attributions of responsibility for group tasks: the …attributions of responsibility for group...

Journal of Personality and Social Psychology1983, Vol. 45, No. 6, 1232-1242

Copyright 1983 by theAmerican Psychological Association, Inc.

Attributions of Responsibility for Group Tasks: The EgocentricBias and the Actor-Observer Difference

Jerry M. Burger and John L. RodmanWake Forest University

In each of three experiments it was demonstrated that under certain conditionsindividuals who work on a task in a dyad will tend to attribute greater responsibilityfor a positive outcome to their partners than to themselves. In Experiment 1 collegestudents working in dyads on a crossword puzzle attributed more responsibility totheir partners than to themselves for an outcome they were led to believe was quitegood, thus contradicting the expected "egocentric bias" effect. In Experiment 2college students working in dyads on the puzzle attributed more responsibility totheir partners than to themselves for a positive outcome when asked immediatelyafter the task to make the attribution. However, subjects attributed greater re-sponsibility to themselves than to their partners when asked to make the attribution3 days later, thus replicating the egocentric bias effect. Half of the college-studentdyads in Experiment 3 believed they were being videotaped while working on thepuzzle, whereas the other half did not. "Videotaped" subjects were found to attributemore responsibility for the positive outcome to themselves than to their partners,whereas the nonvideotaped subjects attributed more responsibility to their partnersthan to themselves when both groups were asked to give their attributions im-mediately after the task. The relationship between the egocentric bias effect andthe actor-observer difference phenomenon is discussed.

Individuals who participate in a group pro-ject tend to give themselves more credit for apositive outcome and recall more of their inputinto the project than do their co-workers. Rossand Sicoly (1979) identified this phenomenonas the "egocentric bias" and provide a seriesof experiments with a variety of subject pop-ulations and situations to demonstrate its per-vasiveness. More recently, Thompson andKelly (1981) provided a series of studies thatdemonstrate this phenomenon among ro-mantically involved heterosexual couples.

Ross and Sicoly provide several explanationsfor the egocentric bias effect. They argue thatpeople may recall more of their contributionsto a project than their partner's and henceattribute more responsibility to themselvesbecause of differential encoding and storage

Data presented in Experiment 1 were taken from themaster's thesis of the second author, who is now at theState University of New York at Albany. Thanks are ex-tended to Teresa Ercoline and Sophie Peden for their helpwith the data collection.

Requests for reprints should be sent to Jerry M. Burger,Department of Psychology, Box 7778, Wake Forest Uni-versity, Winston-Salem, North Carolina 27109.

of these personal responses, because of moreaccessible retrieval from memory, because ofgreater access to personal knowledge andstrategies, and because of a motivation to seethemselves as the primary causal agent relativeto their partners.

The latter explanation, that the egocentricbias results in part for reasons of enhancingone's self-esteem or seeing oneself as a caus-ative force, was the original concern of thepresent series of investigations. The originalhypotheses were concerned with differences inthe use of the egocentric bias by persons ofdifferent depression levels. Several recent in-vestigations have suggested that depressed andnondepressed persons may not use attribu-tional distortions in a similar manner (e.g.,Alloy & Abramson, 1979, 1982; Golin, Terrell,& Johnson, 1977; Golin, Terrell, Weitz, &Drost, 1979; Lewinsohn, Mischel, Chaplain,& Barton, 1980; Rozensky, Rehm, Pry, &Roth, 1977). These investigations generallyhave found that nondepressed persons tend tobias their attributions to put themselves in amore favorable light, whereas depressed in-dividuals fail to make these attributional dis-tortions.

1232

ATTRIBUTIONS FOR GROUP TASKS 1233

It was anticipated that depressed collegestudents working on a two-person projectwould be less likely to demonstrate the ego-centric bias than would nondepressed students.If, as Ross and Sicoly (1979) suggest, one rea-son for the egocentric bias is a motivation toenhance one's self-esteem and to see oneselfas an important causative agent, then de-pressed persons, who often possess low self-esteem and a lack of motivation to controlevents (Beck, 1972) should be less likely toemploy the egocentric bias. The original di-rection of this research, therefore, was to ex-amine the influence of depression on the useof the egocentric bias. It should be noted,however, that the intriguing findings in Ex-periment 1 caused the investigators to lookmore closely at the egocentric bias effect itselfand to abandon the questions concerning de-presssion in the subsequent experiments.

Experiment 1

MethodSubjects. One hundred twenty men and women un-

dergraduate psychology students participated in the ex-periment in exchange for class credit. Fifty-six of thesesubjects had scores on a depression inventory that allowedtheir selection into the final sample.

Procedure. Subjects participated in the experiment ingroups of 20. At the beginning of the session a femaleexperimenter announced that data were being collectedfor experiments being run by two different faculty mem-bers. It was announced that the first part of the experi-mental session would consist of filling out some personalityquestionnaires. The female experimenter distributed andcollected the Beck Depression Inventory (BDI; Beck, 1972).The experimenter then left the room as a second exper-imenter, a male, distributed a second questionnaire, theAdjective Check List (ACL; Gough & Heilbrun, 1980).

The ACL was administered to allow the first experi-menter enough time to score the BDIs and to assign subjectsto partners based on these scores. That is, the subjectswith the two highest scores in the session were paired forthe experimental task, then the next two highest scoreswere, paired, and so forth. This procedure was employedto ensure that each member of the dyad worked with apartner of a similar depression level. A few minutes aftersubjects began work on the ACL the second experimenterstepped out of the room to receive the pairing list. Hethen collected the ACLs and introduced the experimentaltask.

It was explained to the subjects that they were to be"randomly" assigned to partners via a random numberstable. The experimenter then had subjects sit together inpairs as he read their names from the list he had concealedin a folder. After subjects were paired the task was intro-duced. Each dyad was given a crossword puzzle and 10minutes to work on the puzzle as a team. After the time

elapsed, subjects were told to total the number of squarescompleted, regardless of how confident they were of thecorrectness of the answer, and to record this amount inthe space provided.

Subjects then were informed that the experimenter hadused this task with many other subjects and that he couldprovide them with some feedback on their performances.Subjects were told that the mean number of squares filledin, correctly Or incorrectly, was 21.62 and that any dyadthat had completed more than 30 squares had done well.These feedback figures were selected to give all subjectsthe impression that they had performed well on the task.All dyads were able to complete more than 30 squares.This feedback was given because of Ross and Sicoly's (1979,Experiment 2) finding that the egocentric bias effect isattenuated when the dyad's performance is poor.

At this point subjects were separated and given a finalquestionnaire. Subjects were assured that their partnerwould not see their questionnaire responses. The ques-tionnaire contained several filler items asking about thetask. Included was a manipulation check item that askedsubjects how well they believed they and their partnerperformed on the task relative to most college students.The main dependent measure item asked subjects to es-timate the percentage they and their partner had contrib-uted to the dyad's overall performance on the puzzle.

Results

Before examining responses, 56 subjectswere selected from the group of 120, whichcomposed the final sample. Dyads with meanBDI scores falling into the upper and lowersixths of the distribution were selected, whichcomposed the depressed and the nondepressedgroups. Subjects also were selected for a slightlydepressed group in a manner that would max-imize the difference in depression level for thethree groups. The mean BDI scores for subjectsin the three groups were as follows: nonde-pressed = 0, slightly depressed = 3.6, and de-pressed = 12.7. It was necessary to include120 subjects in the original sample in orderto identify enough subjects with high depres-sion scores. Only 16 subjects (eight dyads) metthe pre-set criterion (BDI ̂ 10 for each sub-ject) for inclusion in the depressed-subject cat-egory.

Subjects indicated that they believed theyhad performed well on the puzzle relative tomost college students (M = 6.73, with 1 =

- "very poorly" and 9 = "very well"). Becausesubjects participated in dyads, subject re-sponses on the attribution measure could notbe considered independent. Therefore, thedyad was employed as the unit of analysis. Ifsubjects accurately perceived the relative con-tribution of the two dyad members, it would

1234 JERRY M. BURGER AND JOHN L. RODMAN

be expected that their totaled percentages of"self" contributions would equal 100%. Whenthe two self-percentages totaled more than100%, the egocentric bias was evidenced inthe dyad. The number of dyads with morethan, less than, and exactly 100% self-per-centage totals within each of the three depres-sion-level groups is presented in Table 1.

As can be seen in the table, the same patternemerges in all three depression categories^ Thatis, there is a tendency for subjects to givethemselves less credit than they deserve, a pat-tern exactly opposite of that predicted fromthe earlier egocentric bias literature. To ex-amine this effect statistically, it wbuld be ex-pected that if variations of dyad scores aboveand below the 100% level were caused solelyby chance, then an equal number of dyadswould have self-percentage totals above the100% level as below this level. However, whencollapsed across depression level, we find theoccurrence of dyads with below-100% scoressignificantly more often than dyads withabove-100% scores, p < .02, using a binomialtest (Siegel, 1956).

Discussion

The results of Experiment 1 clearly wereunexpected. Not only was the anticipated ego-centric bias effect not replicated, but a sig-nificant tendency for subjects to give them-selves less credit than they give their partnerswas uncovered. Because of this surprisingfinding, and because depression level did notappear to affect the attributions, the directionof the research project was changed. The newconcern was with understanding why the sub-jects in Experiment 1 tended to give themselvesless credit than they gave their partners,whereas the results of the other egocentric biasstudies reported in the literature reveal theopposite pattern.

Experiment 2The first step in understanding the different

outcomes of the experiments was to comparethe procedures employed in Experiment 1 withthose of other egocentric bias studies. Whencomparing Experiment 1 with the five studiesreported by Ross and Sicoly (1979) and withthe three studies reported by Thompson andKelley (1981), one important procedural dif-ference emerges. In all eight of these studiesthat had uncovered the egocentric bias effect,subjects were asked about events that eitherhad occurred over a long period of time orhad occurred a minimum of one or 2 daysearlier. In contrast, Experiment 1 asked sub-jects to give attributions immediately aftertheir performance.

Therefore, Experiment 2 was designed toexamine the effect of a time lag between per-formance on the dyad task and presentationof the attribution question. It was predictedthat subjects asked about the task immediatelyafter its completion would tend to credit theirpartners with more of a contribution than theywould credit themselves, but that subjectsasked about the task 3 days after their per-formance would tend to give themselves morecredit than they would give their partners.

Method

Subjects. One hundred and ten undergraduate studentsserved as subjects in exchange for class credit. Ten subjectshad to be dropped from the final sample because eitherthey or their partner failed to attend the second half ofthe experiment. There were 54 men and 46 women in thefinal sample.

Procedure. The experiment was conducted in groupsof 26 to. 30 persons. The experimenter explained that theresearchers were concerned with how people worked ongroup tasks. The experimenter then read aloud dyad pair-ings as each subject was randomly paired with anotherindividual from the sign-up sheet. The experimenter ver-ified at the time that the subjects did not know each other.In the few instances in which subjects said they were ac-quainted with one another, another partner was randomly

Table 1Number of Dyad Se.lf-Percentage Totals by Depression Level

Self-percentage totals Nondepressed Slightly depressed Depressed Totals

More than 100%100%Less than 100%

58

15


selected from the list. Because subjects were directed tosign up for the experiment at a time restricted to eitherall-men or all-women subjects, only same-sex dyads werecreated. Each dyad then was given the same crosswordpuzzle used in Experiment 1 and given 15 minutes towork, on the task.

Half of the male and the female experimental sessionshad been preselected for the <>nm«fe«e-attribution con-dition and half for the rfe/ayerf-attribution condition. Afterthe IS minutes elapsed, subjects in the immediate-attri-bution condition were given a "feedback" sheet containingsome norms for the number of puzzle squares completed,supposedly taken from earlier participants in the experi-ment. Subjects were instructed to count their completedsquares and to indicate the percentile group their perfor-mance was equivalent to. As in Experiment 1, the bogusnorm table had been created to give subjects the impressionthey had done quite well on the task. All subjects hadscores that led them to believe they were among at leastthe highest 8% of those who had participated in the ex-periment. Subjects were then separated and given a ques-tionnaire to complete. Once again, subjects were assuredthat their partner would not see their responses. The ques-tionnaire was identical to the one used in Experiment 1with one addition. Subjects were asked to list as manywords as they could, recall that they had contributed tosolving the puzzle and as many as they recalled that theirpartners had contributed. These items were added to ex-amine the role of memory retrieval in the egocentric bias,as hypothesized by Ross and Sicoly (1979).

Subjects in the delayed-attribution condition signed upfor the experiment with the understanding that it consistedof two sessions. At the end of the 15-minute puzzle time,subjects were informed that the experiment would be con-tinued at the next session. The second session was scheduledfor 3 days after the first session. On arrival at the secondsession, subjects were instructed to separate from theirpartners. All subjects then were given the same bogus feed-back sheet used for the immediate-attribution subjects,with their score and percentile indicated. Again, all subjectswere told they and their partner had performed as well asat least the top 8% of the participants. Subjects then wereadministered the questionnaire after being assured theirpartner would not see their responses.

Results

Once again, all subjects believed they hadperformed well on the crossword puzzle (M =7.00, with 1 = "very poorly" and 9 = "verywell"). As in Experiment 1, to ensure the in-dependence of data points, the dyad was em-ployed as the unit of analysis. The percentagesof contribution each member of the dyad at-tributed to himself or herself were added to-gether. The number of dyads with self-per-centage totals above, below, and equal to 100%were compared within each of the two exper-imental conditions. These results are presentedin Table 2. As can be seen from the table,similar patterns were found for male and fe-

male dyads; thus the sex variable was collapsedfor the data analyses.

The number of dyads with self-percentagetotals above and below 100% were comparedfor each of the experimental conditions. Therewas a tendency to replicate the results of Ex-periment 1 in the immediate-attribution con-dition. Dyad members tended to give collec-tively more credit to their partners than tothemselves more often than they gave morecredit to themselves than their partners, p <.07, using a binomial test. However, dyads inthe delayed-attribution condition demon-strated the egocentric bias effect. These dyadmembers gave collectively more credit tothemselves than to their partners more oftenthan they gave more credit to their partnersthan to themselves, p < .05, using a binomialtest.

The number of words subjects could recallthat they and their partner had contributedalso was assessed. Because subjects worked indyads, each subject's responses could not beassumed to be independent from other sub-jects' responses. To resolve this problem, thedifference between self- and partner-words re-called within each dyad was employed as theunit of analysis. Within each dyad the numberof words each subject recalled he or she hadcontributed was compared with the numberof words the partner recalled the subject hadcontributed. The number of words the partnerrecalled the subject contributing was sub-tracted from the number of words the subjectrecalled himself or herself contributing. Thetwo difference scores were then totaled to cre-ate a self-to-partner difference score for eachdyad.

The self-to-partner difference score was thencompared across the two experimental con-ditions. It was found that subjects in both theimmediate- and the delayed-attribution groupstended to recall more words they had con-tributed than words that had been contributedby their partner (self-to-partner differencescore M = 1.96 for the immediate conditionand .88 for the delayed condition). The im-mediate-condition dyads produced signifi-cantly larger self-to-partner difference scoresthan did subjects in the delayed condition,;(48) = 2.46, p < .02. This difference, however,simply may reflect the larger number of totalitems recalled by subjects in the immediate-


Table 2Number of Dyad Self-Percentage Totals for Immediate and Delayed Conditions

Self-percentage totals


Men

338

Immediate

Women

244

Total

57

12

Men

824

Delayed

Women

652

Total

1476

condition dyads. For example, a subject in theimmediate condition typically may have re-called four of his or her own contributions andonly two of the partner's words, whereas asubject in the delayed condition typically mayhave recalled two of the words he or she con-tributed and only one of the partner's words.However, because of the larger number of totalwords recalled, it would be difficult to say thatthe tendency to recall one's own contributionsmore often than those of the partner was nec-essarily stronger in the immediate conditionin this example despite the larger differencebetween self- and partner-words recalled.

To overcome this difficulty, the total numberof words both members of each dyad recalledfor themselves was compared with the totalnumber of words both members recalled fortheir partner. The number of dyads in whichthe self-recall total was greater than the part-ner-recall total was compared with the numberof dyads in which the partner-recall total waslarger than the self-recall total. Thus becausethis comparison examines the number ofdyads within each condition exhibiting thiseffect rather than makes a comparison of thenumber of recalled items between the twoconditions, differences in the total amount ofitems recalled by immediate- and delayed-condition subjects should not influence thisstatistic. The results of this computation arepresented in Table 3. A tendency for membersof a dyad to recall words each had contributedmore often than words their partner had con-tributed was found significantly more oftenthan was the tendency to recall more partner-contributed words in both the immediate(p < .05) and the delayed (p < .001) conditions,(using a binomial test). It can be seen in thetable that this effect was strong in both con-ditions, which suggests that the difference un-covered with the self-to-partner difference

score may reflect the difference in total wordsrecalled and therefore may be somewhat mis-leading.

Some further analyses were conducted tobetter understand the relationship between theword-recall measure and the attribution mea-sure. First, the self-to-partner difference scorewas correlated with the total self-percentagesfor each dyad. Positive but modest correlationswere found in both the immediate (r = .21)and the delayed (r = .24) conditions. This sug-gests that in both conditions subjects' recallof specific events may have partially deter-mined the attributions for the performancebut that this relationship may not be a strongone. Looking at this same relationship in adifferent manner, subjects who recalled morewords they had contributed than words con-tributed by their partners gave themselvesmore credit for the dyad's performance thandid subjects who recalled more partner wordsthan self words in both the immediate (Ms =52.8, and 45.0, respectively) and the delayed(Ms = 56.5 and 47.9, respectively) conditions,although both of these comparisons fall shortof statistical significance.

Discussion

The results of the attribution measure fitnicely with the predictions. When subjects

Table 3Comparison of Dyad Total Words RecalledContributed by Self Versus Total Words RecalledContributed by Partner in Immediate andDelayed Conditons

Dyad totalsof recalled words

Self-totals > partner totalsSelf-totals = partner totalsSelf-totals < partner totals

Immediate

1835

Delayed

1752


were asked immediately after the dyad task toassess the relative contribution of their partnersand themselves, subjects tended to give morecredit to their partners than to themselves.This finding is consistent with the results ofExperiment 1. However, when subjects' attri-butions were assessed 3 days after the task,subjects were more likely to credit themselvesthan their partners. This finding is consistentwith earlier research demonstrating; the ego-centric bias effect.

The results from the word-recall data, how-ever, present another aspect to this phenom-enon. Subjects in both the immediate and thedelayed conditions tended to recall more wordsthey had contributed to the solving of the puz-zle than they could recall their partner con-tributing. This suggests that the reversal of theegocentric bias found in the immediate con-dition may not be completely accounted forthrough the memory storage and retrievalmechanisms that Ross and Sicoly (1979) useto explain the egocentric bias effect. That is,if the perceived contributions were mere re-flections of specific instances (i.e., words) re-called, it would be expected that the attribu-tions would covary strongly with the propor-tions of self-to-partner words recalled. Thefailure to find such a relationship suggests theremay be a difference between one's generalimpression of who contributed how much andthe specific events recalled.

Experiment 3Experiments 1 and 2 demonstrated the ten-

dency for individuals to give their partnersmore credit than they give themselves for agroup project when asked, about their relativecontribution immediately after the task but togive themselves more credit than they give theirpartners when asked about the task some timelater. Why might this procedural differenceaffect subjects' attributions? One explanationcan be found in the large body of literatureexamining the differences between actors' andobservers' attributions of causality (cf. Jones& Nisbett, 1971; Watson, 1982). It typicallyhas been observed that actors tend to attributecausality for their behavior to situationalcauses, whereas persons observing the actor'sperformance (observers) tend to attribute cau-sality for that same behavior to personal qual-ities of the actor. This effect is explained in

part by the differences in perceptual fields ofthe two individuals (Jones & Nisbett, 1971).That is, the observer focuses his or her atten-tion on the actor, whereas the actor focuseshis or her attention on the environment. Bothactor and observer therefore simply may beassigning causality to that which is most salientfor each. Several investigations (Arkin & Du-val, 1975; Storms, 1973; Taylor &Fiske, 1975)provide support for the perceptual field in-terpretation of the actor-observer effect.

When applied to the experiments reportedhere, it appears that the failure to replicatethe egocentric bias effect in Experiment 1 andin the immediate condition in Experiment 2not only makes sense, but perhaps should havebeen predicted from the actor-observer lit-erature. That is, for each of the participantsin the dyad the partner is a more salient featurein the individual's perceptual field than arethe individual's own actions. Thus, when askedto evaluate the relative contribution of theirpartners and themselves, we might expect thatindividuals would attribute more responsibilityfor the dyad's performance to the partner thanto themselves.

Still, the difference between the immediate-and the delayed-attribution conditions re-mains. It is possible that the actor-observerperceptual field effect operates for only a shortperiod of time and that the salience of theenvironmental variable (in this case, the part-ner) dissipates over time. When subjects areasked immediately after a group task to assesstheir and their partners' contributions to thetask, the partner's performance still may behighly salient. Thus, it would be expected thatthese individuals are more likely to attributecausality for the group's performance to thepartner than to their own actions, which atthat point probably are less-salient features.However, over time salience of the partner inone's recalled perception of the event may dis-sipate more rapidly than the recall of one'sown contributions. At some later time, perhapsas short as a few hours for a fairly uneventfulexperience such as working on a crosswordpuzzle, the salience of the partner's behaviormay be reduced enough to allow the mecha-nisms responsible for the egocentric bias tooperate.

This tendency for recall of one's own per-formance to not fade as rapidly as recall of


the partner's performance may be caused bythe differences in the coding and the retrievalof memory, knowledge of personal strategies,and self-enhancing motivations that Ross andSicoly (1979) outlined as causes for the ego-centric bias. For example, it is likely that be-cause we store information about our own be-havior in a more accessible manner than wedo information about others, when we areasked about a group task days later, our owncontributions are more readily recalled thanare the contributions of our partners.

However, one aspect of the data from Ex-periment 2 does not fit cleanly into this in-terpretation. Subjects in the immediate-attri-bution condition recalled more of their owncontributions than their partner's despite at-tributing more of the performance to the part-ner. This finding suggests that a general per-ception of relative contribution, rather thanthe recall of specific events, may be responsiblefor the attribution. That is, subjects may havehad the impression that their partners con-tributed most of the words but could not recallprecisely what those words were. By the timeof the delayed attribution, the salience of thegeneral impression may have faded consid-erably, but some recall of the subjects' ownspecific actions, and not those of the partner,remained.

A pair of studies by Moore, Sherrod, Liu,and Underwood (1979) provides some supportfor this interpretation. These researchers foundthat actors shifted attributions for their be-haviors from situational to dispositional causesover a 3-week period. Moore et al. provideevidence for two processes underlying this shift:(a) a change in focus of attention from thesituation to the self and (b) a difficulty in re-calling situational information because of lessefficient processing than used for dispositionalinformation. Similarly, in Experiment 2 pre-sented here, subjects in the delayed conditionsmay have found the salience of the situation(i.e., the partner) reduced 3 days after the task,thus resulting in the reduced credit to thepartner. However, these subjects may have beenable to recall their own behavior well enoughto give themselves credit for the dyad's per-formance.1

If the above interpretation is correct, thenit would be expected that manipulating thesalience of an actor and his or her partner in

the perceptual field for someone involved ina group task would result in changes in theactor's perceived contributions to the task forhimself or herself and the partner. Experiment3 was designed to test this hypothesis.

The methodology for the experiment istaken largely from a study by Arkin and Duval(1975). These researchers were interested inthe effects of altering the participants' focusof attention on the actor-observer attributionphenomenon. The actor's focus of attentionwas altered from the environment to himselfor herself by aiming a videotape camera at thesubject and informing him or her that the ses-sion was to be videotaped. The procedure wasdesigned to increase feelings of objective self-awareness (Duval & Wicklund, 1972). Thatis, because the subjects believed they werebeing videotaped, their focus of attention pre-sumably was changed from the environmentto themselves. Consequently, Arkin and Duvalfound that under certain conditions videotapedactors, who were focusing on themselves, wereless likely to attribute the cause of their per-formance to the situation than were actors whowere not told they were being videotaped.

In a similar manner, half of the dyads work-ing on a group project in Experiment 3 weretold they were being videotaped as they per-formed the task. The other half was not toldthis. It was expected that the videotaped sub-jects would change their focus of attentionfrom their environments (i.e., their partners)to themselves and subsequently would be morelikely to assess their own contributions to thetask as being larger than those of their partnerswhen asked to make this attribution imme-diately after the task.

Method

Subjects. Thirty men and 30 women undergraduatesserved in the experiment for class credit.

Procedure. Subjects signed up for the experiment insame-sexed pairs. Subjects were instructed to not sign upto participate at the same time as someone they were

1 Note that temporal effects on the actor-observer dif-ference probably are more complex than a simple shifttoward dispositional attributions by actors over time. Millerand Porter (1980), for example, report several studies un-covering the opposite pattern—that actors increased theirsituational attributions for their behaviors over time.


acquainted with. The experimenter verified that the subjectsdid not know each other'before beginning the experiment.

Both subjects were seated together at a large table inthe experimental room, A television monitor, videotapeequipment, and a television camera were located approx-imately 5 feet (4.57 m) in front of the subjects. The monitorfaced the subjects and the camera was pointed directly atthem.

The experimenter explained that she was interested inhow people worked on group tasks. Subjects were told theywould have 15 minutes to work together on the samecrossword puzzle used in the previous experiments. In thefilm condition, subjects were told they would be videotaped

, while they performed the task so that their behaviors couldbe analyzed in detail later. The experimenter then removedthe lens cap, turned on the equipment, and allowed thesubjects to see themselves'on the television monitor beforeturning the monitor away from the subjects. In the no-film condition, subjects were informed that they were tohave been videotaped, but that the equipment was notworking today. The experimenter left the lens cap on thetelevision camera in this condition. Thus, as in the Arkinand Diival (1975) experiment, whatever distracting effectsthe mere presence of the camera and equipment may havehad was kept identical in both conditions.

Subjects then were given 15 minutes to work on thepuzzle. At the end of this time subjects were given thesame feedback sheet used in Experiment 2, which sup-posedly informed them that they had performed as wellas at least the top 8% of those who had participated inthe experiment earlier. At this point subjects were separatedand asked to complete the same questionnaire used inExperiment 2. As in that experiment, the major dependentvariable was found in an item asking subjects to estimatethe relative percentage of the task performance contributedby themselves and by their partners. The questionnairealso contained the items asking subjects to list as manywords as they could remember that they and their partnershad contributed. Following this, subjects were fully de-briefed and dismissed.'

Results

As in the other two experiments, subjectsbelieved they had performed very well on thepuzzle, (M = 7.10, with 1 = "very poorly"and 9 = "very well"). Once again the dyadwas employed as the unit of analysis. The per-.centages of contribution each member of thedyad attributed to himself or herself wereadded together to form the major dependentvariable. The number of dyads with self-totalsgreater than, equal to, and less than 100% arepresented in Table 4. As can be seen in thetable, a similar pattern was obtained for bothmale and female subjects, and thus the sexvariable was collapsed for the data analysis.

A statistical procedure similar to that usedin the previous two experiments was employed.The number of dyads with self-percentage to-

tals above and below 100% were comparedwithin each of the two experimental condi-tions. As predicted, dyads in the film conditionwere significantly more likely collectively togive themselves more responsibility for the taskperformance more often than they gave theirpartners more credit, p < .03, using a binomialtest. In contrast, dyads in the no-film conditiontended collectively to give their partners moreresponsibility more often than they gavethemselves more credit, p < .07, using a bi-nomial test.

The self-to-partner difference score for thenumber of words recalled contributed by one-self and one's partner again was computed foreach dyad. As in Experiment 2, the numberof words each subject recalled the partner con-tributing was subtracted from the numbersubjects remembered themselves contributing.The total of the two self-minus-partner scoresfor each dyad served as the dependent variable.It was found that dyads in the film conditionhad significantly larger self-to-partner differ-ence scores than did subjects in the no-filmcondition, Ms = 1.15 versus .41, respectively;f(28) = 2.17, p < .05. Thus subjects in thefilm condition tended to recall more wordsthey had contributed relative to their partnerssignificantly more than did subjects in the no-film condition.

As in Experiment 2, the self-to-partner dif-ference score was correlated with the self-per-centage total within each condition (r = .30for film, r = .20 for no-film). The mean self-percentage values for subjects recalling moreself-generated than partner-generated wordsversus those with the opposite word-recall pat-tern could not be compared in the film con-dition, because no subject in this conditionrecalled more words contributed by his or herpartner than contributed by himself or herself.In the no-film condition, the subjects recallingmore words that they had contributed did givethemselves more credit for the dyad's perfor-mance (M - 53.6) than did subjects recallingmore words contributed by their partner(Af = 44.0).

Discussion

The results of Experiment 3 provide ad-ditional insight into the causes of the egocen-tric bias phenomenon and the conditions under


which it can and cannot be expected to occur.The findings also help to define the relationshipbetween the egocentric bias and the actor-observer difference. Although these two phe-nomena initially appear to be in contradictionof each other, the results of Experiment 3 helpto define the differences in mechanisms un-derlying the two attributional biases and theconditions under which each may be found.

Consistent with past research on the actor-observer difference, when subjects performedin front of a "live" television camera they weremore likely to attribute greater responsibilityfor the group performance to themselves thanto their partners. This finding follows fromobjective self-awareness theory and researchthat suggests these subjects became self-con-scious in this situation. Because their attentionwas focused on themselves instead of the sit-uation, these subjects were more likely to recalltheir own performance and were more likelyto credit themselves for the group's successfuloutcome than to give their partner's credit.However, when there was no operating tele-vision camera pointed at the subjects, thus notincreasing the subjects' awareness of them-selves, the focus of attention and subsequentattribution of responsibility was given to thepartner. These findings are consistent with alarge number of investigations (cf. Taylor &Fiske, 1978) that find a tendency for individ-uals to attribute causality to that which is sa-lient in their perceptual fields.

The findings, however, are not necessarilyinconsistent with the explanation of the ego-centric bias provided by Ross and Sicoly(1979). These researchers suggested that whenasked about relative contribution to the task,dyad members will attempt to recall the sit-uation and base their responses on these rec-ollections. Differences in the storage and re-trieval of the material, perhaps influenced by

various motivational sources, affect the recall.One additional variable that seems to influencethis recall is the extent to which individualsfocus attention on themselves or their partnersduring the task performance. Consistent withthe above explanation, subjects in the film (self-aware) condition were significantly more likelyto recall words they had contributed relativeto their partner's contribution than were sub-jects in the no-film condition.

The use of the videotape camera to producethe subjects' self-awareness, however, allowsfor the possibility that other subject responseswere created in the film condition. Most no-tably, the subjects' belief that their behaviorwas to be taped, presumably to be evaluatedby others later, may have aroused impression-management concerns (cf. Baumeister, 1982).These subjects may have been more concernedabout how others would evaluate them thanwere subjects in the no-film condition. Al-though such concerns may have caused sub-jects to give themselves more credit than theygave their partners, perhaps causing them toabandon false modesty, this seems unlikely.Subjects desiring to present themselves in afavorable light more likely would want to ap-pear modest and show a nonegotistical eval-uation of the dyad's performance. In a relevantinvestigation, Wells, Petty, Harkins, Kagehiro,and Harvey (1977) found that actors indeedtended to attribute less causality to themselvesfor a successful oucome when they anticipateddiscussing their performance with a researchtrainee who supposedly had watched themthrough a one-way window than when no dis-cussion was anticipated. This suggests thatimpression-management concerns should havedecreased rather than increased attributionsto oneself in the film condition, and thusimpression management probably cannot ac-count for the findings.

Table 4Number of Dyad Self-Percentage Totals for Film and No-Film Conditions

Self-percentage totals


Men

412

Film

Women

530

Total

942

Men

134

No film

Women

205

Total

339


General Discussion

The three experiments presented here dem-onstrate an important qualification to the ego-centric bias phenomenon first described byRoss and Sicoly (1979). In each study subjectswho participated in a group task without aself-awareness manipulation (i.e., a televisioncamera) and who were asked immediately af-terward about their and their partner's relativecontribution to the task tended to give theirpartners more credit for the dyad's perfor-mance than they gave themselves. This effectdisappeared, and in fact reversed, when sub-jects were asked about the dyad's performance3 days later or when subjects were made self-aware during their performance by telling sub-jects they were being videotaped.

The three experiments bring into focus thequestion of how the egocentric bias phenom-enon and the actor-observer attribution effectdiffer. Although further research is needed toclarify the entire process, the available datasuggest that under normal circumstances in-dividuals working within a dyad focus theirattention on their partners. If asked about theirimpressions of the relative contribution bythemselves and their partner immediately afterthe task, the salience of the partner's perfor-mance will cause individuals to assign moreresponsibility to the partners than to them-selves. This is consistent with much researchon the actor-observer difference, which alsofinds this effect for attributions of causalityabout the actor's performance. However, overa period of time the salience of the partner'sperformance appears to fade from memoryand the actor recalls his or her own contri-butions more readily than the contributionsof the partner. This may be because the in-dividual's own actions are coded in memoryin such a manner as to make them less sus-ceptible to memory decay. At some point, per-haps a matter of days or even hours for rel-atively noneventful tasks such as the crosswordpuzzle employed here, the salience of the part-ner's performance may have dissipated to theextent that it no longer causes the individualto assign much responsibility for the perfor-mance to the partner. However, perhaps be-cause the individual's own contributions stillare easily recalled from memory, he or she islikely to attribute more responsibility for the

dyad's performance to himself or herself thanto the partner.

Further investigation is needed, however, atseveral points in this description. First, thedifference between the general impression ofthe partner's performance and the actual recallof that performance needs to be clarified. Assuggested in Experiment 2, individuals mayhave a general sense that their partner or othersin a group have done most of the work, butmay not be able to recall specifically what theseother people have done. A related issue in-volves obtaining a better understanding of thedifferences in the storage and retrieval of one'sown actions versus the actions of others andidentifying the variables affecting the proposeddifference in rates of memory decay for thetwo types of information. Finally, because theyappear to result in contradicting effects, theconditions under which the egocentric bias andthe actor-observer effect are likely to be foundneed to be outlined more fully.

ReferencesAlloy, L. B., & Abrarason, L. Y. Judgment of contingency

in depressed and nondepressed students: Sadder butwiser? Journal of Experimental Psychology: General,1979, 108, 441-485.

Alloy, L. B., & Abramson, L. Y. Learned helplessness,depression, and the illusion of control. Journal of Per-sonality and Social Psychology, 1982, 42, 1114-1126.

Arkin, R. M, & Duval, S. Focus of attention and causalattributions of actors and observers. Journal of Exper-imental Social Psychology, 1975, 11, 427-438.

Baumeister, R. F. A self-presentational view of social phe-nomena. Psychological Bulletin, 1982, 91, 3-26.

Beck, A. T. The diagnosis and management of depression.Philadelphia: University of Pennsylvania Press, 1972.

Duyal, S., & Wicklund, R. A. A theory of objective self-awareness. New York: Academic Press, 1972.

Golin, S., Terrell, F., & Johnson, B. Depression and theillusion of control. Journal of Abnormal Psychology,1977, 86, 440-442.

Golin, S., Terrell, F., Weitz, J., & Drost, P. L. The illusionof control among depressed patients. Journal of Ab-normal Psychology, 1979,55,454-457.

Gough, H. G., & Heilbrun, A. B. The Adjective CheckList Manual. Palo Alto, Calif.: Consulting PsychologistsPress, 1980.

Jones, E. E., & Nisbett, R. E. The actor and the observer:Divergent perceptions of the causes of behavior. Mor-ristown, N.J.: General Learning Press, 1971.

Lewinsohn, P. M., Mischel, W., Chaplain, W., & Barton,R. Social competence and depression: The role of illusoryself-perceptions. Journal of Abnormal Psychology, 1980,89, 203-212.

Miller, P, T, & Porter, C. A. Effects of temporal perspectiveon the attribution process. Journal of Personality andSocial Psychology, 1980, 39, 532-541.


Moore, B. S., Sherrod, D. R., Liu, T. J., & Underwood, Taylor, S. E., & Fiske, S. T. Salience, attention, and attri-B. The dispositional shift in attribution over time. Jour- bution: Top of the head phenomena. In L. Berkowitznal of Experimental Social Psychology, 1979, 75, 553- (Ed.), Advances in experimental social psychology (Vol.569. 11). New York: Academic Press, 1978.

Ross, M., & Sicoly, F. Egocentric biases in availability and Thompson, S. C, & Kelley, H. H. Judgments of respon-attribution. Journal of 'Personality and Social Psychology, sibility for activities in close relationships. Journal of1979, 37, 322-336. Personality and Social Psychology, 1981, 41, 469-477.

Rozensky, R. H., Rehm, L. P., Pry, G., & Roth, D. Depres- Watson, D. The actor and the observer: How are theirsion and self-reinforcement behavior in hospitalized pa- perceptions of causality divergent? Psychological Bul-tients. Journal of Behavior Theaapy and Experimental letin, 1982, 92, 682-700.Psychiatry, 1977, 8, 35-38. Wells, G. L., Petty, R. E., Harkins, S. G., Kagehiro, D.,

Siegel, S. Nonparametric statistics for the behavioral sci- & Harvey, J. H. Anticipated discussion of interpretationences. New York: McGraw-Hill, 1956. eliminates actor-observer differences in the attribution

Storms, M. D. Videotape and the attribution process: Re- of causality. Sociometry, 1977, 40, 247-253.versing actors' and observers' points of view. Journal ofPersonality and Social Psychology, 1973,27, 165-175.

Taylor, S. E., & Fiske, S. T. Point of view and perceptionsof causality. Journal of Personality and Social Psychology, Received December 24, 19821975, 32, 439-445. Revision received May 2, 1983 •

Instructions to Authors

APA policy prohibits an author from submitting the same manuscript for concurrent con-sideration by two or more journals. Authors should prepare manuscripts according to thePublication Manual of the American Psychological Association (3rd edition). All manuscriptsare subject to editing for sexist language. Manuscripts must include an abstract of 100-150words. Instructions on tables, figures, references, metrics, and typing (all copy must be double-spaced) appear in the Manual. Articles not prepared according to the Manual will not bereviewed. For further information on content, authors should refer to the editorial in the March1979 issue of this journal (Vol. 37, No. 3, pp. 468-469).

The reference citation for any article in any JPSP section follows APA's standard referencestyle for journal articles; that is, authors, year of publication, article title, journal title, volumenumber, and page numbers. The citation does not include the section title.

Manuscripts should be submitted in quadruplicate (the original and three photocopies) tothe appropriate section editor. All copies should be clear, readable, and on paper of good quality.Editors' addresses appear on the inside front cover of the journal. Authors should keep a copyof the manuscript to guard against loss.

Blind review is optional and must be specifically requested when a manuscript is first submitted.In those cases, authors' names and affiliations must appear only on a separate title page, includedwith each copy of the manuscript. Footnotes containing the identity of authors or their affiliationsmust be on a separate sheet.

Section editors reserve the right to redirect papers among themselves as appropriate unlessan author specifically requests otherwise. Rejection by one section editor is considered rejectionby all, therefore a manuscript rejected by one section editor should not be submitted to another.

attributions of responsibility for group tasks: the …attributions of responsibility for group...

Documents