Attribution of Responsibility and Valence of Outcome in Relation to Initial Confidence and Success and Failure of Self and Other

Journal of Personality and Social Psychology1971, Vol. 18, No. 2, 173-188

ATTRIBUTION OF RESPONSIBILITY AND VALENCE OFOUTCOME IN RELATION TO INITIAL CONFIDENCE

AND SUCCESS AND FAILURE OF SELFAND OTHER1

N. T. FEATHER2 AND J. G. SIMON

Flinders University of South Australia

Male and female subjects in like-sex pairs first worked at 5 practice anagramsand then at IS test anagrams. For both practice and test anagrams, difficultylevel was manipulated so that one-half of the subjects did well and one-halfdid poorly. After the practice items the subjects announced the number ofanagrams they had solved. Before beginning the test items, the subjects rated thedegree of confidence about passing the test both for self and other. After thesubjects were told if they had passed or failed, they rated the degree to whichthey felt performance was due to ability (internal attribution) or luck (ex-ternal attribution) for both self and other. They also provided recall measuresand satisfaction ratings for self and other. Results indicated that before theybegan the test items, the subjects were more confident of the success of theother than of their own success. For both self and other, the unexpected out-come was more often attributed to variable environmental factors (good or badluck) than the expected outcome. Other's success was more often attributed toability than own success, while other's failure was more often attributed tobad luck than own failure. There was a positivity bias in recall, favoring theother. Contrast effects occurred for satisfaction ratings. As usual, task per-formance was a dominant factor influencing confidence and satisfaction ratings.Results are discussed in terms of models involving Heider's principle of balanceand his analysis of the causes of action, in terms of positivity biases in socialperception, and as indicating effects of the social context of performance uponattribution and valence.

When a person succeeds or fails at a task, formance was due to ability (internal attri-the degree to which he attributes responsibil- bution) or luck (external attribution). Theity for the outcome to ability or luck depends results of the study provided clear evidenceupon his initial expectations of success. In a that attribution of responsibility for the out-recent study (Feather, 1969), subjects come was a function of subjects' initial ex-worked anonymously at an anagrams test pectations of success at the task. If they wereconsisting of 10 items of approximately 50% initially confident, then they tended to attrib-difficulty, each presented for 30 seconds, ute success to ability and failure to bad luck.They were told that to pass the test they But if they were initially unconfident, thenwould have to solve S anagrams or more; they tended to attribute success to good luckotherwise, they failed. Before commencing and failure to lack of ability. However, cotn-they rated the degree to which they were con- plete attribution of responsibility for successfident that they could pass the test. These or failure either to internal or external fac-ratings were taken as measures of their initial tors was the exception rather than the rule,expectations of success. On completing the Most subjects attributed their performance totest and noting success or failure, subjects a mixture of ability and luck. The resultsrated the degree to which they felt their per- also indicated that scores close to the passing

i This study was conducted as an honors project mark were more likely to be attributed to ex-by the second author under the supervision of the ternal factors (good or bad luck) than werefirst author (Simon, 1969). scores at either extreme.

a Requests for reprints should be sent to N. T. HQW ft ^ b interpreted? TwoFeather, Department of Psychology, School of Social J ^Sciences, Flinders University of South Australia, Bed- models were proposed, each based upon theford Park, South Australia S042. seminal contributions of Heider (1958). The

173

174 N. T. FEATHKK AND J. G. SIMON

first model invoked the familiar principle ofbalance to predict that good outcomes (suc-cess) would be attributed to the self whenthere was positive self-evaluation, but wouldbe disowned (attributed to external factors)when there was negative self-evaluation. Incontrast, bad outcomes (failure) would beattributed to the self when there was negativeself-evaluation, but would be disowned (at-tributed to external factors) when there waspositive self-evaluation. This balance modelwas represented in terms of signed digraphs(see Feather, 1969, p. 130). The results ofthe study described above fit this model ifit is assumed that the self-evaluation as repre-sented in the model is self-evaluation withrespect to the particular task to be performed,that high expectations of success connote posi-tive self-evaluation, and that low expectationsof success connote negative self-evaluationwith respect to that task.

The second model employed Heider's(1958, Ch. 4) naive analysis of action, inwhich the outcome of an action is related tofactors that reside in the person and factorsthat reside in the environment, and wherethese factors may be stable (dispositional)or variable (see also Weiner, Frieze, Kukla,Reed, Rest, & Rosenbaum, 1971). It was as-sumed that on the basis of his own past expe-rience, a person comes to develop a relative!}'stable estimate of his own ability in regard totasks of a particular type. Inferences aboutability would develop over occasions that per-mitted variations in the situational context ofperformance and where, despite these varia-tions, the person's performance evidenced acertain consistency in regard to objective cri-teria, that is, when it is possible to rule outenvironmental causes of variations in outcomes(see also Kelley, 1967). A person's specific ex-pectation of success at a task would belargely a function both of this relatively sta-ble estimate of ability and of his perceptionof task difficulty. Information about taskdifficulty may be conveyed by obvious situa-tional characteristics (e.g., complexity) and/or by the performance of others (e.g., thepercentage of people able to reach a certaincriterion). His expectation of success couldalso be affected by other factors that influ-ence an outcome, such as opportunity, luck,

and effort. For example, the person mightconsider his chances to be high because hefeels lucky and/or because he intends to exerthimself more than usual. But these factorsare more variable in contrast to the relativelystable dispositional properties of ability andtask difficulty, and their role would seem tobe more in explaining outcomes (see below)than in influencing expectations.

It was assumed in the second model thatwhen a person's expectation of success is con-firmed by an outcome, his assumptions aboutall the causes of the outcome will also beconfirmed. But when the person's expectationof success is strongly disconfirmed by an out-come, then he is more likely to appeal tospecific variable factors such as luck and/oreffort to explain the outcome, since the otherproperties that underlie performance (abilityand task difficulty) are assumed to be rela-tively stable (dispositional). The assumptionthat these two factors are stable would bejustified under conditions where the estimateof ability was rooted in a consistent patternof past performance, and where informationabout task difficulty was veridical and thetask was not altered. In the case of an unex-pected outcome, therefore, success or failurewould be attributed either to an increase ordecrease in effort over the normal require-ment, to good or bad luck, or to a mixture ofboth effort and luck. Although the study de-scribed above did not obtain information con-cerning effort or lack of it (it assumed thatnormal effort was expended), the results werecertainly consistent with the implication thatthe unexpected outcome would be more likelyto be attributed to variable factors such asgood or bad luck than the expected outcome.

The two models described above should notbe taken as mutually exclusive approaches tothe understanding of the process of attribut-ing responsibility for outcome. On the con-trary, conceptual demarcations between thetwo models are by no means clear-cut in thisarea of research, consisting often of littlemore than differences in emphasis or termi-nology. The balance model relates the direc-tion of attribution to certain preferred orbalanced states, and enables one to predictthe conditions under which success or failurewill be seen as owned or disowned by the self

ATTRIBUTION OF RESPONSIBILITY AND VALENCE OF OUTCOME 175

(internal versus external attribution). Thebalance model also implies that one couldcreate tension by setting up conditions thatproduce unbalanced structures (e.g., by mak-ing it difficult for a person with a high ex-pectation of success to attribute failure toexternal circumstances). If this were done,one would expect the person to react in sucha way as to restore balance. The model basedupon the naive analysis of action is rathermore general and less concerned with moti-vational processes than is the balance model.Its important contribution is its classificationof the causes of action into those that arepersonal or internal versus those that are en-vironmental or external, and those that arestable or dispositional versus those that arevariable. The person is assumed to engage ina sort of cognitive induction in assigning thecauses of action. Both models were supportedby the results of the study described above.

The investigation which is the focus of thepresent report extended the analysis of attri-bution to situations involving another person,that is, both self- and other-attributions wereinvestigated. Very little research has beenconducted on how a person attributes thecauses of another person's success or failure.What sorts of differences might one expect inattributions made for self and other in rela-tion to success or failure at a task? A recentinvestigation by Jones and his colleagues(Jones, Rock, Shaver, Goethals, & Ward,1968) indicated that some self-other differ-ences (e.g., primacy and recency effects ofinformation about task performance) mightdepend upon the fact that one has more in-formation on which to base attribution whenit applies to one's own behavior rather thanto the behavior of another person. In addi-tion, other investigations have indicated per-vasive positivity biases in social perception.For example, De Soto and Kuethe (1959)showed that in the absence of any informa-tion about another person, the subjectiveprobability that that person would be likedwas greater than ,50. A number of studiesconcerning the learning and recall of balancedand unbalanced structures have indicatedthat positive relations have an advantageover negative relations (see Feather, 1970).It is possible that such biases might

also operate when attributions are made inregard to another person. For example, onemight be more ready to see another person'ssuccess as due to positively regarded factorssuch as ability and/or effort than as due togood luck or an easy task. And one might ex-cuse the other person's failure as due to badluck or a difficult task than as due to nega-tively valued factors such as inability and/orlack of effort. It is also possible that posi-tivity biases might act to distort upward aperson's recall of another's performance rela-tive to the other's recall of that performance.

These speculations were explored in thepresent study, which focused upon attribu-tions related to self and other as a functionof performance outcome (success or failure)and initial expectation of success. The studywas based upon the two models presented inthe Feather (1969) paper and describedabove. The same type of prediction was madewith respect to initial expectation of success,namely, that for both self and other, the un-expected success and the unexpected failurewould be more likely to be attributed to vari-able factors such as good or bad luck thanwould the expected success and the expectedfailure. The present study, however, enableda comparison to be made between attribu-tions concerned with subjects' own perform-ances and those related to the performanceof others. Finally, in contrast to the earlierstudy, a deliberate attempt was made to ma-nipulate success and failure at the task ex-perimentally, rather than to rely upon subjectself-selection as before. Selective control in-volves the risk that unknown factors maybe confounded with success and failure,though the results of the previous study weresufficiently strong to render alternative inter-pretations of the obtained differences unlikely(see Feather, 1969, Footnote 6).

While the primary focus of the presentstudy was on attribution behavior, informa-tion concerning the valence of the outcome(attractiveness and repulsiveness of successand failure) was also collected. The mainpredictions here were quite straightforward,namely, that successful outcomes will be moreattractive than unsuccessful outcomes, thatsuccessful outcomes will be more attractivewhen initial expectations of success are low

176 N, T. FEATHER AND J. G. SIMON

than when they are high, and that unsuccess-ful outcomes will be more repulsive when in-itial expectations of success are high thanwhen they are low. These predictions are con-sistent with a valence-difficulty model pre-sented in Feather (1969) and with the resultsof other previous research (Feather, 1963,1967, 1968b). The design of the presentstudy also afforded the opportunity of study-ing the valence of one's own outcome in rela-tion to the performance of the other person.Thus, how attractive one finds one's ownsuccess or how repulsive one finds one's ownfailure might be influenced by whether theother person succeeds or fails at the task,

METHOD

Subjects and Basic ProcedureThe subjects were 130 students (66 male, 64 fe-

male) in the introductory psychology course atFlinders University of South Australia. They weretested during a 2-week period in April-May 1969.One pair of male subjects was eliminated becausea member of this pair expressed suspicion about theexperimental manipulations.

Subjects were tested in like-sex pairs and workedindependently and simultaneously at a set of 20anagrams. They were seated side by side at separatesmall desks facing the experimenter.3 They weretold that the experiment was designed to study theperformance of university students under test con-ditions, and that only limitations of time preventedindividual testing. Each session took approximately30 minutes.

Expectations of success at the task (high versuslow) were experimentally manipulated by varyingthe difficulty level of the first 5 "practice" anagrams.The outcome on the last 15 "test" items (success orfailure) was similarly manipulated by varying thedifficulty level of the anagrams. Ratings of initialconfidence were collected after the S practice items.Ratings of attribution and valence and a measureof recall were obtained after the 15 test items hadbeen completed.

Since there were two levels of induced expecta-tion of success (high or low) for both self and other,and two levels of performance outcome (success orfailure) for both self and other, there were 16 dif-ferent experimental conditions altogether. For ex-ample, one of these conditions involved a successfulperson with a high expectation of success pairedwith an unsuccessful other with a low expectationof success, etc. Sixty-four male and 64 female sub-jects were randomly assigned to the 16 experimentalconditions in like-sex pairs, with four pairs (twomale pairs, two female pairs) being assigned to eachcondition.

3 The experimenter was John Simon.

Performance TaskInduction of expectations: Practice Hems. The first

five anagrams were designated as practice anagrams,and each anagram was printed on a separate pageof a small booklet. Subjects were told that they hadto rearrange each group of letters so that they madea meaningful English word. They were given 30seconds to work at each anagram. The difficultylevel of these practice items was experimentally ma-nipulated, so that half of the subjects did well onthe items and the other half did poorly. The ex-perimental manipulation was achieved by using twodifferent booklets, one containing easy items (EVOLTR,WITHNI, POLIEC, WADNET, DAMAEc) and the othercontaining difficult or insoluble items (NARCEN,HITGKN", UFTRUE, MORBEP, GGAWiL). All of the differ-ent possible combinations of success and failure atthe practice items were allowed for in the experi-mental design. Both members of a pair could suc-ceed at the practice items, both members of a paircould fail at the practice items, or one membercould succeed and the other fail at the practiceitems. These experimental manipulations were as-sumed to induce widely different expectations ofsuccess at the task, with prior success determining ahigh expectation and prior failure a low expectation.

Manipulation of outcome: Test items. After com-pleting the practice booklet, subjects were instructedto count the number of anagrams they had solvedin the time allowed. Each member of the pair thenstated his score aloud. Hence, each subject knewhow many practice items the other subject had cor-rectly solved. Each subject was then given a set of25 white cards, each card with 1 anagram printedin the center. They were told that each had thesame set of 25 anagrams, that the anagrams variedwidely in difficulty, and that the deck had beenrandomly shuffled. They were then told that theywere required to attempt 15 of the total of 25 ana-grams in the deck and that they would have 30seconds to work on each anagram. At the end of 30seconds, each subject was to hand his card to theexperimenter when it would be checked and the re-sult for each subject (correct or incorrect) an-nounced. After that, each subject would turn overhis next card, work at the next anagram for 30 sec-onds, hand it to the experimenter for checking andannouncement of result, etc., for the full set of 15anagrams. Subjects were told that if they solved 8or more anagrams, then they passed the test; ifthey scored 7 or less, then they failed the test.

In fact, two different decks of cards were em-ployed, although the subjects were led to believe thateach had the same deck randomly shuffled. One deckwas heavily weighted with easy anagrams to inducesuccess at the task. The other deck was weightedwith difficult and insoluble items to induce failureat the task.4 The decks were distributed so that halfof the subjects succeeded at the task and the otherhalf failed. Again, all of the different possible com-binations of success and failure were allowed for in

4 Copies of the test anagrams can be obtainedfrom the first author,

ATTRIBUTION OF RESPONSIBILITY AND VALENCE OF OUTCOME 177

the experimental design. Both members of a paircould pass the test, both could fai l , or one membercould pass and the other fail the test. The reasonsfor using 25 anagrams of which only 15 were at-tempted by subjects were twofold: to increase thecredibility of the deception involved in the tasksituation and to increase the salience of the luck/opportunity factor relative to task difficulty.

Initial confidence. As the subjects read the instruc-tions, the experimenter ostentatiously shuffled theanagram decks and placed each deck, face down infront of the subject. Before they began the test, thesubjects were asked to rate "how confident you arethat you can pass the test." They indicated this byputting a cross on a S-inch scale with the statement,"Not confident at all" at one extreme, "Very con-fident" at the other extreme, and "Moderately con-fident" in the middle. Next they were asked to rate"how confident you are that the other person canpass this test," again by placing a cross on an iden-tical 5-inch scale.

These ratings were made privately and were as-sumed to reflect the subjects' expectations of successboth for self and for other. It was assumed thatthese expectations would be influenced by knowledgeof one's own and other's performance on the prac-tice items, success determining relatively high con-fidence, and failure determining relatively low con-fidence. Subjects were given no other informationregarding the difficulty of the anagrams test otherthan that ostensibly provided by the practice items.The expectation ratings for self and other werescored from 0 to 10 in the direction of increasingconfidence. This score will be referred to as the in-itial confidence score.

After subjects had provided their confidence rat-ings, they commenced working on the 15 test ana-grams. At the end of each 30-second trial, cardswere collected from each pair in counterbalancedorder, and the result of each item (correct or in-correct) for each subject was announced aloud bythe experimenter. After the fifteenth anagram, theexperimenter publicly informed each subject of hisoverall result on the test ("You have passed/failedthe test"), but the subject's actual number of itemscorrect was not announced.

Postperjormance QuestionnaireAttribution ratings. After each subject's overall

result (pass or fail) was publicly announced, thesubjects were required to answer a postperformancequestionnaire which took one of four forms corre-sponding to the outcomes of self and other on thetest, namely, pass-pass, pass-fail, fail-pass, and fail-fail. For example, if both subjects passed the test(the pass-pass condition), the instructions were asfollows:

How a person does at tasks like the one youhave just completed depends upon a number offactors. At the one extreme there are some peoplewho are just lucky enough to get the right com-bination of letters quickly in the time allowed.They happen to bit upon the right combination

of letters largely by chance. Or they arc luckyenough to select a lot of very easy anagrams fromtheir deck in the first place. If they were to re-peat a similar test, they might not do so well.

At the other extreme there are people who suc-ceed because they have the skill and ability. Goodluck isn't really involved for these people. If theywere to repeat a similar test, they would probablydo just as well because they have the ability.

Consider your own performance on the test. Inyour case, do you consider that your performancewas mainly clue to good luck, mainly due to skilland ability, or reflected some mixture of good luckand ability?

Subjects put a cross on a 5-inch scale, with thestatement, "Mainly due to good luck" at one ex-treme of the scale, the statement, "Mainly due toability" at the other extreme, and the statement,"50% luck, 50% ability" in the middle. These rat-ings were assumed to reflect external (good luck)versus internal (ability) attribution for the self andwere scored 0-10 in the direction of external attri-bution.

Next, subjects in the pass-pass condition were in-structed as follows:

Now consider the other person's performance onthe test. In his/her case do you consider that his/her performance was mainly due to good luck,mainly due to skill and ability, or reflected somemixture of good luck and ability?

Subjects put a cross on a 5-inch scale identical tothat described above. These ratings were assumedto reflect internal versus external attribution for theother, and were scored as for self-attribution.

If both subjects failed the test (the fail-fail con-dition), the instructions were as follows:

How a person does at tasks like the one youhave just completed depends upon a number offactors. At the one extreme there are some peoplewho are just not lucky enough to get the rightcombination of letters in the time allowed. Orthey are unlucky enough to select a lot of verydifficult anagrams from their pack in the firstplace. Their failure is mainly due to bad luck. Ifthey were to repeat a similar test, they might dobetter.

At the other extreme are people who fail be-cause they lack the necessary skill and ability. Badluck isn't really involved for these people. If theywere to repeat a similar test, they would probablydo just as poorly because they lack the necessaryability.

Consider your own performance on the test. Inyour case do you consider that your performancewas mainly due to bad luck, mainly due to lackof skill and ability, or reflected some mixture ofbad luck and lack of ability?

Subjects put a cross on a 5-inch scale, with thestatement, "Mainly due to bad luck" at one extremeof the scale; "Mainly due to lack of ability" at theother extreme; and the statement, "50% bad luck,

178 N. T. FEATHER AND J. G. SIMON

TABLE 1

MEAN INITIAL CONFIDENCE SCORES FOR MALES ANDFEMALES IN RELATION TO INDUCED

EXPECTATION OF SUCCESS

Inducedexpectancy

for self

HighHighLowLow

Inducedexpectancy

for other

HighLowHighLow

Column M

Males

Confidenceratings

Self

S.136.692.002.003.95

Other

6.314.446.563.445.19

Females

Confidenceratings

Self

3.885.383.003.193.86

Othcr

6.194.007.314.445.49

Note.—K = 16 for all conditions. Initial confidence wasscored in the positive direction.

50% inability" in the middle. These ratings wereassumed to reflect external (bad luck) versus inter-nal (inability) attribution for the self, and werescored 0-10 in the direction of external attribution.

Next, subjects in the fail-fail condition were in-structed as follows:

Now consider the other person's performance onthe test. In his/her case do you consider that his/her performance was mainly due to bad luck,mainly due to lack of skill and ability, or reflectedsome mixture of bad luck and lack of ability?

Subjects put a cross on a 5-inch scale identical tothat described above. These ratings were assumedto reflect internal versus external attribution for theother and were scored as for self-attribution.

Instructions for subjects who were successful inthe presence of an unsuccessful other (the pass-failcondition) were identical to those for self-ratingwhen both subjects passed and other-rating whenboth subjects failed, except for minor variations inthe opening statements. These subjects also putcrosses on the appropriate 5-inch scales ("good luck-ability" for self; "bad luck-lack of ability" forother). Instructions for subjects who were unsuc-cessful in the presence of a successful other (thefail-pass condition) were identical to those for self-rating when both subjects failed and other-ratingwhen both subjects passed, except for minor varia-tions in the opening statements. These subjects alsoput crosses on the appropriate S-inch scales ("badluck-lack of ability" for self; "good luck-ability"for other). All ratings were scored from 0 to 10 inthe direction of external attribution.

Recall of performance level. In each of the fourpostperformance questionnaires, the subjects wereasked, "Approximately how many anagrams did youcorrectly solve?" They were also asked, "Approxi-mately how many anagrams did the other personcorrectly solve?" They answered each question byputting a circle around one numeral in a row ofnumerals from 1 to 15.

Valence of outcome. Subjects were also asked ineach of the four postperformance questionnaires torate how satisfied they were with their own scoreon the anagrams test and how satisfied they thoughtthe other person would be with his/her score on theanagrams test. They answered each question byputting a cross on a 5-inch scale, with the statement,"Not satisfied at all" at one extreme of the scale;the statement, "Moderately satisfied" in the middle;and the statement, "Extremely satisfied" at the otherextreme. Satisfaction ratings were scored 0-10 inthe direction of increasing satisfaction with per-formance.

Finally, subjects were asked to write down anyprior information and any comments they had con-cerning the experiment. On the basis of these com-ments, one pair of male subjects was excluded fromthe analysis because one member of the pair indi-cated strong suspicion of the manipulations.

Form of AnalysisInitial confidence scores, attribution scores, and

valence scores were separately analyzed using a 2 X2 X 2 X 2 X 2 X 2 analysis of variance with sex, ownoutcome (success or failure), other outcome (suc-cess or failure), own expectancy induction (high orlow), other expectancy induction (high or low), andrating (self or other) as factors in the analysis,and repeated measures on the last factor (Winer,1962). Because of the complexity of these analyses,complete analysis of variance tables are not pre-sented, but the relevant significant effects will beabstracted.

RESULTSAnalysts of Initial Confidence Scores

Table 1 presents the mean initial confi-dence scores for males and females (self-rating and other-rating) in relation to inducedexpectation of success for self and other.

Table 1 indicates that taken overall, sub-jects were more confident of the success ofthe other than of their own success. The maineffect of rating (self versus other) was highlysignificant (F = 70.71, df = 1/96, p < .001).In addition Table 1 shows that the experi-mental inductions influenced ratings of initialconfidence. In particular, there was a highlysignificant Own Expectancy X Rating inter-action (F = 73.84, df = 1/96, p < .001).The expectancy induction for self affected theinitial confidence scores for self but not forother. In a corresponding way the Other Ex-pectancy X Rating interaction was highlysignificant (F = 95.80, df = 1/96, p < .001).The expectancy induction for other affectedthe initial confidence scores for other but not

ATTRIBUTION OF RESPONSIBILITY AND VALENCE OP OUTCOME 179

TABLE 2

MEAN ATTRIBUTION SCORES FOR MALES AND FEMALES IN RELATION TO INDUCEDEXPECTATION or SUCCESS AND PERFORMANCE OUTCOME

Inducedexpectancy

for self

HighHighHighHighHighHighHighHighLowLowLowLowLowLowLowLow

Inducedexpectancy

for other

HighHighHighHighLowLowLowLowHighHighHighHighLowLowLowLow

Performanceoutcomefor self

SuccessSuccessFailureFailureSuccessSuccessFailureFailureSuccessSuccessFailureFailureSuccessSuccessFailureFailure

Performanceoutcomefor other

SuccessFailureSuccessFailureSuccessFailureSuccessFailureSuccessFailureSuccessFailureSuccessFailureSuccessFailure

Males

Attribution ratings

Self

4.SO6.2S4.754.754.504.506.505.005.758.75

.754.505.754.003.751.50

Other

3.007.504.755.505.004.255.504.254.008.75

06.755.255.003.005.00

Females

Attribution ratings

Self

6.255.754.504.004.754.256.253.756.256.255.003.255.755.753.505.25

Other

3,757.253.505.004.754.756.003.753.506.752.756.505.005.753.005.25

Note.—» = 4 for all conditions. Attribution was scored in the external direction.

for self. Thus, the experimental manipulationsthat were designed to produce differences ininitial confidence were successful. Ratings ofinitial confidence were higher when the per-son being rated (self or other) had donewell at the five practice anagrams than whenhe had done poorly, but subjects tended tobe more confident of other's success than ownsuccess when ratings were averaged over allconditions.

Females did not have generally lower initialconfidence scores than males as they did inthe previous study (Feather, 1969). The maineffect of sex was not significant. However, theSex X Own Expectancy interaction was sig-nificant (^=12.80, d}=l/96, p<.001).The confidence ratings were higher for malesthan for females when the experimental pro-cedure was designed to induce a high expecta-tion of success for self, but lower for malesthan for females when it was designed to in-duce a low expectation of success for self.

Analysis of Attribution Scores

Table 2 presents the mean attributionscores for males and females (self-rating andother-rating) in relation both to induced ex-pectation of success for self and other andperformance outcome for self and other.

The results emerging from the analysis ofvariance that were most relevant to predic-tions were a significant Own Outcome XOwn Expectancy X Rating interaction (F =4.44, df = 1/96, p < .05) and a significantOther Outcome X Other Expectancy X Rat-ing interaction (F = 11.59, df - 1/96, p<.001). The basis of the former interaction isshown in Figure 1; the basis for the latterinteraction is shown in Figure 2. Figure 3presents the attribution results from the previ-ous study (Feather, 1969).

Figure 1 shows that if a person succeededat the task, attribution ratings for one's suc-cess were more external when the procedurewas designed to induce a low expectation ofsuccess than when it was designed to induce ahigh expectation, that is, one's own unex-pected success was more likely to be attrib-uted to good luck than one's expected success.If a person failed at the task, however, attri-bution ratings for one's failure were moreexternal when the procedure was designed toinduce a high expectation of success thanwhen it was designed to induce a low expecta-tion, that is, one's own unexpected failurewas more likely to be attributed to bad luckthan one's expected failure. These results areconsistent with those of the previous study,

180 N. T. FEATHER AND J. G. SIMON

External 7-00

6'00

CDon

5'00

UJ4'00

Internal 3 oo _

Own Success — Own Attribulion Rating

Own Failure — Own Attribution Rating

Own Success —Other Attribution RatingFailure — Other Attribution Rating

IHIGH LOW

INDUCED EXPECTANCY FOR SELFFIG. 1. Mean attribution scores (own and other) in relation to own outcome

and induced expectancy for self,

although, in contrast to the previous results,there was little difference in attribution rat-ings for one's success and failure when initialexpectations of success could be assumed tobe high—both means were close to S, imply-ing neither internal nor external attribution(see Feather, 1969, and Figure 3 for compari-

son). When initial expectations of successcould be assumed to be low, however, one'ssuccess was attributed to good luck and one'sfailure to lack of ability.

The same trends were evident (though lessstrong) when attribution ratings for other'soutcomes were considered in relation to own

Externol 7^00

zoP 6-00

CD

5-00

2

UJ

Internal 3^00

4 »i Other Success - Other Attribution RatingA A -A Other Failure —Other Attribution Rating\ A AOther Success - Own Attribution Rating

s^A Mother Failure — Own Attribution Rating

\\

1 1

HIGH LOW

INDUCED EXPECTANCY FOR OTHER

FIG. 2. Mean attribution scores (own and other) in relation to other outcomeand induced expectancy for other.

ATTRIBUTION OF RESPONSIBILITY AND VALENCE OF OUTCOME

External 7-00

181

2OH- 6-00

Sa:

< 5'00

LU4'00 _

Internal

. Own Success - Own Attribution Rating

Own Failure - Own Attribution Rating

(N=36)

(N=44)

J_HIGH LOW

INITIAL CONFIDENCE FOR SELF

FIG. 3. Mean attribution scores in relation to own outcome and initial con-fidence for self (adapted from Feather, 1969).

expectancy induction and own performanceoutcome (see Figure 1). Thus, the results im-ply that when a subject's own outcome onthe test items was unexpected, he was morelikely to attribute the outcome to good or badluck than when the outcome was expected.These results were in line with prediction,

Figure 2 shows that if the other personsucceeded at the task, attribution ratings forother's success were more external when theprocedure was designed to induce a low ex-pectation of success for other than when itwas designed to induce a high expectation;that is, another's unexpected success wasmore likely to be attributed to good luck thananother's expected success. If the other personfailed at the task, however, attribution rat-ings for other's failure was more externalwhen the procedure was designed to producea high expectation of success for other thanwhen it was designed to induce a low expecta-tion; that is, another's unexpected failure wasmore likely to be attributed to bad luck thananother's expected failure. These results arein line with prediction, although it should benoted that there was little difference in at-tribution ratings for another's success wheninitial expectations of success could be as-sumed to be low—both means were slightlyless than 5, implying a slight tendency towardinternal attribution. When initial expecta-tions of success could be assumed to be high,

however, another's success was attributed toability and another's failure to bad luck.

Opposite trends were evident when attribu-tion ratings for one's own outcomes were con-sidered in relation to other expectancy induc-tion and other performance outcome (see Fig-ure 2 ) , but these effects were not very strong.Thus, the results imply that when another'soutcome on the test items was seen to be un-expected, the other's outcome was more likelyto be attributed to good or bad luck thanwhen it was seen as expected. These resultsare also in line with prediction.

Consideration of Figure 1 and Figure 2separately could miss an important effectwhich is evident in Figure 4. This figure con-siders attribution ratings for self only in rela-tion to own expectancy induction and ownoutcome and attribution ratings for other onlyin relation to other's expectancy inductionand other's outcome.

Figure 4 shows quite clearly that when theother person succeeded at the task, there wasa greater tendency to attribute his success tointernal factors (ability) than when one'sown outcome was success. But when the otherperson failed at the task, there was a greatertendency to attribute his failure to externalfactors (bad luck) than when one's own out-come was failure. These effects were inde-pendent of the level of induced expectancy.For both levels of expectancy, the attribution

182 N. T. FEATHER AND J. G. SIMON

External T O O

2 6-00

o

5-00

<UJ

400

Internal 3-00 _^

Q 0 Own Success-Own Attr ibulion Rating

Q Q Own Failure — Own Attr ibut ion Rating

A A Other Success—Othe r Attribution Rating\A Bother Failure —Other Attribution Rating

\

HIGH LOW

INDUCED EXPECTANCY FOR SELF OR OTHER

FIG. 4. Mean attribution scores (own or other) in relation to own or otheroutcome and induced expectancy for self or other.

ratings for other's success were more internalthan the attribution ratings for one's ownsuccess. And for both levels of expectancy,the attribution ratings for other's failure weremore external than the attribution ratings forone's own failure.

In the analysis of variance, the followinginteractions relevant to the above differenceswere significant or approached significance:Own Outcome X Rating (F = 3.18, dj = \/96, p < .10) and Other Outcome X Rating(F = 42.76, df - 1/96, p < .001). In regardto the former interaction, subjects tended torate their own success more externally thantheir own failure, in comparison to other-rat-ings. With regard to the latter interaction,subjects tended to rate another's success moreinternally than another's failure, in compari-son to self-ratings.

Finally, in contrast to the results of theprevious study (Feather, 1969), females didnot have higher attribution scores than males,that is, they were not more external in theirattributions than were males. The main effectof sex was not significant.

In summary, therefore, the analysis of at-tribution ratings indicated that two sorts ofeffects occurred in the social situation underconsideration. As predicted for both self andother, the unexpected outcome was morelikely to be attributed to external factors

(good or bad luck) than the expected out-come. But overlaying this effect was anothernew finding of considerable interest. Successfor the other person was more likely to beseen as reflecting ability (internal attribution)than success for oneself; failure for the otherperson was more likely to be seen as reflectingbad luck (external attribution) than failurefor oneself.

Analysis oj Performance Recall Scores

Table 3 presents the mean performancelevels on the test items for subjects in thevarious outcome conditions and the meanperformance recall scores on the test itemsfor the person given both by himself and bythe other for the various outcome conditions.

Over all outcome conditions, there was atendency for the person to distort recall of hisown performance downward and for the otherto distort recall of the person's performanceupward. The former tendency was strongestwhen the person's outcome was success, thelatter, when his outcome was failure. Table 3also presents the results of t tests applied tothe recall data. Individual comparisons be-tween the recall scores given by the personand the corresponding recall scores given bythe other reveal that except for the success-fail condition, mean recall scores of the other

ATTRIBUTION OF RESPONSIBILITY AND VALENCE OP OUTCOME 183

TABLE 3MEAN PERFORMANCE LEVELS FOR PERSONS, MEAN RECALL PERFORMANCE LEVELS FOR

PERSONS GIVEN BY SELF AND OTHER, AND INDIVIDUAL COMPARISONSBETWEEN MEAN RECALL SCORES

Outcome condition

All successful personsSuccess (self )-succcss (other)Success (self )-fail (other)All unsuccessful personsFai 1 (self )-success (other)Fail (self)-fail (other)

All persons

Mean performancelevel for person

10.9810.4411.534.113.844.387.55

Mean recall scoreof person for person

10.419.88

10.944.033.814.257.22

Mean recall scoreof other for person

10.9110.5911.194.925.094.757.91

t.

2.56**2.55**

.995.39***6.05***2.11*5.40***

» All tests were one-tailed and for correlated samples.* t < .025.

**p < .01.*** p < .0005.

were significantly greater than those of theperson.5

Analysis of Valence Scores

Table 4 presents mean ratings of satisfac-tion with performance outcome for males andfemales (self-rating and other-rating) in rela-tion to both induced expectation of success

5 As in previous studies conducted by the firstauthor (e.g., Feather, 1968a), performance on thetest items tended to be lower following initial fail-ure on the practice items (M = 7.03) than followinginitial success (M = 8.06). This effect occurred ir-respective of whether or not the person passed orfailed the test items.

for self and other and performance outcomefor self and other.

The analysis of variance indicated that themain effect of own outcome (F — 49.44, dj- 1/96, p< .001) and the main effect ofother outcome (F — 87.67, dj — 1/96, p <.001) were statistically significant. Subjects'ratings of satisfaction over all conditions werehigher when one succeeded (M = 4.69) thanwhen one failed (M — 3.13), and higher whenother succeeded (M = 4.9S) than when otherfailed (M = 2.88). Of more interest, however,were the significant Own Outcome X OtherOutcome (F = 15.94, dj = 1/96, p < .001),

TABLE 4

MEAN SATISFACTION RATINGS FOR MALES AND FEMALES IN RELATION TO INDUCEDEXPECTATION OP SUCCESS AND PERFORMANCE OUTCOME

Inducedexpectancy

for self

HighHighHighHighHighHighHighHighLowLowLowLowLowLowLowLow

Inducedexpectancj'

for other

HighHighHighHighLowLowLowLowHighHighHighHighLowLowLowLow

Performanceoutcomefor self

SuccessSuccessFailureFailureSuccessSuccessFailureFailureSuccessSuccessFailureFailureSuccessSuccessFailureFailure

Performanceoutcomefor other

SuccessFailureSuccessFailureSuccessFailureSuccessFailureSuccessFailureSuccessFailureSuccessFailureSuccessFailure

Males

Satisfaction ratings

Self

5.004.252.501.756.506.002.251.504.506.001.502.005.255.252.001.25

Other

4.503.507.001.505.253.007.252.255.502,258.251.755.501.256.003.75

Females

Satisfaction ratings

Self

5.507.251.501.505.506.751.00.50

4.507.50

.251.754.257.001.00.75

Other

6.501.257.501.505.25

.758.75

.755.752.009.502.255.251.507.501.75

Note.—ft = 4 for a]j conditions. Satisfaction was scored in the positive direction.

184 N. T. FEATHER AND J. G. SIMON

8-00

o6-00

2

$ 4-00

2<UJS 2-00

. t j Moles - Own Satisfaction Rating0- 0 Males — Other Satisfaction Rating

O Q F°r""1'" — r""" Satisfaction RatingO- — -O Females - Other Satisfaction Rating

SUCCESS FAILURE

OWN OUTCOME

Fig. 5. Mean satisfaction scores for males and femalesin relation to own outcome.

Own Outcome X Rating (7? =167.55, df =1/96, p < .001), and Other Outcome X Rat-ing interactions (F - 152.37, df = 1/96, p <.001). The first of these interactions was dueto the fact that ratings of satisfaction tendedto be especially high when both self and othersucceeded (M = S.28), and especially lowwhen both self and other failed (M = 1.66).Ratings for the own-success-other-failure con-dition (M — 4.09) and the own-failure-other-success condition (M = 4.61) were intermedi-ate.

The significant Own Outcome X Ratinginteraction was due to the fact that whereasa person rated his own satisfaction with out-come higher when he succeeded than when hefailed, the effects on the ratings of the otherperson's outcome were in the reverse direc-tion: satisfaction ratings for the other's out-come were higher when self failed than whenself succeeded. Similarly, the significant OtherOutcome X Rating interaction was due to thefact that whereas a person rated other's satis-faction with outcome higher when the othersucceeded than when the other failed, theeffects on one's own ratings were in the re-verse direction: satisfaction ratings for one'sown outcome were higher when other failedthan when other succeeded. These effects weremagnified for females when compared tomales. The Sex X Own Outcome X Rating in-teraction (F = 5.95, df = 1/96, p < .05) andthe Sex X Other Outcome X Rating interac-tion (F = 14.36, df - 1/96, p < .001) wereboth statistically significant. These unusualeffects are indicated in Figures 5 and 6. Theysuggest that a complex process of social com-parison occurred in the experimental situa-tion, that subjects tended to displace satis-faction ratings for others' outcomes awayfrom their own when self had succeeded orfailed, that subjects tended to displace satis-faction ratings for their own outcomes away

8-00

6-00

CO

CO 4'00

UJ2 2-00

QM.. i ,3 Males — Own Satisfact ion Rating

g --- 9 Males — Other Satisfaction Rating

\ QI Q Females— Own Satisfaction Rating

\O --- O Females — Other Satisfaction Ratinge

iK^\\

D

SUCCESS

OTHER

FAILURE

OUTCOME

FIG. 6. Mean satisfaction scores for males and females in relation to otheroutcome.

ATTRIBUTION OF RESPONSIBILITY AND VALENCE or OUTCOME 185

from those for others' outcomes when otherhad succeeded or failed, and that these effectswere more likely to occur among females thanamong males.

Figures 5 and 6 also show that females weremore responsive to the success and failure ex-periences: the effect of own success and ownfailure on own ratings was greater for femalesthan for males; and the effect of other successand other failure on other ratings was greaterfor females than for males.

None of the predicted effects involving ex-pectation of success occurred. Satisfaction rat-ings were not higher when success was associ-ated with a low initial expectation of successthan with a high initial expectation of success.Nor were satisfaction ratings lower when fail-ure was associated with a high initial expecta-tion of success than with a low initial expecta-tion of success. The dominant factor influen-cing satisfaction ratings was outcome, and ex-pectancy effects were submerged.

DISCUSSION

The results presented above replicate thoseof the previous study (Feather, 1969) since,in line with prediction, a person wasmore likely to attribute an unexpected out-come to variable, environmental factors (goodor bad luck) than an expected outcome. Thisresult occurred both in relation to self andother. It is important to note, however, thatthe results of the present study did not fullysupport the stronger hypothesis, that the un-expected outcome would be attributed to ex-ternal factors (good or bad luck) and the ex-pected outcome to internal factors (ability orlack of ability). Mean attribution ratings forexpected success and unexpected failure wereclose to S (a mixture of ability and luck)when attributions concerning own outcome inrelation to own expectancy were considered(see Figure 1 when own expectancy was high).Mean attribution ratings for expected failureand unexpected success were slightly below 5when attributions concerning other outcomeand other expectancy were considered (seeFigure 2 when other expectancy was low). Inboth of these cases, there was no clear ten-dency for scores to move toward the internalor external poles of the scale; ratings tended

to be intermediate. When own expectationswere low, however, and other expectationswere high, attribution ratings diverged, andthe stronger hypothesis was supported.

The stronger hypothesis would be consistentwith the implications of the balance model. Interms of the model, good outcomes (success)would be attributed to self when there waspositive self-evaluation (high expectancy), butwould be disowned when there was negativeself-evaluation (low expectancy); bad out-comes (failure) would be attributed to selfwhen there was negative self-evaluation (lowexpectancy), but would be disowned whenthere was positive self-evaluation (high expec-tancy). The results of the previous study,which did not involve a coacting partner,were fully consistent with these implications(see Figure 3), but the results of the presentstudy were not consistent with these implica-tions when own expectations were high andother expectations were low. Hence, the pres-ent results are not altogether compatible withthe balance model, although the former resultsare. In view of these findings, it may be nec-essary to supplement the balance model bymaking allowance for other forms of cogni-tive bias that may be elicited in a social situ-ation, for example, the positivity bias notedin the present study.

The requirements of the naive action modelare not as stringent as those of the balancemodel. The naive analysis of action predictsthat when a person's expectation of success isdisconfirmed by an outcome, then he will tendto appeal to variable factors (luck and/oreffort) to explain the outcome, since the otherpossible causes of outcomes (those involvingability and task difficulty) are assumed to bestable. Hence, the weaker hypothesis, that theunexpected outcome will be more likely to beattributed to variable factors than the ex-pected outcome, follows from this analysis,and this weaker hypothesis was supported bythe results of both studies so far conducted.Both sets of results, then, are essentially com-patible with the implications from the naiveaction formulation.

A further variable factor that might beconsidered as influencing the outcome whenan expectation is disconfirmed is effort. Where-as luck may be considered as a variable en-

186 N. T. FEATHER AND J. G. SIMON

vironmental factor, however, effort can beclassified as a variable personal factor (seeWeiner et al, 1971). Future studies might alsoinclude measures of the perceived amount ofeffort expended, since subjects might accountfor unexpected outcomes not only in terms ofgood or bad luck, but also in terms of an in-crease or decrease in the amount of effort ex-pended in relation to the average effort nor-mally required.

The important contribution of the presentstudy, however, is the firm indication that thesocial context of performance has markedeffects on attribution behavior. A person wasmore likely to attribute another's success toability than his own success, and was morelikely to attribute his own failure to inabilitythan another's failure. Moreover, there was astrong and pervasive tendency for the recallscore given by another to exceed the cor-responding score given by oneself. Further-more a person was more optimistic about thechances of success for the other than for him-self (see Table 1). In this sense the presentdata are consistent with those of other studieswhich have indicated positivity efforts in so-cial recall, interpersonal attraction, etc. (DeSoto & Kuethe, 19S9; Feather, 1966, 1970).Presumably, the presence of a coactingpartner and the public announcement of re-sults on both the practice items and the testitems created a situation in which socialcomparison could occur and where social mo-tives were elicited. A person might feel thatattributing success to ability would be boast-ful and that attributing failure to bad luckwould involve "making excuses," that a bet-ter self-presentation would occur if one tendedto devalue one's success and to admit to one'sfailure when another was present. In contrast,one might be more willing to acknowledge thesuccess of another as due to ability, since thiswould imply approval of the other, and to ex-cuse the failure of another as due to bad luck,since this reaction would avoid negative criti-cism of the other (one does not like to beaccused of lack of ability). Hence, it might beargued that a desire to create and maintainpositive relationships with the other by react-ing to own and other outcomes in a mannerdesigned to win the approval of the other wasa general motivation underlying the effects of

the social context discovered in the presentstudy.

The difficulty with the above argument,however, is that the self-other effects occurredeven though attribution ratings were not pub-licly announced in the presence of the other.But the ratings were public, in the sense thatthey would be seen by the experimenter, andit is this regard that a more adequate expla-nation of the self-other results suggests itself.Thus, it is possible that subjects were intentupon gaining the approval of the experimenterand were conscious that their behavior wasbeing evaluated by the experimenter, that is,that there was some degree of evaluationapprehension in the situation (Rosenberg,1965). By this reasoning, subjects tended torespond in ways that they thought would besocially acceptable to the experimenter (hu-mility, generosity).

The remaining self-other results might alsobe explained in terms of the subject's concernwith self-presentation. A subject who rateshimself as very confident about succeedingmight also run the perceived risk of beinglabeled as boastful and cocksure, and he mighttherefore tend to reduce his estimate whenother people are present, particularly when anexperimenter is perceived to be evaluatingthat estimate. No such risks are involved whenone rates another's chances. In this case ahigh estimate could be taken to mean thatone has magnanimous faith in the other'sability. The tendency for subjects to distortrecall of their own performance downward andanother's performance upward may be inter-preted in a similar manner. Again, the otherperson is upgraded, the self, devalued, andthe subject presents himself to the experi-menter as being characterized by modesty andgenerosity.

The social context was also important ininfluencing how subjects rated their own andthe other's satisfaction with outcome. Satis-faction ratings for the other's outcome werehigher when self failed than when self suc-ceeded; satisfaction ratings for own outcomewere higher when the other failed than whenthe other succeeded. These results remind oneof the contrast effects frequently obtained instudies of social judgment (Sherif, Sherif, &Nebergall, 196S). There was no evidence that

ATTRIBUTION OP RESPONSIBILITY AND VALENCE OF OUTCOME 187

the unexpected success was rated as moresatisfying than the expected success or thatthe unexpected failure was rated as more re-pulsive than the expected failure. Perhaps thepredicted relationships between expectationand valence failed to appear because the ex-pectations themselves were based upon per-formance on the practice items, and subjectsmay have reacted to performance on bothpractice and test items in making their satis-faction ratings. In fact, the satisfaction rat-ings were relatively low in the low-expecta-tion-failure conditions where maximal failureoccurred, and were relatively high in the high-expectation-success conditions where maximalsuccess occurred (see Table 4). Thus, it ap-pears that task performance overshadowedother factors in influencing satisfaction rat-ings. The dominant effect of sheer frequencyof success and failure in influencing subjects'ratings of satisfaction or disappointment withoutcome has been noted before (Feather,1965).

The results of the present study indicatedthat females were rather more responsive tosuccess and failure experiences than weremales, in that their satisfaction ratings werehigher following success and lower followingfailure when they rated with respect to ownoutcome and when they rated with respect toother outcome (see Figures 5 and 6). In theprevious study (Feather, 1969), females whofailed rated themselves as less satisfied withtheir performance than did males who failed(p < .05), but there was no significant dif-ference between males and females in theratings following success. Also, in the formerstudy, females had lower initial confidencescores than did males and higher externalattribution scores than did males. These twosex differences did not occur in the presentstudy. The failure to find lower expectancyestimates in females than in males is espe-cially puzzling, since this difference has beenfound in other studies as well (e.g., Crandall,1969; Feather, 1968a). In the Feather (1969)study it was argued that this sex difference,together with the higher external attributionscores and the higher feelings of inadequacyscores that were also found among females,may indicate the way females conceive of the

feminine role, that is, that the female shouldappear modest and dependent, and that as-sertiveness and self-confidence are masculinetraits. But this conclusion was based upon theresults of studies where males and femalesworked at the same task in a large mixedgroup under conditions that could be per-ceived as competitive. Under such conditionsthe motive to avoid success ("fear of suc-cess") may be elicited among females (Hor-ner, 1969), and this motive could be associ-ated with lower confidence estimates andhigher external attribution.0 In the presentstudy, however, only like-sex pairs were in-volved in task performance, and fear of suc-cess may not have been an important factor,since females could not see themselves as incompetition with males. If this interpretationis correct, it implies that the former differ-ences would be obtained in a competitive situ-ation where the pairs of subjects involve amale and a female, and where subjects realizethat their results will be made public.

More generally, it would be of interest infuture studies to vary the degree of competi-tion involved in the performance situation.Under competitive conditions the positivitybiases noted above might not occur. Instead,one might attempt to upgrade the self and todevalue the other person when one is in com-petition with him.

In conclusion, the results of the presentstudy have advanced our understanding ofattribution and valence in two main ways. Inthe first place, they underline the need forcareful analysis of the social context of per-formance in future studies of attribution andvalence. Furthermore, they strengthen thegeneralization that it is the unexpected out-come rather than the expected outcome thattends to be attributed to variable factors(both external and internal) which may un-derlie performance.

6 Werner and Kukla (1970) have argued that sub-jects high in resultant achievement motivation (highmotive to achieve success, low motive to avoid fail-ure) will be more likely to attribute success inachievement-oriented situations to themselves thansubjects classified as low in resultant achievementmotivation (low motive to achieve success, high mo-tive to avoid failure).

188 N. T. FEATHER AND J. G. SIMON

REFERENCES

CRANDAIX, V. C. Sex differences in expectancy of in-tellectual and academic reinforcement. In C. P.Smith (Ed.), Achievement-related motives in chil-dren. New York: Russell Sage Foundation, 1969.

DE SOTO, C. B., & KUETHE, J. L. Subjective proba-bilities of interpersonal relationships. Journal ofAbnormal and Social Psychology, 1959, 59, 290-294.

FEATHER, N. T. Mowrer's revised two-factor theoryand the motive-expectancy-value model. Psycho-logical Review, 1963, 70, 500-515.

FEATHER, N. T. Performance at a difficult task inrelation to initial expectation of success, test anxi-ety, and need achievement. Journal of Personality,196S, 33, 200-217.

FEATHER, N. T. The prediction of interpersonal at-traction: Effects of sign and strength of relationsin different structures. Human Relations, 1966, 19,213-237.

FEATHER, N. T. Valence of outcome and expectationof success in relation to task difficulty and per-ceived locus of control. Journal of Personality andSocial Psychology, 1967, 7, 372-386.

FEATHER, N. T. Change in confidence following suc-cess or failure as a predictor of subsequent per-formance. Journal of Personality mid Social Psy-chology, 1968, 9, 38-46. (a)

FEATHER, N. T. Valence of success and failure inrelation to task difficulty: Past research and re-cent progress. Australian Journal of Psychology,1968, 20, 111-122. (b)

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. Balancing and positivity effects insocial recall. Journal of Personality, 1970, 38, 602-62S.

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

HORNER, M. Fail: Bright women. Psychology Today,1969, 3, 36-62.

JONES, E. E., ROCK, L., SHAVER, K. G., GOETHALS, G.R., & WARD, L. M. Pattern of performance andability attribution: An unexpected primacy effect.Journal oj Personality and Social Psychology,1968, 10, 317-340.

KELLY, H. H. Attribution theory in social psychol-ogy. Nebraska Symposium on Motivation, 1967,15, 192-240.

ROSENBERG, M. J. When dissonance fails: On elimi-nating evaluation apprehension from attitudemeasurement. Journal of Personality and SocialPsychology, 1965, 1, 28-43.

SHERIF, C. W., SHERIF, M., & NEBERGALL, R. Atti-tude and attitude change: The social judgment-involvement approach. Philadelphia: Saunders,1965.

SIMOX, J. G. Attribution of responsibility for ownand other's task outcome in relation to expectationof success and task outcome for self and other.Unpublished honors thesis, Flinders University ofSouth Australia, 1969.

WEINER, B., FRIEZE, I., KUKLA, A., REED, I., REST, S.A., & ROSENEAUM, R. M. Perceiving the causes ofsuccess and failure. In E. E. Jones, D. Kanouse, H.H. Kelley, R. E. Nisbett, S. Valins, & B. Weiner(Eds.), Attribution: Perceiving the causes of be-havior. New York: McCaleb-Seiler, 1971, in press.

WEIXER, B., & KUKLA, A. An attributional analysisof achievement motivation. Journal of Personalityand Social Psychology, 1970, 15, 1-20.

WINER, B. J. Statistical principles in experimentaldesign. New York: McGraw-Hill, 1962.

(Received May 25, 1970)