). High-social-anxiety subjects assumed significantly

32 R. ARKIN, A. APPELMAN, AND J. BURGER

Dispositional

StandardizedIndex ofDispositionalAttribution

Situational

No Bogus Pipeline

Success Failure

Figure 1. Mean dispositional attribution as a func-tion of social anxiety and outcome.

more responsibility for the failure outcomethan did low-social-anxiety subjects (p <.04), whereas low-social-anxiety subjects as-sumed somewhat more responsibility for thesuccess outcome than did high-social-anxietysubjects ( n s ) .

An interaction of outcome and measure-ment technique also emerged in the analysisof variance, F(\, 72) = 4.12, p < .05. As canbe seen in Figure 2, high- and low-social-anxiety subjects collectively assumed morepersonal responsibility for success than forfailure only in the no-bogus-pipeline condi-tions (p < .02), assuming only slightly moreresponsibility for success than for failure whentheir judgments were taken via the bogus-pipeline (ns). Within success and within fail-ure the differences between the bogus-pipelineand no-pipeline conditions were not signifi-cant. However, an inspection of the means inTable 3 reveals that the effect of the mea-surement technique manipulation was muchgreater for low- than for high-social-anxietysubjects. The interaction of outcome andmeasurement technique was significant amonglow-, F( 1,36) = 5.77,/> < .03, but not amonghigh-social-anxiety subjects (F < I ) . Low-social-anxiety subjects assumed significantlyless personal responsibility for the failureoutcome in the no-bogus-pipeline than in the

bogus-pipeline conditions (p < .03), whereasthey assumed slightly more personal responsi-bility for the success outcome in the no-bogus-pipeline conditions than in the bogus-pipeline conditions ( n s ) . Although an outcomemain effect emerged in the overall 2 X 2 X 2analysis of variance, F(l, 72) = 13.55, p <.0004, this effect was also entirely attributa-ble to the low-, F(l, 36) = 33.31, p < .0001,rather than the high-social-anxiety subjects(F< 1).

General Discussion

The results of Experiment 2 clarify thefindings of Experiment 1 considerably. Thetendency of high-social-anxiety subjects toassume more personal responsibility for fail-ure than for success was essentially repli-cated (most prominently when subjects'attributions were recorded via the boguspipeline rather than the traditional paper-and-pencil technique). These data argue thathigh-social-anxiety subjects privately acceptedthe "unflattering, modest" attributions theyreported and suggest that subjects in theimmediate evaluation conditions of Experi-ment 1 had been responding honestly ratherthan self-presenting when they assumed

Dispositional 2 r

StandardizedIndex ofDispositional 0Attribution

Situational -2

0.23

High Social Anxiety

Low Social Anxiety

Success Failure

Figure 2. Mean dispositional attribution as a func-tion of measurement technique and outcome.

SELF-SERVING BIAS 33

greater personal responsibility for failurethan for success.

High-social-anxiety subjects' perception ofthemselves as more personally responsible forfailure than for success may derive fromtheir severely debilitating reactions to evalu-ative social situations.8 Brockner (1979) re-cently found that high-social-anxiety individ-uals were overly self-preoccupied when theywere confronted with an evaluative contextand were therefore unable to devote suf-ficient attention to the task to perform aswell as subjects low in social anxiety (whowould not experience such a debilitating re-action to social evaluation). It thereforeseems reasonable to assume that high-social-anxiety subjects in the present experiments,like Brockner's subjects, were overly anxiousor self-preoccupied and that they would haveperformed less well than their low-social-anxiety counterparts, had the experimentaltask been a real one. (High-social-anxietysubjects did report feeling more nervous atthe outset of the experiment than did low-social-anxiety subjects.)

There was also considerable evidence ofself-presentation uncovered in the two ex-periments reported here, however. High- andlow-social-anxiety subjects alike rated theirperformance better in the no-bogus-pipelinethan in the bogus-pipeline conditions of Ex-periment 2 and reported a much more flat-tering causal interpretation of events in theno-pipeline than in the pipeline conditions.However, this attenuation of the self-servingattribution bias in the bogus-pipeline condi-tions was due primarily to low-social-anxietyparticipants, who assumed significantly lesspersonal responsibility for the failure andslightly more responsibility for the successoutcome in the no-bogus-pipeline than in thebogus-pipeline conditions. High-social-anxietyparticipants were much less affected by themeasurement technique manipulation.

This failure to uncover effects of the mea-surement technique among high-social-anxi-ety participants could reflect a ceiling effect.That is, the evaluation by the committee ofexperts may have been sufficient to causehigh-social-anxiety participants to endorsepublicly their private conception of the truenature of events. High-social-anxiety partici-

pants may have been highly sensitive andresponsive to the ability of the experts todistinguish honest from dishonest responsesand perhaps challenge publicly ratings thatdid not appear to match reality. Thus, be-cause subjects in every condition of Experi-ment 2 expected to interact with the com-mittee of experts, it makes sense that high-social-anxiety participants' attributions shouldhave been affected little, if at all, by themeasurement technique manipulation. In con-trast, low-social-anxiety participants, whowere not expected to be as sensitive to thethreat of embarrassment posed by the com-mittee of experts, may have been less in-clined to be accurate rather than personallyflattering in the face of the impending evalu-ation than were high-social-anxiety subjects.In sum, less "cost orientation" among low-than among high-social-anxiety individualscould account for the greater impact of themeasurement technique manipulation onthem.

Alternatively, it may be that low-social-anxiety participants were not merely less"cost oriented" than high-social-anxiety sub-jects, but were actually more actively "re-ward oriented" when they expected to inter-act with the committee. In Experiment 1,low-social-anxiety subjects presented them-selves in a somewhat more favorable lightwhen the evaluation was scheduled to takeplace than when it was canceled, suggestingthat their self-presentation concerns mayhave been aroused by the social evaluation.A reward orientation toward the committeeevaluation could have led low-social-anxietyparticipants to present themselves favorablywhen it was both possible and useful to doso (immediate evaluation conditions of Ex-periment 1; no-bogus-pipeline conditions of

6 Brockner and Hulton (1978) found no signifi-cant attributional differences between individualshigh and low in self-esteem, a construct that Brock-ner (1979) has argued is comparable to, if notisomorphic with, social anxiety. However, becausethe procedures in the Brockner and Hulton studywere most comparable to the delayed evaluationconditions of Experiment 1 reported here, their find-ings are entirely consistent with the present find-ings. Moreover, Ickes and Layden (1978), Fitch(1970), and others have found such effects.

34 R. ARKIN, A. APPELMAN, AND J. BURGER

Experiment 2) but not when it was not pos-sible (bogus-pipeline conditions of Experi-ment 2) or not useful (delayed evaluationconditions of Experiment 1) to do so.

In fairness, however, the self-serving bias(i.e., individuals assuming more responsi-bility for success than for failure) was evi-dent in the responses of low-social-anxietysubjects in the bogus-pipeline conditions. Ifit is assumed that the bogus-pipeline mea-surement technique does substantially reduce,if not entirely eliminate, distortion and dis-simulation in verbal responses, then it mustbe concluded that low-social-anxiety individ-uals were portraying themselves honestly inassuming somewhat greater responsibility forthe success than for the failure outcome (andthat high-social-anxiety subjects were por-traying themselves honestly in assumingsomewhat greater responsibility for the fail-ure than for the success outcome). Of course,low-social-anxiety individuals' (honest) rat-ings of themselves as more personally respon-sible for success than for failure could reflectany number of processes (see Arkin & Maru-yama, 1979, p. 92; Ickes & Layden, 1978).But the present data argue minimally thatwhatever the exact process responsible, thesesubjects' attributions reflect their privatelyaccepted perceptions. Thus, among low-social-anxiety subjects it is only the sizable exag-geration of the self-serving bias in the no-bogus-pipeline conditions that seems to bedue unambiguously to self-presentation.

pipeline conditions; delayed evaluation con-ditions) than when they were precluded fromdoing so (bogus-pipeline conditions; imme-diate evaluation conditions). In short, par-ticipants portrayed themselves in a far moremodest fashion, apparently reporting theirprivate and personal view of events, whenan overly positive portrayal of events waslikely to be quite embarrassing (because itwould be revealed to be untrue) than whenthis likelihood was small. Finally, comple-menting earlier research (e.g., Lepper, 1970),there was evidence that high-social-anxietysubjects were "cost oriented" and some hintsthat low-social-anxiety subjects were "rewardoriented" in their attempts to win the ap-proval and avoid the disapproval of others(i.e., self-present). Specifically, high-social-anxiety participants responded to social eval-uation with a modest presentation of self,reflecting an attempt to avoid embarrassmentrather than win approval, whereas low-social-anxiety participants tended to respond to so-cial evaluation with a favorable presentationof self, perhaps reflecting an attempt to winapproval rather than avoid embarrassment.

Reference Note

1. Arkin, R. M., Petty, R. E., Burger, J. M., Pum-mill, R. A., & Reaves, B. Anticipated discussion,bystander status and knowledgeability, and theelimination oj the actor-observer divergence. Pa-per presented at the meeting of the MidwesternPsychological Association, Chicago, May 1978.

Conclusions

The present studies generally confirm thehypothesis that the self-serving bias in causalattribution and its occasional but systematicreversal are in part a reflection of self-pre-sentation, or a desire to maintain or createa favorable impression. Specifically, whetherindividuals privately believed that they weremore personally responsible for success thanfor failure (low-social-anxiety individuals) ormore personally responsible for failure thanfor success (high-social-anxiety individuals)in the present context, they portrayed them-selves in a far more personally flattering waywhen it was possible to do so (no-bogus-

References

Arkin, R. M., Gleason, J. M., & Johnston, S. Effectsof perceived choice, expected outcome, and ob-served outcome of an action on the causal attri-butions of actors. Journal of Experimental SocialPsychology, 1976, 12, 151-158.

Arkin, R. M., & Maruyama, G. M. Attribution,affect, and college exam performance. Journal ofEducational Psychology, 1979, 71, 85-93.

Bradley, G. Self-serving biases in the attributionprocess: A recxamination of the fact or fictionquestion. Journal oj Personality and Social Psy-chology, 1978, 36, S6-71.

Brockner, J. Self-esteem, self-consciousness, and taskperformance: Replications, extensions, and pos-sible explanations. Journal of Personality andSocial Psychology, 1979, 37, 447-461.

Brockner, J., Hulton, A. J. B. How to reverse thevicious cycle of self-esteem: The importance of

SELF-SERVING BIAS 35

attentional focus. Journal of Experimental SocialPsychology, 1978,15, 564-578.

Carlsmith, J. M., Lepper, M. R., & Landauer, T. K.Children's obedience to adult requests: Interactiveeffects of anxiety, arousal, and apparent punitive-ness of the adult. Journal of Personality andSocial Psychology, 1974, 30, 822-828.

Federoff, N. A., & Harvey, J. H. Focus of atten-tion, self-esteem, and attribution of causality.Journal of Research in Personality, 1976, 10,336-345.

Fenigstein, A., Scheier, M. F., & Buss, A. H. Pri-vate and public self-consciousness: Assessment andtheory. Journal of Consulting and Clinical Psy-chology, 1975, 43, 522-527.

Fitch, G. Effects of self-esteem, perceived perform-ance, and choice on causal attributions. Journalof Personality and Social Psychology, 1970, 16,311-315.

Frcedman, J. L., Wellington, S. A., & Bless, E.,Compliance without pressure: The effect of guilt.Journal of Personality and Social Psychology,1967, 7, 117-124.

Gaes, G. G., Quigley-Fernandez, B., & Tedeschi, J.T. Unclogging the bogus pipeline: A critical re-analysis of the Cherry, Byrne & Mitchell study.Journal of Research in Personality, 1978, 12,189-192.

Harvey, J. H., Arkin, R. M., Gleason, J. M., &Johnston, S. Effects of expected and observedoutcome of an action on the differential causalattributions of actor and observer. Journal ofPersonality, 1974, 42, 62-77.

Hass, R. G., & Mann, R. W. Anticipatory beliefchange: Persuasion or impression management.Journal of Personality and Social Psychology,1976, 34, 105-111.

Heider, F. The psychology of interpersonal rela-tions. New York: Wiley, 1958.

Ickes, W., & Layden, M. A. Attributional styles.In J. H. Harvey, W. Ickes, & R. F. Kidd (Eds.),New directions in attributional research (Vol. 2).Hillsdale, N.J.: Erlbaum, 1978.

Janis, I. L., & Mann, L. Decision making: A psy-chological analysis of conflict, choice, and com-mitment. New York: Free Press, 1977.

Jellison, J. M., & Arkin, R. M. Social comparisonof abilities: A self presentation approach to de-cision making in groups. In J. M. Suls & R. L.Miller (Eds.), Social comparison processes: Theo-retical and empirical perspectives. Washington,D.C.: Hemisphere, 1977.

Jones, E. E., & Sigall, H. The bogus pipeline: Anew paradigm for measuring affect and attitude.Psychological Bulletin, 1971, 76, 349-364.

Jones, E. E., & Wortman, C. Ingratiation: An at-

tributional approach. New York: General Learn-ing Press, 1973.

Krantz, D. S. The social context of obesity re-search: Another perspective on its place in thefield of social psychology. Personality and SocialPsychology Bulletin, 1978, 4, 177-184.

Lepper, M. R. Anxiety and experimenter valenceas determinants of social reinforcer effectiveness.Journal of Personality and Social Psychology,1970, 16, 704-709.

McPeek, R. W., & Cialdini, R. B. Social anxiety,emotion, and helping behavior. Motivation andEmotion, 1977, ;, 225-233.

Miller, D. T., & Ross, M. Self-serving biases inthe attribution of causality: Fact or ficton? Psy-chological Bulletin, 1975, 82, 213-225.

Quigley-Fernandez, B., & Tedeschi, J. T. The boguspipeline as a lie detector: Two validity studies.Journal of Personality and Social Psychology,1978, 36, 247-256.

Reis, H. T., & Gruzen, J. On mediating equity,equality, and self-interest: The role of self-pre-sentation in social exchange. Journal of Experi-mental Social Psychology, 1976, 5, 487-503.

Ross, L. The intuitive psychologist and his short-comings. Distortions in the attribution process.In L. Berkowitz (Ed.), Advances in experimentalsocial psychology (Vol. 10). New York: AcademicPress, 1977.

Snyder, M. Self-monitoring. In L. Berkowitz (Ed.),Advances in experimental social psychology (Vol.12). New York: Academic Press, in press.

Snyder, M. L., Stephan, W. G., & Rosenfield, D.Attributional egotism. In J. H. Harvey, W. Ickes,& R. F. Kidd (Eds.), New directions in attri-butional research (Vol. 2). Hillsdale, N.J.: Erl-baum, 1978.

Tedeschi, J. T., Schlenker, B. R., & Bonoma, T. V.Cognitive dissonance: Private ratiocination orpublic spectacle? American Psychologist, 1971,26, 685-695.

Weary, G. An examination of affect and egotismas mediators of bias in causal attributions. Jour-nal of Personality and Social Psychology, in press.

Weiner, B., & Sierad, J. Misattribution for failureand enhancement of achievement strivings. Journalof Personality and Social Psychology, 1975, 31,415-421.

Wells, G. L., Petty, R. E., Harkins, S. G., Kage-hiro, D., & Harvey, J. H. Anticipated discussionof interpretation eliminates actor-observer differ-ences in the attribution of causality. Sociometry,1977, 40, 247-253.

Received October 23, 1978