Self-Serving Biases in the Attribution of Causality:...

Psychological Bulletin1975, Vol. 82, No. 2, 213-225

Self-Serving Biases in the Attribution of Causality:Fact or Fiction?

Dale T. Miller and Michael RossUniversity of Waterloo, Ontario, Canada

A review of the evidence for and against the proposition that self-serving biasesaffect attributions of causality indicated that there is little empirical supportfor the proposition in its most general form. The literature provides somesupport for the contention that individuals engage in self-enhancing attributionsunder conditions of success, but only minimal evidence was found to suggestthat individuals engage in self-protective attributions under conditions offailure. Moreover, it was proposed that the self-enhancing effect may not bedue to motivational distortion, but rather to the tendency of people (a) toexpect their behavior to produce success, (b) to discern a closer covariationbetween behavior and outcomes in the case of increasing success than in thecase of constant failure, and (c) to misconstrue the meaning of contingency.

We are prone to alter our perception of causality soas to protect or enhance our self esteem. We attributesuccess to our own dispositions and failure to ex-ternal forces. (Hastorf, Schneider, & Polefka, 1970,p. 73)

The principle described in the foregoingobservation has received considerable atten-tion in social psychology. Variously labeledego-defensive, ego-protective, or ego-biased at-tribution, discussions of the phenomenon canbe found in source books (Heider, 1958;Kelley, 1967), text books (Hastorf et al.,1970; Weiner, 1973), and innumerable jour-nal articles (e.g., Beckman, 1973; Schopler &Layton, 1972a, 1972b; Wolosin, Sherman, &Till, 1973).

The concept of a self-serving attributionprocess owes its existence, in part at least, toHeider's (1958) "naive analysis of action"model. According to Heider, the selection ofan acceptable causal attribution depends ontwo factors: "(1) The reason has to fit thewishes of the person and (2) the datum hasto be plausibly derived from the reason" (p.172). Thus, assuming it is plausible to do so,

The writing of this article was supported by aCanada Council Doctoral Fellowship to Dale Millerand a Canada Council Research Grant to MichaelRoss. The authors would like to thank Don DiTecco,Dick Eiser, John Holmes, and Mel Lerner for theircomments on an earlier version of this manuscript.

Requests for reprints should be sent to Dale T.Miller, who is now at the Department of Psychology,University of Western Ontario, London N6A 3K7,Ontario, Canada.

we try to explain our behavior in terms that"flatter us" and "put us in a good light."This formulation is, of course, quite consistentwith social psychological theories predicatedupon self-esteem maintenance (e.g., balancetheory, dissonance theory). The principle ofego-biased attribution is also compatible withKelley's (1971) notion of "effective control."Kelley, postulating that it is important forindividuals to be able to exercise control overtheir environment, described the relationshipbetween the need for effective control andthe attribution process as follows: "The at-tribution to self of success and the attributionto external factors of failure provides for thecontinuation of control attempts" (p. 23).

While most discussions of self-serving biasesin perception of causality have assumed theirexistence, some authors have questioned theprinciple's theoretical and empirical under-pinnings (Bern, 1972; Kelley, 1971). No-where in the extant literature, however, hasthe principle's empirical base been criticallyexamined; and it is to this task that we ad-dress ourselves in succeeding portions of thepresent article. The major thrust of this ex-amination is to compare the explanatorypower of the self-serving attributional analy-sis with that of a nonmotivational, informa-tion-processing analysis. It should be notedat this time that the biased-attribution prin-ciple is double-edged and to receive full sup-port it must be shown that people indulge

213

214 DALE T. MILLER AND MICHAEL ROSS

both in self-protective attributions under con-• ditions of failure and in self-enhancing attri-butions under conditions of success.1

The empirical data that are most oftenadduced in support of the existence of self-serving attributional biases come from twogeneral traditions. The first is represented bya series of interpersonal influence studieswhich examined the relationship betweencausal ascription and the success of influenceattempts. Studies of the relationship betweencausal ascriptions and outcomes in skill-oriented performance tasks provide the secondsource of relevant data.

INTERPERSONAL INFLUENCE RESEARCHIn the experiments described in this section,

subjects are asked to instruct another personon how to perform at a particular task. Thesubjects are informed that following theirintervention, the target person either suc-ceeded or failed at the activity. The subjects'perceptions of the causal determinants of thetarget person's performance are then assessed.

Of the research employing an interpersonalinfluence paradigm, three studies involved aquasi teacher-student relationship. The proto-typic experiment in this area (Johnson,Feigenbaum, & Weiby, 1964) required edu-cational psychology students to teach arith-metic concepts to fourth-grade boys. Eachteacher explained to two pupils how to multi-ply by 10 and then saw that Pupil A haddone well but that Pupil B had done poorly.

1 The scope of the present review has been limitedto studies which assess causal attributions followingsuccess and failure. The concept of self-protectivebiases is also integral to a body of literature whichfocuses on causal attributions for another's accidentalharm doing (Chaikin & Darley, 1973; Shaver,1970a, 1970b; Walster, 1966, 1967). Defensive at-tribution, as the phenomenon has been labeled, im-plies that an individual often distorts his ascriptionsof responsibility for an accident in order to protecthimself against the threatening possibility that (a)a similar accident could happen to him or (b) atsome future time he could himself be responsiblefor someone's suffering. Unlike the defensive attribu-tion literature, the studies described in the currentreview center on the individual's causal attributionsfor the consequences of his own behavior. It isinteresting to note that the empirical and theoreticalbases of the defensive attribution phenomenon havealso recently been questioned (Fishbein & Ajzen,1973; Kelley, 1973; Ross & DiTecco, in press; Vidmar& Crinklaw, 1974).

(Actually there were no pupils and the feed-back was programmed.) Subjects then taughttheir pupils to multiply by 20. On receivingtheir subsequent work sheets, subjects learnedthat A had continued to do well and B hadeither (a) continued to do badly or (b) im-proved to do as well as A. When asked toaccount for the performances of the two pupilsin an open-ended format, the teachers tendedto accept responsibility for improved per-formance but to blame continued low per-formance on B himself.

While this study is ubiquitously cited as ademonstration of self-serving biases in attribu-tion, a careful examination of the data anddesign fails to bear this out. To begin with,it is not obvious why a motivational conceptof any description needs to be invoked toexplain these data. Since all subjects simul-taneously taught a student who performed ina consistently successful manner, they knewthat their behavior did not predictably pro-duce low performance. With this informationbefore them, it would seem only rational todraw the inference that a poor performanceby B was not due solely to their influence.Further support for a nonmotivational inter-pretation comes from a consideration ofKelley's (1971) "covariation principle." Ac-cording to Kelley's model of attribution, "aneffect is attributed to the one of its possiblecauses with which, over time, it covaries" (p.3). Within the present context, therefore, itis conceivable that insofar as the teacher triedharder or varied her instructional methodafter B's initial poor performance, the in-formational pattern available to her wouldshow a strong positive covariation betweenher own behavior and that of the improvingstudent, but a negative or noncovariance be-tween her own behavior and that of the con-sistently poor student (Kelley, 1967). By thecovariation principle, the former would war-rant a self-attribution and the latter wouldnot.

Another feature orthe data which suggeststhat the subjects' attributions could representlogical inferences, and not ego-colored ones,is the finding that subjects did not creditthemselves with the consistently high per-formance of Child A. Child A's performancetended to be attributed to his eagerness, moti-

BIASES IN ATTRIBUTION 215

vation, ability, and so on. In other words,subjects took credit for a successful student'sperformance only when it was possible toperceive some correspondence between theirefforts and the student's performance.2

In a study reported by Beckman (1970),female educational students were again re-quired to teach arithmetic concepts and sym-bols to children. As in the Johnson et al.(1964) study, teachers instructed two stu-dents at a time, one of whom performed con-sistently well over a series of four trials.Beckman's design, however, diverged fromthe Johnson et al. study in two importantrespects. First, causal attributions were col-lected for both ascending and descendingperformance patterns. Thus, B's performancewas varied according to one of three pat-terns: (a) low initial and low final perform-ance, (b) low initial and high final perform-ance, or (c) high initial and low final per-formance. The second manner in which Beck-man's study differed from the earlier one wasthat she included observer subjects. Thesesubjects were asked to make causal attribu-tions on the basis of written transcripts ofthe teacher-student interaction.

The results of this study were generallyconsistent with those reported by Johnsonet al. (1964). Teachers tended to accept re-sponsibility for a student's performance whenthe student improved. When the student per-formed poorly, responsibility was attributedto various causes external to the teacher, suchas the child's motivation or situational factors.Interestingly, this was true whether the pat-tern for failure was low-low or high-low.

Further evidence relevant to the phenome-non of biased attribution comes from a com-parison of the causal attributions made byparticipant and observer subjects. The as-sumption underlying the inclusion of observ-ers in this experiment was that their attribu-tions would not be ego colored and wouldserve as a baseline against which to comparethe ego-biased attributions of the teachers.The only condition in which teachers andobservers differed significantly was the low-

2 It should be noted that Johnson et al. (1964)did not actually report the subjects' perceptions ofA. Consequently, the present description is basedsolely on those authors' discussions of the data.

high condition. Teachers attributed more cau-sal responsibility to self in this condition thandid observers.

To summarize, Beckman's study, like theJohnson et al. (1964) study, reported littleevidence to indicate a biased-attribution phe-nomenon. The fact that teachers creditedthemselves more than observers did in thelow-high condition suggests that somethinglike self-enhancing attribution may be oper-ative, but there is certainly no evidence inthis study to suggest that teachers engaged inego-protective attributions. Even the presenceof a self-enhancing attributional bias (as-sumed from the existence of participant-ob-server differences) can be questioned, how-ever. The major problem of interpretationconcerns the fact that observers did not actu-ally view the interaction but were providedonly with a stylized transcript. (Moreover,the transcript was somewhat misleading sinceit incorrectly implied that teachers and stu-dents were in face-to-face contact with eachother). Without the same evidence that wasavailable to the teachers, observers may havebeen unable to appreciate the contingenciesbetween the teacher's efforts and the student'sperformance. To return to Kelley's (1971) co-variance model, the attributional patterns ofthe observers may not have concurred withthose of the teachers because the observerdid not have the covariance data that theteacher herself had. It should be noted thatthis is a very difficult problem to resolve.While participant-observer comparisons seemto represent an effective means of testing forthe presence of self-serving attributions, onecan never be certain that participants andobservers are receiving, or attending to, thesame information in forming causal ascrip-tions (Jones & Nisbett, 1971).

A second study by Beckman (1973) im-proved upon her earlier one by having ob-server subjects view directly the teacher-student interaction. While there were onlyminor variations in design from Beckman'sfirst to her second study, the pattern of re-sults was strikingly different. Contrary to thefirst study, participants and observers did notdiffer in their causal attributions for the low-high condition. More importantly, participantsubjects assumed more personal responsibility


in the high-low condition than in the low-high condition. Observer judgments showedno such differentiation. In sum, the secondBeckman study not only failed to replicatethe apparent ego-enhancing effect in the pre-vious study, but actually found evidenceagainst defensive attribution; that is, teachersaccepted more responsibility for failure thanfor success (high-low versus low-high).3

The final study which examined the processof self-serving attributional biases within aninterpersonal influence context (Schopler &Layton, 1972a) employed a social judgmenttask. Each subject was paired with a partnerwho had previously performed either well orpoorly on the task. During the next set oftrials, subjects were asked to offer advice totheir partner while he performed the judg-ment test. The degree of interaction betweenthe adviser and the partner was held constantby limiting the adviser's communication tothe sending of recommendations which thepartner was free to accept or reject. Aftercompleting the second set of trials the sub-ject learned how well his partner had per-formed. The partner's performance was pre-sented as having remained the same (low-low or high-high) or as having changed (low-high or high-low).

The data from this study are consistentwith some aspects of the results obtained inresearch already described. As in Johnsonet al. (1964), subjects accepted less responsi-bility for the other's performance when it re-mained low (low-low) than when it improved(low-high). Unlike in the Johnson et al.study, however, subjects seemed to assignthemselves a high degree of responsibility inthe continuous success condition (high-high).Moreover, as in the Beckman (1973) experi-ment, subjects accepted considerable responsi-bility for a decrease in performance (high-low). (Examination of the interaction meanssuggests that the high-low condition yieldedsubstantially more self-attribution of responsi-bility than the low-low condition and notappreciably less than the low-high or high-high conditions). This self-attribution of re-sponsibility for a decrease in performance

3 An unpublished study by Ross, Bierbrauer, andPolly (Note 1), which employed actual school teach-ers in a similar situation, also failed to find evidencefor self-serving causal attributions.

makes it difficult to argue that either self-enhancing or self-protective attributions wereoperative. Nevertheless, it is necessary to con-sider why the high-high condition yieldedmore self-attribution of responsibility thanthe low-low condition. Schopler and Layton(1972b) suggested the possibility that thisdifference may be due, in part, to the subject'sexpectations:From A's point of view his own interventions maysimply appear to be more predictive of success thanfailure. Indeed, . . . A's interventions are typicallyintended to produce success and may actually bemore predictive of success than failure, (p. 10)

Support for this speculation comes from thework of Parducci (1963, 1965, 1968) on the"range-frequency compromise principle." Ex-amining a wide range of outcome judgments,Parducci has consistently found that peopleexpect the majority of outcomes to be positive.The work of Marks (1951) and Irwin (1953)also indicates that individuals tend to under-estimate the probability of a negative event.

On the whole, the interpersonal influencestudies seem to provide little evidence to sup-port the assertion that perceptions of causalityare distorted in the service of self-protectionor self-enhancement. The only finding to re-ceive considerable empirical support is thatpeople are more likely to perceive that theyhave influenced a target person's improve-ment than that they have produced his re-peated failure. As noted earlier, however, thisfinding need not be interpreted in motivationalterms. The self-enhancing effect is readilyinterpretable by an information-processingmodel which assumes that attributions aredetermined by intentions, expectations, andthe perceived covariation between behaviorand outcome.

RESEARCH ON ACHIEVEMENT TASKSDo people take responsibility for their suc-

cesses at a skill task, while attributing theirfailures to external factors such as bad luck,other persons, or the difficulty of the taskitself? Much recent research has addresseditself directly to this question, though therange of tasks employed has been relativelynarrow. The use of an anagrams task hasbeen particularly popular (Davis & Davis,1972; Feather, 1969; Feather & Simon,197la, 197Ib), although pattern recognition


(Chaikin, 1971), number estimation (Fitch,1970), simulation games (Streufert & Streu-fert, 19.69; Wolosin et al., 1973), and bogussocial perceptiveness tests (Davis & Davis,1972; Wortman, Costanzo, & Witt, 1973)have also been used. The choice of task islimited by the fact that it is necessary torandomly assign subjects to success and fail-ure experiences. In the anagrams task this isusually accomplished by varying the diffi-culty of the anagrams given to the subjects.In the other tasks, subjects are provided withfalse feedback as to their degree of successat the activity.

A review of the literature on skill tasksdoes not suggest a straightforward answer tothe question posed above. The predicted suc-cess-failure effects on self-attributions ofcausality were obtained unambiguously inonly a small number of experiments (Streu-fert & Streufert, 1969; Wolosin et al., 1973;Wortman et al., 1973). The remaining studiesshowed that causal attributions were depen-dent on an interaction between success andfailure and certain situational and personalityfactors.

First, let us consider the research thatprovides the strongest evidence for self-serv-ing biases in the attribution of causality.Streufert and Streufert (1969) required pairsof subjects to play an international simulationgame against another team. Programmedfeedback indicated that subjects either gradu-ally improved or became worse at the game.In these circumstances, subjects attributedgreater responsibility to their own team forsuccess than for failure. Similarly, Wortmanet al. (1973) found that subjects who al-legedly succeeded at a social perceptivenesstask were less likely to attribute their per-formance to luck than were failing subjects.However, since the Streufert and Streufertand Wortman et al. experiments did not in-clude a condition that received intermediateoutcomes, it is impossible to determine thedirectionality of this effect. That is, the re-sults may indicate either enhanced responsi-bility for success or avoidance of responsi-bility for failure, or both.

The two experiments conducted by Wolosinet al. (1973) did include intermediate out-

come conditions.4 Subjects played either acooperative (Experiment 1) or competitive(Experiment 2) game with another person.The outcome of this game, from the subject'spoint of view, was either successful, unsuc-cessful, or neutral. Subjects were then askedto distribute responsibility for the outcomeamong self, other, and situation. In the co-operative experiment, subjects who receivedfailing outcomes attributed slightly, thoughnonsignificantly, more responsibility to selfthan did subjects who received neutral out-comes (32.6% versus 29.8%); the most re-sponsibility was assigned to self in the suc-cess condition (38.7%). In th§ competitiveexperiment the pattern of means suggests bothavoidance of responsibility for failure and in-creased acceptance of responsibility for suc-cess (success: 42.7%; neutral: 35.3%; fail-ure: 27.6%), though only the success-failuredifference was statistically significant.5

In summary, there are four studies (Streu-fert & Streufert, 1969; Wolosin et al., 1973;Wortman et alv 1973) which provide rela-tively unambiguous support for the hypothesisthat attributions of causality are directly af-fected by task performance. In only two ofthese studies (Wolosin et al., 1973; Experi-ments 1 and 2), however, is it possible toassess the directionality of the effect. One ofthese studies (Experiment 2) suggests thatpeople are inclined to make self-enhancingattributions under conditions of success, butneither provides strong evidence to suggestthat people avoid personal responsibility

4 The present authors would like to thank Wolosinet al. for providing the data required to compute thestatistical significance of these comparisons.

B It is not clear that an intermediate outcome con-dition completely relieves the interpretational am-biguity which surrounds this phenomenon. For in-stance, how is the intermediate outcome experiencedin the individual's phenomenology? The valence ofneutral outcomes may be located somewhere betweensuccess and failure, but it is also possible that indi-viduals may not discriminate between a neutral andfailing outcome, given a desire for and an expectationof success. If the latter were the case, then a com-parison of causal ascriptions for failure and neutraloutcomes would be insensitive to a potential self-protective bias. A neutral outcome condition is onlyvaluable to the extent that the outcomes are evalu-ated as truly neutral in the subject's phenomenology.Thus, while we stress the conceptual value of inter-mediate outcome conditions, their empirical validityshould be assessed in future research.


under conditions of failure. Moreover, asnoted earlier, the self-enhancing effect mayreflect the tendency of people to anticipateand work toward success rather than failure.

It is also possible that people actually per-ceive a greater relationship between behaviorand positive outcomes that between behaviorand negative outcomes. Support for this alter-native explanation may be derived from twostudies which examined adults' understandingof the concept of correlation. Smedslund(1963) required nursing students to estimatethe relation between a symptom (s) and adiagnosis (d). The subjects based their deci-sions on data which indicated varying fre-quencies of the following categories: s present,d present; s absent, d absent; s present, dabsent; s absent, d present. Smedslund foundthat the subjects' judgments of a relationshipdepended almost solely on the frequency ofthe s present, d present category. The fre-quency and distribution of the remainingcategories were virtually ignored.

In a similar vein, Jenkins and Ward (196S)asked subjects to judge the degree of con-tingency between their responses and put-comes at a task which required subjects toguess. which of two possible responses wascorrect. The major finding paralleled thatobtained by Smedslund (1963): The amountof judged control was determined by the fre-quency of correct guesses, rather than theactual programmed contingency between out-comes and responses.

These data indicate that adults do notunderstand the meaning of statistical con-tingency. As Jenkins and Ward observed,"The baseline against which subjects assesstheir performance appears to be one of zerooccurrence of the event in interest in theabsence of control" (1965, p. 17). Estimatesof amount of control are thus based on thefrequency of positive instances, with the fre-quency of negative instances being uninforma-tive. The attributional results obtained onskill tasks are consistent with this primitivenotion of contingency. Positive instances ofthe co-occurrence of the response and the de-sired outcome (i.e., success) induce percep-tions of self-control. Negative instances (i.e.,failure), on the other hand, are less informa-

tive and, hence, yield neither stable personalnor environmental attributions.

INTERACTIVE VARIABLES

This section reviews research which hasinvestigated factors that interact with attribu-tions of causality following success or failureat achievement tasks.

Performance Expectancies

Feather and his colleagues (Feather, 1969;Feather & Simon, 197la, 197Ib) have ex-amined the effect of expectations for successand failure on subjects' attributions of cau-sality for low and high levels of performance.Expectations of success either were based onsubjects' own estimates of their anagram-solving ability (Feather, 1969) or manipu-lated by providing subjects with easy or diffi-cult practice problems (Feather & Simon,1971 a, 1971b). This research suggests thatan unexpected outcome—whether it be suc-cess or failure—is more likely to be attrib-uted to external factors than an expected out-come. Gilmor and Minton (1974) have repli-cated these results using an experimentalparadigm similar to that of Feather (1969).Additional relevant evidence comes fromFitch's (1970) experiment in which chronic,low self-esteem subjects (who may anticipatefailure at most tasks) were found to acceptmore -responsibility for failure than high self-esteem subjects at a task in which they wererequired to estimate the number of dots ona briefly exposed slide. In the Fitch study,however, low self-esteem subjects did notdiffer from high self-esteem subjects in theircausal attributions for success.

The research conducted by Feather (1969)suggests that expectations of performancelevel are more important determinants of at-tributions of causality than the actual levelof performance attained. Feather interpretedthis effect in terms of balance theory (Heider,1958). A balance theory formulation assumesthat positive outcomes (success) will be at-tributed to self when there is positive self-evaluation with respect to the performancetask (high expectations of success), but willbe attributed to external factors when thereis negative evaluation (low expectation ofsuccess). Similarly, with failure, negative self-


evaluations produce internal attributions, andpositive self-evaluations yield external attri-butions.

A recent experiment by McMahan (1973)suggests a refinement of the balance theoryanalysis. McMahon found that expected suc-cess or failure tended to be attributed to abil-ity, a relatively stable dispositional propertyof the person. On the other hand, unexpectedperformance increased attributions to morevariable causes such as luck and effort. Theimportant finding here is that effort, an in-ternal cause, may be used to account forunexpected success or failure. The reliabilityof this result is in doubt, however, as it wasnot obtained in a Feather and Simon (1971b)study which also assessed attributions to ef-fort. Nevertheless, it seems likely that thestability dimension is independent of the in-ternal-external dimension and may prove tobe more critical in determining causal infer-ences that the latter (Feather & Simon,1971a, 1971b; Weiner, Frieze, Kukla, Reed,Rest, and Rosenbaum, 1971).

Simple-Complex Conceptual Structures

An information-processing theory proposedby Schroder, Driver, and Streufert (1967)suggests that people differ in terms of thedegree of complexity they perceive in theirenvironment. For example, the attitudes ofsimple subjects tend to be based on unidimen-sional interpersonal perceptions, while com-plex subjects tend to perceive others in amore multidimensional fashion (Streufert &Streufert, 1969). Streufert and Streufert havesuggested that this complexity dimension mayhave important implications for attributionsof causality. Specifically, complex personsshould be less likely than simple individualsto assign responsibility to a single causalfactor for either a successful or failing per-formance. Thus, complex individuals shouldshow less of a tendency to attribute failuresolely to external factors and to attribute suc-cess solely to internal factors than shouldsimple individuals. In the Streufert and Streu-fert study, both complex and simple subjectstended to accept^ more responsibility for suc-cess than failure. However, as expected, thiseffect was more pronounced for simple thanfor complex subjects.

Internal Versus External Locus of ControlOne individual difference dimension which

might be expected to influence the nature ofcausal attributions is internal-external con-trol of reinforcement. Individuals occupyingthe internal end of this dimension perceivereinforcement as contingent upon their be-havior, whereas externals perceive reinforce-ment as more dependent on external factorssuch as luck (Lefcourt, 19.66; Rotter, 1966).Of major interest here is the suggestion thatsome individuals who obtain external scoreson the internal-external dimension may havedeveloped this orientation as a defense sinceit enables them to maintain self-esteem byattributing failures to events beyond theircontrol (Rotter, 1966).

If an external orientation does serve a de-fensive function, it might be expected thatexternals would be more likely than internalsto attribute failure to external factors. Con-versely, the causal ascriptions of internals andexternals for a successful performance shouldbe similar, since success is presumably non-threatening. This hypothesis was tested intwo studies conducted by Davis and Davis(1972). In the first study, highly internal andhighly external subjects were induced to per-form either well or poorly on an anagrams testand were then asked to indicate the extent towhich their scores were based on skill or luck.The second study manipulated subjects' per-formance at a "social sensitivity test" andrequired subjects to evaluate the degree towhich their scores were determined by ability,effort, luck, or situational factors. EmployingHeider's (1958) classification system, abilityand effort were classified as internal causes,and luck and situational factors as externalcauses.

The results provide some support for inter-preting an external orientation as defensive.In both studies externals were more likelythan internals to attribute failure to bad luckrather than to lack of ability. No differencesoccurred between the two groups under suc-cess conditions. However, these results werenot replicated by Gilmor and Minton (1974)and Lefcourt (Note 2), who also used ananagrams test. Gilmor and Minton obtainedan internal-external locus of control by per-formance interaction, indicating that internals


accepted more responsibility for success thandid externals; a nonsignificant trend in theopposite direction was found under failure.(It should be noted that unlike the Davis andDavis (1972) study, a successful-failure out-come was determined by the subject's actualperformance at the task). Finally, Lefcourt(Note 2) found that externals accepted lessresponsibility for both success and failurethan did internals in a design in which allsubjects were induced to both succeed andfail on repeated administrations of an ana-grams test.

The contradictory nature of these resultsindicates that the relationship between in-ternal-external locus of control and attribu-tions of causality for success and failure re-mains to be determined. Procedural differ-ences between the studies must be examinedand further research conducted if the incon-sistencies are to be explained. At the presenttime, research on the internal-external dimen-sion does not provide strong evidence forself-serving biases in causal attributions.

Achievement Motivation

Weiner and Kukla (1970) have investi-gated the relation of achievement motivationto perceptions of causality. In one study,these authors compared subjects above andbelow the median in achievement motivationin their causal ascriptions for hypotheticalsuccesses and failures. The data indicated thatthe tendency to make internal attributionsfor success was greater for the high- thanfor the low-achievement-oriented males. How-ever, these two groups did not differ signifi-cantly in their causal attributions for failure,and none of the differences between femalesubjects reached statistical significance.

In a second study, Weiner and Kukla hadmale subjects perform a skill task which re-quired them to guess the next number in asequence. At the end of the task, subjectswere asked to estimate how many of theirtotal number of correct responses were dueto skill rather than lucky guessing and howmuch effort they had expended at the task.For the purpose of statistical analyses, sub-jects who scored in the upper one third ofthe outcome distribution were defined as suc-cessful, and subjects in the lower one third

as failing. The results indicated that maleshigh in achievement motivation were morelikely to attribute success to ability and fail-ure to lack of effort than males low in achieve-ment motivation.

In summary, the study of personality vari-ables indicates that there are individual dif-ferences in the tendency to accept responsi-bility for success and failure. It is possiblethat individual differences of this nature re-flect motivationally induced distortions ofcausality, but it is also possible that these dif-ferences simply reflect differential perform-ance expectancies (cf. Feather, 1969). Forexample, people high in achievement motiva-tion may be motivated to distort their per-ceptions of causality in a manner which en-hances their feelings of mastery and achieve-ment, or, alternatively, their achievement-oriented dispositions may render them morelikely to expect success than people low inachievement motivation. Indirect support forthis latter possibility is provided by Kukla(1972), who found that following either suc-cess or failure at a task, subjects high inachievement motivation rated themselves ashigher in ability than did subjects in the low-motivation group.

It is also possible that individual differ-ences in causal attribution stem from a differ-ential concern with the covariation betweenbehavior and the consequences of behavior.To stay with the dimension of achievementmotivation, people high in achievement moti-vation may very well be more sensitive tothe covariation between behavior and conse-quences than people scoring low on thisdimension, and this may account for thetendency of the former to assume more creditfor success relative to failure than do thelatter.

In any event, there appear to be individualdifferences in the attribution of causality,and future research should attempt to assessthe relative explanatory values of motiva-tional and nonmotivational interpretations ofthese differences.

COMPARISONS WITH OTHERS

Attributions of causality for success andfailure at skill tasks have rarely been exam-ined in a social context. This is surprising


because it would seem that a social situationis well suited for determining whether biasesand misperceptions do occur. For example, dowe attribute more causality to ourselves fora successful performance at a task than wewould to another person who performedequally well? Alternatively, do we attributemore responsibility for failure to anotherperson than to self, given equal performancelevels? Once again, however, any confidencethat we place in interpersonal comparisons asa basis for assessing motivational biases mustbe tempered by the recognition that the co-variation between behavior and outcomes islikely to be considerably more salient for selfthan for others.

Feather and Simon (1971a) devised a situ-ation in which two subjects worked simul-taneously but independently at anagramstests. Comparisons of attributions made toself and another showed that the other's suc-cess was more often attributed to ability thanwas one's own success, while the other's fail-ure was more often attributed to such ex-ternal factors as bad luck than was one's ownfailure. These results are directly opposite tothose predicted by a self-enhancement hy-pothesis.

Wortman et al. (1973) obtained resultsconsistent with those of Feather and Simon(1971a). In their experiment, subjects workedindependently at a social perceptiveness test,at the end of which they were asked to makecausal inferences for their own performanceand that of a highly successful other. Sub-jects indicated that their own performancewas due more to luck and less to ability thanthe other's performance. Further evidenceagainst the presence of self-serving biasescomes from the fact that subjects' own per-formance levels did not affect their causalinterpretation of the others' (always success-ful) performance.

The Feather and Simon (1971a) and Wort-man et al. (1973) studies differed from otherrelevant research (Streufert & Streufert,1969; Wolosin et al., 1973) in two importantrespects. Most importantly, the outcomes ofthe participants in these latter studies wereinterdependent. In the competitive set (Streu-fert & Streufert, 1969; Wolosin et al., 1973,Experiment 2) there could only be one win-

ner, and in the cooperative set (Wolosin et al.,1973, Experiment 1) both participants re-ceived the same outcome. In contrast, Featherand Simon (1971a) had subjects work inde-pendently at a task in which their outcomeswere in no way related. Second, whereasFeather and Simon (1971a) and Wortmanet al. (1973) assessed causal attributions onbipolar scales, the other studies required sub-jects to assign percentages of responsibility toself (or own team), other (or opposing team),and situation (or causes other than the de-cisions made by the other team), with thesum of the three percentages always equaling100%. Thus, the Streufert and Streufert andWolosin et al. procedures seem likely to makeinterpersonal comparisons of task skill ex-tremely salient. As a result, these latterstudies might be expected to provide a par-ticularly facilitative setting for self-servingdistortions in attributions of causality.

The results of these studies do yield somesupport for a biased attribution process.Streufert and Streufert (1969) found thatapproximately equal responsibility was as-signed to own and other team for failure byone's own team. When one's own team suc-ceeded, however, more causality was attrib-uted to it and less to the decisions of theother team. Wolosin et al. (1973) also foundthat subjects attributed greater responsibilityto themselves for success than to their part-ners. This was true in both the competitiveand cooperative games. In the failure condi-tion, Wolosin et al. expected subjects in thecooperative game to attribute responsibilityto their partner. In the competitive game,however, it was expected that subjects wouldblame the situation since attributing responsi-bility to the partner would imply that he wasbetter at the game. Neither of these predic-tions was strongly substantiated. In both thecooperative and competitive games, failureproduced approximately equal attributions ofresponsibility to self and others—a findingconsistent with Streufert and Streufert's re-sults. In the competitive game, failure didcause subjects to attribute more responsibilityto the situation than to self or another. How-ever, between-group comparisons suggest thatsubjects who received failing outcomes didnot attribute any more responsibility to the


situation than did subjects who received neu-tral outcomes. Moreover, Streufert and Streu-fert reported no tendency for failing subjectsto increase situational attributions. In fact,Streufert and Streufert's successful subjectsattributed significantly more responsibilityto causes external to the decisions made byeither team than did failing subjects^

In summary, studies in which the outcomesof the participants were independent yieldedno indication of self-serving biases in inter-personal attributions (Feather & Simon,1971a; Wortman et al., 1973). Three studiesin which the outcomes of the participantswere directly related did appear to obtainself-enhancing attributions because success atthe task tended to be self-attributed. On theother hand, subjects showed no strong tend-ency to deny responsibility for failure byattributing it to the other person (s) or thesituation. Moreover, to return to an earlierpoint, even self-attribution of success neednot be due to motivational biases. A person islikely to be more aware of the covariationbetween his own responses and the resultantoutcomes than he is of the relation betweenthe outcomes and behavior of the other par-ticipant (s). Hence, he may be expected toattribute greater responsibility to himself forsuccess. Constant failure may be less likelyto be self-attributed because the response-outcome covariation may be less apparent.Changes in responding are not associated withvariation in outcomes when subjects continu-ally fail at the task (Kelley, 1967; Wolosinet al., 1973).

Another interesting finding is that self-attributions of causality seem to be affectedby the degree of interdependence of out-comes. Since independent and interdependentsituations have not been directly contrastedexperimentally, any conclusions are specula-tive at present. A comparison of the Featherand Simon (1971a) study with those con-ducted by Streufert and Streufert (1969) andWolosin et al. (1973) does suggest, however,that increasing the interdependence of out-comes increases the likelihood that subjectswill assign more responsibility to themselvesfor successful outcomes. A comparison of theChaikin (1971) and Streufert and Streufertstudies lends credence to this interpretation.

The Chaikin experiment, involving only asingle subject playing a skill game at eachsession, failed to obtain significant effects forattributions of causality to self following de-scending versus ascending success at the task.This contrasts sharply with the very strongresults obtained by Streufert and Streufertfor the same patterns of performance in acompetitive, interdependent game situation.

Why might interdependence lead to an in-creased tendency for individuals to take creditfor success? One possibility is that interde-pendence increases the subject's motivation todo well, to perform better than the otherparticipant. A strong desire for success incombination with a high degree of effort wouldbe expected to result in individuals perceivingthemselves as personally responsible for suc-cess. In contrast, success achieved under con-ditions of lower levels of effort may be at-tributed, in part, to external factors such asgood luck.

Is the Failure to Obtain Self-Protective Biasesdue to the Nature of the Experimental Tasks?

As the present review has shown, there isvery little experimental evidence to supportthe hypothesis that people in general attemptto avoid blame for negative outcomes by at-tributing responsibility to external factors.Possibly, the tasks employed in these studiesfailed to arouse the level of ego involvementnecessary to elicit self-protective reactions.Most of the experimental activities wererelatively trivial (anagrams, guessing games,etc.), and failure could not be expected toprovide a telling blow to the ego. On theother hand, almost any task may appear im-portant by virtue of the fact that it is em-ployed in an experimental setting. As studiesof evaluation apprehension (Rosenberg, 1969)have indicated, experimental subjects wantvery much to appear intelligent in the eyesof the experimenter.

Furthermore, research by Lerner (1970)indicates that people often do, in fact, acceptresponsibility for nontrivial negative conse-quences. Pairs of subjects reported to thelaboratory for an experiment in which one ofthem would receive- strong electric shock,while the other would be in a control condi-tion. The decision as to who would receive


shock was determined by which of two slipsof paper the first subject selected in a randomdraw. The interesting conditions here were:(a) Subject A drew a slip assigning SubjectB to the shock and (b) Subject B drew aslip assigning himself to the shock. Subject Awas later asked to assign responsibility forhis own and B's fate to either self, other, orthe experimenter. When Subject A drew theslip, we might have expected him to assignresponsibility to the experimenter and henceavoid blame for committing the other subjectto shock. Moreover, this would be quite arational attribution. Was it not the experi-menter who determined that one of the sub-jects would receive shock and arranged thedraw to designate which one it would be?Nevertheless, approximately 47% of the sub-jects who drew the slip committing B toshock accepted responsibility for their ownand B's fate, as compared to less than 10%when the shock victim drew the slip himself.Thus, subjects do not seem to be stronglymotivated to deny responsibility for theother's fate. An examination of the "trans-gression-compliance" literature also indicatesthat people are very willing, perhaps too will-ing, to accept responsibility for another per-son's negative fate (Carlsmith & Gross, 1969;Darlington & Macker, 1966; Freedman, Wal-lington, & Bless, 1967; Rawlings, 1968;Regan, 1971; Wallace & Sadalla, 1966).

It seems likely, therefore, that the trivial-ness of the tasks employed in many of theexperimental investigations is not responsiblefor the preponderance of negative results.Nevertheless, this possibility should be ex-plored in future research in which the poten-tial seriousness of failure for the subject isvaried. In doing so, however, it will be im-portant to control subjects' expectations ofsuccess in the different achievement situations(cf. Feather, 1969).

Finally, it should be pointed out that thereis some evidence to suggest that people aremotivated to avoid feelings of responsibilityfor failure. Research examining the relation-ship between performance expectations and ex-pended effort, for example, has consistentlyfound that people who are led to expect fail-ure on a particular task are inclined to expendless effort on that task than people expecting

success (Archibald, 1974; Diggory, 196.6;Diggory, Klein, & Cohen, 1964; Feather,1963). One interpretation of this finding isthat people who expect failure are inclined toexpend low degrees of effort so that they willbe able to legitimately locate causality fortheir anticipated failure in sources which arenonthreatening to their self-esteem (i.e., per-sonal effort).

Once again, however, a very different proc-ess may underlie this phenomenon. In theface of anticipated failure, individuals maysimply feel that it does not make sense totry hard because their efforts will likely gofor nought. Support for the argument that dif-ferences in expended effort under failure andsuccess conditions reflect rational decisionmaking comes from a study (Weiner, Heck-hausen, Meyer, & Cook, 1972) in which sub-jects were asked to indicate "the best per-formance strategy" in tasks of varying diffi-culty. When there was little chance of suc-cess, subjects indicated that effort was oflittle importance and advocated a "low-effort"strategy. As the likelihood of success in-creased, the perceived importance of effortgenerally increased, although under high ex-pectations of success, effort also ceased to beregarded as an important factor.

SELF-ENHANCING BIASES IN THEATTRIBUTION PROCESS

While evidence for self-protective attribu-tional biases is minimal, there are at leastsome data consistent with a self-enhancementposition. In a number of the experiments re-viewed above, successful outcomes led togreater self-attribution of performance thaneither failing or neutral outcomes. As notedearlier, the self-enhancement effect need notbe explained in motivational terms. Conceiv-ably, subjects are simply more likely to per-ceive a relationship between their behaviorand its outcome when they succeed than whenthey fail. This may occur for any or all of thefollowing reasons: (a) People are more likelyto accept responsibility for expected outcomesthan for unexpected outcomes (Feather, 1969;Feather & Simon, 1971a, 1971b), and, ingeneral, people intend and expect success notfailure (Parducci, 1963, 1965, 1968); (b) acovariation between behavior and outcome is


more likely to be perceived under conditionsof increasing success than under conditions ofconstant failure, where changes in behaviorare not perceived to be associated withchanges in .outcome; and (c) an erroneousconception of contingency leads people to as-sociate control primarily with the occurrenceof the desired outcome (cf. Jenkins & Ward,1965; Smedslund, 1963).

In conclusion, existing data seem readilyinterpreted in information-processing terms.It would be clearly premature, however, todeny the possibility of self-serving causal at-tributions. An examination of the literature onperceptual defense may be instructive in thisregard. Early, methodologically flawed re-search supported the perceptual defense hy-pothesis; later evidence contradicted the hy-pothesis on logical and empirical grounds andsupported a rational, decision-making ex-planation. The most recent evidence, however,has once again supported the perceptual de-fense hypothesis (Dixon, 1971;. Erdelyi,1974). It is our belief that the self-servingbias hypothesis may suffer a similar fate, forit is too intuitively appealing to be summarilyabandoned. The challenge remains for futureresearchers to assess the relative explanatoryvalues of the motivational and nonmotiva-tional interpretations of asymmetrical causalattributions.

REFERENCE NOTES

1. Ross, L., Bierbrauer, G., & Polly, S. Attributionof educational outcomes by professional and non-professional instructors. Unpublished manuscript,Stanford University, 1974.

2. Lefcourt, H. M. Causal attributions as a junctionof locus of control, initial confidence and per-formance outcomes. Unpublished manuscript,University of Waterloo, 1974.

REFERENCESArchibald, W. P. Alternative explanations for self-

fulfilling prophecy. Psychological Bulletin, 1974, 81,74-84.

Beckman, L. Effects of students' performance onteachers' and observers' attributions of causality.Journal of Educational Psychology, 1970,61, 76-82.

Beckman, L. Teachers' and observers' perceptions ofcausality for a child's performance. Journal ofEducational Psychology, 1973, 65, 198-204.

Bern, D. J. Self-perception theory. In L. Berkowitz(Ed.), Advances in experimental social psychology(Vol. 6). New York: Academic Press, 1972.

Carlsmith, J. M., & Gross, A. Some effects of guilton compliance. Journal of Personality and SocialPsychology, 1969, 11, 232-239.

Chaikin, A. J. The effects of four outcome scheduleson persistence, liking for the task, and attributionsof causality. Journal of Personality, 1971, 39, 512-526.

Chaikin, A. L., & Darley, J. M. Victim or perpetra-tor? Defensive attribution of responsibility and theneed for order and justice. Journal of Personalityand Social Psychology, 1973, 25, 268-275.

Darlington, R. B., & Macker, C. E. Displacement ofguilt-produced altruistic behavior. Journal of Per-sonality and Social Psychology, 1966, 4, 442-443.

Davis, W. L., & Davis, D. E. Internal-external con^trol and attribution of responsibility for successand failure. Journal of Personality, 1972, 40, 123-136.

Diggory, J. S. Self-evaluation: Concepts and studies.New York: Wiley, 1966.

Diggory, J. C., Klein, S. J., & Cohen, M. Muscle-action potentials and estimated probability of suc-cess. Journal of Experimental Psychology, 1964,(55,449-455.

Dixon, N. F. Subliminal perception: The nature ofa controversy. London: McGraw-Hill, 1971.

Erdelyi, M. H. A new look at the new look: Per-ceptual defense and vigilance. Psychological Re-view, 1974, 81, 1-25.

Feather, N. T. Persistence at a difficult task withan alternative task of intermediate difficulty. Jour-nal of Abnormal and Social Psychology, 1963, 66,604-609.

Feather, N. T. Attribution of responsibility andvalence of success and failure in relation to initialconfidence and task performance. Journal of Per-sonality and Social Psychology, 1969, 13, 129-144.

Feather, N. T., & Simon, J. G. Attribution of re-sponsibility and valence of outcome in relation toinitial confidence and success and failure of selfand other. Journal of Personality and Social Psy-chology, 1971, 18, 173-188. (a)

Feather, N. T., & Simon, J. G. Causal attributionsfor success and failure in relation to expectationof success based upon selective or manipulativecontrol. Journal of Personality, 1971, 39, 527-541. (b)

Fishbein, M., & Ajzen, I. Attribution of responsibility:A theoretical note. Journal of Experimental SocialPsychology, 1973, 9, 148-153.

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

Freedman, J. L., Wallington, S. A., & Bless, E. Com-pliance without pressure: The effect of guilt. Jour-nal of Personality and Social Psychology, 1967, 7,117-124.

Gilmor, T. M., & Minton, H. L. Internal versus ex-ternal attribution of task performance as a func-tion of locus of control, initial confidence and suc-cess-failure outcome. Journal of Personality, 1974,42,159-174.

Hastorf, A., Schneider, D., & Polefka, J. Person per-ception. Reading, Mass.: Addison-Wesley, 1970.

Heider, F. The psychology of interpersonal relations.New York: Wiley, 1958.


Irwin, F. W. Stated expectations as functions ofprobability and desirability of outcomes. Journalof Personality, 1953, 21, 329-335.

Jenkins, H. M., & Ward, W. C. Judgment of con-tingency between responses and outcomes. Psy-chological Monographs, 1965, 79(1, Whole No.594).

Johnson, T. J., Feigenbaum, R-., & Weiby, M. Somedeterminants and consequences of the teacher'sperception of causation. Journal of EducationalPsychology, 1964, 55, 237-246.

Jones, E. E., & Nisbett, R. E. The actor and theobserver: Divergent perceptions of the causes ofbehavior. New York: General Learning Press, 1971.

Kelley, H. H. Attribution theory in social psychology.In D. Levine (Ed.), Nebraska Symposium onMotivation (Vol. IS). Lincoln: University of Ne-braska Press, 1967.

Kelley, H. H. Attribution in social interaction. NewYork: General Learning Press, 1971.

Kelley, H. H. The processes of causal attribution.American Psychologist, 1973, 28, 107-128.

Kukla, A. Attribution determinants of achievement-related behavior. Journal of Personality and SocialPsychology, 1972, 21, 166-174.

Lefcourt, H. M. Internal versus external control ofreinforcement: A review. Psychological Bulletin,1966, 65, 206-220.

Lerner, M. J. The desire for justice and reactions tovictims. In J. Macaulay & L. Berkowitz (Eds.),Altruism and helping behavior. New York: Aca-demic Press, 1970.

Marks, R. W. The effect of probability, desirability,and privilege on the stated expectations of chil-dren. Journal of Personality, 1951, 19, 431-465.

McMahan, I. D. Relationships between causal at-tributions and expectancy of success. Journal ofPersonality and Social Psychology, 1973, 28, 108-114.

Parducci, A. Range-frequency compromise in judg-ment. Psychological Monographs, 1963, 77(2, WholeNo. 565).

Parducci, A. Category judgment: A range-frequencymodel. Psychological Review, 1965, 72, 407-418.

Parducci, A. The relativism of absolute judgments,Scientific American, 1968, 219, 84-90.

Rawlings, E. I. Witnessing harm to other: A reassess-ment of the role of guilt in altruistic behavior.Journal of Personality and Social Psychology,1968,10, 377-380.

Regan, J. W. Guilt, perceived injustice, and altruisticbehavior. Journal of Personality and Social Psy-chology, 1971, 18, 124-132.

Rosenberg, M. J. The conditions and consequencesof evaluation apprehension. In R. Rosenthal & R.L. Rosnow (Eds.), Artifact in behavioral research.New York: Academic Press, 1969.

Ross, M., & DiTecco, D. An attributional analysisof moral judgments. Journal of Social Issues, inpress.

Rotter, J. B. Generalized expectancies for internalversus external control of reinforcement. Psycho-logical Monographs, 1966, 80(1, Whole No. 609).

Schopler, J., & Layton, B. Attribution of interper-sonal power and influence. New York: GeneralLearning Press, 1972. (a)

Schopler, J., & Layton, B. Determinants of the self-attribution of having influenced another person.Journal of Personality and Social Psychology,1972, 22, 326-332. (b)

Schroder, H. M., Driver, M. J., & Streufert, S.Human information processing. New York: Holt,Rinehart & Winston, 1967.

Shaver, K. G. Defensive attribution: Effects of sever-ity and relevance on the responsibility assignedfor an accident. Journal of Personality and SocialPsychology, 1970, 14, 101-113. (a)

Shaver, K. G. Redress and conscientiousness in theattribution of responsibility for accidents. Journalof Experimental Social Psychology, 1970, 6, 100-110. (b)

Smedslund, J. The concept of correlation in adults.Scandinavian Journal of Psychology, 1963, 4, 165-173.

Streufert, S., & Streufert, S. C. Effects of conceptualstructure, failure, and success on attribution ofcausality and interpersonal attitudes. Journal ofPersonality and Social Psychology, 1969, 11, 138-147.

Vidmar, N., & Crinklaw, L. D. Attributing responsi-bility for an accident: A methodological and con-ceptual critique. Canadian Journal of BehaviouralScience, 1974, 6, 112-130.

Wallace, J., & Sadalla, E. Behavioral consequencesof transgression: I. The effects of social recognition.Journal of Experimental Research in Personality,1966,1, 187-194.

Walster, E. Assignment of responsibility for an acci-dent. Journal of Personality and Social Psychology,1966, 3, 73-79.

Walster, E. "Second-guessing" important events.Human Relations, 1967, 20, 239-250.

Weiner, B. Theories of motivation. Chicago: RandMcNally, 1973.

Weiner, B., Frieze, I., Kukla, A., Reed, L., Rest, S.,& Rosenbaum, R. Perceiving the causes of successand failure. New York: General Learning Press,1971.

Weiner, B., Heckhausen, H., Meyer, W. U., & Cook,R. E. Causal ascriptions and achievement behavior:A conceptual analysis of effort and reanalysis oflocus of control. Journal of Personality and SocialPsychology, 1972, 21, 239-248.

Weiner, B., & Kukla, A. An attributional analysis ofachievement motivation. Journal of Personality andSocial Psychology, 1970, 15, 1-20.

Wolosin, R. J., Sherman, S. J., & Till, A. Effects ofcooperation and competition on responsibility at-tribution after success and failure. Journal of Ex-perimental Social Psychology, 1973, 9, 220-235.

Wortman, C. B., Costanzo, P. R., & Witt, T. R.Effect of anticipated performance on the attribu-tions of causality to self and others. Journal ofPersonality and Social Psychology, 1973, 27, 372-381.

(Received April 22, 1974)

Self-Serving Biases in the Attribution of Causality:...

Documents

Transcript of Self-Serving Biases in the Attribution of Causality:...