Cha n an d Yba rraINT ER ACTION GOAL S AND MEM ORY

INTERACTION GOALS AND SOCIALINFORMATION PROCESSINGUNDERESTIMATING ONErsquoS PARTNERS BUTOVERESTIMATING ONErsquoS OPPONENTS

Emily Chan and Oscar YbarraUniversity of Michigan

The present experiments examined how the interaction goals of cooperation andcompetition affected social information processing Study 1a and 1b evaluatedwhether people prefer to overestimateor underestimate another personrsquos strengthswhen assessing a partner or an opponent The findings indicated that people wereinclined to underestimate how good their partners were but to overestimate howgood their opponents were In Study 2 consistent with the strategy selections fromStudy 1 the results showed that participants anticipating cooperation with anotherstudent rememberedbest information thatwas diagnostic of negativequalities thanpositive qualities In contrast participants expecting to compete with another stu-dent remembered best information that was diagnostic of positive qualities thannegative qualities In Study 3 participants had a chance to actively seek out infor-mation about a potential partner or opponent by selecting a subset of their behav-iors to verify The results provided a validation of the results from Study 2 Thefindings were discussed in terms of their implications for interpersonal and inter-group perception

When people meet others for the first time what do they notice and whatinformation do they emphasize Most likely if they were to meet somepeople at a bar and casually chatted with them about the latest NASCAR

Social Cognition Vol 20 No 5 2002 pp 409-439

409

Emily Chan Department of Psychology University of Michigan Oscar Ybarra Depart-ment of Psychology University of Michigan

We thank the members of the Social Cognition Lab for their helpful comments Parts ofthis paper have been presented at the annual meeting of the Society of Personality and So-cial Psychology Nashville TN 2000 and also at the annual meeting of the MidwesternPsychological Society Chicago IL 2000 This research has been supported by a BarbourFellowship and the Hough Summer Fellowship to the first author from the University ofMichigan

Correspondence concerning this article should be addressed to Emily Chan or OscarYbarra Department of Psychology 525 East University Ann Arbor MI 48109 E-mailechanumichedu or oybarraumichedu

race they would in all likelihood remember different information thanif they had interviewed these people for postndashdoctoral research posi-tions Information pertaining to intellectual creativity and academic po-tential should be more relevant when interacting with postndashdoctoralcandidates than when talking to race car fans However for the carndashrac-ing fans other information such as the types of cars they drive their fa-vorite race teams and their knowledge about cars would be relevantinformation

Peoplersquos interaction goals can be quite specific as the above examplesillustrate However some goals are broader in nature and occur across awide range of contexts Examples of such goals include cooperating andcompeting with other people Cooperation and competition are an inte-gral part of everyday interaction ranging from childrenrsquos play sportsand games work and business interactions to intergroup relations andworld politics Thus it is of interest and importance to examine howthese broader goals affect social information processing Before thinkingabout such outcomes though it is important to briefly outline how otherresearchers have approached the study of cooperationand competition

Some research has shown for example that competition enhances thesalience of intergroup distinctions leading to greater bias and an in-crease in the perception of outgroup homogeneity (cf Brewer 1979) Co-operation and competition have also been investigated in the context ofdecision making and negotiation in mixed motive games (eg Bornsteinamp Rapoport 1988Hertel amp Fiedler 1994 Insko Schopler Hoyle Dardisamp Graetz 1990 Parks Henager amp Scamahorn 1996) Findings haveshown for example that public appeal (Rosen amp Haagen 1998) discus-sion within groups (Bornstein amp Rapoport 1988) and increasing the sa-lience of positive connotations of cooperation (Hertel amp Fiedler 1994) allincrease cooperation On the other hand anonymity and low risk of so-cial exclusion decrease cooperation (Kerr 1999) These studies focus ondelineating the conditions that foster cooperation and competition butdo not address the social information processing strategies associatedwith these interaction goals

ANTICIPATED INTERACTION COOPERATION COMPETITIONAND SOCIAL INFORMATION PROCESSING

The mere anticipation of having to interact with others has powerful ef-fects on social cognition For example research has shown that expect-

410 CHAN AND YBARRA

ing to meet and interact with another person leads to more extensiveinformation processing compared to control conditions (eg memoriza-tion condition) (Devine Sedikides amp Fuhrman 1989 Srull amp Brand1983)

Another small body of research that has explored the interaction goalsof cooperation and competition has shown that these interaction goalsaffect a targetrsquos attractiveness leading to increased attention under an-ticipated cooperation but a decrease under anticipated competition(Tesser amp Danheiser 1978) The interaction goals of cooperation andcompetition have also been shown to lead to more thorough processingof target information (Neuberg amp Fiske 1987 Ruscher amp Fiske 1990)Neuberg and Fiske (1987) showed that participants who expected to co-operate with a previously hospitalized schizophrenic person did not en-gage in categoryndashbased processing but individuated the partnerRuscher and Fiske (1990) presented participants with a competitive in-teraction and led them to form either a positive or negative compe-tencendashbased expectancy Then they allowed the participants to readinformation about their interaction partner that was either consistentinconsistent or irrelevant to the expectancy They found that partici-pants engaged in individuating processes such as showing increased at-tention to inconsistencies and forming more varied impressions of thepartner

THE PRESENT CONCEPTUALIZATION

The current research approaches the study of cooperation and compe-tition in a slightly different manner and construes them as guidingprinciples in peoplersquos selection of information processing strategywhich leads to overestimating or underestimating othersrsquo qualitiesMuch research in person perception and decision making suggests thatpeople are risk aversemdashpotential losses loom larger than potentialgains (eg Kahneman amp Tversky 1979) Thus when learning about afuture partner under a cooperation goal or an opponent under a com-petition goal a perceiver might selectively process information that ismost indicative of potential losses This would be adaptive because be-ing aware of potential losses and dangers would allow the perceiver toprepare in advance to minimize the chance of loss and if the riskseemed too large to avoid the cooperation or competition to preventloss

INTERACTION GOALS AND MEMORY 411

Specifically a perceiver who is sizing up a future partner can minimizepotential losses by avoiding the overestimationof the partnerrsquos strengthsThis can be done by being sensitive to information diagnostic of the part-nerrsquos negative qualities For example if a person finds out that it is easy fora partner to lose enthusiasm after initial involvement with a project he orshe could anticipate the problem and attempt to reduce its negative im-pact on outcomes In addition the perceiver would also be motivated toattend to such weaknesses to decide how much to trust and rely on thepartner If the perceiver learned that the partner was weak in some skillhe or she could try to strengthen that skill in the self or at least be vigilantto any task that involved that skill (Williams amp Karau 1991 WittenbaumVaughan amp Stasser 1998) The cost of overestimating how good the part-ner is and failing to notice the partnerrsquos weaknesses is immense if the per-son mistakenly thinks that the partner is hardworking and talented andtherefore becomes complacent the likelihood of failure is increased

The interaction goal of competition puts the perceiver in a differentposition A perceiver who is sizing up a future opponent can minimizepotential losses by avoiding the underestimation of the opponentrsquosstrengths This can be done by being sensitive to information diagnosticof the opponentrsquos positive qualities For example if a person finds outthat his opponent is highly skilled and hardworking he could anticipatethe challenge and attempt to reduce his chance of losing by workingeven harder and improving his own skills Underestimating the oppo-nent would undoubtedly increase the risk of defeat because theperceiver would be underndashprepared for the competition

Study 1a and 1b first tested the critical assumptions regarding how theinteraction goal of cooperation would lead people to avoid overestimat-ing how good their partners are whereas the interaction goal of compe-tition would lead people to avoid underestimating how good theiropponents are In a subsequent validation of our analysis Study 2 exam-ined how these interaction goals affected memory for behavioral infor-mation from different trait domains Finally Study 3 examined how theinteraction goals affected the nature and amount of information peoplesought about their future partnerscompetitors

STUDY 1A

We reasoned that people who are evaluating a potential partner shouldtry to avoid the mistake of overestimating the partnerrsquos positive qualities

412 CHAN AND YBARRA

because being generally risk averse they should focus on potentiallosses that could arise from cooperating with an inferior partner Thusthey should attempt to gather negative dispositional information aboutthe partner People who are evaluating a potential opponent should tryto avoid the mistake of underestimating the opponentrsquos positive quali-ties because being risk averse they should focus on potential losses thatcould arise from competing with a superior opponent Thus theyshould attempt to gather positive dispositional information about theopponent To examine these predictions participants in the first studywere presented a fictitious scenario in which a person in the scenario hadto learn about someone else with whom he was to cooperate or competeParticipants were then asked to evaluate the targetrsquos strategy indicatinghow costly it would be for the target to overestimate or underestimatehow good his partner or opponent was

METHOD

Design and ParticipantsEighteen students between 19ndash25 years old volunteered for the studywithout payment They were approached on a university campus andasked if they would fill out a short questionnaire

Materials and ProcedureParticipants were informed that the study was a short psychology ques-tionnaire dealing with interpersonal perception They were given aquestionnaire containing either the cooperation or the competition sce-nario In the cooperation condition the scenario read

Person X is about to meet someone with whom he will have to work to-gether on a task Person X has a few moments to find out what the other per-son is like After the meeting Person X will have to decide what strategy totake during the cooperation in order to achieve the most successful out-come

In the competition condition the scenario read

Person X is about to meet someone with whom he will have to compete on atask Person X has a few moments to find out what the other person is likeAfter the meeting Person X will have to decide what strategy to take duringthe competition in order to achieve the most successful outcome

INTERACTION GOALS AND MEMORY 413

Participants then answered 2 questions Those in the cooperation condi-tion answered ldquoHow costly is it for Person X to mistakenly overestimatehow good the partner isrdquo and ldquoHow costly is it for Person X to mistak-enly underestimate how good the partner isrdquo on a 1 (not at all costly) to7 (extremely costly) scale Those in the competition condition answeredthe same items with the word ldquocompetitorrdquo replacing the word ldquopart-nerrdquo The presentation order of the scenarios was counterbalanced Af-ter they completed the questionnaires the participants were debriefedand thanked for their participation

RESULTS AND DISCUSSION

Participantsrsquo responses were submitted to a 2 (goal competition coop-eration) acute 2 (estimation overestimation underestimation) acute 2 (questionorder) mixed ANOVA with the second factor varying within subjectThere was no main effect or interaction associated with question orderand thus the order factorwas dropped from the analysis It was expectedthat overestimating how good onersquos partner is and underestimatinghow good onersquos opponent is would be deemed undesirable and this wasindeed the case The only significant effect was the goal acute estimation in-teraction F(1 16) = 2918 p lt 001 When expecting to cooperate peoplejudged that it was more costly to overestimate (M = 556) than underesti-mate (M = 289)onersquos partner F(1 16) = 3011p lt 001 In contrastpeoplejudged that it was more costly to underestimate (M = 574) than overesti-mate (M = 294) onersquos opponent F(1 16) = 1580 p = 001

The findings from Study 1a suggest that for cooperation people deemit preferable to avoid overestimating how good the partner is and forcompetition to avoid underestimating how good the opponent is Onecould argue that these results were obtained because participants wereexplicitly instructed to consider how costly it was to overestimate or un-derestimate the other person By focusing peoplersquos attention on the cost-liness of their decisions we might have led the participants to be morelossndashfocused than they would have been otherwise Another feature ofthis study that requires further examination is whether asking partici-pants to imagine the scenario from anotherrsquos perspective instead ofimagining themselves in the scenario might have affected the resultsMuch social psychological research has found differences between peo-plersquos psychological processes when they are thinking about themselvesversus others For example people tend to make situational causal attri-

414 CHAN AND YBARRA

butions for themselves but dispositional causal attributions for others(Jones amp Nisbett 1972) People pay more attention to unobservable andunintentional events when thinking about themselves but observableand intentional events when thinking about others (Malle amp Pearce2001) Peoplersquos representations of themselves in memory are character-ized by their own thoughts and feelings whereas their representationsof others are characterized by actions and appearances (McGuire ampMcGuire 1986 Prentice 1990) Although it is not perfectly clear how itis possible that the findings from Study 1a might not occur if participantswere asked to think of how they themselves would react To rule outthese alternative accounts we conducted Study 1b to replicate the re-sults of Study 1a

STUDY 1B

The current study was identical to Study 1a save for two features Firstwe asked the participants to judge how overestimating or underestimat-ing the other person would affect their chances of success instead of ex-plicitly pointing to the costliness of the decision as in Study 1a Secondwe asked participants to imagine themselves instead of others in scenar-ios where they either had to cooperate or compete with another person

METHOD

Design and ParticipantsTwenty students between 19ndash25 years old participated in this studyThey were approached on a university campus and asked to fill out ashort questionnaire

Materials and ProcedureParticipants were informed that the study was a short psychology ques-tionnaire dealing with interpersonal perception They were asked toimagine themselves in the following scenarios

In the cooperation condition the scenario read

You are about to meet Person X and both of you will have to work togetherYou have a few moments to find out more about Person X After finding outmore about what Person X is like you will have to decide what strategy to

INTERACTION GOALS AND MEMORY 415

take during the cooperation in order to achieve the most successful out-come

In the competition condition the scenario read

You are about to meet Person X and both of you will have to compete againsteach other You will have a few moments to find out more about Person XAfter finding out more about what Person X is like you will have to decidewhat strategy to take during the competition in order to achieve the mostsuccessful outcome

Participants then answered 2 questions Those in the cooperation condi-tion answered ldquoIf you were to overestimate how good your partner ishow would it affect your chances of achieving a successful outcomerdquoand ldquoIf you were to underestimate how good your partner is howwould it affect your chances of achieving a successful outcomerdquo on a 1(reduce chance of success) to 7 (increase chance of success) scale Thequestions in the competition condition were the same with the wordldquoopponentrdquo replacing the word ldquopartnerrdquo The presentation order of thescenarios was counterbalanced After they completed the question-naires the participants were debriefed and thanked for their participa-tion

RESULTS AND DISCUSSION

Participantsrsquo responses were submitted to a 2 (goal competition coop-eration) acute 2 (estimation overestimation underestimation) acute 2 (questionorder) mixed ANOVA with the second factor varying within subjectThere was neither a main effect nor interaction associated with questionorder so it was dropped from the analysis It was expected that peoplewould judge that overestimating their partners and underestimatingtheir opponents would reduce their chances of success The results wereconsistent with the hypothesis The only significant effect was the goal acuteestimation interaction F(1 19) = 4532 p lt 001When expecting to coop-erate people judged that their chances of success would be lower if theyoverestimated their partners (M = 200)than if they underestimated theirpartners (M = 410) F(1 19) = 2000 p lt 001 In contrast people judgedthat their chances of success would be higher if they overestimated theiropponents (M = 430) than if they underestimated their opponents (M =260) F(1 19) = 4185p lt 001These findings thus confirm those of Study

416 CHAN AND YBARRA

1a and indicate that people may be inclined to adopt different informa-tion processing strategies depending on their interaction goals

If it really is the case that people when expecting to cooperate arepoised to process and be vigilant about a partnerrsquos negative qualities(underestimation) it would be expected that in a different cognitive taskthose inclinations would also be expressed For example in learningabout a potential partner who is described with negative and positivequalities we should find that people expecting to cooperate are morelikely to elaborate information diagnostic of the partnerrsquos negative thanpositive qualities Similar outcomes should be expected under antici-pated competition If people who expect to compete are inclined to un-cover an opponentrsquos positive qualities (overestimation) then it would beexpected that in a related cognitive task they would be more likely toelaborate information diagnostic of the opponentrsquos positive than nega-tive qualities This is the issue we examined in the second study

STUDY 2

Trait concepts differ in the degree to which they are negativendashdiagnosticor positivendashdiagnostic according to the schematic model of dispositionalattribution (Reeder 1985 Reeder amp Brewer 1979 ) and thecuendashdiagnosticity model of social perception (Skowronski amp Carlston1987 1989) The lay causal theories that people use to understand behav-iors related to negativendashdiagnostic traits (eg moralityndashrelated traits)indicate that negative behaviors are caused by dispositional factors butthat positive behaviors are caused by situational factors or a combina-tion of both (Ybarra amp Stephan 1999 Ybarra 2002) In contrast the laycausal theories people use to understand behaviors related to posi-tivendashdiagnostic traits (eg competencendashrelated traits but also othertraits as we will describe presently) indicate that positive behaviors arecaused by dispositional factors but that negative behaviors are causedby situational factors (Reeder amp Fulks 1980 but see Ybarra 2001 2002for a different perspective)

If a cooperation interaction goal leads people to emphasize a futurepartnerrsquos negative qualities then perceivers should focus on informa-tion related to negativendashdiagnostic traits As a result their informationprocessing (eg attention to and elaboration of information) should beguided by the lay causal theories underlying such traits (negative behav-iors are caused by dispositional causes and positive behaviors are

INTERACTION GOALS AND MEMORY 417

caused by situational causes) (Ybarra amp Stephan 1996 Ybarra ampStephan 1999) Thus people with a cooperation interaction goal shouldbe likely to remember information consistent rather than inconsistentwith the casual theory for the negativendashdiagnostic trait domain (iegood recall for positive behaviors caused by situational factors and neg-ative behaviors caused by dispositional factors) Behaviors related topositivendashdiagnostic traits should be processed less well because this in-formation is of little relevance to the goal of uncovering the partnerrsquosweaknesses Therefore it was expected that perceivers under a coopera-tion goal would not differentiate in memory between theory consistentand inconsistent behavioral information related to positivendashdiagnostictraits1

By comparison if competition leads people to emphasize a future op-ponentrsquos positive qualities then perceivers should focus on informationrelated to positivendashdiagnostic traits As a result their information pro-cessing (eg elaboration of information) should be guided by the laycausal theories underlying such traits (positive behaviors caused bydispositional causes and negative behaviors caused by situationalcauses) Thus people with a competition interaction goal should betterremember information consistent rather than inconsistent with the ca-sual theory for the positivendashdiagnostic trait domain (ie good recall forpositive behaviors caused by dispositional factors and negative valencebehaviors caused by situational factors) Behaviors from the nega-tivendashdiagnostic trait domain should not be well processed because thisinformationis of little relevance to the goal of uncovering the opponentrsquosstrengths Therefore it was expected that perceivers under a competi-tion goal would not show discrimination in their memory between the-ory consistent and inconsistent behavioral information related tonegativendashdiagnostic traits

A relevant question for sake of comparison is what kind of informa-tion would a simple impression formation goal lead perceivers to em-phasize and remember Some research on impression formation

418 CHAN AND YBARRA

1 The strategies of underestimating onersquos partner and overestimating onersquos opponentdo not have to be deliberate and intentional As Study 1b demonstrated even people whowere not consciously thinking about avoiding costs emerged with the same preferences forunderestimating partners and overestimating opponents Once the interaction goals trig-ger a preference for underndash or overestimating the other person the differences in informa-tion processing and memory may occur as a result of peoplersquos lay causal theories guidingthe processing of available information

suggests that people who are forming impressions focus more on the af-fective and general evaluative qualities of others (Levy amp Dugan 1960Ybarra 2001 Zajonc 1980) and that they tend to make such evaluationsbased mainly on negativendashdiagnostic (moralityndashrelated) information(Wojciszke Bazinska amp Jaworski 1998) Thus according to this per-spective people under a general impression formation goal should onlyattend to information related to the negativendashdiagnostic traits similar topeople with a cooperation goal (ie good recall for positive behaviorscaused by situational factors and negative behaviors caused bydispositional factors) Behaviors from the positivendashdiagnostic trait do-main should not be well processed because this information is of littlerelevance to the goal of forming general evaluative impressions There-fore it was expected that perceivers under an impression goal would notdiscriminate in their memory between theory consistent and inconsis-tent behavioral information related to the positivendashdiagnostic traitsThetheoretical model is summarized in Figure 1

Depending on the experimental group to which they were assignedparticipants were told that they would either cooperate or compete withanother participant Then they learned about the other participantrsquos be-haviors which were related to traits in the negativendashdiagnostic trait do-

INTERACTION GOALS AND MEMORY 419

FIGURE 1 Summary of theoretical model

main (hon esty h elpfu lness and fr iendliness ) and thepositivendashdiagnostic trait domain (hardworking image scoring) Imagescoring refers to people engaging in costly positive behaviors to signal toothers that they are valuable community members (Nowak amp Sigmund1998Wedekind amp Milinski 2000) Controlparticipants learned the samematerial but they were asked to form an impression of the person andwere told that they would not interact with the other participant

METHOD

Design and ParticipantsNinetyndashone students participated in the study for course credit Theywere randomly assigned to the cooperation competition or control (im-pression formation) condition All of the participants were presentedwith behaviors with a positive or negative valence from the negativendashdi-agnostic and positivendashdiagnostic trait domains and these behaviorswere linked to either explicit dispositional or situational attributionsThus the overall design of the study was a 3 (interaction goal coopera-tion competition control) acute 2 (trait domain positivendashdiagnostic nega-tivendashdiagnostic) acute 2 (behavior valence positive negative) acute 2(attribution dispositional situational) mixed design with the first fac-tor varying between participants and the latter three factors varyingwithin participants The participants were randomly assigned to condi-tions and were run in nonndashinteracting groups of two to six

Stimulus MaterialsThe behavioral information that the participants processed was pre-sented on a cassette tape allegedly recorded by another participant Thegender of the speaker on the tape and that of the participant were alwaysthe same Each tape contained 20 behavioral statements half of whichwere negative and half positive in valence Also half of the behaviorswere linked to dispositional attributions and half to situational attribu-tions Twelve of the statements pertained to negativendashdiagnostic traits(eg ldquoI entered a concert through the fire exitrdquo [negative valence] ldquoI sentmy sister a present and a birthday bouquetrdquo [positive valence]) Eight ofthe statements pertained to the positivendashdiagnostic traits (eg ldquoI missedmorning lecturesrdquo [negative valence] ldquoI reported a crimerdquo [positive va-lence])

420 CHAN AND YBARRA

Our choice of the traits hardworking and image scoring was deliberatebecause it allows us to show that perceiverrsquos attention to a targetrsquosstrengths (overestimation) is not confounded with emphasizing infor-mation about the targetrsquos competence For example it is usually as-sumed that there is an inverse relationship between the amount of worka person engages in and the presence of the related ability (Darley ampGoethals 1980 Jones 1989) but being hardworking should still signal apositive quality a strength In addition image scoring refers to peopleengaging in costly positive behaviors such as extreme levels of philan-thropy to enhance their reputation and status (Wedekind amp Milinski2000 McAndrew 2002) It is completely unrelated to competence but itsdiagnosticity is in line with that of a positivendashdiagnostic trait

The behavior stems (no attributions) were pretested to ensure thatthey were indeed regarded as positive or negativendashdiagnostic Thirtyparticipants rated each behavior on how common they are For exampleone item queried them as to ldquoWhat percentage of college students doyou think enter a concert through the fire exitrdquo (6ndashpoint scale with 0ndash1010ndash30 30ndash50 50ndash70 70ndash90 and 90ndash100) For positivendashdiagnos-tic traits behaviors that are positive in valence should be less common(and hence more diagnostic of the actorrsquos dispositions) than behaviorsthat are negative in valence whereas for negativendashdiagnostic traits be-haviors negative in valence should be less common (and hence more di-agnostic of the actorrsquos dispositions) than behaviors positive in valenceResults showed an interaction effect that confirmed our classification ofthe behaviors into the respective positive and negativendashdiagnostic cate-gories F(1 29) = 4913 p lt 001 For the positivendashdiagnostic traits posi-tive valence behaviors (M = 281) were judged to be enacted less oftenthan negative valence behaviors (M = 376) F(1 29) = 1961 p lt 001 Forthe negativendashdiagnostic traits negative valence behaviors (M = 294)were judged to be enacted less often than positive valence behaviors (M= 331) F(1 29) = 7103 p = 012

Two sets of materials both containing the same 20 behavior stemswere created Each behavior stem was matched with a dispositional at-tribution in one set and a situational attribution in the other set For ex-ample in set A the behaviors ldquoI didnrsquot help my dad with yard workbecause I wanted to stay inside to watch TVrdquo [negative behavior in nega-tivendashdiagnostic domain with dispositional attribution] ldquoI sent my sistera present and a birthday bouquet because my grandmother asked me todo itrdquo [positive behavior in negativendashdiagnostic domain with situational

INTERACTION GOALS AND MEMORY 421

attribution] ldquoI missed morning lectures because my roommate playedpranks with my alarm clockrdquo [negative behavior in positivendashdiagnosticdomain with situational attribution] and ldquoI worked as a research assis-tant because I am very interested in the subject matterrdquo [positive behav-ior in positive diagnostic domain with dispositional attribution] wouldbe counterbalanced in set B with different causal attributions These in-cluded ldquoI didnrsquot help my dad with yard work because he couldnrsquot findthe keys to the tool shedrdquo [situational attribution] ldquoI sent my sister apresent and a birthday bouquet because I wanted to give her a pleasantsurpriserdquo [dispositional attribution] ldquoI missed morning lecture becauseI donrsquot feel like waking up for themrdquo [dispositional attribution] and ldquoIworked as a research assistant because my professor said I had to do itrdquo[situational attribution] As will be described below the valence of thebehaviors and attributions (dispositional vs external) were manipu-lated independently to ensure that subsequent differences in recall oftheoryndashconsistent or inconsistent information could not simply be dueto an artifact based on the memorability of the particular behaviors

The two sets of stimuli were presented to 17 participants for pretest-ing One participant misunderstood how to use the response scale andwas dropped from the analysis Participants were first presented withthe behavior stems and were asked to rate how favorable each behaviorwas on a 7ndashpoint scale with higher values indicating greaterfavorability Then they were presented with the same behavior stemslinked to either a dispositional or situational attribution and were askedto judge the locus of causation for the behaviors on a 7ndashpoint scale withhigher values indicating more dispositional causation The responseswere submitted to a 2 (valence)acute 2 (attribution)withinndashsubject ANOVAFor the valence judgments the positive behaviors in both stimulus sets(Mset A = 544 Mset B = 559) were judged as more favorable than the nega-tive behaviors (Mset A = 284Mset B = 215) Fset A (1 8)= 3416 p lt 0004Fset B

(1 8) = 9782 p lt 0001 For the second set of judgments a main effect forattribution was obtained for both sets of stimuli Behaviors linked todispositional attributions (Mset A = 575 Mset B = 646) were judged to bemore dispositionally caused than behaviors linked to situational attribu-tions (M set A = 235 Mset B = 298) F set A (1 8) = 4574 p lt 0001 Fset B (1 8) =15826 p lt 0001 No other main effects or interaction effects were ob-tained The two sets of stimuli were presented under three different ran-domization schemes

422 CHAN AND YBARRA

PROCEDURE

Upon arrival at the lab the participants were provided the cover storyfor the study in which they were told that the study would examine howpersonal information would affect subsequent interactions betweenstrangers They were told that another set of participants was recordingsome personal information on a tape for them to listen to To increase thecredibility of the tape contents participants were told that the partici-pants recording the tape had been asked to provide distinct personal in-formation about ldquothings that they have done in the past few weeksrdquoAdditional information about the experiment including whether theywere in the cooperation competition or control (impression formation)condition was given to each participant in a sealed envelope so that theexperimenter could remain blind to the conditions Experimental groupparticipants were told that after they listened to a tape with personal in-formation recorded by another participant they would get a chance tomeet them in either a cooperative or competitive setting Control partici-pants were informed that they had been randomly assigned to the con-trol group and would listen to some information about a previousparticipant instead of having an interaction with them

Participants in the cooperation condition read that they would engagein a ldquoCommunity Simulation Gamerdquo in which they and their assignedpartners ldquowill work together to try to maximize the number of points foryour team The team with the highest points will receive a prizerdquo Thecooperative nature of the interaction was emphasized with repeated useof the word ldquopartnerrdquo In contrastparticipants in the competition condi-tion read that they would engage in a ldquoWall Street Simulation Gamerdquo inwhich they would ldquocompete against [their] competitor to maximize[their] own profit (points) and minimize their opponentrsquos profit Theperson with the highest points will win a prizerdquo The competitive aspectof the upcoming interaction was emphasized further with repeated useof the words ldquocompetitorrdquo and ldquoopponentrdquo This type of reward struc-ture manipulation has been used to study cooperation and competitionby other researchers (Deutsch 1973)

Participants in the control condition read that they would listen to atape that had been recorded in a previous experimental session Theywere told to form an impression of the person using the informationfrom the tape as part of a ldquostandardizationrdquo procedure for the experi-ment To bolster the believability of the cover story participants then

INTERACTION GOALS AND MEMORY 423

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

race they would in all likelihood remember different information thanif they had interviewed these people for postndashdoctoral research posi-tions Information pertaining to intellectual creativity and academic po-tential should be more relevant when interacting with postndashdoctoralcandidates than when talking to race car fans However for the carndashrac-ing fans other information such as the types of cars they drive their fa-vorite race teams and their knowledge about cars would be relevantinformation

Peoplersquos interaction goals can be quite specific as the above examplesillustrate However some goals are broader in nature and occur across awide range of contexts Examples of such goals include cooperating andcompeting with other people Cooperation and competition are an inte-gral part of everyday interaction ranging from childrenrsquos play sportsand games work and business interactions to intergroup relations andworld politics Thus it is of interest and importance to examine howthese broader goals affect social information processing Before thinkingabout such outcomes though it is important to briefly outline how otherresearchers have approached the study of cooperationand competition

Some research has shown for example that competition enhances thesalience of intergroup distinctions leading to greater bias and an in-crease in the perception of outgroup homogeneity (cf Brewer 1979) Co-operation and competition have also been investigated in the context ofdecision making and negotiation in mixed motive games (eg Bornsteinamp Rapoport 1988Hertel amp Fiedler 1994 Insko Schopler Hoyle Dardisamp Graetz 1990 Parks Henager amp Scamahorn 1996) Findings haveshown for example that public appeal (Rosen amp Haagen 1998) discus-sion within groups (Bornstein amp Rapoport 1988) and increasing the sa-lience of positive connotations of cooperation (Hertel amp Fiedler 1994) allincrease cooperation On the other hand anonymity and low risk of so-cial exclusion decrease cooperation (Kerr 1999) These studies focus ondelineating the conditions that foster cooperation and competition butdo not address the social information processing strategies associatedwith these interaction goals

ANTICIPATED INTERACTION COOPERATION COMPETITIONAND SOCIAL INFORMATION PROCESSING

The mere anticipation of having to interact with others has powerful ef-fects on social cognition For example research has shown that expect-

410 CHAN AND YBARRA

ing to meet and interact with another person leads to more extensiveinformation processing compared to control conditions (eg memoriza-tion condition) (Devine Sedikides amp Fuhrman 1989 Srull amp Brand1983)

Another small body of research that has explored the interaction goalsof cooperation and competition has shown that these interaction goalsaffect a targetrsquos attractiveness leading to increased attention under an-ticipated cooperation but a decrease under anticipated competition(Tesser amp Danheiser 1978) The interaction goals of cooperation andcompetition have also been shown to lead to more thorough processingof target information (Neuberg amp Fiske 1987 Ruscher amp Fiske 1990)Neuberg and Fiske (1987) showed that participants who expected to co-operate with a previously hospitalized schizophrenic person did not en-gage in categoryndashbased processing but individuated the partnerRuscher and Fiske (1990) presented participants with a competitive in-teraction and led them to form either a positive or negative compe-tencendashbased expectancy Then they allowed the participants to readinformation about their interaction partner that was either consistentinconsistent or irrelevant to the expectancy They found that partici-pants engaged in individuating processes such as showing increased at-tention to inconsistencies and forming more varied impressions of thepartner

THE PRESENT CONCEPTUALIZATION

The current research approaches the study of cooperation and compe-tition in a slightly different manner and construes them as guidingprinciples in peoplersquos selection of information processing strategywhich leads to overestimating or underestimating othersrsquo qualitiesMuch research in person perception and decision making suggests thatpeople are risk aversemdashpotential losses loom larger than potentialgains (eg Kahneman amp Tversky 1979) Thus when learning about afuture partner under a cooperation goal or an opponent under a com-petition goal a perceiver might selectively process information that ismost indicative of potential losses This would be adaptive because be-ing aware of potential losses and dangers would allow the perceiver toprepare in advance to minimize the chance of loss and if the riskseemed too large to avoid the cooperation or competition to preventloss

INTERACTION GOALS AND MEMORY 411

Specifically a perceiver who is sizing up a future partner can minimizepotential losses by avoiding the overestimationof the partnerrsquos strengthsThis can be done by being sensitive to information diagnostic of the part-nerrsquos negative qualities For example if a person finds out that it is easy fora partner to lose enthusiasm after initial involvement with a project he orshe could anticipate the problem and attempt to reduce its negative im-pact on outcomes In addition the perceiver would also be motivated toattend to such weaknesses to decide how much to trust and rely on thepartner If the perceiver learned that the partner was weak in some skillhe or she could try to strengthen that skill in the self or at least be vigilantto any task that involved that skill (Williams amp Karau 1991 WittenbaumVaughan amp Stasser 1998) The cost of overestimating how good the part-ner is and failing to notice the partnerrsquos weaknesses is immense if the per-son mistakenly thinks that the partner is hardworking and talented andtherefore becomes complacent the likelihood of failure is increased

The interaction goal of competition puts the perceiver in a differentposition A perceiver who is sizing up a future opponent can minimizepotential losses by avoiding the underestimation of the opponentrsquosstrengths This can be done by being sensitive to information diagnosticof the opponentrsquos positive qualities For example if a person finds outthat his opponent is highly skilled and hardworking he could anticipatethe challenge and attempt to reduce his chance of losing by workingeven harder and improving his own skills Underestimating the oppo-nent would undoubtedly increase the risk of defeat because theperceiver would be underndashprepared for the competition

Study 1a and 1b first tested the critical assumptions regarding how theinteraction goal of cooperation would lead people to avoid overestimat-ing how good their partners are whereas the interaction goal of compe-tition would lead people to avoid underestimating how good theiropponents are In a subsequent validation of our analysis Study 2 exam-ined how these interaction goals affected memory for behavioral infor-mation from different trait domains Finally Study 3 examined how theinteraction goals affected the nature and amount of information peoplesought about their future partnerscompetitors

STUDY 1A

We reasoned that people who are evaluating a potential partner shouldtry to avoid the mistake of overestimating the partnerrsquos positive qualities

412 CHAN AND YBARRA

because being generally risk averse they should focus on potentiallosses that could arise from cooperating with an inferior partner Thusthey should attempt to gather negative dispositional information aboutthe partner People who are evaluating a potential opponent should tryto avoid the mistake of underestimating the opponentrsquos positive quali-ties because being risk averse they should focus on potential losses thatcould arise from competing with a superior opponent Thus theyshould attempt to gather positive dispositional information about theopponent To examine these predictions participants in the first studywere presented a fictitious scenario in which a person in the scenario hadto learn about someone else with whom he was to cooperate or competeParticipants were then asked to evaluate the targetrsquos strategy indicatinghow costly it would be for the target to overestimate or underestimatehow good his partner or opponent was

METHOD

Design and ParticipantsEighteen students between 19ndash25 years old volunteered for the studywithout payment They were approached on a university campus andasked if they would fill out a short questionnaire

Materials and ProcedureParticipants were informed that the study was a short psychology ques-tionnaire dealing with interpersonal perception They were given aquestionnaire containing either the cooperation or the competition sce-nario In the cooperation condition the scenario read

Person X is about to meet someone with whom he will have to work to-gether on a task Person X has a few moments to find out what the other per-son is like After the meeting Person X will have to decide what strategy totake during the cooperation in order to achieve the most successful out-come

In the competition condition the scenario read

Person X is about to meet someone with whom he will have to compete on atask Person X has a few moments to find out what the other person is likeAfter the meeting Person X will have to decide what strategy to take duringthe competition in order to achieve the most successful outcome

INTERACTION GOALS AND MEMORY 413

Participants then answered 2 questions Those in the cooperation condi-tion answered ldquoHow costly is it for Person X to mistakenly overestimatehow good the partner isrdquo and ldquoHow costly is it for Person X to mistak-enly underestimate how good the partner isrdquo on a 1 (not at all costly) to7 (extremely costly) scale Those in the competition condition answeredthe same items with the word ldquocompetitorrdquo replacing the word ldquopart-nerrdquo The presentation order of the scenarios was counterbalanced Af-ter they completed the questionnaires the participants were debriefedand thanked for their participation

RESULTS AND DISCUSSION

Participantsrsquo responses were submitted to a 2 (goal competition coop-eration) acute 2 (estimation overestimation underestimation) acute 2 (questionorder) mixed ANOVA with the second factor varying within subjectThere was no main effect or interaction associated with question orderand thus the order factorwas dropped from the analysis It was expectedthat overestimating how good onersquos partner is and underestimatinghow good onersquos opponent is would be deemed undesirable and this wasindeed the case The only significant effect was the goal acute estimation in-teraction F(1 16) = 2918 p lt 001 When expecting to cooperate peoplejudged that it was more costly to overestimate (M = 556) than underesti-mate (M = 289)onersquos partner F(1 16) = 3011p lt 001 In contrastpeoplejudged that it was more costly to underestimate (M = 574) than overesti-mate (M = 294) onersquos opponent F(1 16) = 1580 p = 001

The findings from Study 1a suggest that for cooperation people deemit preferable to avoid overestimating how good the partner is and forcompetition to avoid underestimating how good the opponent is Onecould argue that these results were obtained because participants wereexplicitly instructed to consider how costly it was to overestimate or un-derestimate the other person By focusing peoplersquos attention on the cost-liness of their decisions we might have led the participants to be morelossndashfocused than they would have been otherwise Another feature ofthis study that requires further examination is whether asking partici-pants to imagine the scenario from anotherrsquos perspective instead ofimagining themselves in the scenario might have affected the resultsMuch social psychological research has found differences between peo-plersquos psychological processes when they are thinking about themselvesversus others For example people tend to make situational causal attri-

414 CHAN AND YBARRA

butions for themselves but dispositional causal attributions for others(Jones amp Nisbett 1972) People pay more attention to unobservable andunintentional events when thinking about themselves but observableand intentional events when thinking about others (Malle amp Pearce2001) Peoplersquos representations of themselves in memory are character-ized by their own thoughts and feelings whereas their representationsof others are characterized by actions and appearances (McGuire ampMcGuire 1986 Prentice 1990) Although it is not perfectly clear how itis possible that the findings from Study 1a might not occur if participantswere asked to think of how they themselves would react To rule outthese alternative accounts we conducted Study 1b to replicate the re-sults of Study 1a

STUDY 1B

The current study was identical to Study 1a save for two features Firstwe asked the participants to judge how overestimating or underestimat-ing the other person would affect their chances of success instead of ex-plicitly pointing to the costliness of the decision as in Study 1a Secondwe asked participants to imagine themselves instead of others in scenar-ios where they either had to cooperate or compete with another person

METHOD

Design and ParticipantsTwenty students between 19ndash25 years old participated in this studyThey were approached on a university campus and asked to fill out ashort questionnaire

Materials and ProcedureParticipants were informed that the study was a short psychology ques-tionnaire dealing with interpersonal perception They were asked toimagine themselves in the following scenarios

In the cooperation condition the scenario read

You are about to meet Person X and both of you will have to work togetherYou have a few moments to find out more about Person X After finding outmore about what Person X is like you will have to decide what strategy to

INTERACTION GOALS AND MEMORY 415

take during the cooperation in order to achieve the most successful out-come

In the competition condition the scenario read

You are about to meet Person X and both of you will have to compete againsteach other You will have a few moments to find out more about Person XAfter finding out more about what Person X is like you will have to decidewhat strategy to take during the competition in order to achieve the mostsuccessful outcome

Participants then answered 2 questions Those in the cooperation condi-tion answered ldquoIf you were to overestimate how good your partner ishow would it affect your chances of achieving a successful outcomerdquoand ldquoIf you were to underestimate how good your partner is howwould it affect your chances of achieving a successful outcomerdquo on a 1(reduce chance of success) to 7 (increase chance of success) scale Thequestions in the competition condition were the same with the wordldquoopponentrdquo replacing the word ldquopartnerrdquo The presentation order of thescenarios was counterbalanced After they completed the question-naires the participants were debriefed and thanked for their participa-tion

RESULTS AND DISCUSSION

Participantsrsquo responses were submitted to a 2 (goal competition coop-eration) acute 2 (estimation overestimation underestimation) acute 2 (questionorder) mixed ANOVA with the second factor varying within subjectThere was neither a main effect nor interaction associated with questionorder so it was dropped from the analysis It was expected that peoplewould judge that overestimating their partners and underestimatingtheir opponents would reduce their chances of success The results wereconsistent with the hypothesis The only significant effect was the goal acuteestimation interaction F(1 19) = 4532 p lt 001When expecting to coop-erate people judged that their chances of success would be lower if theyoverestimated their partners (M = 200)than if they underestimated theirpartners (M = 410) F(1 19) = 2000 p lt 001 In contrast people judgedthat their chances of success would be higher if they overestimated theiropponents (M = 430) than if they underestimated their opponents (M =260) F(1 19) = 4185p lt 001These findings thus confirm those of Study

416 CHAN AND YBARRA

1a and indicate that people may be inclined to adopt different informa-tion processing strategies depending on their interaction goals

If it really is the case that people when expecting to cooperate arepoised to process and be vigilant about a partnerrsquos negative qualities(underestimation) it would be expected that in a different cognitive taskthose inclinations would also be expressed For example in learningabout a potential partner who is described with negative and positivequalities we should find that people expecting to cooperate are morelikely to elaborate information diagnostic of the partnerrsquos negative thanpositive qualities Similar outcomes should be expected under antici-pated competition If people who expect to compete are inclined to un-cover an opponentrsquos positive qualities (overestimation) then it would beexpected that in a related cognitive task they would be more likely toelaborate information diagnostic of the opponentrsquos positive than nega-tive qualities This is the issue we examined in the second study

STUDY 2

Trait concepts differ in the degree to which they are negativendashdiagnosticor positivendashdiagnostic according to the schematic model of dispositionalattribution (Reeder 1985 Reeder amp Brewer 1979 ) and thecuendashdiagnosticity model of social perception (Skowronski amp Carlston1987 1989) The lay causal theories that people use to understand behav-iors related to negativendashdiagnostic traits (eg moralityndashrelated traits)indicate that negative behaviors are caused by dispositional factors butthat positive behaviors are caused by situational factors or a combina-tion of both (Ybarra amp Stephan 1999 Ybarra 2002) In contrast the laycausal theories people use to understand behaviors related to posi-tivendashdiagnostic traits (eg competencendashrelated traits but also othertraits as we will describe presently) indicate that positive behaviors arecaused by dispositional factors but that negative behaviors are causedby situational factors (Reeder amp Fulks 1980 but see Ybarra 2001 2002for a different perspective)

If a cooperation interaction goal leads people to emphasize a futurepartnerrsquos negative qualities then perceivers should focus on informa-tion related to negativendashdiagnostic traits As a result their informationprocessing (eg attention to and elaboration of information) should beguided by the lay causal theories underlying such traits (negative behav-iors are caused by dispositional causes and positive behaviors are

INTERACTION GOALS AND MEMORY 417

caused by situational causes) (Ybarra amp Stephan 1996 Ybarra ampStephan 1999) Thus people with a cooperation interaction goal shouldbe likely to remember information consistent rather than inconsistentwith the casual theory for the negativendashdiagnostic trait domain (iegood recall for positive behaviors caused by situational factors and neg-ative behaviors caused by dispositional factors) Behaviors related topositivendashdiagnostic traits should be processed less well because this in-formation is of little relevance to the goal of uncovering the partnerrsquosweaknesses Therefore it was expected that perceivers under a coopera-tion goal would not differentiate in memory between theory consistentand inconsistent behavioral information related to positivendashdiagnostictraits1

By comparison if competition leads people to emphasize a future op-ponentrsquos positive qualities then perceivers should focus on informationrelated to positivendashdiagnostic traits As a result their information pro-cessing (eg elaboration of information) should be guided by the laycausal theories underlying such traits (positive behaviors caused bydispositional causes and negative behaviors caused by situationalcauses) Thus people with a competition interaction goal should betterremember information consistent rather than inconsistent with the ca-sual theory for the positivendashdiagnostic trait domain (ie good recall forpositive behaviors caused by dispositional factors and negative valencebehaviors caused by situational factors) Behaviors from the nega-tivendashdiagnostic trait domain should not be well processed because thisinformationis of little relevance to the goal of uncovering the opponentrsquosstrengths Therefore it was expected that perceivers under a competi-tion goal would not show discrimination in their memory between the-ory consistent and inconsistent behavioral information related tonegativendashdiagnostic traits

A relevant question for sake of comparison is what kind of informa-tion would a simple impression formation goal lead perceivers to em-phasize and remember Some research on impression formation

418 CHAN AND YBARRA

1 The strategies of underestimating onersquos partner and overestimating onersquos opponentdo not have to be deliberate and intentional As Study 1b demonstrated even people whowere not consciously thinking about avoiding costs emerged with the same preferences forunderestimating partners and overestimating opponents Once the interaction goals trig-ger a preference for underndash or overestimating the other person the differences in informa-tion processing and memory may occur as a result of peoplersquos lay causal theories guidingthe processing of available information

suggests that people who are forming impressions focus more on the af-fective and general evaluative qualities of others (Levy amp Dugan 1960Ybarra 2001 Zajonc 1980) and that they tend to make such evaluationsbased mainly on negativendashdiagnostic (moralityndashrelated) information(Wojciszke Bazinska amp Jaworski 1998) Thus according to this per-spective people under a general impression formation goal should onlyattend to information related to the negativendashdiagnostic traits similar topeople with a cooperation goal (ie good recall for positive behaviorscaused by situational factors and negative behaviors caused bydispositional factors) Behaviors from the positivendashdiagnostic trait do-main should not be well processed because this information is of littlerelevance to the goal of forming general evaluative impressions There-fore it was expected that perceivers under an impression goal would notdiscriminate in their memory between theory consistent and inconsis-tent behavioral information related to the positivendashdiagnostic traitsThetheoretical model is summarized in Figure 1

Depending on the experimental group to which they were assignedparticipants were told that they would either cooperate or compete withanother participant Then they learned about the other participantrsquos be-haviors which were related to traits in the negativendashdiagnostic trait do-

INTERACTION GOALS AND MEMORY 419

FIGURE 1 Summary of theoretical model

main (hon esty h elpfu lness and fr iendliness ) and thepositivendashdiagnostic trait domain (hardworking image scoring) Imagescoring refers to people engaging in costly positive behaviors to signal toothers that they are valuable community members (Nowak amp Sigmund1998Wedekind amp Milinski 2000) Controlparticipants learned the samematerial but they were asked to form an impression of the person andwere told that they would not interact with the other participant

METHOD

Design and ParticipantsNinetyndashone students participated in the study for course credit Theywere randomly assigned to the cooperation competition or control (im-pression formation) condition All of the participants were presentedwith behaviors with a positive or negative valence from the negativendashdi-agnostic and positivendashdiagnostic trait domains and these behaviorswere linked to either explicit dispositional or situational attributionsThus the overall design of the study was a 3 (interaction goal coopera-tion competition control) acute 2 (trait domain positivendashdiagnostic nega-tivendashdiagnostic) acute 2 (behavior valence positive negative) acute 2(attribution dispositional situational) mixed design with the first fac-tor varying between participants and the latter three factors varyingwithin participants The participants were randomly assigned to condi-tions and were run in nonndashinteracting groups of two to six

Stimulus MaterialsThe behavioral information that the participants processed was pre-sented on a cassette tape allegedly recorded by another participant Thegender of the speaker on the tape and that of the participant were alwaysthe same Each tape contained 20 behavioral statements half of whichwere negative and half positive in valence Also half of the behaviorswere linked to dispositional attributions and half to situational attribu-tions Twelve of the statements pertained to negativendashdiagnostic traits(eg ldquoI entered a concert through the fire exitrdquo [negative valence] ldquoI sentmy sister a present and a birthday bouquetrdquo [positive valence]) Eight ofthe statements pertained to the positivendashdiagnostic traits (eg ldquoI missedmorning lecturesrdquo [negative valence] ldquoI reported a crimerdquo [positive va-lence])

420 CHAN AND YBARRA

Our choice of the traits hardworking and image scoring was deliberatebecause it allows us to show that perceiverrsquos attention to a targetrsquosstrengths (overestimation) is not confounded with emphasizing infor-mation about the targetrsquos competence For example it is usually as-sumed that there is an inverse relationship between the amount of worka person engages in and the presence of the related ability (Darley ampGoethals 1980 Jones 1989) but being hardworking should still signal apositive quality a strength In addition image scoring refers to peopleengaging in costly positive behaviors such as extreme levels of philan-thropy to enhance their reputation and status (Wedekind amp Milinski2000 McAndrew 2002) It is completely unrelated to competence but itsdiagnosticity is in line with that of a positivendashdiagnostic trait

The behavior stems (no attributions) were pretested to ensure thatthey were indeed regarded as positive or negativendashdiagnostic Thirtyparticipants rated each behavior on how common they are For exampleone item queried them as to ldquoWhat percentage of college students doyou think enter a concert through the fire exitrdquo (6ndashpoint scale with 0ndash1010ndash30 30ndash50 50ndash70 70ndash90 and 90ndash100) For positivendashdiagnos-tic traits behaviors that are positive in valence should be less common(and hence more diagnostic of the actorrsquos dispositions) than behaviorsthat are negative in valence whereas for negativendashdiagnostic traits be-haviors negative in valence should be less common (and hence more di-agnostic of the actorrsquos dispositions) than behaviors positive in valenceResults showed an interaction effect that confirmed our classification ofthe behaviors into the respective positive and negativendashdiagnostic cate-gories F(1 29) = 4913 p lt 001 For the positivendashdiagnostic traits posi-tive valence behaviors (M = 281) were judged to be enacted less oftenthan negative valence behaviors (M = 376) F(1 29) = 1961 p lt 001 Forthe negativendashdiagnostic traits negative valence behaviors (M = 294)were judged to be enacted less often than positive valence behaviors (M= 331) F(1 29) = 7103 p = 012

Two sets of materials both containing the same 20 behavior stemswere created Each behavior stem was matched with a dispositional at-tribution in one set and a situational attribution in the other set For ex-ample in set A the behaviors ldquoI didnrsquot help my dad with yard workbecause I wanted to stay inside to watch TVrdquo [negative behavior in nega-tivendashdiagnostic domain with dispositional attribution] ldquoI sent my sistera present and a birthday bouquet because my grandmother asked me todo itrdquo [positive behavior in negativendashdiagnostic domain with situational

INTERACTION GOALS AND MEMORY 421

attribution] ldquoI missed morning lectures because my roommate playedpranks with my alarm clockrdquo [negative behavior in positivendashdiagnosticdomain with situational attribution] and ldquoI worked as a research assis-tant because I am very interested in the subject matterrdquo [positive behav-ior in positive diagnostic domain with dispositional attribution] wouldbe counterbalanced in set B with different causal attributions These in-cluded ldquoI didnrsquot help my dad with yard work because he couldnrsquot findthe keys to the tool shedrdquo [situational attribution] ldquoI sent my sister apresent and a birthday bouquet because I wanted to give her a pleasantsurpriserdquo [dispositional attribution] ldquoI missed morning lecture becauseI donrsquot feel like waking up for themrdquo [dispositional attribution] and ldquoIworked as a research assistant because my professor said I had to do itrdquo[situational attribution] As will be described below the valence of thebehaviors and attributions (dispositional vs external) were manipu-lated independently to ensure that subsequent differences in recall oftheoryndashconsistent or inconsistent information could not simply be dueto an artifact based on the memorability of the particular behaviors

The two sets of stimuli were presented to 17 participants for pretest-ing One participant misunderstood how to use the response scale andwas dropped from the analysis Participants were first presented withthe behavior stems and were asked to rate how favorable each behaviorwas on a 7ndashpoint scale with higher values indicating greaterfavorability Then they were presented with the same behavior stemslinked to either a dispositional or situational attribution and were askedto judge the locus of causation for the behaviors on a 7ndashpoint scale withhigher values indicating more dispositional causation The responseswere submitted to a 2 (valence)acute 2 (attribution)withinndashsubject ANOVAFor the valence judgments the positive behaviors in both stimulus sets(Mset A = 544 Mset B = 559) were judged as more favorable than the nega-tive behaviors (Mset A = 284Mset B = 215) Fset A (1 8)= 3416 p lt 0004Fset B

(1 8) = 9782 p lt 0001 For the second set of judgments a main effect forattribution was obtained for both sets of stimuli Behaviors linked todispositional attributions (Mset A = 575 Mset B = 646) were judged to bemore dispositionally caused than behaviors linked to situational attribu-tions (M set A = 235 Mset B = 298) F set A (1 8) = 4574 p lt 0001 Fset B (1 8) =15826 p lt 0001 No other main effects or interaction effects were ob-tained The two sets of stimuli were presented under three different ran-domization schemes

422 CHAN AND YBARRA

PROCEDURE

Upon arrival at the lab the participants were provided the cover storyfor the study in which they were told that the study would examine howpersonal information would affect subsequent interactions betweenstrangers They were told that another set of participants was recordingsome personal information on a tape for them to listen to To increase thecredibility of the tape contents participants were told that the partici-pants recording the tape had been asked to provide distinct personal in-formation about ldquothings that they have done in the past few weeksrdquoAdditional information about the experiment including whether theywere in the cooperation competition or control (impression formation)condition was given to each participant in a sealed envelope so that theexperimenter could remain blind to the conditions Experimental groupparticipants were told that after they listened to a tape with personal in-formation recorded by another participant they would get a chance tomeet them in either a cooperative or competitive setting Control partici-pants were informed that they had been randomly assigned to the con-trol group and would listen to some information about a previousparticipant instead of having an interaction with them

Participants in the cooperation condition read that they would engagein a ldquoCommunity Simulation Gamerdquo in which they and their assignedpartners ldquowill work together to try to maximize the number of points foryour team The team with the highest points will receive a prizerdquo Thecooperative nature of the interaction was emphasized with repeated useof the word ldquopartnerrdquo In contrastparticipants in the competition condi-tion read that they would engage in a ldquoWall Street Simulation Gamerdquo inwhich they would ldquocompete against [their] competitor to maximize[their] own profit (points) and minimize their opponentrsquos profit Theperson with the highest points will win a prizerdquo The competitive aspectof the upcoming interaction was emphasized further with repeated useof the words ldquocompetitorrdquo and ldquoopponentrdquo This type of reward struc-ture manipulation has been used to study cooperation and competitionby other researchers (Deutsch 1973)

Participants in the control condition read that they would listen to atape that had been recorded in a previous experimental session Theywere told to form an impression of the person using the informationfrom the tape as part of a ldquostandardizationrdquo procedure for the experi-ment To bolster the believability of the cover story participants then

INTERACTION GOALS AND MEMORY 423

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

ing to meet and interact with another person leads to more extensiveinformation processing compared to control conditions (eg memoriza-tion condition) (Devine Sedikides amp Fuhrman 1989 Srull amp Brand1983)

Another small body of research that has explored the interaction goalsof cooperation and competition has shown that these interaction goalsaffect a targetrsquos attractiveness leading to increased attention under an-ticipated cooperation but a decrease under anticipated competition(Tesser amp Danheiser 1978) The interaction goals of cooperation andcompetition have also been shown to lead to more thorough processingof target information (Neuberg amp Fiske 1987 Ruscher amp Fiske 1990)Neuberg and Fiske (1987) showed that participants who expected to co-operate with a previously hospitalized schizophrenic person did not en-gage in categoryndashbased processing but individuated the partnerRuscher and Fiske (1990) presented participants with a competitive in-teraction and led them to form either a positive or negative compe-tencendashbased expectancy Then they allowed the participants to readinformation about their interaction partner that was either consistentinconsistent or irrelevant to the expectancy They found that partici-pants engaged in individuating processes such as showing increased at-tention to inconsistencies and forming more varied impressions of thepartner

THE PRESENT CONCEPTUALIZATION

The current research approaches the study of cooperation and compe-tition in a slightly different manner and construes them as guidingprinciples in peoplersquos selection of information processing strategywhich leads to overestimating or underestimating othersrsquo qualitiesMuch research in person perception and decision making suggests thatpeople are risk aversemdashpotential losses loom larger than potentialgains (eg Kahneman amp Tversky 1979) Thus when learning about afuture partner under a cooperation goal or an opponent under a com-petition goal a perceiver might selectively process information that ismost indicative of potential losses This would be adaptive because be-ing aware of potential losses and dangers would allow the perceiver toprepare in advance to minimize the chance of loss and if the riskseemed too large to avoid the cooperation or competition to preventloss

INTERACTION GOALS AND MEMORY 411

Specifically a perceiver who is sizing up a future partner can minimizepotential losses by avoiding the overestimationof the partnerrsquos strengthsThis can be done by being sensitive to information diagnostic of the part-nerrsquos negative qualities For example if a person finds out that it is easy fora partner to lose enthusiasm after initial involvement with a project he orshe could anticipate the problem and attempt to reduce its negative im-pact on outcomes In addition the perceiver would also be motivated toattend to such weaknesses to decide how much to trust and rely on thepartner If the perceiver learned that the partner was weak in some skillhe or she could try to strengthen that skill in the self or at least be vigilantto any task that involved that skill (Williams amp Karau 1991 WittenbaumVaughan amp Stasser 1998) The cost of overestimating how good the part-ner is and failing to notice the partnerrsquos weaknesses is immense if the per-son mistakenly thinks that the partner is hardworking and talented andtherefore becomes complacent the likelihood of failure is increased

The interaction goal of competition puts the perceiver in a differentposition A perceiver who is sizing up a future opponent can minimizepotential losses by avoiding the underestimation of the opponentrsquosstrengths This can be done by being sensitive to information diagnosticof the opponentrsquos positive qualities For example if a person finds outthat his opponent is highly skilled and hardworking he could anticipatethe challenge and attempt to reduce his chance of losing by workingeven harder and improving his own skills Underestimating the oppo-nent would undoubtedly increase the risk of defeat because theperceiver would be underndashprepared for the competition

Study 1a and 1b first tested the critical assumptions regarding how theinteraction goal of cooperation would lead people to avoid overestimat-ing how good their partners are whereas the interaction goal of compe-tition would lead people to avoid underestimating how good theiropponents are In a subsequent validation of our analysis Study 2 exam-ined how these interaction goals affected memory for behavioral infor-mation from different trait domains Finally Study 3 examined how theinteraction goals affected the nature and amount of information peoplesought about their future partnerscompetitors

STUDY 1A

We reasoned that people who are evaluating a potential partner shouldtry to avoid the mistake of overestimating the partnerrsquos positive qualities

412 CHAN AND YBARRA

because being generally risk averse they should focus on potentiallosses that could arise from cooperating with an inferior partner Thusthey should attempt to gather negative dispositional information aboutthe partner People who are evaluating a potential opponent should tryto avoid the mistake of underestimating the opponentrsquos positive quali-ties because being risk averse they should focus on potential losses thatcould arise from competing with a superior opponent Thus theyshould attempt to gather positive dispositional information about theopponent To examine these predictions participants in the first studywere presented a fictitious scenario in which a person in the scenario hadto learn about someone else with whom he was to cooperate or competeParticipants were then asked to evaluate the targetrsquos strategy indicatinghow costly it would be for the target to overestimate or underestimatehow good his partner or opponent was

METHOD

Design and ParticipantsEighteen students between 19ndash25 years old volunteered for the studywithout payment They were approached on a university campus andasked if they would fill out a short questionnaire

Materials and ProcedureParticipants were informed that the study was a short psychology ques-tionnaire dealing with interpersonal perception They were given aquestionnaire containing either the cooperation or the competition sce-nario In the cooperation condition the scenario read

Person X is about to meet someone with whom he will have to work to-gether on a task Person X has a few moments to find out what the other per-son is like After the meeting Person X will have to decide what strategy totake during the cooperation in order to achieve the most successful out-come

In the competition condition the scenario read

Person X is about to meet someone with whom he will have to compete on atask Person X has a few moments to find out what the other person is likeAfter the meeting Person X will have to decide what strategy to take duringthe competition in order to achieve the most successful outcome

INTERACTION GOALS AND MEMORY 413

Participants then answered 2 questions Those in the cooperation condi-tion answered ldquoHow costly is it for Person X to mistakenly overestimatehow good the partner isrdquo and ldquoHow costly is it for Person X to mistak-enly underestimate how good the partner isrdquo on a 1 (not at all costly) to7 (extremely costly) scale Those in the competition condition answeredthe same items with the word ldquocompetitorrdquo replacing the word ldquopart-nerrdquo The presentation order of the scenarios was counterbalanced Af-ter they completed the questionnaires the participants were debriefedand thanked for their participation

RESULTS AND DISCUSSION

Participantsrsquo responses were submitted to a 2 (goal competition coop-eration) acute 2 (estimation overestimation underestimation) acute 2 (questionorder) mixed ANOVA with the second factor varying within subjectThere was no main effect or interaction associated with question orderand thus the order factorwas dropped from the analysis It was expectedthat overestimating how good onersquos partner is and underestimatinghow good onersquos opponent is would be deemed undesirable and this wasindeed the case The only significant effect was the goal acute estimation in-teraction F(1 16) = 2918 p lt 001 When expecting to cooperate peoplejudged that it was more costly to overestimate (M = 556) than underesti-mate (M = 289)onersquos partner F(1 16) = 3011p lt 001 In contrastpeoplejudged that it was more costly to underestimate (M = 574) than overesti-mate (M = 294) onersquos opponent F(1 16) = 1580 p = 001

The findings from Study 1a suggest that for cooperation people deemit preferable to avoid overestimating how good the partner is and forcompetition to avoid underestimating how good the opponent is Onecould argue that these results were obtained because participants wereexplicitly instructed to consider how costly it was to overestimate or un-derestimate the other person By focusing peoplersquos attention on the cost-liness of their decisions we might have led the participants to be morelossndashfocused than they would have been otherwise Another feature ofthis study that requires further examination is whether asking partici-pants to imagine the scenario from anotherrsquos perspective instead ofimagining themselves in the scenario might have affected the resultsMuch social psychological research has found differences between peo-plersquos psychological processes when they are thinking about themselvesversus others For example people tend to make situational causal attri-

414 CHAN AND YBARRA

butions for themselves but dispositional causal attributions for others(Jones amp Nisbett 1972) People pay more attention to unobservable andunintentional events when thinking about themselves but observableand intentional events when thinking about others (Malle amp Pearce2001) Peoplersquos representations of themselves in memory are character-ized by their own thoughts and feelings whereas their representationsof others are characterized by actions and appearances (McGuire ampMcGuire 1986 Prentice 1990) Although it is not perfectly clear how itis possible that the findings from Study 1a might not occur if participantswere asked to think of how they themselves would react To rule outthese alternative accounts we conducted Study 1b to replicate the re-sults of Study 1a

STUDY 1B

The current study was identical to Study 1a save for two features Firstwe asked the participants to judge how overestimating or underestimat-ing the other person would affect their chances of success instead of ex-plicitly pointing to the costliness of the decision as in Study 1a Secondwe asked participants to imagine themselves instead of others in scenar-ios where they either had to cooperate or compete with another person

METHOD

Design and ParticipantsTwenty students between 19ndash25 years old participated in this studyThey were approached on a university campus and asked to fill out ashort questionnaire

Materials and ProcedureParticipants were informed that the study was a short psychology ques-tionnaire dealing with interpersonal perception They were asked toimagine themselves in the following scenarios

In the cooperation condition the scenario read

You are about to meet Person X and both of you will have to work togetherYou have a few moments to find out more about Person X After finding outmore about what Person X is like you will have to decide what strategy to

INTERACTION GOALS AND MEMORY 415

take during the cooperation in order to achieve the most successful out-come

In the competition condition the scenario read

You are about to meet Person X and both of you will have to compete againsteach other You will have a few moments to find out more about Person XAfter finding out more about what Person X is like you will have to decidewhat strategy to take during the competition in order to achieve the mostsuccessful outcome

Participants then answered 2 questions Those in the cooperation condi-tion answered ldquoIf you were to overestimate how good your partner ishow would it affect your chances of achieving a successful outcomerdquoand ldquoIf you were to underestimate how good your partner is howwould it affect your chances of achieving a successful outcomerdquo on a 1(reduce chance of success) to 7 (increase chance of success) scale Thequestions in the competition condition were the same with the wordldquoopponentrdquo replacing the word ldquopartnerrdquo The presentation order of thescenarios was counterbalanced After they completed the question-naires the participants were debriefed and thanked for their participa-tion

RESULTS AND DISCUSSION

Participantsrsquo responses were submitted to a 2 (goal competition coop-eration) acute 2 (estimation overestimation underestimation) acute 2 (questionorder) mixed ANOVA with the second factor varying within subjectThere was neither a main effect nor interaction associated with questionorder so it was dropped from the analysis It was expected that peoplewould judge that overestimating their partners and underestimatingtheir opponents would reduce their chances of success The results wereconsistent with the hypothesis The only significant effect was the goal acuteestimation interaction F(1 19) = 4532 p lt 001When expecting to coop-erate people judged that their chances of success would be lower if theyoverestimated their partners (M = 200)than if they underestimated theirpartners (M = 410) F(1 19) = 2000 p lt 001 In contrast people judgedthat their chances of success would be higher if they overestimated theiropponents (M = 430) than if they underestimated their opponents (M =260) F(1 19) = 4185p lt 001These findings thus confirm those of Study

416 CHAN AND YBARRA

1a and indicate that people may be inclined to adopt different informa-tion processing strategies depending on their interaction goals

If it really is the case that people when expecting to cooperate arepoised to process and be vigilant about a partnerrsquos negative qualities(underestimation) it would be expected that in a different cognitive taskthose inclinations would also be expressed For example in learningabout a potential partner who is described with negative and positivequalities we should find that people expecting to cooperate are morelikely to elaborate information diagnostic of the partnerrsquos negative thanpositive qualities Similar outcomes should be expected under antici-pated competition If people who expect to compete are inclined to un-cover an opponentrsquos positive qualities (overestimation) then it would beexpected that in a related cognitive task they would be more likely toelaborate information diagnostic of the opponentrsquos positive than nega-tive qualities This is the issue we examined in the second study

STUDY 2

Trait concepts differ in the degree to which they are negativendashdiagnosticor positivendashdiagnostic according to the schematic model of dispositionalattribution (Reeder 1985 Reeder amp Brewer 1979 ) and thecuendashdiagnosticity model of social perception (Skowronski amp Carlston1987 1989) The lay causal theories that people use to understand behav-iors related to negativendashdiagnostic traits (eg moralityndashrelated traits)indicate that negative behaviors are caused by dispositional factors butthat positive behaviors are caused by situational factors or a combina-tion of both (Ybarra amp Stephan 1999 Ybarra 2002) In contrast the laycausal theories people use to understand behaviors related to posi-tivendashdiagnostic traits (eg competencendashrelated traits but also othertraits as we will describe presently) indicate that positive behaviors arecaused by dispositional factors but that negative behaviors are causedby situational factors (Reeder amp Fulks 1980 but see Ybarra 2001 2002for a different perspective)

If a cooperation interaction goal leads people to emphasize a futurepartnerrsquos negative qualities then perceivers should focus on informa-tion related to negativendashdiagnostic traits As a result their informationprocessing (eg attention to and elaboration of information) should beguided by the lay causal theories underlying such traits (negative behav-iors are caused by dispositional causes and positive behaviors are

INTERACTION GOALS AND MEMORY 417

caused by situational causes) (Ybarra amp Stephan 1996 Ybarra ampStephan 1999) Thus people with a cooperation interaction goal shouldbe likely to remember information consistent rather than inconsistentwith the casual theory for the negativendashdiagnostic trait domain (iegood recall for positive behaviors caused by situational factors and neg-ative behaviors caused by dispositional factors) Behaviors related topositivendashdiagnostic traits should be processed less well because this in-formation is of little relevance to the goal of uncovering the partnerrsquosweaknesses Therefore it was expected that perceivers under a coopera-tion goal would not differentiate in memory between theory consistentand inconsistent behavioral information related to positivendashdiagnostictraits1

By comparison if competition leads people to emphasize a future op-ponentrsquos positive qualities then perceivers should focus on informationrelated to positivendashdiagnostic traits As a result their information pro-cessing (eg elaboration of information) should be guided by the laycausal theories underlying such traits (positive behaviors caused bydispositional causes and negative behaviors caused by situationalcauses) Thus people with a competition interaction goal should betterremember information consistent rather than inconsistent with the ca-sual theory for the positivendashdiagnostic trait domain (ie good recall forpositive behaviors caused by dispositional factors and negative valencebehaviors caused by situational factors) Behaviors from the nega-tivendashdiagnostic trait domain should not be well processed because thisinformationis of little relevance to the goal of uncovering the opponentrsquosstrengths Therefore it was expected that perceivers under a competi-tion goal would not show discrimination in their memory between the-ory consistent and inconsistent behavioral information related tonegativendashdiagnostic traits

A relevant question for sake of comparison is what kind of informa-tion would a simple impression formation goal lead perceivers to em-phasize and remember Some research on impression formation

418 CHAN AND YBARRA

1 The strategies of underestimating onersquos partner and overestimating onersquos opponentdo not have to be deliberate and intentional As Study 1b demonstrated even people whowere not consciously thinking about avoiding costs emerged with the same preferences forunderestimating partners and overestimating opponents Once the interaction goals trig-ger a preference for underndash or overestimating the other person the differences in informa-tion processing and memory may occur as a result of peoplersquos lay causal theories guidingthe processing of available information

suggests that people who are forming impressions focus more on the af-fective and general evaluative qualities of others (Levy amp Dugan 1960Ybarra 2001 Zajonc 1980) and that they tend to make such evaluationsbased mainly on negativendashdiagnostic (moralityndashrelated) information(Wojciszke Bazinska amp Jaworski 1998) Thus according to this per-spective people under a general impression formation goal should onlyattend to information related to the negativendashdiagnostic traits similar topeople with a cooperation goal (ie good recall for positive behaviorscaused by situational factors and negative behaviors caused bydispositional factors) Behaviors from the positivendashdiagnostic trait do-main should not be well processed because this information is of littlerelevance to the goal of forming general evaluative impressions There-fore it was expected that perceivers under an impression goal would notdiscriminate in their memory between theory consistent and inconsis-tent behavioral information related to the positivendashdiagnostic traitsThetheoretical model is summarized in Figure 1

Depending on the experimental group to which they were assignedparticipants were told that they would either cooperate or compete withanother participant Then they learned about the other participantrsquos be-haviors which were related to traits in the negativendashdiagnostic trait do-

INTERACTION GOALS AND MEMORY 419

FIGURE 1 Summary of theoretical model

main (hon esty h elpfu lness and fr iendliness ) and thepositivendashdiagnostic trait domain (hardworking image scoring) Imagescoring refers to people engaging in costly positive behaviors to signal toothers that they are valuable community members (Nowak amp Sigmund1998Wedekind amp Milinski 2000) Controlparticipants learned the samematerial but they were asked to form an impression of the person andwere told that they would not interact with the other participant

METHOD

Design and ParticipantsNinetyndashone students participated in the study for course credit Theywere randomly assigned to the cooperation competition or control (im-pression formation) condition All of the participants were presentedwith behaviors with a positive or negative valence from the negativendashdi-agnostic and positivendashdiagnostic trait domains and these behaviorswere linked to either explicit dispositional or situational attributionsThus the overall design of the study was a 3 (interaction goal coopera-tion competition control) acute 2 (trait domain positivendashdiagnostic nega-tivendashdiagnostic) acute 2 (behavior valence positive negative) acute 2(attribution dispositional situational) mixed design with the first fac-tor varying between participants and the latter three factors varyingwithin participants The participants were randomly assigned to condi-tions and were run in nonndashinteracting groups of two to six

Stimulus MaterialsThe behavioral information that the participants processed was pre-sented on a cassette tape allegedly recorded by another participant Thegender of the speaker on the tape and that of the participant were alwaysthe same Each tape contained 20 behavioral statements half of whichwere negative and half positive in valence Also half of the behaviorswere linked to dispositional attributions and half to situational attribu-tions Twelve of the statements pertained to negativendashdiagnostic traits(eg ldquoI entered a concert through the fire exitrdquo [negative valence] ldquoI sentmy sister a present and a birthday bouquetrdquo [positive valence]) Eight ofthe statements pertained to the positivendashdiagnostic traits (eg ldquoI missedmorning lecturesrdquo [negative valence] ldquoI reported a crimerdquo [positive va-lence])

420 CHAN AND YBARRA

Our choice of the traits hardworking and image scoring was deliberatebecause it allows us to show that perceiverrsquos attention to a targetrsquosstrengths (overestimation) is not confounded with emphasizing infor-mation about the targetrsquos competence For example it is usually as-sumed that there is an inverse relationship between the amount of worka person engages in and the presence of the related ability (Darley ampGoethals 1980 Jones 1989) but being hardworking should still signal apositive quality a strength In addition image scoring refers to peopleengaging in costly positive behaviors such as extreme levels of philan-thropy to enhance their reputation and status (Wedekind amp Milinski2000 McAndrew 2002) It is completely unrelated to competence but itsdiagnosticity is in line with that of a positivendashdiagnostic trait

The behavior stems (no attributions) were pretested to ensure thatthey were indeed regarded as positive or negativendashdiagnostic Thirtyparticipants rated each behavior on how common they are For exampleone item queried them as to ldquoWhat percentage of college students doyou think enter a concert through the fire exitrdquo (6ndashpoint scale with 0ndash1010ndash30 30ndash50 50ndash70 70ndash90 and 90ndash100) For positivendashdiagnos-tic traits behaviors that are positive in valence should be less common(and hence more diagnostic of the actorrsquos dispositions) than behaviorsthat are negative in valence whereas for negativendashdiagnostic traits be-haviors negative in valence should be less common (and hence more di-agnostic of the actorrsquos dispositions) than behaviors positive in valenceResults showed an interaction effect that confirmed our classification ofthe behaviors into the respective positive and negativendashdiagnostic cate-gories F(1 29) = 4913 p lt 001 For the positivendashdiagnostic traits posi-tive valence behaviors (M = 281) were judged to be enacted less oftenthan negative valence behaviors (M = 376) F(1 29) = 1961 p lt 001 Forthe negativendashdiagnostic traits negative valence behaviors (M = 294)were judged to be enacted less often than positive valence behaviors (M= 331) F(1 29) = 7103 p = 012

Two sets of materials both containing the same 20 behavior stemswere created Each behavior stem was matched with a dispositional at-tribution in one set and a situational attribution in the other set For ex-ample in set A the behaviors ldquoI didnrsquot help my dad with yard workbecause I wanted to stay inside to watch TVrdquo [negative behavior in nega-tivendashdiagnostic domain with dispositional attribution] ldquoI sent my sistera present and a birthday bouquet because my grandmother asked me todo itrdquo [positive behavior in negativendashdiagnostic domain with situational

INTERACTION GOALS AND MEMORY 421

attribution] ldquoI missed morning lectures because my roommate playedpranks with my alarm clockrdquo [negative behavior in positivendashdiagnosticdomain with situational attribution] and ldquoI worked as a research assis-tant because I am very interested in the subject matterrdquo [positive behav-ior in positive diagnostic domain with dispositional attribution] wouldbe counterbalanced in set B with different causal attributions These in-cluded ldquoI didnrsquot help my dad with yard work because he couldnrsquot findthe keys to the tool shedrdquo [situational attribution] ldquoI sent my sister apresent and a birthday bouquet because I wanted to give her a pleasantsurpriserdquo [dispositional attribution] ldquoI missed morning lecture becauseI donrsquot feel like waking up for themrdquo [dispositional attribution] and ldquoIworked as a research assistant because my professor said I had to do itrdquo[situational attribution] As will be described below the valence of thebehaviors and attributions (dispositional vs external) were manipu-lated independently to ensure that subsequent differences in recall oftheoryndashconsistent or inconsistent information could not simply be dueto an artifact based on the memorability of the particular behaviors

The two sets of stimuli were presented to 17 participants for pretest-ing One participant misunderstood how to use the response scale andwas dropped from the analysis Participants were first presented withthe behavior stems and were asked to rate how favorable each behaviorwas on a 7ndashpoint scale with higher values indicating greaterfavorability Then they were presented with the same behavior stemslinked to either a dispositional or situational attribution and were askedto judge the locus of causation for the behaviors on a 7ndashpoint scale withhigher values indicating more dispositional causation The responseswere submitted to a 2 (valence)acute 2 (attribution)withinndashsubject ANOVAFor the valence judgments the positive behaviors in both stimulus sets(Mset A = 544 Mset B = 559) were judged as more favorable than the nega-tive behaviors (Mset A = 284Mset B = 215) Fset A (1 8)= 3416 p lt 0004Fset B

(1 8) = 9782 p lt 0001 For the second set of judgments a main effect forattribution was obtained for both sets of stimuli Behaviors linked todispositional attributions (Mset A = 575 Mset B = 646) were judged to bemore dispositionally caused than behaviors linked to situational attribu-tions (M set A = 235 Mset B = 298) F set A (1 8) = 4574 p lt 0001 Fset B (1 8) =15826 p lt 0001 No other main effects or interaction effects were ob-tained The two sets of stimuli were presented under three different ran-domization schemes

422 CHAN AND YBARRA

PROCEDURE

Upon arrival at the lab the participants were provided the cover storyfor the study in which they were told that the study would examine howpersonal information would affect subsequent interactions betweenstrangers They were told that another set of participants was recordingsome personal information on a tape for them to listen to To increase thecredibility of the tape contents participants were told that the partici-pants recording the tape had been asked to provide distinct personal in-formation about ldquothings that they have done in the past few weeksrdquoAdditional information about the experiment including whether theywere in the cooperation competition or control (impression formation)condition was given to each participant in a sealed envelope so that theexperimenter could remain blind to the conditions Experimental groupparticipants were told that after they listened to a tape with personal in-formation recorded by another participant they would get a chance tomeet them in either a cooperative or competitive setting Control partici-pants were informed that they had been randomly assigned to the con-trol group and would listen to some information about a previousparticipant instead of having an interaction with them

Participants in the cooperation condition read that they would engagein a ldquoCommunity Simulation Gamerdquo in which they and their assignedpartners ldquowill work together to try to maximize the number of points foryour team The team with the highest points will receive a prizerdquo Thecooperative nature of the interaction was emphasized with repeated useof the word ldquopartnerrdquo In contrastparticipants in the competition condi-tion read that they would engage in a ldquoWall Street Simulation Gamerdquo inwhich they would ldquocompete against [their] competitor to maximize[their] own profit (points) and minimize their opponentrsquos profit Theperson with the highest points will win a prizerdquo The competitive aspectof the upcoming interaction was emphasized further with repeated useof the words ldquocompetitorrdquo and ldquoopponentrdquo This type of reward struc-ture manipulation has been used to study cooperation and competitionby other researchers (Deutsch 1973)

Participants in the control condition read that they would listen to atape that had been recorded in a previous experimental session Theywere told to form an impression of the person using the informationfrom the tape as part of a ldquostandardizationrdquo procedure for the experi-ment To bolster the believability of the cover story participants then

INTERACTION GOALS AND MEMORY 423

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

Specifically a perceiver who is sizing up a future partner can minimizepotential losses by avoiding the overestimationof the partnerrsquos strengthsThis can be done by being sensitive to information diagnostic of the part-nerrsquos negative qualities For example if a person finds out that it is easy fora partner to lose enthusiasm after initial involvement with a project he orshe could anticipate the problem and attempt to reduce its negative im-pact on outcomes In addition the perceiver would also be motivated toattend to such weaknesses to decide how much to trust and rely on thepartner If the perceiver learned that the partner was weak in some skillhe or she could try to strengthen that skill in the self or at least be vigilantto any task that involved that skill (Williams amp Karau 1991 WittenbaumVaughan amp Stasser 1998) The cost of overestimating how good the part-ner is and failing to notice the partnerrsquos weaknesses is immense if the per-son mistakenly thinks that the partner is hardworking and talented andtherefore becomes complacent the likelihood of failure is increased

The interaction goal of competition puts the perceiver in a differentposition A perceiver who is sizing up a future opponent can minimizepotential losses by avoiding the underestimation of the opponentrsquosstrengths This can be done by being sensitive to information diagnosticof the opponentrsquos positive qualities For example if a person finds outthat his opponent is highly skilled and hardworking he could anticipatethe challenge and attempt to reduce his chance of losing by workingeven harder and improving his own skills Underestimating the oppo-nent would undoubtedly increase the risk of defeat because theperceiver would be underndashprepared for the competition

Study 1a and 1b first tested the critical assumptions regarding how theinteraction goal of cooperation would lead people to avoid overestimat-ing how good their partners are whereas the interaction goal of compe-tition would lead people to avoid underestimating how good theiropponents are In a subsequent validation of our analysis Study 2 exam-ined how these interaction goals affected memory for behavioral infor-mation from different trait domains Finally Study 3 examined how theinteraction goals affected the nature and amount of information peoplesought about their future partnerscompetitors

STUDY 1A

We reasoned that people who are evaluating a potential partner shouldtry to avoid the mistake of overestimating the partnerrsquos positive qualities

412 CHAN AND YBARRA

because being generally risk averse they should focus on potentiallosses that could arise from cooperating with an inferior partner Thusthey should attempt to gather negative dispositional information aboutthe partner People who are evaluating a potential opponent should tryto avoid the mistake of underestimating the opponentrsquos positive quali-ties because being risk averse they should focus on potential losses thatcould arise from competing with a superior opponent Thus theyshould attempt to gather positive dispositional information about theopponent To examine these predictions participants in the first studywere presented a fictitious scenario in which a person in the scenario hadto learn about someone else with whom he was to cooperate or competeParticipants were then asked to evaluate the targetrsquos strategy indicatinghow costly it would be for the target to overestimate or underestimatehow good his partner or opponent was

METHOD

Design and ParticipantsEighteen students between 19ndash25 years old volunteered for the studywithout payment They were approached on a university campus andasked if they would fill out a short questionnaire

Materials and ProcedureParticipants were informed that the study was a short psychology ques-tionnaire dealing with interpersonal perception They were given aquestionnaire containing either the cooperation or the competition sce-nario In the cooperation condition the scenario read

Person X is about to meet someone with whom he will have to work to-gether on a task Person X has a few moments to find out what the other per-son is like After the meeting Person X will have to decide what strategy totake during the cooperation in order to achieve the most successful out-come

In the competition condition the scenario read

Person X is about to meet someone with whom he will have to compete on atask Person X has a few moments to find out what the other person is likeAfter the meeting Person X will have to decide what strategy to take duringthe competition in order to achieve the most successful outcome

INTERACTION GOALS AND MEMORY 413

Participants then answered 2 questions Those in the cooperation condi-tion answered ldquoHow costly is it for Person X to mistakenly overestimatehow good the partner isrdquo and ldquoHow costly is it for Person X to mistak-enly underestimate how good the partner isrdquo on a 1 (not at all costly) to7 (extremely costly) scale Those in the competition condition answeredthe same items with the word ldquocompetitorrdquo replacing the word ldquopart-nerrdquo The presentation order of the scenarios was counterbalanced Af-ter they completed the questionnaires the participants were debriefedand thanked for their participation

RESULTS AND DISCUSSION

Participantsrsquo responses were submitted to a 2 (goal competition coop-eration) acute 2 (estimation overestimation underestimation) acute 2 (questionorder) mixed ANOVA with the second factor varying within subjectThere was no main effect or interaction associated with question orderand thus the order factorwas dropped from the analysis It was expectedthat overestimating how good onersquos partner is and underestimatinghow good onersquos opponent is would be deemed undesirable and this wasindeed the case The only significant effect was the goal acute estimation in-teraction F(1 16) = 2918 p lt 001 When expecting to cooperate peoplejudged that it was more costly to overestimate (M = 556) than underesti-mate (M = 289)onersquos partner F(1 16) = 3011p lt 001 In contrastpeoplejudged that it was more costly to underestimate (M = 574) than overesti-mate (M = 294) onersquos opponent F(1 16) = 1580 p = 001

The findings from Study 1a suggest that for cooperation people deemit preferable to avoid overestimating how good the partner is and forcompetition to avoid underestimating how good the opponent is Onecould argue that these results were obtained because participants wereexplicitly instructed to consider how costly it was to overestimate or un-derestimate the other person By focusing peoplersquos attention on the cost-liness of their decisions we might have led the participants to be morelossndashfocused than they would have been otherwise Another feature ofthis study that requires further examination is whether asking partici-pants to imagine the scenario from anotherrsquos perspective instead ofimagining themselves in the scenario might have affected the resultsMuch social psychological research has found differences between peo-plersquos psychological processes when they are thinking about themselvesversus others For example people tend to make situational causal attri-

414 CHAN AND YBARRA

butions for themselves but dispositional causal attributions for others(Jones amp Nisbett 1972) People pay more attention to unobservable andunintentional events when thinking about themselves but observableand intentional events when thinking about others (Malle amp Pearce2001) Peoplersquos representations of themselves in memory are character-ized by their own thoughts and feelings whereas their representationsof others are characterized by actions and appearances (McGuire ampMcGuire 1986 Prentice 1990) Although it is not perfectly clear how itis possible that the findings from Study 1a might not occur if participantswere asked to think of how they themselves would react To rule outthese alternative accounts we conducted Study 1b to replicate the re-sults of Study 1a

STUDY 1B

The current study was identical to Study 1a save for two features Firstwe asked the participants to judge how overestimating or underestimat-ing the other person would affect their chances of success instead of ex-plicitly pointing to the costliness of the decision as in Study 1a Secondwe asked participants to imagine themselves instead of others in scenar-ios where they either had to cooperate or compete with another person

METHOD

Design and ParticipantsTwenty students between 19ndash25 years old participated in this studyThey were approached on a university campus and asked to fill out ashort questionnaire

Materials and ProcedureParticipants were informed that the study was a short psychology ques-tionnaire dealing with interpersonal perception They were asked toimagine themselves in the following scenarios

In the cooperation condition the scenario read

You are about to meet Person X and both of you will have to work togetherYou have a few moments to find out more about Person X After finding outmore about what Person X is like you will have to decide what strategy to

INTERACTION GOALS AND MEMORY 415

take during the cooperation in order to achieve the most successful out-come

In the competition condition the scenario read

You are about to meet Person X and both of you will have to compete againsteach other You will have a few moments to find out more about Person XAfter finding out more about what Person X is like you will have to decidewhat strategy to take during the competition in order to achieve the mostsuccessful outcome

Participants then answered 2 questions Those in the cooperation condi-tion answered ldquoIf you were to overestimate how good your partner ishow would it affect your chances of achieving a successful outcomerdquoand ldquoIf you were to underestimate how good your partner is howwould it affect your chances of achieving a successful outcomerdquo on a 1(reduce chance of success) to 7 (increase chance of success) scale Thequestions in the competition condition were the same with the wordldquoopponentrdquo replacing the word ldquopartnerrdquo The presentation order of thescenarios was counterbalanced After they completed the question-naires the participants were debriefed and thanked for their participa-tion

RESULTS AND DISCUSSION

Participantsrsquo responses were submitted to a 2 (goal competition coop-eration) acute 2 (estimation overestimation underestimation) acute 2 (questionorder) mixed ANOVA with the second factor varying within subjectThere was neither a main effect nor interaction associated with questionorder so it was dropped from the analysis It was expected that peoplewould judge that overestimating their partners and underestimatingtheir opponents would reduce their chances of success The results wereconsistent with the hypothesis The only significant effect was the goal acuteestimation interaction F(1 19) = 4532 p lt 001When expecting to coop-erate people judged that their chances of success would be lower if theyoverestimated their partners (M = 200)than if they underestimated theirpartners (M = 410) F(1 19) = 2000 p lt 001 In contrast people judgedthat their chances of success would be higher if they overestimated theiropponents (M = 430) than if they underestimated their opponents (M =260) F(1 19) = 4185p lt 001These findings thus confirm those of Study

416 CHAN AND YBARRA

1a and indicate that people may be inclined to adopt different informa-tion processing strategies depending on their interaction goals

If it really is the case that people when expecting to cooperate arepoised to process and be vigilant about a partnerrsquos negative qualities(underestimation) it would be expected that in a different cognitive taskthose inclinations would also be expressed For example in learningabout a potential partner who is described with negative and positivequalities we should find that people expecting to cooperate are morelikely to elaborate information diagnostic of the partnerrsquos negative thanpositive qualities Similar outcomes should be expected under antici-pated competition If people who expect to compete are inclined to un-cover an opponentrsquos positive qualities (overestimation) then it would beexpected that in a related cognitive task they would be more likely toelaborate information diagnostic of the opponentrsquos positive than nega-tive qualities This is the issue we examined in the second study

STUDY 2

Trait concepts differ in the degree to which they are negativendashdiagnosticor positivendashdiagnostic according to the schematic model of dispositionalattribution (Reeder 1985 Reeder amp Brewer 1979 ) and thecuendashdiagnosticity model of social perception (Skowronski amp Carlston1987 1989) The lay causal theories that people use to understand behav-iors related to negativendashdiagnostic traits (eg moralityndashrelated traits)indicate that negative behaviors are caused by dispositional factors butthat positive behaviors are caused by situational factors or a combina-tion of both (Ybarra amp Stephan 1999 Ybarra 2002) In contrast the laycausal theories people use to understand behaviors related to posi-tivendashdiagnostic traits (eg competencendashrelated traits but also othertraits as we will describe presently) indicate that positive behaviors arecaused by dispositional factors but that negative behaviors are causedby situational factors (Reeder amp Fulks 1980 but see Ybarra 2001 2002for a different perspective)

If a cooperation interaction goal leads people to emphasize a futurepartnerrsquos negative qualities then perceivers should focus on informa-tion related to negativendashdiagnostic traits As a result their informationprocessing (eg attention to and elaboration of information) should beguided by the lay causal theories underlying such traits (negative behav-iors are caused by dispositional causes and positive behaviors are

INTERACTION GOALS AND MEMORY 417

caused by situational causes) (Ybarra amp Stephan 1996 Ybarra ampStephan 1999) Thus people with a cooperation interaction goal shouldbe likely to remember information consistent rather than inconsistentwith the casual theory for the negativendashdiagnostic trait domain (iegood recall for positive behaviors caused by situational factors and neg-ative behaviors caused by dispositional factors) Behaviors related topositivendashdiagnostic traits should be processed less well because this in-formation is of little relevance to the goal of uncovering the partnerrsquosweaknesses Therefore it was expected that perceivers under a coopera-tion goal would not differentiate in memory between theory consistentand inconsistent behavioral information related to positivendashdiagnostictraits1

By comparison if competition leads people to emphasize a future op-ponentrsquos positive qualities then perceivers should focus on informationrelated to positivendashdiagnostic traits As a result their information pro-cessing (eg elaboration of information) should be guided by the laycausal theories underlying such traits (positive behaviors caused bydispositional causes and negative behaviors caused by situationalcauses) Thus people with a competition interaction goal should betterremember information consistent rather than inconsistent with the ca-sual theory for the positivendashdiagnostic trait domain (ie good recall forpositive behaviors caused by dispositional factors and negative valencebehaviors caused by situational factors) Behaviors from the nega-tivendashdiagnostic trait domain should not be well processed because thisinformationis of little relevance to the goal of uncovering the opponentrsquosstrengths Therefore it was expected that perceivers under a competi-tion goal would not show discrimination in their memory between the-ory consistent and inconsistent behavioral information related tonegativendashdiagnostic traits

A relevant question for sake of comparison is what kind of informa-tion would a simple impression formation goal lead perceivers to em-phasize and remember Some research on impression formation

418 CHAN AND YBARRA

1 The strategies of underestimating onersquos partner and overestimating onersquos opponentdo not have to be deliberate and intentional As Study 1b demonstrated even people whowere not consciously thinking about avoiding costs emerged with the same preferences forunderestimating partners and overestimating opponents Once the interaction goals trig-ger a preference for underndash or overestimating the other person the differences in informa-tion processing and memory may occur as a result of peoplersquos lay causal theories guidingthe processing of available information

suggests that people who are forming impressions focus more on the af-fective and general evaluative qualities of others (Levy amp Dugan 1960Ybarra 2001 Zajonc 1980) and that they tend to make such evaluationsbased mainly on negativendashdiagnostic (moralityndashrelated) information(Wojciszke Bazinska amp Jaworski 1998) Thus according to this per-spective people under a general impression formation goal should onlyattend to information related to the negativendashdiagnostic traits similar topeople with a cooperation goal (ie good recall for positive behaviorscaused by situational factors and negative behaviors caused bydispositional factors) Behaviors from the positivendashdiagnostic trait do-main should not be well processed because this information is of littlerelevance to the goal of forming general evaluative impressions There-fore it was expected that perceivers under an impression goal would notdiscriminate in their memory between theory consistent and inconsis-tent behavioral information related to the positivendashdiagnostic traitsThetheoretical model is summarized in Figure 1

Depending on the experimental group to which they were assignedparticipants were told that they would either cooperate or compete withanother participant Then they learned about the other participantrsquos be-haviors which were related to traits in the negativendashdiagnostic trait do-

INTERACTION GOALS AND MEMORY 419

FIGURE 1 Summary of theoretical model

main (hon esty h elpfu lness and fr iendliness ) and thepositivendashdiagnostic trait domain (hardworking image scoring) Imagescoring refers to people engaging in costly positive behaviors to signal toothers that they are valuable community members (Nowak amp Sigmund1998Wedekind amp Milinski 2000) Controlparticipants learned the samematerial but they were asked to form an impression of the person andwere told that they would not interact with the other participant

METHOD

Design and ParticipantsNinetyndashone students participated in the study for course credit Theywere randomly assigned to the cooperation competition or control (im-pression formation) condition All of the participants were presentedwith behaviors with a positive or negative valence from the negativendashdi-agnostic and positivendashdiagnostic trait domains and these behaviorswere linked to either explicit dispositional or situational attributionsThus the overall design of the study was a 3 (interaction goal coopera-tion competition control) acute 2 (trait domain positivendashdiagnostic nega-tivendashdiagnostic) acute 2 (behavior valence positive negative) acute 2(attribution dispositional situational) mixed design with the first fac-tor varying between participants and the latter three factors varyingwithin participants The participants were randomly assigned to condi-tions and were run in nonndashinteracting groups of two to six

Stimulus MaterialsThe behavioral information that the participants processed was pre-sented on a cassette tape allegedly recorded by another participant Thegender of the speaker on the tape and that of the participant were alwaysthe same Each tape contained 20 behavioral statements half of whichwere negative and half positive in valence Also half of the behaviorswere linked to dispositional attributions and half to situational attribu-tions Twelve of the statements pertained to negativendashdiagnostic traits(eg ldquoI entered a concert through the fire exitrdquo [negative valence] ldquoI sentmy sister a present and a birthday bouquetrdquo [positive valence]) Eight ofthe statements pertained to the positivendashdiagnostic traits (eg ldquoI missedmorning lecturesrdquo [negative valence] ldquoI reported a crimerdquo [positive va-lence])

420 CHAN AND YBARRA

Our choice of the traits hardworking and image scoring was deliberatebecause it allows us to show that perceiverrsquos attention to a targetrsquosstrengths (overestimation) is not confounded with emphasizing infor-mation about the targetrsquos competence For example it is usually as-sumed that there is an inverse relationship between the amount of worka person engages in and the presence of the related ability (Darley ampGoethals 1980 Jones 1989) but being hardworking should still signal apositive quality a strength In addition image scoring refers to peopleengaging in costly positive behaviors such as extreme levels of philan-thropy to enhance their reputation and status (Wedekind amp Milinski2000 McAndrew 2002) It is completely unrelated to competence but itsdiagnosticity is in line with that of a positivendashdiagnostic trait

The behavior stems (no attributions) were pretested to ensure thatthey were indeed regarded as positive or negativendashdiagnostic Thirtyparticipants rated each behavior on how common they are For exampleone item queried them as to ldquoWhat percentage of college students doyou think enter a concert through the fire exitrdquo (6ndashpoint scale with 0ndash1010ndash30 30ndash50 50ndash70 70ndash90 and 90ndash100) For positivendashdiagnos-tic traits behaviors that are positive in valence should be less common(and hence more diagnostic of the actorrsquos dispositions) than behaviorsthat are negative in valence whereas for negativendashdiagnostic traits be-haviors negative in valence should be less common (and hence more di-agnostic of the actorrsquos dispositions) than behaviors positive in valenceResults showed an interaction effect that confirmed our classification ofthe behaviors into the respective positive and negativendashdiagnostic cate-gories F(1 29) = 4913 p lt 001 For the positivendashdiagnostic traits posi-tive valence behaviors (M = 281) were judged to be enacted less oftenthan negative valence behaviors (M = 376) F(1 29) = 1961 p lt 001 Forthe negativendashdiagnostic traits negative valence behaviors (M = 294)were judged to be enacted less often than positive valence behaviors (M= 331) F(1 29) = 7103 p = 012

Two sets of materials both containing the same 20 behavior stemswere created Each behavior stem was matched with a dispositional at-tribution in one set and a situational attribution in the other set For ex-ample in set A the behaviors ldquoI didnrsquot help my dad with yard workbecause I wanted to stay inside to watch TVrdquo [negative behavior in nega-tivendashdiagnostic domain with dispositional attribution] ldquoI sent my sistera present and a birthday bouquet because my grandmother asked me todo itrdquo [positive behavior in negativendashdiagnostic domain with situational

INTERACTION GOALS AND MEMORY 421

attribution] ldquoI missed morning lectures because my roommate playedpranks with my alarm clockrdquo [negative behavior in positivendashdiagnosticdomain with situational attribution] and ldquoI worked as a research assis-tant because I am very interested in the subject matterrdquo [positive behav-ior in positive diagnostic domain with dispositional attribution] wouldbe counterbalanced in set B with different causal attributions These in-cluded ldquoI didnrsquot help my dad with yard work because he couldnrsquot findthe keys to the tool shedrdquo [situational attribution] ldquoI sent my sister apresent and a birthday bouquet because I wanted to give her a pleasantsurpriserdquo [dispositional attribution] ldquoI missed morning lecture becauseI donrsquot feel like waking up for themrdquo [dispositional attribution] and ldquoIworked as a research assistant because my professor said I had to do itrdquo[situational attribution] As will be described below the valence of thebehaviors and attributions (dispositional vs external) were manipu-lated independently to ensure that subsequent differences in recall oftheoryndashconsistent or inconsistent information could not simply be dueto an artifact based on the memorability of the particular behaviors

The two sets of stimuli were presented to 17 participants for pretest-ing One participant misunderstood how to use the response scale andwas dropped from the analysis Participants were first presented withthe behavior stems and were asked to rate how favorable each behaviorwas on a 7ndashpoint scale with higher values indicating greaterfavorability Then they were presented with the same behavior stemslinked to either a dispositional or situational attribution and were askedto judge the locus of causation for the behaviors on a 7ndashpoint scale withhigher values indicating more dispositional causation The responseswere submitted to a 2 (valence)acute 2 (attribution)withinndashsubject ANOVAFor the valence judgments the positive behaviors in both stimulus sets(Mset A = 544 Mset B = 559) were judged as more favorable than the nega-tive behaviors (Mset A = 284Mset B = 215) Fset A (1 8)= 3416 p lt 0004Fset B

(1 8) = 9782 p lt 0001 For the second set of judgments a main effect forattribution was obtained for both sets of stimuli Behaviors linked todispositional attributions (Mset A = 575 Mset B = 646) were judged to bemore dispositionally caused than behaviors linked to situational attribu-tions (M set A = 235 Mset B = 298) F set A (1 8) = 4574 p lt 0001 Fset B (1 8) =15826 p lt 0001 No other main effects or interaction effects were ob-tained The two sets of stimuli were presented under three different ran-domization schemes

422 CHAN AND YBARRA

PROCEDURE

Upon arrival at the lab the participants were provided the cover storyfor the study in which they were told that the study would examine howpersonal information would affect subsequent interactions betweenstrangers They were told that another set of participants was recordingsome personal information on a tape for them to listen to To increase thecredibility of the tape contents participants were told that the partici-pants recording the tape had been asked to provide distinct personal in-formation about ldquothings that they have done in the past few weeksrdquoAdditional information about the experiment including whether theywere in the cooperation competition or control (impression formation)condition was given to each participant in a sealed envelope so that theexperimenter could remain blind to the conditions Experimental groupparticipants were told that after they listened to a tape with personal in-formation recorded by another participant they would get a chance tomeet them in either a cooperative or competitive setting Control partici-pants were informed that they had been randomly assigned to the con-trol group and would listen to some information about a previousparticipant instead of having an interaction with them

Participants in the cooperation condition read that they would engagein a ldquoCommunity Simulation Gamerdquo in which they and their assignedpartners ldquowill work together to try to maximize the number of points foryour team The team with the highest points will receive a prizerdquo Thecooperative nature of the interaction was emphasized with repeated useof the word ldquopartnerrdquo In contrastparticipants in the competition condi-tion read that they would engage in a ldquoWall Street Simulation Gamerdquo inwhich they would ldquocompete against [their] competitor to maximize[their] own profit (points) and minimize their opponentrsquos profit Theperson with the highest points will win a prizerdquo The competitive aspectof the upcoming interaction was emphasized further with repeated useof the words ldquocompetitorrdquo and ldquoopponentrdquo This type of reward struc-ture manipulation has been used to study cooperation and competitionby other researchers (Deutsch 1973)

Participants in the control condition read that they would listen to atape that had been recorded in a previous experimental session Theywere told to form an impression of the person using the informationfrom the tape as part of a ldquostandardizationrdquo procedure for the experi-ment To bolster the believability of the cover story participants then

INTERACTION GOALS AND MEMORY 423

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

because being generally risk averse they should focus on potentiallosses that could arise from cooperating with an inferior partner Thusthey should attempt to gather negative dispositional information aboutthe partner People who are evaluating a potential opponent should tryto avoid the mistake of underestimating the opponentrsquos positive quali-ties because being risk averse they should focus on potential losses thatcould arise from competing with a superior opponent Thus theyshould attempt to gather positive dispositional information about theopponent To examine these predictions participants in the first studywere presented a fictitious scenario in which a person in the scenario hadto learn about someone else with whom he was to cooperate or competeParticipants were then asked to evaluate the targetrsquos strategy indicatinghow costly it would be for the target to overestimate or underestimatehow good his partner or opponent was

METHOD

Design and ParticipantsEighteen students between 19ndash25 years old volunteered for the studywithout payment They were approached on a university campus andasked if they would fill out a short questionnaire

Materials and ProcedureParticipants were informed that the study was a short psychology ques-tionnaire dealing with interpersonal perception They were given aquestionnaire containing either the cooperation or the competition sce-nario In the cooperation condition the scenario read

Person X is about to meet someone with whom he will have to work to-gether on a task Person X has a few moments to find out what the other per-son is like After the meeting Person X will have to decide what strategy totake during the cooperation in order to achieve the most successful out-come

In the competition condition the scenario read

Person X is about to meet someone with whom he will have to compete on atask Person X has a few moments to find out what the other person is likeAfter the meeting Person X will have to decide what strategy to take duringthe competition in order to achieve the most successful outcome

INTERACTION GOALS AND MEMORY 413

Participants then answered 2 questions Those in the cooperation condi-tion answered ldquoHow costly is it for Person X to mistakenly overestimatehow good the partner isrdquo and ldquoHow costly is it for Person X to mistak-enly underestimate how good the partner isrdquo on a 1 (not at all costly) to7 (extremely costly) scale Those in the competition condition answeredthe same items with the word ldquocompetitorrdquo replacing the word ldquopart-nerrdquo The presentation order of the scenarios was counterbalanced Af-ter they completed the questionnaires the participants were debriefedand thanked for their participation

RESULTS AND DISCUSSION

Participantsrsquo responses were submitted to a 2 (goal competition coop-eration) acute 2 (estimation overestimation underestimation) acute 2 (questionorder) mixed ANOVA with the second factor varying within subjectThere was no main effect or interaction associated with question orderand thus the order factorwas dropped from the analysis It was expectedthat overestimating how good onersquos partner is and underestimatinghow good onersquos opponent is would be deemed undesirable and this wasindeed the case The only significant effect was the goal acute estimation in-teraction F(1 16) = 2918 p lt 001 When expecting to cooperate peoplejudged that it was more costly to overestimate (M = 556) than underesti-mate (M = 289)onersquos partner F(1 16) = 3011p lt 001 In contrastpeoplejudged that it was more costly to underestimate (M = 574) than overesti-mate (M = 294) onersquos opponent F(1 16) = 1580 p = 001

The findings from Study 1a suggest that for cooperation people deemit preferable to avoid overestimating how good the partner is and forcompetition to avoid underestimating how good the opponent is Onecould argue that these results were obtained because participants wereexplicitly instructed to consider how costly it was to overestimate or un-derestimate the other person By focusing peoplersquos attention on the cost-liness of their decisions we might have led the participants to be morelossndashfocused than they would have been otherwise Another feature ofthis study that requires further examination is whether asking partici-pants to imagine the scenario from anotherrsquos perspective instead ofimagining themselves in the scenario might have affected the resultsMuch social psychological research has found differences between peo-plersquos psychological processes when they are thinking about themselvesversus others For example people tend to make situational causal attri-

414 CHAN AND YBARRA

butions for themselves but dispositional causal attributions for others(Jones amp Nisbett 1972) People pay more attention to unobservable andunintentional events when thinking about themselves but observableand intentional events when thinking about others (Malle amp Pearce2001) Peoplersquos representations of themselves in memory are character-ized by their own thoughts and feelings whereas their representationsof others are characterized by actions and appearances (McGuire ampMcGuire 1986 Prentice 1990) Although it is not perfectly clear how itis possible that the findings from Study 1a might not occur if participantswere asked to think of how they themselves would react To rule outthese alternative accounts we conducted Study 1b to replicate the re-sults of Study 1a

STUDY 1B

The current study was identical to Study 1a save for two features Firstwe asked the participants to judge how overestimating or underestimat-ing the other person would affect their chances of success instead of ex-plicitly pointing to the costliness of the decision as in Study 1a Secondwe asked participants to imagine themselves instead of others in scenar-ios where they either had to cooperate or compete with another person

METHOD

Design and ParticipantsTwenty students between 19ndash25 years old participated in this studyThey were approached on a university campus and asked to fill out ashort questionnaire

Materials and ProcedureParticipants were informed that the study was a short psychology ques-tionnaire dealing with interpersonal perception They were asked toimagine themselves in the following scenarios

In the cooperation condition the scenario read

You are about to meet Person X and both of you will have to work togetherYou have a few moments to find out more about Person X After finding outmore about what Person X is like you will have to decide what strategy to

INTERACTION GOALS AND MEMORY 415

take during the cooperation in order to achieve the most successful out-come

In the competition condition the scenario read

You are about to meet Person X and both of you will have to compete againsteach other You will have a few moments to find out more about Person XAfter finding out more about what Person X is like you will have to decidewhat strategy to take during the competition in order to achieve the mostsuccessful outcome

Participants then answered 2 questions Those in the cooperation condi-tion answered ldquoIf you were to overestimate how good your partner ishow would it affect your chances of achieving a successful outcomerdquoand ldquoIf you were to underestimate how good your partner is howwould it affect your chances of achieving a successful outcomerdquo on a 1(reduce chance of success) to 7 (increase chance of success) scale Thequestions in the competition condition were the same with the wordldquoopponentrdquo replacing the word ldquopartnerrdquo The presentation order of thescenarios was counterbalanced After they completed the question-naires the participants were debriefed and thanked for their participa-tion

RESULTS AND DISCUSSION

Participantsrsquo responses were submitted to a 2 (goal competition coop-eration) acute 2 (estimation overestimation underestimation) acute 2 (questionorder) mixed ANOVA with the second factor varying within subjectThere was neither a main effect nor interaction associated with questionorder so it was dropped from the analysis It was expected that peoplewould judge that overestimating their partners and underestimatingtheir opponents would reduce their chances of success The results wereconsistent with the hypothesis The only significant effect was the goal acuteestimation interaction F(1 19) = 4532 p lt 001When expecting to coop-erate people judged that their chances of success would be lower if theyoverestimated their partners (M = 200)than if they underestimated theirpartners (M = 410) F(1 19) = 2000 p lt 001 In contrast people judgedthat their chances of success would be higher if they overestimated theiropponents (M = 430) than if they underestimated their opponents (M =260) F(1 19) = 4185p lt 001These findings thus confirm those of Study

416 CHAN AND YBARRA

1a and indicate that people may be inclined to adopt different informa-tion processing strategies depending on their interaction goals

If it really is the case that people when expecting to cooperate arepoised to process and be vigilant about a partnerrsquos negative qualities(underestimation) it would be expected that in a different cognitive taskthose inclinations would also be expressed For example in learningabout a potential partner who is described with negative and positivequalities we should find that people expecting to cooperate are morelikely to elaborate information diagnostic of the partnerrsquos negative thanpositive qualities Similar outcomes should be expected under antici-pated competition If people who expect to compete are inclined to un-cover an opponentrsquos positive qualities (overestimation) then it would beexpected that in a related cognitive task they would be more likely toelaborate information diagnostic of the opponentrsquos positive than nega-tive qualities This is the issue we examined in the second study

STUDY 2

Trait concepts differ in the degree to which they are negativendashdiagnosticor positivendashdiagnostic according to the schematic model of dispositionalattribution (Reeder 1985 Reeder amp Brewer 1979 ) and thecuendashdiagnosticity model of social perception (Skowronski amp Carlston1987 1989) The lay causal theories that people use to understand behav-iors related to negativendashdiagnostic traits (eg moralityndashrelated traits)indicate that negative behaviors are caused by dispositional factors butthat positive behaviors are caused by situational factors or a combina-tion of both (Ybarra amp Stephan 1999 Ybarra 2002) In contrast the laycausal theories people use to understand behaviors related to posi-tivendashdiagnostic traits (eg competencendashrelated traits but also othertraits as we will describe presently) indicate that positive behaviors arecaused by dispositional factors but that negative behaviors are causedby situational factors (Reeder amp Fulks 1980 but see Ybarra 2001 2002for a different perspective)

If a cooperation interaction goal leads people to emphasize a futurepartnerrsquos negative qualities then perceivers should focus on informa-tion related to negativendashdiagnostic traits As a result their informationprocessing (eg attention to and elaboration of information) should beguided by the lay causal theories underlying such traits (negative behav-iors are caused by dispositional causes and positive behaviors are

INTERACTION GOALS AND MEMORY 417

caused by situational causes) (Ybarra amp Stephan 1996 Ybarra ampStephan 1999) Thus people with a cooperation interaction goal shouldbe likely to remember information consistent rather than inconsistentwith the casual theory for the negativendashdiagnostic trait domain (iegood recall for positive behaviors caused by situational factors and neg-ative behaviors caused by dispositional factors) Behaviors related topositivendashdiagnostic traits should be processed less well because this in-formation is of little relevance to the goal of uncovering the partnerrsquosweaknesses Therefore it was expected that perceivers under a coopera-tion goal would not differentiate in memory between theory consistentand inconsistent behavioral information related to positivendashdiagnostictraits1

By comparison if competition leads people to emphasize a future op-ponentrsquos positive qualities then perceivers should focus on informationrelated to positivendashdiagnostic traits As a result their information pro-cessing (eg elaboration of information) should be guided by the laycausal theories underlying such traits (positive behaviors caused bydispositional causes and negative behaviors caused by situationalcauses) Thus people with a competition interaction goal should betterremember information consistent rather than inconsistent with the ca-sual theory for the positivendashdiagnostic trait domain (ie good recall forpositive behaviors caused by dispositional factors and negative valencebehaviors caused by situational factors) Behaviors from the nega-tivendashdiagnostic trait domain should not be well processed because thisinformationis of little relevance to the goal of uncovering the opponentrsquosstrengths Therefore it was expected that perceivers under a competi-tion goal would not show discrimination in their memory between the-ory consistent and inconsistent behavioral information related tonegativendashdiagnostic traits

A relevant question for sake of comparison is what kind of informa-tion would a simple impression formation goal lead perceivers to em-phasize and remember Some research on impression formation

418 CHAN AND YBARRA

1 The strategies of underestimating onersquos partner and overestimating onersquos opponentdo not have to be deliberate and intentional As Study 1b demonstrated even people whowere not consciously thinking about avoiding costs emerged with the same preferences forunderestimating partners and overestimating opponents Once the interaction goals trig-ger a preference for underndash or overestimating the other person the differences in informa-tion processing and memory may occur as a result of peoplersquos lay causal theories guidingthe processing of available information

suggests that people who are forming impressions focus more on the af-fective and general evaluative qualities of others (Levy amp Dugan 1960Ybarra 2001 Zajonc 1980) and that they tend to make such evaluationsbased mainly on negativendashdiagnostic (moralityndashrelated) information(Wojciszke Bazinska amp Jaworski 1998) Thus according to this per-spective people under a general impression formation goal should onlyattend to information related to the negativendashdiagnostic traits similar topeople with a cooperation goal (ie good recall for positive behaviorscaused by situational factors and negative behaviors caused bydispositional factors) Behaviors from the positivendashdiagnostic trait do-main should not be well processed because this information is of littlerelevance to the goal of forming general evaluative impressions There-fore it was expected that perceivers under an impression goal would notdiscriminate in their memory between theory consistent and inconsis-tent behavioral information related to the positivendashdiagnostic traitsThetheoretical model is summarized in Figure 1

Depending on the experimental group to which they were assignedparticipants were told that they would either cooperate or compete withanother participant Then they learned about the other participantrsquos be-haviors which were related to traits in the negativendashdiagnostic trait do-

INTERACTION GOALS AND MEMORY 419

FIGURE 1 Summary of theoretical model

main (hon esty h elpfu lness and fr iendliness ) and thepositivendashdiagnostic trait domain (hardworking image scoring) Imagescoring refers to people engaging in costly positive behaviors to signal toothers that they are valuable community members (Nowak amp Sigmund1998Wedekind amp Milinski 2000) Controlparticipants learned the samematerial but they were asked to form an impression of the person andwere told that they would not interact with the other participant

METHOD

Design and ParticipantsNinetyndashone students participated in the study for course credit Theywere randomly assigned to the cooperation competition or control (im-pression formation) condition All of the participants were presentedwith behaviors with a positive or negative valence from the negativendashdi-agnostic and positivendashdiagnostic trait domains and these behaviorswere linked to either explicit dispositional or situational attributionsThus the overall design of the study was a 3 (interaction goal coopera-tion competition control) acute 2 (trait domain positivendashdiagnostic nega-tivendashdiagnostic) acute 2 (behavior valence positive negative) acute 2(attribution dispositional situational) mixed design with the first fac-tor varying between participants and the latter three factors varyingwithin participants The participants were randomly assigned to condi-tions and were run in nonndashinteracting groups of two to six

Stimulus MaterialsThe behavioral information that the participants processed was pre-sented on a cassette tape allegedly recorded by another participant Thegender of the speaker on the tape and that of the participant were alwaysthe same Each tape contained 20 behavioral statements half of whichwere negative and half positive in valence Also half of the behaviorswere linked to dispositional attributions and half to situational attribu-tions Twelve of the statements pertained to negativendashdiagnostic traits(eg ldquoI entered a concert through the fire exitrdquo [negative valence] ldquoI sentmy sister a present and a birthday bouquetrdquo [positive valence]) Eight ofthe statements pertained to the positivendashdiagnostic traits (eg ldquoI missedmorning lecturesrdquo [negative valence] ldquoI reported a crimerdquo [positive va-lence])

420 CHAN AND YBARRA

Our choice of the traits hardworking and image scoring was deliberatebecause it allows us to show that perceiverrsquos attention to a targetrsquosstrengths (overestimation) is not confounded with emphasizing infor-mation about the targetrsquos competence For example it is usually as-sumed that there is an inverse relationship between the amount of worka person engages in and the presence of the related ability (Darley ampGoethals 1980 Jones 1989) but being hardworking should still signal apositive quality a strength In addition image scoring refers to peopleengaging in costly positive behaviors such as extreme levels of philan-thropy to enhance their reputation and status (Wedekind amp Milinski2000 McAndrew 2002) It is completely unrelated to competence but itsdiagnosticity is in line with that of a positivendashdiagnostic trait

The behavior stems (no attributions) were pretested to ensure thatthey were indeed regarded as positive or negativendashdiagnostic Thirtyparticipants rated each behavior on how common they are For exampleone item queried them as to ldquoWhat percentage of college students doyou think enter a concert through the fire exitrdquo (6ndashpoint scale with 0ndash1010ndash30 30ndash50 50ndash70 70ndash90 and 90ndash100) For positivendashdiagnos-tic traits behaviors that are positive in valence should be less common(and hence more diagnostic of the actorrsquos dispositions) than behaviorsthat are negative in valence whereas for negativendashdiagnostic traits be-haviors negative in valence should be less common (and hence more di-agnostic of the actorrsquos dispositions) than behaviors positive in valenceResults showed an interaction effect that confirmed our classification ofthe behaviors into the respective positive and negativendashdiagnostic cate-gories F(1 29) = 4913 p lt 001 For the positivendashdiagnostic traits posi-tive valence behaviors (M = 281) were judged to be enacted less oftenthan negative valence behaviors (M = 376) F(1 29) = 1961 p lt 001 Forthe negativendashdiagnostic traits negative valence behaviors (M = 294)were judged to be enacted less often than positive valence behaviors (M= 331) F(1 29) = 7103 p = 012

Two sets of materials both containing the same 20 behavior stemswere created Each behavior stem was matched with a dispositional at-tribution in one set and a situational attribution in the other set For ex-ample in set A the behaviors ldquoI didnrsquot help my dad with yard workbecause I wanted to stay inside to watch TVrdquo [negative behavior in nega-tivendashdiagnostic domain with dispositional attribution] ldquoI sent my sistera present and a birthday bouquet because my grandmother asked me todo itrdquo [positive behavior in negativendashdiagnostic domain with situational

INTERACTION GOALS AND MEMORY 421

attribution] ldquoI missed morning lectures because my roommate playedpranks with my alarm clockrdquo [negative behavior in positivendashdiagnosticdomain with situational attribution] and ldquoI worked as a research assis-tant because I am very interested in the subject matterrdquo [positive behav-ior in positive diagnostic domain with dispositional attribution] wouldbe counterbalanced in set B with different causal attributions These in-cluded ldquoI didnrsquot help my dad with yard work because he couldnrsquot findthe keys to the tool shedrdquo [situational attribution] ldquoI sent my sister apresent and a birthday bouquet because I wanted to give her a pleasantsurpriserdquo [dispositional attribution] ldquoI missed morning lecture becauseI donrsquot feel like waking up for themrdquo [dispositional attribution] and ldquoIworked as a research assistant because my professor said I had to do itrdquo[situational attribution] As will be described below the valence of thebehaviors and attributions (dispositional vs external) were manipu-lated independently to ensure that subsequent differences in recall oftheoryndashconsistent or inconsistent information could not simply be dueto an artifact based on the memorability of the particular behaviors

The two sets of stimuli were presented to 17 participants for pretest-ing One participant misunderstood how to use the response scale andwas dropped from the analysis Participants were first presented withthe behavior stems and were asked to rate how favorable each behaviorwas on a 7ndashpoint scale with higher values indicating greaterfavorability Then they were presented with the same behavior stemslinked to either a dispositional or situational attribution and were askedto judge the locus of causation for the behaviors on a 7ndashpoint scale withhigher values indicating more dispositional causation The responseswere submitted to a 2 (valence)acute 2 (attribution)withinndashsubject ANOVAFor the valence judgments the positive behaviors in both stimulus sets(Mset A = 544 Mset B = 559) were judged as more favorable than the nega-tive behaviors (Mset A = 284Mset B = 215) Fset A (1 8)= 3416 p lt 0004Fset B

(1 8) = 9782 p lt 0001 For the second set of judgments a main effect forattribution was obtained for both sets of stimuli Behaviors linked todispositional attributions (Mset A = 575 Mset B = 646) were judged to bemore dispositionally caused than behaviors linked to situational attribu-tions (M set A = 235 Mset B = 298) F set A (1 8) = 4574 p lt 0001 Fset B (1 8) =15826 p lt 0001 No other main effects or interaction effects were ob-tained The two sets of stimuli were presented under three different ran-domization schemes

422 CHAN AND YBARRA

PROCEDURE

Upon arrival at the lab the participants were provided the cover storyfor the study in which they were told that the study would examine howpersonal information would affect subsequent interactions betweenstrangers They were told that another set of participants was recordingsome personal information on a tape for them to listen to To increase thecredibility of the tape contents participants were told that the partici-pants recording the tape had been asked to provide distinct personal in-formation about ldquothings that they have done in the past few weeksrdquoAdditional information about the experiment including whether theywere in the cooperation competition or control (impression formation)condition was given to each participant in a sealed envelope so that theexperimenter could remain blind to the conditions Experimental groupparticipants were told that after they listened to a tape with personal in-formation recorded by another participant they would get a chance tomeet them in either a cooperative or competitive setting Control partici-pants were informed that they had been randomly assigned to the con-trol group and would listen to some information about a previousparticipant instead of having an interaction with them

Participants in the cooperation condition read that they would engagein a ldquoCommunity Simulation Gamerdquo in which they and their assignedpartners ldquowill work together to try to maximize the number of points foryour team The team with the highest points will receive a prizerdquo Thecooperative nature of the interaction was emphasized with repeated useof the word ldquopartnerrdquo In contrastparticipants in the competition condi-tion read that they would engage in a ldquoWall Street Simulation Gamerdquo inwhich they would ldquocompete against [their] competitor to maximize[their] own profit (points) and minimize their opponentrsquos profit Theperson with the highest points will win a prizerdquo The competitive aspectof the upcoming interaction was emphasized further with repeated useof the words ldquocompetitorrdquo and ldquoopponentrdquo This type of reward struc-ture manipulation has been used to study cooperation and competitionby other researchers (Deutsch 1973)

Participants in the control condition read that they would listen to atape that had been recorded in a previous experimental session Theywere told to form an impression of the person using the informationfrom the tape as part of a ldquostandardizationrdquo procedure for the experi-ment To bolster the believability of the cover story participants then

INTERACTION GOALS AND MEMORY 423

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

Participants then answered 2 questions Those in the cooperation condi-tion answered ldquoHow costly is it for Person X to mistakenly overestimatehow good the partner isrdquo and ldquoHow costly is it for Person X to mistak-enly underestimate how good the partner isrdquo on a 1 (not at all costly) to7 (extremely costly) scale Those in the competition condition answeredthe same items with the word ldquocompetitorrdquo replacing the word ldquopart-nerrdquo The presentation order of the scenarios was counterbalanced Af-ter they completed the questionnaires the participants were debriefedand thanked for their participation

RESULTS AND DISCUSSION

Participantsrsquo responses were submitted to a 2 (goal competition coop-eration) acute 2 (estimation overestimation underestimation) acute 2 (questionorder) mixed ANOVA with the second factor varying within subjectThere was no main effect or interaction associated with question orderand thus the order factorwas dropped from the analysis It was expectedthat overestimating how good onersquos partner is and underestimatinghow good onersquos opponent is would be deemed undesirable and this wasindeed the case The only significant effect was the goal acute estimation in-teraction F(1 16) = 2918 p lt 001 When expecting to cooperate peoplejudged that it was more costly to overestimate (M = 556) than underesti-mate (M = 289)onersquos partner F(1 16) = 3011p lt 001 In contrastpeoplejudged that it was more costly to underestimate (M = 574) than overesti-mate (M = 294) onersquos opponent F(1 16) = 1580 p = 001

The findings from Study 1a suggest that for cooperation people deemit preferable to avoid overestimating how good the partner is and forcompetition to avoid underestimating how good the opponent is Onecould argue that these results were obtained because participants wereexplicitly instructed to consider how costly it was to overestimate or un-derestimate the other person By focusing peoplersquos attention on the cost-liness of their decisions we might have led the participants to be morelossndashfocused than they would have been otherwise Another feature ofthis study that requires further examination is whether asking partici-pants to imagine the scenario from anotherrsquos perspective instead ofimagining themselves in the scenario might have affected the resultsMuch social psychological research has found differences between peo-plersquos psychological processes when they are thinking about themselvesversus others For example people tend to make situational causal attri-

414 CHAN AND YBARRA

butions for themselves but dispositional causal attributions for others(Jones amp Nisbett 1972) People pay more attention to unobservable andunintentional events when thinking about themselves but observableand intentional events when thinking about others (Malle amp Pearce2001) Peoplersquos representations of themselves in memory are character-ized by their own thoughts and feelings whereas their representationsof others are characterized by actions and appearances (McGuire ampMcGuire 1986 Prentice 1990) Although it is not perfectly clear how itis possible that the findings from Study 1a might not occur if participantswere asked to think of how they themselves would react To rule outthese alternative accounts we conducted Study 1b to replicate the re-sults of Study 1a

STUDY 1B

The current study was identical to Study 1a save for two features Firstwe asked the participants to judge how overestimating or underestimat-ing the other person would affect their chances of success instead of ex-plicitly pointing to the costliness of the decision as in Study 1a Secondwe asked participants to imagine themselves instead of others in scenar-ios where they either had to cooperate or compete with another person

METHOD

Design and ParticipantsTwenty students between 19ndash25 years old participated in this studyThey were approached on a university campus and asked to fill out ashort questionnaire

Materials and ProcedureParticipants were informed that the study was a short psychology ques-tionnaire dealing with interpersonal perception They were asked toimagine themselves in the following scenarios

In the cooperation condition the scenario read

You are about to meet Person X and both of you will have to work togetherYou have a few moments to find out more about Person X After finding outmore about what Person X is like you will have to decide what strategy to

INTERACTION GOALS AND MEMORY 415

take during the cooperation in order to achieve the most successful out-come

In the competition condition the scenario read

You are about to meet Person X and both of you will have to compete againsteach other You will have a few moments to find out more about Person XAfter finding out more about what Person X is like you will have to decidewhat strategy to take during the competition in order to achieve the mostsuccessful outcome

Participants then answered 2 questions Those in the cooperation condi-tion answered ldquoIf you were to overestimate how good your partner ishow would it affect your chances of achieving a successful outcomerdquoand ldquoIf you were to underestimate how good your partner is howwould it affect your chances of achieving a successful outcomerdquo on a 1(reduce chance of success) to 7 (increase chance of success) scale Thequestions in the competition condition were the same with the wordldquoopponentrdquo replacing the word ldquopartnerrdquo The presentation order of thescenarios was counterbalanced After they completed the question-naires the participants were debriefed and thanked for their participa-tion

RESULTS AND DISCUSSION

Participantsrsquo responses were submitted to a 2 (goal competition coop-eration) acute 2 (estimation overestimation underestimation) acute 2 (questionorder) mixed ANOVA with the second factor varying within subjectThere was neither a main effect nor interaction associated with questionorder so it was dropped from the analysis It was expected that peoplewould judge that overestimating their partners and underestimatingtheir opponents would reduce their chances of success The results wereconsistent with the hypothesis The only significant effect was the goal acuteestimation interaction F(1 19) = 4532 p lt 001When expecting to coop-erate people judged that their chances of success would be lower if theyoverestimated their partners (M = 200)than if they underestimated theirpartners (M = 410) F(1 19) = 2000 p lt 001 In contrast people judgedthat their chances of success would be higher if they overestimated theiropponents (M = 430) than if they underestimated their opponents (M =260) F(1 19) = 4185p lt 001These findings thus confirm those of Study

416 CHAN AND YBARRA

1a and indicate that people may be inclined to adopt different informa-tion processing strategies depending on their interaction goals

If it really is the case that people when expecting to cooperate arepoised to process and be vigilant about a partnerrsquos negative qualities(underestimation) it would be expected that in a different cognitive taskthose inclinations would also be expressed For example in learningabout a potential partner who is described with negative and positivequalities we should find that people expecting to cooperate are morelikely to elaborate information diagnostic of the partnerrsquos negative thanpositive qualities Similar outcomes should be expected under antici-pated competition If people who expect to compete are inclined to un-cover an opponentrsquos positive qualities (overestimation) then it would beexpected that in a related cognitive task they would be more likely toelaborate information diagnostic of the opponentrsquos positive than nega-tive qualities This is the issue we examined in the second study

STUDY 2

Trait concepts differ in the degree to which they are negativendashdiagnosticor positivendashdiagnostic according to the schematic model of dispositionalattribution (Reeder 1985 Reeder amp Brewer 1979 ) and thecuendashdiagnosticity model of social perception (Skowronski amp Carlston1987 1989) The lay causal theories that people use to understand behav-iors related to negativendashdiagnostic traits (eg moralityndashrelated traits)indicate that negative behaviors are caused by dispositional factors butthat positive behaviors are caused by situational factors or a combina-tion of both (Ybarra amp Stephan 1999 Ybarra 2002) In contrast the laycausal theories people use to understand behaviors related to posi-tivendashdiagnostic traits (eg competencendashrelated traits but also othertraits as we will describe presently) indicate that positive behaviors arecaused by dispositional factors but that negative behaviors are causedby situational factors (Reeder amp Fulks 1980 but see Ybarra 2001 2002for a different perspective)

If a cooperation interaction goal leads people to emphasize a futurepartnerrsquos negative qualities then perceivers should focus on informa-tion related to negativendashdiagnostic traits As a result their informationprocessing (eg attention to and elaboration of information) should beguided by the lay causal theories underlying such traits (negative behav-iors are caused by dispositional causes and positive behaviors are

INTERACTION GOALS AND MEMORY 417

caused by situational causes) (Ybarra amp Stephan 1996 Ybarra ampStephan 1999) Thus people with a cooperation interaction goal shouldbe likely to remember information consistent rather than inconsistentwith the casual theory for the negativendashdiagnostic trait domain (iegood recall for positive behaviors caused by situational factors and neg-ative behaviors caused by dispositional factors) Behaviors related topositivendashdiagnostic traits should be processed less well because this in-formation is of little relevance to the goal of uncovering the partnerrsquosweaknesses Therefore it was expected that perceivers under a coopera-tion goal would not differentiate in memory between theory consistentand inconsistent behavioral information related to positivendashdiagnostictraits1

By comparison if competition leads people to emphasize a future op-ponentrsquos positive qualities then perceivers should focus on informationrelated to positivendashdiagnostic traits As a result their information pro-cessing (eg elaboration of information) should be guided by the laycausal theories underlying such traits (positive behaviors caused bydispositional causes and negative behaviors caused by situationalcauses) Thus people with a competition interaction goal should betterremember information consistent rather than inconsistent with the ca-sual theory for the positivendashdiagnostic trait domain (ie good recall forpositive behaviors caused by dispositional factors and negative valencebehaviors caused by situational factors) Behaviors from the nega-tivendashdiagnostic trait domain should not be well processed because thisinformationis of little relevance to the goal of uncovering the opponentrsquosstrengths Therefore it was expected that perceivers under a competi-tion goal would not show discrimination in their memory between the-ory consistent and inconsistent behavioral information related tonegativendashdiagnostic traits

A relevant question for sake of comparison is what kind of informa-tion would a simple impression formation goal lead perceivers to em-phasize and remember Some research on impression formation

418 CHAN AND YBARRA

1 The strategies of underestimating onersquos partner and overestimating onersquos opponentdo not have to be deliberate and intentional As Study 1b demonstrated even people whowere not consciously thinking about avoiding costs emerged with the same preferences forunderestimating partners and overestimating opponents Once the interaction goals trig-ger a preference for underndash or overestimating the other person the differences in informa-tion processing and memory may occur as a result of peoplersquos lay causal theories guidingthe processing of available information

suggests that people who are forming impressions focus more on the af-fective and general evaluative qualities of others (Levy amp Dugan 1960Ybarra 2001 Zajonc 1980) and that they tend to make such evaluationsbased mainly on negativendashdiagnostic (moralityndashrelated) information(Wojciszke Bazinska amp Jaworski 1998) Thus according to this per-spective people under a general impression formation goal should onlyattend to information related to the negativendashdiagnostic traits similar topeople with a cooperation goal (ie good recall for positive behaviorscaused by situational factors and negative behaviors caused bydispositional factors) Behaviors from the positivendashdiagnostic trait do-main should not be well processed because this information is of littlerelevance to the goal of forming general evaluative impressions There-fore it was expected that perceivers under an impression goal would notdiscriminate in their memory between theory consistent and inconsis-tent behavioral information related to the positivendashdiagnostic traitsThetheoretical model is summarized in Figure 1

Depending on the experimental group to which they were assignedparticipants were told that they would either cooperate or compete withanother participant Then they learned about the other participantrsquos be-haviors which were related to traits in the negativendashdiagnostic trait do-

INTERACTION GOALS AND MEMORY 419

FIGURE 1 Summary of theoretical model

main (hon esty h elpfu lness and fr iendliness ) and thepositivendashdiagnostic trait domain (hardworking image scoring) Imagescoring refers to people engaging in costly positive behaviors to signal toothers that they are valuable community members (Nowak amp Sigmund1998Wedekind amp Milinski 2000) Controlparticipants learned the samematerial but they were asked to form an impression of the person andwere told that they would not interact with the other participant

METHOD

Design and ParticipantsNinetyndashone students participated in the study for course credit Theywere randomly assigned to the cooperation competition or control (im-pression formation) condition All of the participants were presentedwith behaviors with a positive or negative valence from the negativendashdi-agnostic and positivendashdiagnostic trait domains and these behaviorswere linked to either explicit dispositional or situational attributionsThus the overall design of the study was a 3 (interaction goal coopera-tion competition control) acute 2 (trait domain positivendashdiagnostic nega-tivendashdiagnostic) acute 2 (behavior valence positive negative) acute 2(attribution dispositional situational) mixed design with the first fac-tor varying between participants and the latter three factors varyingwithin participants The participants were randomly assigned to condi-tions and were run in nonndashinteracting groups of two to six

Stimulus MaterialsThe behavioral information that the participants processed was pre-sented on a cassette tape allegedly recorded by another participant Thegender of the speaker on the tape and that of the participant were alwaysthe same Each tape contained 20 behavioral statements half of whichwere negative and half positive in valence Also half of the behaviorswere linked to dispositional attributions and half to situational attribu-tions Twelve of the statements pertained to negativendashdiagnostic traits(eg ldquoI entered a concert through the fire exitrdquo [negative valence] ldquoI sentmy sister a present and a birthday bouquetrdquo [positive valence]) Eight ofthe statements pertained to the positivendashdiagnostic traits (eg ldquoI missedmorning lecturesrdquo [negative valence] ldquoI reported a crimerdquo [positive va-lence])

420 CHAN AND YBARRA

Our choice of the traits hardworking and image scoring was deliberatebecause it allows us to show that perceiverrsquos attention to a targetrsquosstrengths (overestimation) is not confounded with emphasizing infor-mation about the targetrsquos competence For example it is usually as-sumed that there is an inverse relationship between the amount of worka person engages in and the presence of the related ability (Darley ampGoethals 1980 Jones 1989) but being hardworking should still signal apositive quality a strength In addition image scoring refers to peopleengaging in costly positive behaviors such as extreme levels of philan-thropy to enhance their reputation and status (Wedekind amp Milinski2000 McAndrew 2002) It is completely unrelated to competence but itsdiagnosticity is in line with that of a positivendashdiagnostic trait

The behavior stems (no attributions) were pretested to ensure thatthey were indeed regarded as positive or negativendashdiagnostic Thirtyparticipants rated each behavior on how common they are For exampleone item queried them as to ldquoWhat percentage of college students doyou think enter a concert through the fire exitrdquo (6ndashpoint scale with 0ndash1010ndash30 30ndash50 50ndash70 70ndash90 and 90ndash100) For positivendashdiagnos-tic traits behaviors that are positive in valence should be less common(and hence more diagnostic of the actorrsquos dispositions) than behaviorsthat are negative in valence whereas for negativendashdiagnostic traits be-haviors negative in valence should be less common (and hence more di-agnostic of the actorrsquos dispositions) than behaviors positive in valenceResults showed an interaction effect that confirmed our classification ofthe behaviors into the respective positive and negativendashdiagnostic cate-gories F(1 29) = 4913 p lt 001 For the positivendashdiagnostic traits posi-tive valence behaviors (M = 281) were judged to be enacted less oftenthan negative valence behaviors (M = 376) F(1 29) = 1961 p lt 001 Forthe negativendashdiagnostic traits negative valence behaviors (M = 294)were judged to be enacted less often than positive valence behaviors (M= 331) F(1 29) = 7103 p = 012

Two sets of materials both containing the same 20 behavior stemswere created Each behavior stem was matched with a dispositional at-tribution in one set and a situational attribution in the other set For ex-ample in set A the behaviors ldquoI didnrsquot help my dad with yard workbecause I wanted to stay inside to watch TVrdquo [negative behavior in nega-tivendashdiagnostic domain with dispositional attribution] ldquoI sent my sistera present and a birthday bouquet because my grandmother asked me todo itrdquo [positive behavior in negativendashdiagnostic domain with situational

INTERACTION GOALS AND MEMORY 421

attribution] ldquoI missed morning lectures because my roommate playedpranks with my alarm clockrdquo [negative behavior in positivendashdiagnosticdomain with situational attribution] and ldquoI worked as a research assis-tant because I am very interested in the subject matterrdquo [positive behav-ior in positive diagnostic domain with dispositional attribution] wouldbe counterbalanced in set B with different causal attributions These in-cluded ldquoI didnrsquot help my dad with yard work because he couldnrsquot findthe keys to the tool shedrdquo [situational attribution] ldquoI sent my sister apresent and a birthday bouquet because I wanted to give her a pleasantsurpriserdquo [dispositional attribution] ldquoI missed morning lecture becauseI donrsquot feel like waking up for themrdquo [dispositional attribution] and ldquoIworked as a research assistant because my professor said I had to do itrdquo[situational attribution] As will be described below the valence of thebehaviors and attributions (dispositional vs external) were manipu-lated independently to ensure that subsequent differences in recall oftheoryndashconsistent or inconsistent information could not simply be dueto an artifact based on the memorability of the particular behaviors

The two sets of stimuli were presented to 17 participants for pretest-ing One participant misunderstood how to use the response scale andwas dropped from the analysis Participants were first presented withthe behavior stems and were asked to rate how favorable each behaviorwas on a 7ndashpoint scale with higher values indicating greaterfavorability Then they were presented with the same behavior stemslinked to either a dispositional or situational attribution and were askedto judge the locus of causation for the behaviors on a 7ndashpoint scale withhigher values indicating more dispositional causation The responseswere submitted to a 2 (valence)acute 2 (attribution)withinndashsubject ANOVAFor the valence judgments the positive behaviors in both stimulus sets(Mset A = 544 Mset B = 559) were judged as more favorable than the nega-tive behaviors (Mset A = 284Mset B = 215) Fset A (1 8)= 3416 p lt 0004Fset B

(1 8) = 9782 p lt 0001 For the second set of judgments a main effect forattribution was obtained for both sets of stimuli Behaviors linked todispositional attributions (Mset A = 575 Mset B = 646) were judged to bemore dispositionally caused than behaviors linked to situational attribu-tions (M set A = 235 Mset B = 298) F set A (1 8) = 4574 p lt 0001 Fset B (1 8) =15826 p lt 0001 No other main effects or interaction effects were ob-tained The two sets of stimuli were presented under three different ran-domization schemes

422 CHAN AND YBARRA

PROCEDURE

Upon arrival at the lab the participants were provided the cover storyfor the study in which they were told that the study would examine howpersonal information would affect subsequent interactions betweenstrangers They were told that another set of participants was recordingsome personal information on a tape for them to listen to To increase thecredibility of the tape contents participants were told that the partici-pants recording the tape had been asked to provide distinct personal in-formation about ldquothings that they have done in the past few weeksrdquoAdditional information about the experiment including whether theywere in the cooperation competition or control (impression formation)condition was given to each participant in a sealed envelope so that theexperimenter could remain blind to the conditions Experimental groupparticipants were told that after they listened to a tape with personal in-formation recorded by another participant they would get a chance tomeet them in either a cooperative or competitive setting Control partici-pants were informed that they had been randomly assigned to the con-trol group and would listen to some information about a previousparticipant instead of having an interaction with them

Participants in the cooperation condition read that they would engagein a ldquoCommunity Simulation Gamerdquo in which they and their assignedpartners ldquowill work together to try to maximize the number of points foryour team The team with the highest points will receive a prizerdquo Thecooperative nature of the interaction was emphasized with repeated useof the word ldquopartnerrdquo In contrastparticipants in the competition condi-tion read that they would engage in a ldquoWall Street Simulation Gamerdquo inwhich they would ldquocompete against [their] competitor to maximize[their] own profit (points) and minimize their opponentrsquos profit Theperson with the highest points will win a prizerdquo The competitive aspectof the upcoming interaction was emphasized further with repeated useof the words ldquocompetitorrdquo and ldquoopponentrdquo This type of reward struc-ture manipulation has been used to study cooperation and competitionby other researchers (Deutsch 1973)

Participants in the control condition read that they would listen to atape that had been recorded in a previous experimental session Theywere told to form an impression of the person using the informationfrom the tape as part of a ldquostandardizationrdquo procedure for the experi-ment To bolster the believability of the cover story participants then

INTERACTION GOALS AND MEMORY 423

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

butions for themselves but dispositional causal attributions for others(Jones amp Nisbett 1972) People pay more attention to unobservable andunintentional events when thinking about themselves but observableand intentional events when thinking about others (Malle amp Pearce2001) Peoplersquos representations of themselves in memory are character-ized by their own thoughts and feelings whereas their representationsof others are characterized by actions and appearances (McGuire ampMcGuire 1986 Prentice 1990) Although it is not perfectly clear how itis possible that the findings from Study 1a might not occur if participantswere asked to think of how they themselves would react To rule outthese alternative accounts we conducted Study 1b to replicate the re-sults of Study 1a

STUDY 1B

The current study was identical to Study 1a save for two features Firstwe asked the participants to judge how overestimating or underestimat-ing the other person would affect their chances of success instead of ex-plicitly pointing to the costliness of the decision as in Study 1a Secondwe asked participants to imagine themselves instead of others in scenar-ios where they either had to cooperate or compete with another person

METHOD

Design and ParticipantsTwenty students between 19ndash25 years old participated in this studyThey were approached on a university campus and asked to fill out ashort questionnaire

Materials and ProcedureParticipants were informed that the study was a short psychology ques-tionnaire dealing with interpersonal perception They were asked toimagine themselves in the following scenarios

In the cooperation condition the scenario read

You are about to meet Person X and both of you will have to work togetherYou have a few moments to find out more about Person X After finding outmore about what Person X is like you will have to decide what strategy to

INTERACTION GOALS AND MEMORY 415

take during the cooperation in order to achieve the most successful out-come

In the competition condition the scenario read

You are about to meet Person X and both of you will have to compete againsteach other You will have a few moments to find out more about Person XAfter finding out more about what Person X is like you will have to decidewhat strategy to take during the competition in order to achieve the mostsuccessful outcome

Participants then answered 2 questions Those in the cooperation condi-tion answered ldquoIf you were to overestimate how good your partner ishow would it affect your chances of achieving a successful outcomerdquoand ldquoIf you were to underestimate how good your partner is howwould it affect your chances of achieving a successful outcomerdquo on a 1(reduce chance of success) to 7 (increase chance of success) scale Thequestions in the competition condition were the same with the wordldquoopponentrdquo replacing the word ldquopartnerrdquo The presentation order of thescenarios was counterbalanced After they completed the question-naires the participants were debriefed and thanked for their participa-tion

RESULTS AND DISCUSSION

Participantsrsquo responses were submitted to a 2 (goal competition coop-eration) acute 2 (estimation overestimation underestimation) acute 2 (questionorder) mixed ANOVA with the second factor varying within subjectThere was neither a main effect nor interaction associated with questionorder so it was dropped from the analysis It was expected that peoplewould judge that overestimating their partners and underestimatingtheir opponents would reduce their chances of success The results wereconsistent with the hypothesis The only significant effect was the goal acuteestimation interaction F(1 19) = 4532 p lt 001When expecting to coop-erate people judged that their chances of success would be lower if theyoverestimated their partners (M = 200)than if they underestimated theirpartners (M = 410) F(1 19) = 2000 p lt 001 In contrast people judgedthat their chances of success would be higher if they overestimated theiropponents (M = 430) than if they underestimated their opponents (M =260) F(1 19) = 4185p lt 001These findings thus confirm those of Study

416 CHAN AND YBARRA

1a and indicate that people may be inclined to adopt different informa-tion processing strategies depending on their interaction goals

If it really is the case that people when expecting to cooperate arepoised to process and be vigilant about a partnerrsquos negative qualities(underestimation) it would be expected that in a different cognitive taskthose inclinations would also be expressed For example in learningabout a potential partner who is described with negative and positivequalities we should find that people expecting to cooperate are morelikely to elaborate information diagnostic of the partnerrsquos negative thanpositive qualities Similar outcomes should be expected under antici-pated competition If people who expect to compete are inclined to un-cover an opponentrsquos positive qualities (overestimation) then it would beexpected that in a related cognitive task they would be more likely toelaborate information diagnostic of the opponentrsquos positive than nega-tive qualities This is the issue we examined in the second study

STUDY 2

Trait concepts differ in the degree to which they are negativendashdiagnosticor positivendashdiagnostic according to the schematic model of dispositionalattribution (Reeder 1985 Reeder amp Brewer 1979 ) and thecuendashdiagnosticity model of social perception (Skowronski amp Carlston1987 1989) The lay causal theories that people use to understand behav-iors related to negativendashdiagnostic traits (eg moralityndashrelated traits)indicate that negative behaviors are caused by dispositional factors butthat positive behaviors are caused by situational factors or a combina-tion of both (Ybarra amp Stephan 1999 Ybarra 2002) In contrast the laycausal theories people use to understand behaviors related to posi-tivendashdiagnostic traits (eg competencendashrelated traits but also othertraits as we will describe presently) indicate that positive behaviors arecaused by dispositional factors but that negative behaviors are causedby situational factors (Reeder amp Fulks 1980 but see Ybarra 2001 2002for a different perspective)

If a cooperation interaction goal leads people to emphasize a futurepartnerrsquos negative qualities then perceivers should focus on informa-tion related to negativendashdiagnostic traits As a result their informationprocessing (eg attention to and elaboration of information) should beguided by the lay causal theories underlying such traits (negative behav-iors are caused by dispositional causes and positive behaviors are

INTERACTION GOALS AND MEMORY 417

caused by situational causes) (Ybarra amp Stephan 1996 Ybarra ampStephan 1999) Thus people with a cooperation interaction goal shouldbe likely to remember information consistent rather than inconsistentwith the casual theory for the negativendashdiagnostic trait domain (iegood recall for positive behaviors caused by situational factors and neg-ative behaviors caused by dispositional factors) Behaviors related topositivendashdiagnostic traits should be processed less well because this in-formation is of little relevance to the goal of uncovering the partnerrsquosweaknesses Therefore it was expected that perceivers under a coopera-tion goal would not differentiate in memory between theory consistentand inconsistent behavioral information related to positivendashdiagnostictraits1

By comparison if competition leads people to emphasize a future op-ponentrsquos positive qualities then perceivers should focus on informationrelated to positivendashdiagnostic traits As a result their information pro-cessing (eg elaboration of information) should be guided by the laycausal theories underlying such traits (positive behaviors caused bydispositional causes and negative behaviors caused by situationalcauses) Thus people with a competition interaction goal should betterremember information consistent rather than inconsistent with the ca-sual theory for the positivendashdiagnostic trait domain (ie good recall forpositive behaviors caused by dispositional factors and negative valencebehaviors caused by situational factors) Behaviors from the nega-tivendashdiagnostic trait domain should not be well processed because thisinformationis of little relevance to the goal of uncovering the opponentrsquosstrengths Therefore it was expected that perceivers under a competi-tion goal would not show discrimination in their memory between the-ory consistent and inconsistent behavioral information related tonegativendashdiagnostic traits

A relevant question for sake of comparison is what kind of informa-tion would a simple impression formation goal lead perceivers to em-phasize and remember Some research on impression formation

418 CHAN AND YBARRA

1 The strategies of underestimating onersquos partner and overestimating onersquos opponentdo not have to be deliberate and intentional As Study 1b demonstrated even people whowere not consciously thinking about avoiding costs emerged with the same preferences forunderestimating partners and overestimating opponents Once the interaction goals trig-ger a preference for underndash or overestimating the other person the differences in informa-tion processing and memory may occur as a result of peoplersquos lay causal theories guidingthe processing of available information

suggests that people who are forming impressions focus more on the af-fective and general evaluative qualities of others (Levy amp Dugan 1960Ybarra 2001 Zajonc 1980) and that they tend to make such evaluationsbased mainly on negativendashdiagnostic (moralityndashrelated) information(Wojciszke Bazinska amp Jaworski 1998) Thus according to this per-spective people under a general impression formation goal should onlyattend to information related to the negativendashdiagnostic traits similar topeople with a cooperation goal (ie good recall for positive behaviorscaused by situational factors and negative behaviors caused bydispositional factors) Behaviors from the positivendashdiagnostic trait do-main should not be well processed because this information is of littlerelevance to the goal of forming general evaluative impressions There-fore it was expected that perceivers under an impression goal would notdiscriminate in their memory between theory consistent and inconsis-tent behavioral information related to the positivendashdiagnostic traitsThetheoretical model is summarized in Figure 1

Depending on the experimental group to which they were assignedparticipants were told that they would either cooperate or compete withanother participant Then they learned about the other participantrsquos be-haviors which were related to traits in the negativendashdiagnostic trait do-

INTERACTION GOALS AND MEMORY 419

FIGURE 1 Summary of theoretical model

main (hon esty h elpfu lness and fr iendliness ) and thepositivendashdiagnostic trait domain (hardworking image scoring) Imagescoring refers to people engaging in costly positive behaviors to signal toothers that they are valuable community members (Nowak amp Sigmund1998Wedekind amp Milinski 2000) Controlparticipants learned the samematerial but they were asked to form an impression of the person andwere told that they would not interact with the other participant

METHOD

Design and ParticipantsNinetyndashone students participated in the study for course credit Theywere randomly assigned to the cooperation competition or control (im-pression formation) condition All of the participants were presentedwith behaviors with a positive or negative valence from the negativendashdi-agnostic and positivendashdiagnostic trait domains and these behaviorswere linked to either explicit dispositional or situational attributionsThus the overall design of the study was a 3 (interaction goal coopera-tion competition control) acute 2 (trait domain positivendashdiagnostic nega-tivendashdiagnostic) acute 2 (behavior valence positive negative) acute 2(attribution dispositional situational) mixed design with the first fac-tor varying between participants and the latter three factors varyingwithin participants The participants were randomly assigned to condi-tions and were run in nonndashinteracting groups of two to six

Stimulus MaterialsThe behavioral information that the participants processed was pre-sented on a cassette tape allegedly recorded by another participant Thegender of the speaker on the tape and that of the participant were alwaysthe same Each tape contained 20 behavioral statements half of whichwere negative and half positive in valence Also half of the behaviorswere linked to dispositional attributions and half to situational attribu-tions Twelve of the statements pertained to negativendashdiagnostic traits(eg ldquoI entered a concert through the fire exitrdquo [negative valence] ldquoI sentmy sister a present and a birthday bouquetrdquo [positive valence]) Eight ofthe statements pertained to the positivendashdiagnostic traits (eg ldquoI missedmorning lecturesrdquo [negative valence] ldquoI reported a crimerdquo [positive va-lence])

420 CHAN AND YBARRA

Our choice of the traits hardworking and image scoring was deliberatebecause it allows us to show that perceiverrsquos attention to a targetrsquosstrengths (overestimation) is not confounded with emphasizing infor-mation about the targetrsquos competence For example it is usually as-sumed that there is an inverse relationship between the amount of worka person engages in and the presence of the related ability (Darley ampGoethals 1980 Jones 1989) but being hardworking should still signal apositive quality a strength In addition image scoring refers to peopleengaging in costly positive behaviors such as extreme levels of philan-thropy to enhance their reputation and status (Wedekind amp Milinski2000 McAndrew 2002) It is completely unrelated to competence but itsdiagnosticity is in line with that of a positivendashdiagnostic trait

The behavior stems (no attributions) were pretested to ensure thatthey were indeed regarded as positive or negativendashdiagnostic Thirtyparticipants rated each behavior on how common they are For exampleone item queried them as to ldquoWhat percentage of college students doyou think enter a concert through the fire exitrdquo (6ndashpoint scale with 0ndash1010ndash30 30ndash50 50ndash70 70ndash90 and 90ndash100) For positivendashdiagnos-tic traits behaviors that are positive in valence should be less common(and hence more diagnostic of the actorrsquos dispositions) than behaviorsthat are negative in valence whereas for negativendashdiagnostic traits be-haviors negative in valence should be less common (and hence more di-agnostic of the actorrsquos dispositions) than behaviors positive in valenceResults showed an interaction effect that confirmed our classification ofthe behaviors into the respective positive and negativendashdiagnostic cate-gories F(1 29) = 4913 p lt 001 For the positivendashdiagnostic traits posi-tive valence behaviors (M = 281) were judged to be enacted less oftenthan negative valence behaviors (M = 376) F(1 29) = 1961 p lt 001 Forthe negativendashdiagnostic traits negative valence behaviors (M = 294)were judged to be enacted less often than positive valence behaviors (M= 331) F(1 29) = 7103 p = 012

Two sets of materials both containing the same 20 behavior stemswere created Each behavior stem was matched with a dispositional at-tribution in one set and a situational attribution in the other set For ex-ample in set A the behaviors ldquoI didnrsquot help my dad with yard workbecause I wanted to stay inside to watch TVrdquo [negative behavior in nega-tivendashdiagnostic domain with dispositional attribution] ldquoI sent my sistera present and a birthday bouquet because my grandmother asked me todo itrdquo [positive behavior in negativendashdiagnostic domain with situational

INTERACTION GOALS AND MEMORY 421

attribution] ldquoI missed morning lectures because my roommate playedpranks with my alarm clockrdquo [negative behavior in positivendashdiagnosticdomain with situational attribution] and ldquoI worked as a research assis-tant because I am very interested in the subject matterrdquo [positive behav-ior in positive diagnostic domain with dispositional attribution] wouldbe counterbalanced in set B with different causal attributions These in-cluded ldquoI didnrsquot help my dad with yard work because he couldnrsquot findthe keys to the tool shedrdquo [situational attribution] ldquoI sent my sister apresent and a birthday bouquet because I wanted to give her a pleasantsurpriserdquo [dispositional attribution] ldquoI missed morning lecture becauseI donrsquot feel like waking up for themrdquo [dispositional attribution] and ldquoIworked as a research assistant because my professor said I had to do itrdquo[situational attribution] As will be described below the valence of thebehaviors and attributions (dispositional vs external) were manipu-lated independently to ensure that subsequent differences in recall oftheoryndashconsistent or inconsistent information could not simply be dueto an artifact based on the memorability of the particular behaviors

The two sets of stimuli were presented to 17 participants for pretest-ing One participant misunderstood how to use the response scale andwas dropped from the analysis Participants were first presented withthe behavior stems and were asked to rate how favorable each behaviorwas on a 7ndashpoint scale with higher values indicating greaterfavorability Then they were presented with the same behavior stemslinked to either a dispositional or situational attribution and were askedto judge the locus of causation for the behaviors on a 7ndashpoint scale withhigher values indicating more dispositional causation The responseswere submitted to a 2 (valence)acute 2 (attribution)withinndashsubject ANOVAFor the valence judgments the positive behaviors in both stimulus sets(Mset A = 544 Mset B = 559) were judged as more favorable than the nega-tive behaviors (Mset A = 284Mset B = 215) Fset A (1 8)= 3416 p lt 0004Fset B

(1 8) = 9782 p lt 0001 For the second set of judgments a main effect forattribution was obtained for both sets of stimuli Behaviors linked todispositional attributions (Mset A = 575 Mset B = 646) were judged to bemore dispositionally caused than behaviors linked to situational attribu-tions (M set A = 235 Mset B = 298) F set A (1 8) = 4574 p lt 0001 Fset B (1 8) =15826 p lt 0001 No other main effects or interaction effects were ob-tained The two sets of stimuli were presented under three different ran-domization schemes

422 CHAN AND YBARRA

PROCEDURE

Upon arrival at the lab the participants were provided the cover storyfor the study in which they were told that the study would examine howpersonal information would affect subsequent interactions betweenstrangers They were told that another set of participants was recordingsome personal information on a tape for them to listen to To increase thecredibility of the tape contents participants were told that the partici-pants recording the tape had been asked to provide distinct personal in-formation about ldquothings that they have done in the past few weeksrdquoAdditional information about the experiment including whether theywere in the cooperation competition or control (impression formation)condition was given to each participant in a sealed envelope so that theexperimenter could remain blind to the conditions Experimental groupparticipants were told that after they listened to a tape with personal in-formation recorded by another participant they would get a chance tomeet them in either a cooperative or competitive setting Control partici-pants were informed that they had been randomly assigned to the con-trol group and would listen to some information about a previousparticipant instead of having an interaction with them

Participants in the cooperation condition read that they would engagein a ldquoCommunity Simulation Gamerdquo in which they and their assignedpartners ldquowill work together to try to maximize the number of points foryour team The team with the highest points will receive a prizerdquo Thecooperative nature of the interaction was emphasized with repeated useof the word ldquopartnerrdquo In contrastparticipants in the competition condi-tion read that they would engage in a ldquoWall Street Simulation Gamerdquo inwhich they would ldquocompete against [their] competitor to maximize[their] own profit (points) and minimize their opponentrsquos profit Theperson with the highest points will win a prizerdquo The competitive aspectof the upcoming interaction was emphasized further with repeated useof the words ldquocompetitorrdquo and ldquoopponentrdquo This type of reward struc-ture manipulation has been used to study cooperation and competitionby other researchers (Deutsch 1973)

Participants in the control condition read that they would listen to atape that had been recorded in a previous experimental session Theywere told to form an impression of the person using the informationfrom the tape as part of a ldquostandardizationrdquo procedure for the experi-ment To bolster the believability of the cover story participants then

INTERACTION GOALS AND MEMORY 423

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

take during the cooperation in order to achieve the most successful out-come

In the competition condition the scenario read

You are about to meet Person X and both of you will have to compete againsteach other You will have a few moments to find out more about Person XAfter finding out more about what Person X is like you will have to decidewhat strategy to take during the competition in order to achieve the mostsuccessful outcome

Participants then answered 2 questions Those in the cooperation condi-tion answered ldquoIf you were to overestimate how good your partner ishow would it affect your chances of achieving a successful outcomerdquoand ldquoIf you were to underestimate how good your partner is howwould it affect your chances of achieving a successful outcomerdquo on a 1(reduce chance of success) to 7 (increase chance of success) scale Thequestions in the competition condition were the same with the wordldquoopponentrdquo replacing the word ldquopartnerrdquo The presentation order of thescenarios was counterbalanced After they completed the question-naires the participants were debriefed and thanked for their participa-tion

RESULTS AND DISCUSSION

Participantsrsquo responses were submitted to a 2 (goal competition coop-eration) acute 2 (estimation overestimation underestimation) acute 2 (questionorder) mixed ANOVA with the second factor varying within subjectThere was neither a main effect nor interaction associated with questionorder so it was dropped from the analysis It was expected that peoplewould judge that overestimating their partners and underestimatingtheir opponents would reduce their chances of success The results wereconsistent with the hypothesis The only significant effect was the goal acuteestimation interaction F(1 19) = 4532 p lt 001When expecting to coop-erate people judged that their chances of success would be lower if theyoverestimated their partners (M = 200)than if they underestimated theirpartners (M = 410) F(1 19) = 2000 p lt 001 In contrast people judgedthat their chances of success would be higher if they overestimated theiropponents (M = 430) than if they underestimated their opponents (M =260) F(1 19) = 4185p lt 001These findings thus confirm those of Study

416 CHAN AND YBARRA

1a and indicate that people may be inclined to adopt different informa-tion processing strategies depending on their interaction goals

If it really is the case that people when expecting to cooperate arepoised to process and be vigilant about a partnerrsquos negative qualities(underestimation) it would be expected that in a different cognitive taskthose inclinations would also be expressed For example in learningabout a potential partner who is described with negative and positivequalities we should find that people expecting to cooperate are morelikely to elaborate information diagnostic of the partnerrsquos negative thanpositive qualities Similar outcomes should be expected under antici-pated competition If people who expect to compete are inclined to un-cover an opponentrsquos positive qualities (overestimation) then it would beexpected that in a related cognitive task they would be more likely toelaborate information diagnostic of the opponentrsquos positive than nega-tive qualities This is the issue we examined in the second study

STUDY 2

Trait concepts differ in the degree to which they are negativendashdiagnosticor positivendashdiagnostic according to the schematic model of dispositionalattribution (Reeder 1985 Reeder amp Brewer 1979 ) and thecuendashdiagnosticity model of social perception (Skowronski amp Carlston1987 1989) The lay causal theories that people use to understand behav-iors related to negativendashdiagnostic traits (eg moralityndashrelated traits)indicate that negative behaviors are caused by dispositional factors butthat positive behaviors are caused by situational factors or a combina-tion of both (Ybarra amp Stephan 1999 Ybarra 2002) In contrast the laycausal theories people use to understand behaviors related to posi-tivendashdiagnostic traits (eg competencendashrelated traits but also othertraits as we will describe presently) indicate that positive behaviors arecaused by dispositional factors but that negative behaviors are causedby situational factors (Reeder amp Fulks 1980 but see Ybarra 2001 2002for a different perspective)

If a cooperation interaction goal leads people to emphasize a futurepartnerrsquos negative qualities then perceivers should focus on informa-tion related to negativendashdiagnostic traits As a result their informationprocessing (eg attention to and elaboration of information) should beguided by the lay causal theories underlying such traits (negative behav-iors are caused by dispositional causes and positive behaviors are

INTERACTION GOALS AND MEMORY 417

caused by situational causes) (Ybarra amp Stephan 1996 Ybarra ampStephan 1999) Thus people with a cooperation interaction goal shouldbe likely to remember information consistent rather than inconsistentwith the casual theory for the negativendashdiagnostic trait domain (iegood recall for positive behaviors caused by situational factors and neg-ative behaviors caused by dispositional factors) Behaviors related topositivendashdiagnostic traits should be processed less well because this in-formation is of little relevance to the goal of uncovering the partnerrsquosweaknesses Therefore it was expected that perceivers under a coopera-tion goal would not differentiate in memory between theory consistentand inconsistent behavioral information related to positivendashdiagnostictraits1

By comparison if competition leads people to emphasize a future op-ponentrsquos positive qualities then perceivers should focus on informationrelated to positivendashdiagnostic traits As a result their information pro-cessing (eg elaboration of information) should be guided by the laycausal theories underlying such traits (positive behaviors caused bydispositional causes and negative behaviors caused by situationalcauses) Thus people with a competition interaction goal should betterremember information consistent rather than inconsistent with the ca-sual theory for the positivendashdiagnostic trait domain (ie good recall forpositive behaviors caused by dispositional factors and negative valencebehaviors caused by situational factors) Behaviors from the nega-tivendashdiagnostic trait domain should not be well processed because thisinformationis of little relevance to the goal of uncovering the opponentrsquosstrengths Therefore it was expected that perceivers under a competi-tion goal would not show discrimination in their memory between the-ory consistent and inconsistent behavioral information related tonegativendashdiagnostic traits

A relevant question for sake of comparison is what kind of informa-tion would a simple impression formation goal lead perceivers to em-phasize and remember Some research on impression formation

418 CHAN AND YBARRA

1 The strategies of underestimating onersquos partner and overestimating onersquos opponentdo not have to be deliberate and intentional As Study 1b demonstrated even people whowere not consciously thinking about avoiding costs emerged with the same preferences forunderestimating partners and overestimating opponents Once the interaction goals trig-ger a preference for underndash or overestimating the other person the differences in informa-tion processing and memory may occur as a result of peoplersquos lay causal theories guidingthe processing of available information

suggests that people who are forming impressions focus more on the af-fective and general evaluative qualities of others (Levy amp Dugan 1960Ybarra 2001 Zajonc 1980) and that they tend to make such evaluationsbased mainly on negativendashdiagnostic (moralityndashrelated) information(Wojciszke Bazinska amp Jaworski 1998) Thus according to this per-spective people under a general impression formation goal should onlyattend to information related to the negativendashdiagnostic traits similar topeople with a cooperation goal (ie good recall for positive behaviorscaused by situational factors and negative behaviors caused bydispositional factors) Behaviors from the positivendashdiagnostic trait do-main should not be well processed because this information is of littlerelevance to the goal of forming general evaluative impressions There-fore it was expected that perceivers under an impression goal would notdiscriminate in their memory between theory consistent and inconsis-tent behavioral information related to the positivendashdiagnostic traitsThetheoretical model is summarized in Figure 1

Depending on the experimental group to which they were assignedparticipants were told that they would either cooperate or compete withanother participant Then they learned about the other participantrsquos be-haviors which were related to traits in the negativendashdiagnostic trait do-

INTERACTION GOALS AND MEMORY 419

FIGURE 1 Summary of theoretical model

main (hon esty h elpfu lness and fr iendliness ) and thepositivendashdiagnostic trait domain (hardworking image scoring) Imagescoring refers to people engaging in costly positive behaviors to signal toothers that they are valuable community members (Nowak amp Sigmund1998Wedekind amp Milinski 2000) Controlparticipants learned the samematerial but they were asked to form an impression of the person andwere told that they would not interact with the other participant

METHOD

Design and ParticipantsNinetyndashone students participated in the study for course credit Theywere randomly assigned to the cooperation competition or control (im-pression formation) condition All of the participants were presentedwith behaviors with a positive or negative valence from the negativendashdi-agnostic and positivendashdiagnostic trait domains and these behaviorswere linked to either explicit dispositional or situational attributionsThus the overall design of the study was a 3 (interaction goal coopera-tion competition control) acute 2 (trait domain positivendashdiagnostic nega-tivendashdiagnostic) acute 2 (behavior valence positive negative) acute 2(attribution dispositional situational) mixed design with the first fac-tor varying between participants and the latter three factors varyingwithin participants The participants were randomly assigned to condi-tions and were run in nonndashinteracting groups of two to six

Stimulus MaterialsThe behavioral information that the participants processed was pre-sented on a cassette tape allegedly recorded by another participant Thegender of the speaker on the tape and that of the participant were alwaysthe same Each tape contained 20 behavioral statements half of whichwere negative and half positive in valence Also half of the behaviorswere linked to dispositional attributions and half to situational attribu-tions Twelve of the statements pertained to negativendashdiagnostic traits(eg ldquoI entered a concert through the fire exitrdquo [negative valence] ldquoI sentmy sister a present and a birthday bouquetrdquo [positive valence]) Eight ofthe statements pertained to the positivendashdiagnostic traits (eg ldquoI missedmorning lecturesrdquo [negative valence] ldquoI reported a crimerdquo [positive va-lence])

420 CHAN AND YBARRA

Our choice of the traits hardworking and image scoring was deliberatebecause it allows us to show that perceiverrsquos attention to a targetrsquosstrengths (overestimation) is not confounded with emphasizing infor-mation about the targetrsquos competence For example it is usually as-sumed that there is an inverse relationship between the amount of worka person engages in and the presence of the related ability (Darley ampGoethals 1980 Jones 1989) but being hardworking should still signal apositive quality a strength In addition image scoring refers to peopleengaging in costly positive behaviors such as extreme levels of philan-thropy to enhance their reputation and status (Wedekind amp Milinski2000 McAndrew 2002) It is completely unrelated to competence but itsdiagnosticity is in line with that of a positivendashdiagnostic trait

The behavior stems (no attributions) were pretested to ensure thatthey were indeed regarded as positive or negativendashdiagnostic Thirtyparticipants rated each behavior on how common they are For exampleone item queried them as to ldquoWhat percentage of college students doyou think enter a concert through the fire exitrdquo (6ndashpoint scale with 0ndash1010ndash30 30ndash50 50ndash70 70ndash90 and 90ndash100) For positivendashdiagnos-tic traits behaviors that are positive in valence should be less common(and hence more diagnostic of the actorrsquos dispositions) than behaviorsthat are negative in valence whereas for negativendashdiagnostic traits be-haviors negative in valence should be less common (and hence more di-agnostic of the actorrsquos dispositions) than behaviors positive in valenceResults showed an interaction effect that confirmed our classification ofthe behaviors into the respective positive and negativendashdiagnostic cate-gories F(1 29) = 4913 p lt 001 For the positivendashdiagnostic traits posi-tive valence behaviors (M = 281) were judged to be enacted less oftenthan negative valence behaviors (M = 376) F(1 29) = 1961 p lt 001 Forthe negativendashdiagnostic traits negative valence behaviors (M = 294)were judged to be enacted less often than positive valence behaviors (M= 331) F(1 29) = 7103 p = 012

Two sets of materials both containing the same 20 behavior stemswere created Each behavior stem was matched with a dispositional at-tribution in one set and a situational attribution in the other set For ex-ample in set A the behaviors ldquoI didnrsquot help my dad with yard workbecause I wanted to stay inside to watch TVrdquo [negative behavior in nega-tivendashdiagnostic domain with dispositional attribution] ldquoI sent my sistera present and a birthday bouquet because my grandmother asked me todo itrdquo [positive behavior in negativendashdiagnostic domain with situational

INTERACTION GOALS AND MEMORY 421

attribution] ldquoI missed morning lectures because my roommate playedpranks with my alarm clockrdquo [negative behavior in positivendashdiagnosticdomain with situational attribution] and ldquoI worked as a research assis-tant because I am very interested in the subject matterrdquo [positive behav-ior in positive diagnostic domain with dispositional attribution] wouldbe counterbalanced in set B with different causal attributions These in-cluded ldquoI didnrsquot help my dad with yard work because he couldnrsquot findthe keys to the tool shedrdquo [situational attribution] ldquoI sent my sister apresent and a birthday bouquet because I wanted to give her a pleasantsurpriserdquo [dispositional attribution] ldquoI missed morning lecture becauseI donrsquot feel like waking up for themrdquo [dispositional attribution] and ldquoIworked as a research assistant because my professor said I had to do itrdquo[situational attribution] As will be described below the valence of thebehaviors and attributions (dispositional vs external) were manipu-lated independently to ensure that subsequent differences in recall oftheoryndashconsistent or inconsistent information could not simply be dueto an artifact based on the memorability of the particular behaviors

The two sets of stimuli were presented to 17 participants for pretest-ing One participant misunderstood how to use the response scale andwas dropped from the analysis Participants were first presented withthe behavior stems and were asked to rate how favorable each behaviorwas on a 7ndashpoint scale with higher values indicating greaterfavorability Then they were presented with the same behavior stemslinked to either a dispositional or situational attribution and were askedto judge the locus of causation for the behaviors on a 7ndashpoint scale withhigher values indicating more dispositional causation The responseswere submitted to a 2 (valence)acute 2 (attribution)withinndashsubject ANOVAFor the valence judgments the positive behaviors in both stimulus sets(Mset A = 544 Mset B = 559) were judged as more favorable than the nega-tive behaviors (Mset A = 284Mset B = 215) Fset A (1 8)= 3416 p lt 0004Fset B

(1 8) = 9782 p lt 0001 For the second set of judgments a main effect forattribution was obtained for both sets of stimuli Behaviors linked todispositional attributions (Mset A = 575 Mset B = 646) were judged to bemore dispositionally caused than behaviors linked to situational attribu-tions (M set A = 235 Mset B = 298) F set A (1 8) = 4574 p lt 0001 Fset B (1 8) =15826 p lt 0001 No other main effects or interaction effects were ob-tained The two sets of stimuli were presented under three different ran-domization schemes

422 CHAN AND YBARRA

PROCEDURE

Upon arrival at the lab the participants were provided the cover storyfor the study in which they were told that the study would examine howpersonal information would affect subsequent interactions betweenstrangers They were told that another set of participants was recordingsome personal information on a tape for them to listen to To increase thecredibility of the tape contents participants were told that the partici-pants recording the tape had been asked to provide distinct personal in-formation about ldquothings that they have done in the past few weeksrdquoAdditional information about the experiment including whether theywere in the cooperation competition or control (impression formation)condition was given to each participant in a sealed envelope so that theexperimenter could remain blind to the conditions Experimental groupparticipants were told that after they listened to a tape with personal in-formation recorded by another participant they would get a chance tomeet them in either a cooperative or competitive setting Control partici-pants were informed that they had been randomly assigned to the con-trol group and would listen to some information about a previousparticipant instead of having an interaction with them

Participants in the cooperation condition read that they would engagein a ldquoCommunity Simulation Gamerdquo in which they and their assignedpartners ldquowill work together to try to maximize the number of points foryour team The team with the highest points will receive a prizerdquo Thecooperative nature of the interaction was emphasized with repeated useof the word ldquopartnerrdquo In contrastparticipants in the competition condi-tion read that they would engage in a ldquoWall Street Simulation Gamerdquo inwhich they would ldquocompete against [their] competitor to maximize[their] own profit (points) and minimize their opponentrsquos profit Theperson with the highest points will win a prizerdquo The competitive aspectof the upcoming interaction was emphasized further with repeated useof the words ldquocompetitorrdquo and ldquoopponentrdquo This type of reward struc-ture manipulation has been used to study cooperation and competitionby other researchers (Deutsch 1973)

Participants in the control condition read that they would listen to atape that had been recorded in a previous experimental session Theywere told to form an impression of the person using the informationfrom the tape as part of a ldquostandardizationrdquo procedure for the experi-ment To bolster the believability of the cover story participants then

INTERACTION GOALS AND MEMORY 423

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

1a and indicate that people may be inclined to adopt different informa-tion processing strategies depending on their interaction goals

If it really is the case that people when expecting to cooperate arepoised to process and be vigilant about a partnerrsquos negative qualities(underestimation) it would be expected that in a different cognitive taskthose inclinations would also be expressed For example in learningabout a potential partner who is described with negative and positivequalities we should find that people expecting to cooperate are morelikely to elaborate information diagnostic of the partnerrsquos negative thanpositive qualities Similar outcomes should be expected under antici-pated competition If people who expect to compete are inclined to un-cover an opponentrsquos positive qualities (overestimation) then it would beexpected that in a related cognitive task they would be more likely toelaborate information diagnostic of the opponentrsquos positive than nega-tive qualities This is the issue we examined in the second study

STUDY 2

Trait concepts differ in the degree to which they are negativendashdiagnosticor positivendashdiagnostic according to the schematic model of dispositionalattribution (Reeder 1985 Reeder amp Brewer 1979 ) and thecuendashdiagnosticity model of social perception (Skowronski amp Carlston1987 1989) The lay causal theories that people use to understand behav-iors related to negativendashdiagnostic traits (eg moralityndashrelated traits)indicate that negative behaviors are caused by dispositional factors butthat positive behaviors are caused by situational factors or a combina-tion of both (Ybarra amp Stephan 1999 Ybarra 2002) In contrast the laycausal theories people use to understand behaviors related to posi-tivendashdiagnostic traits (eg competencendashrelated traits but also othertraits as we will describe presently) indicate that positive behaviors arecaused by dispositional factors but that negative behaviors are causedby situational factors (Reeder amp Fulks 1980 but see Ybarra 2001 2002for a different perspective)

If a cooperation interaction goal leads people to emphasize a futurepartnerrsquos negative qualities then perceivers should focus on informa-tion related to negativendashdiagnostic traits As a result their informationprocessing (eg attention to and elaboration of information) should beguided by the lay causal theories underlying such traits (negative behav-iors are caused by dispositional causes and positive behaviors are

INTERACTION GOALS AND MEMORY 417

caused by situational causes) (Ybarra amp Stephan 1996 Ybarra ampStephan 1999) Thus people with a cooperation interaction goal shouldbe likely to remember information consistent rather than inconsistentwith the casual theory for the negativendashdiagnostic trait domain (iegood recall for positive behaviors caused by situational factors and neg-ative behaviors caused by dispositional factors) Behaviors related topositivendashdiagnostic traits should be processed less well because this in-formation is of little relevance to the goal of uncovering the partnerrsquosweaknesses Therefore it was expected that perceivers under a coopera-tion goal would not differentiate in memory between theory consistentand inconsistent behavioral information related to positivendashdiagnostictraits1

By comparison if competition leads people to emphasize a future op-ponentrsquos positive qualities then perceivers should focus on informationrelated to positivendashdiagnostic traits As a result their information pro-cessing (eg elaboration of information) should be guided by the laycausal theories underlying such traits (positive behaviors caused bydispositional causes and negative behaviors caused by situationalcauses) Thus people with a competition interaction goal should betterremember information consistent rather than inconsistent with the ca-sual theory for the positivendashdiagnostic trait domain (ie good recall forpositive behaviors caused by dispositional factors and negative valencebehaviors caused by situational factors) Behaviors from the nega-tivendashdiagnostic trait domain should not be well processed because thisinformationis of little relevance to the goal of uncovering the opponentrsquosstrengths Therefore it was expected that perceivers under a competi-tion goal would not show discrimination in their memory between the-ory consistent and inconsistent behavioral information related tonegativendashdiagnostic traits

A relevant question for sake of comparison is what kind of informa-tion would a simple impression formation goal lead perceivers to em-phasize and remember Some research on impression formation

418 CHAN AND YBARRA

1 The strategies of underestimating onersquos partner and overestimating onersquos opponentdo not have to be deliberate and intentional As Study 1b demonstrated even people whowere not consciously thinking about avoiding costs emerged with the same preferences forunderestimating partners and overestimating opponents Once the interaction goals trig-ger a preference for underndash or overestimating the other person the differences in informa-tion processing and memory may occur as a result of peoplersquos lay causal theories guidingthe processing of available information

suggests that people who are forming impressions focus more on the af-fective and general evaluative qualities of others (Levy amp Dugan 1960Ybarra 2001 Zajonc 1980) and that they tend to make such evaluationsbased mainly on negativendashdiagnostic (moralityndashrelated) information(Wojciszke Bazinska amp Jaworski 1998) Thus according to this per-spective people under a general impression formation goal should onlyattend to information related to the negativendashdiagnostic traits similar topeople with a cooperation goal (ie good recall for positive behaviorscaused by situational factors and negative behaviors caused bydispositional factors) Behaviors from the positivendashdiagnostic trait do-main should not be well processed because this information is of littlerelevance to the goal of forming general evaluative impressions There-fore it was expected that perceivers under an impression goal would notdiscriminate in their memory between theory consistent and inconsis-tent behavioral information related to the positivendashdiagnostic traitsThetheoretical model is summarized in Figure 1

Depending on the experimental group to which they were assignedparticipants were told that they would either cooperate or compete withanother participant Then they learned about the other participantrsquos be-haviors which were related to traits in the negativendashdiagnostic trait do-

INTERACTION GOALS AND MEMORY 419

FIGURE 1 Summary of theoretical model

main (hon esty h elpfu lness and fr iendliness ) and thepositivendashdiagnostic trait domain (hardworking image scoring) Imagescoring refers to people engaging in costly positive behaviors to signal toothers that they are valuable community members (Nowak amp Sigmund1998Wedekind amp Milinski 2000) Controlparticipants learned the samematerial but they were asked to form an impression of the person andwere told that they would not interact with the other participant

METHOD

Design and ParticipantsNinetyndashone students participated in the study for course credit Theywere randomly assigned to the cooperation competition or control (im-pression formation) condition All of the participants were presentedwith behaviors with a positive or negative valence from the negativendashdi-agnostic and positivendashdiagnostic trait domains and these behaviorswere linked to either explicit dispositional or situational attributionsThus the overall design of the study was a 3 (interaction goal coopera-tion competition control) acute 2 (trait domain positivendashdiagnostic nega-tivendashdiagnostic) acute 2 (behavior valence positive negative) acute 2(attribution dispositional situational) mixed design with the first fac-tor varying between participants and the latter three factors varyingwithin participants The participants were randomly assigned to condi-tions and were run in nonndashinteracting groups of two to six

Stimulus MaterialsThe behavioral information that the participants processed was pre-sented on a cassette tape allegedly recorded by another participant Thegender of the speaker on the tape and that of the participant were alwaysthe same Each tape contained 20 behavioral statements half of whichwere negative and half positive in valence Also half of the behaviorswere linked to dispositional attributions and half to situational attribu-tions Twelve of the statements pertained to negativendashdiagnostic traits(eg ldquoI entered a concert through the fire exitrdquo [negative valence] ldquoI sentmy sister a present and a birthday bouquetrdquo [positive valence]) Eight ofthe statements pertained to the positivendashdiagnostic traits (eg ldquoI missedmorning lecturesrdquo [negative valence] ldquoI reported a crimerdquo [positive va-lence])

420 CHAN AND YBARRA

Our choice of the traits hardworking and image scoring was deliberatebecause it allows us to show that perceiverrsquos attention to a targetrsquosstrengths (overestimation) is not confounded with emphasizing infor-mation about the targetrsquos competence For example it is usually as-sumed that there is an inverse relationship between the amount of worka person engages in and the presence of the related ability (Darley ampGoethals 1980 Jones 1989) but being hardworking should still signal apositive quality a strength In addition image scoring refers to peopleengaging in costly positive behaviors such as extreme levels of philan-thropy to enhance their reputation and status (Wedekind amp Milinski2000 McAndrew 2002) It is completely unrelated to competence but itsdiagnosticity is in line with that of a positivendashdiagnostic trait

The behavior stems (no attributions) were pretested to ensure thatthey were indeed regarded as positive or negativendashdiagnostic Thirtyparticipants rated each behavior on how common they are For exampleone item queried them as to ldquoWhat percentage of college students doyou think enter a concert through the fire exitrdquo (6ndashpoint scale with 0ndash1010ndash30 30ndash50 50ndash70 70ndash90 and 90ndash100) For positivendashdiagnos-tic traits behaviors that are positive in valence should be less common(and hence more diagnostic of the actorrsquos dispositions) than behaviorsthat are negative in valence whereas for negativendashdiagnostic traits be-haviors negative in valence should be less common (and hence more di-agnostic of the actorrsquos dispositions) than behaviors positive in valenceResults showed an interaction effect that confirmed our classification ofthe behaviors into the respective positive and negativendashdiagnostic cate-gories F(1 29) = 4913 p lt 001 For the positivendashdiagnostic traits posi-tive valence behaviors (M = 281) were judged to be enacted less oftenthan negative valence behaviors (M = 376) F(1 29) = 1961 p lt 001 Forthe negativendashdiagnostic traits negative valence behaviors (M = 294)were judged to be enacted less often than positive valence behaviors (M= 331) F(1 29) = 7103 p = 012

Two sets of materials both containing the same 20 behavior stemswere created Each behavior stem was matched with a dispositional at-tribution in one set and a situational attribution in the other set For ex-ample in set A the behaviors ldquoI didnrsquot help my dad with yard workbecause I wanted to stay inside to watch TVrdquo [negative behavior in nega-tivendashdiagnostic domain with dispositional attribution] ldquoI sent my sistera present and a birthday bouquet because my grandmother asked me todo itrdquo [positive behavior in negativendashdiagnostic domain with situational

INTERACTION GOALS AND MEMORY 421

attribution] ldquoI missed morning lectures because my roommate playedpranks with my alarm clockrdquo [negative behavior in positivendashdiagnosticdomain with situational attribution] and ldquoI worked as a research assis-tant because I am very interested in the subject matterrdquo [positive behav-ior in positive diagnostic domain with dispositional attribution] wouldbe counterbalanced in set B with different causal attributions These in-cluded ldquoI didnrsquot help my dad with yard work because he couldnrsquot findthe keys to the tool shedrdquo [situational attribution] ldquoI sent my sister apresent and a birthday bouquet because I wanted to give her a pleasantsurpriserdquo [dispositional attribution] ldquoI missed morning lecture becauseI donrsquot feel like waking up for themrdquo [dispositional attribution] and ldquoIworked as a research assistant because my professor said I had to do itrdquo[situational attribution] As will be described below the valence of thebehaviors and attributions (dispositional vs external) were manipu-lated independently to ensure that subsequent differences in recall oftheoryndashconsistent or inconsistent information could not simply be dueto an artifact based on the memorability of the particular behaviors

The two sets of stimuli were presented to 17 participants for pretest-ing One participant misunderstood how to use the response scale andwas dropped from the analysis Participants were first presented withthe behavior stems and were asked to rate how favorable each behaviorwas on a 7ndashpoint scale with higher values indicating greaterfavorability Then they were presented with the same behavior stemslinked to either a dispositional or situational attribution and were askedto judge the locus of causation for the behaviors on a 7ndashpoint scale withhigher values indicating more dispositional causation The responseswere submitted to a 2 (valence)acute 2 (attribution)withinndashsubject ANOVAFor the valence judgments the positive behaviors in both stimulus sets(Mset A = 544 Mset B = 559) were judged as more favorable than the nega-tive behaviors (Mset A = 284Mset B = 215) Fset A (1 8)= 3416 p lt 0004Fset B

(1 8) = 9782 p lt 0001 For the second set of judgments a main effect forattribution was obtained for both sets of stimuli Behaviors linked todispositional attributions (Mset A = 575 Mset B = 646) were judged to bemore dispositionally caused than behaviors linked to situational attribu-tions (M set A = 235 Mset B = 298) F set A (1 8) = 4574 p lt 0001 Fset B (1 8) =15826 p lt 0001 No other main effects or interaction effects were ob-tained The two sets of stimuli were presented under three different ran-domization schemes

422 CHAN AND YBARRA

PROCEDURE

Upon arrival at the lab the participants were provided the cover storyfor the study in which they were told that the study would examine howpersonal information would affect subsequent interactions betweenstrangers They were told that another set of participants was recordingsome personal information on a tape for them to listen to To increase thecredibility of the tape contents participants were told that the partici-pants recording the tape had been asked to provide distinct personal in-formation about ldquothings that they have done in the past few weeksrdquoAdditional information about the experiment including whether theywere in the cooperation competition or control (impression formation)condition was given to each participant in a sealed envelope so that theexperimenter could remain blind to the conditions Experimental groupparticipants were told that after they listened to a tape with personal in-formation recorded by another participant they would get a chance tomeet them in either a cooperative or competitive setting Control partici-pants were informed that they had been randomly assigned to the con-trol group and would listen to some information about a previousparticipant instead of having an interaction with them

Participants in the cooperation condition read that they would engagein a ldquoCommunity Simulation Gamerdquo in which they and their assignedpartners ldquowill work together to try to maximize the number of points foryour team The team with the highest points will receive a prizerdquo Thecooperative nature of the interaction was emphasized with repeated useof the word ldquopartnerrdquo In contrastparticipants in the competition condi-tion read that they would engage in a ldquoWall Street Simulation Gamerdquo inwhich they would ldquocompete against [their] competitor to maximize[their] own profit (points) and minimize their opponentrsquos profit Theperson with the highest points will win a prizerdquo The competitive aspectof the upcoming interaction was emphasized further with repeated useof the words ldquocompetitorrdquo and ldquoopponentrdquo This type of reward struc-ture manipulation has been used to study cooperation and competitionby other researchers (Deutsch 1973)

Participants in the control condition read that they would listen to atape that had been recorded in a previous experimental session Theywere told to form an impression of the person using the informationfrom the tape as part of a ldquostandardizationrdquo procedure for the experi-ment To bolster the believability of the cover story participants then

INTERACTION GOALS AND MEMORY 423

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

caused by situational causes) (Ybarra amp Stephan 1996 Ybarra ampStephan 1999) Thus people with a cooperation interaction goal shouldbe likely to remember information consistent rather than inconsistentwith the casual theory for the negativendashdiagnostic trait domain (iegood recall for positive behaviors caused by situational factors and neg-ative behaviors caused by dispositional factors) Behaviors related topositivendashdiagnostic traits should be processed less well because this in-formation is of little relevance to the goal of uncovering the partnerrsquosweaknesses Therefore it was expected that perceivers under a coopera-tion goal would not differentiate in memory between theory consistentand inconsistent behavioral information related to positivendashdiagnostictraits1

By comparison if competition leads people to emphasize a future op-ponentrsquos positive qualities then perceivers should focus on informationrelated to positivendashdiagnostic traits As a result their information pro-cessing (eg elaboration of information) should be guided by the laycausal theories underlying such traits (positive behaviors caused bydispositional causes and negative behaviors caused by situationalcauses) Thus people with a competition interaction goal should betterremember information consistent rather than inconsistent with the ca-sual theory for the positivendashdiagnostic trait domain (ie good recall forpositive behaviors caused by dispositional factors and negative valencebehaviors caused by situational factors) Behaviors from the nega-tivendashdiagnostic trait domain should not be well processed because thisinformationis of little relevance to the goal of uncovering the opponentrsquosstrengths Therefore it was expected that perceivers under a competi-tion goal would not show discrimination in their memory between the-ory consistent and inconsistent behavioral information related tonegativendashdiagnostic traits

A relevant question for sake of comparison is what kind of informa-tion would a simple impression formation goal lead perceivers to em-phasize and remember Some research on impression formation

418 CHAN AND YBARRA

1 The strategies of underestimating onersquos partner and overestimating onersquos opponentdo not have to be deliberate and intentional As Study 1b demonstrated even people whowere not consciously thinking about avoiding costs emerged with the same preferences forunderestimating partners and overestimating opponents Once the interaction goals trig-ger a preference for underndash or overestimating the other person the differences in informa-tion processing and memory may occur as a result of peoplersquos lay causal theories guidingthe processing of available information

suggests that people who are forming impressions focus more on the af-fective and general evaluative qualities of others (Levy amp Dugan 1960Ybarra 2001 Zajonc 1980) and that they tend to make such evaluationsbased mainly on negativendashdiagnostic (moralityndashrelated) information(Wojciszke Bazinska amp Jaworski 1998) Thus according to this per-spective people under a general impression formation goal should onlyattend to information related to the negativendashdiagnostic traits similar topeople with a cooperation goal (ie good recall for positive behaviorscaused by situational factors and negative behaviors caused bydispositional factors) Behaviors from the positivendashdiagnostic trait do-main should not be well processed because this information is of littlerelevance to the goal of forming general evaluative impressions There-fore it was expected that perceivers under an impression goal would notdiscriminate in their memory between theory consistent and inconsis-tent behavioral information related to the positivendashdiagnostic traitsThetheoretical model is summarized in Figure 1

Depending on the experimental group to which they were assignedparticipants were told that they would either cooperate or compete withanother participant Then they learned about the other participantrsquos be-haviors which were related to traits in the negativendashdiagnostic trait do-

INTERACTION GOALS AND MEMORY 419

FIGURE 1 Summary of theoretical model

main (hon esty h elpfu lness and fr iendliness ) and thepositivendashdiagnostic trait domain (hardworking image scoring) Imagescoring refers to people engaging in costly positive behaviors to signal toothers that they are valuable community members (Nowak amp Sigmund1998Wedekind amp Milinski 2000) Controlparticipants learned the samematerial but they were asked to form an impression of the person andwere told that they would not interact with the other participant

METHOD

Design and ParticipantsNinetyndashone students participated in the study for course credit Theywere randomly assigned to the cooperation competition or control (im-pression formation) condition All of the participants were presentedwith behaviors with a positive or negative valence from the negativendashdi-agnostic and positivendashdiagnostic trait domains and these behaviorswere linked to either explicit dispositional or situational attributionsThus the overall design of the study was a 3 (interaction goal coopera-tion competition control) acute 2 (trait domain positivendashdiagnostic nega-tivendashdiagnostic) acute 2 (behavior valence positive negative) acute 2(attribution dispositional situational) mixed design with the first fac-tor varying between participants and the latter three factors varyingwithin participants The participants were randomly assigned to condi-tions and were run in nonndashinteracting groups of two to six

Stimulus MaterialsThe behavioral information that the participants processed was pre-sented on a cassette tape allegedly recorded by another participant Thegender of the speaker on the tape and that of the participant were alwaysthe same Each tape contained 20 behavioral statements half of whichwere negative and half positive in valence Also half of the behaviorswere linked to dispositional attributions and half to situational attribu-tions Twelve of the statements pertained to negativendashdiagnostic traits(eg ldquoI entered a concert through the fire exitrdquo [negative valence] ldquoI sentmy sister a present and a birthday bouquetrdquo [positive valence]) Eight ofthe statements pertained to the positivendashdiagnostic traits (eg ldquoI missedmorning lecturesrdquo [negative valence] ldquoI reported a crimerdquo [positive va-lence])

420 CHAN AND YBARRA

Our choice of the traits hardworking and image scoring was deliberatebecause it allows us to show that perceiverrsquos attention to a targetrsquosstrengths (overestimation) is not confounded with emphasizing infor-mation about the targetrsquos competence For example it is usually as-sumed that there is an inverse relationship between the amount of worka person engages in and the presence of the related ability (Darley ampGoethals 1980 Jones 1989) but being hardworking should still signal apositive quality a strength In addition image scoring refers to peopleengaging in costly positive behaviors such as extreme levels of philan-thropy to enhance their reputation and status (Wedekind amp Milinski2000 McAndrew 2002) It is completely unrelated to competence but itsdiagnosticity is in line with that of a positivendashdiagnostic trait

The behavior stems (no attributions) were pretested to ensure thatthey were indeed regarded as positive or negativendashdiagnostic Thirtyparticipants rated each behavior on how common they are For exampleone item queried them as to ldquoWhat percentage of college students doyou think enter a concert through the fire exitrdquo (6ndashpoint scale with 0ndash1010ndash30 30ndash50 50ndash70 70ndash90 and 90ndash100) For positivendashdiagnos-tic traits behaviors that are positive in valence should be less common(and hence more diagnostic of the actorrsquos dispositions) than behaviorsthat are negative in valence whereas for negativendashdiagnostic traits be-haviors negative in valence should be less common (and hence more di-agnostic of the actorrsquos dispositions) than behaviors positive in valenceResults showed an interaction effect that confirmed our classification ofthe behaviors into the respective positive and negativendashdiagnostic cate-gories F(1 29) = 4913 p lt 001 For the positivendashdiagnostic traits posi-tive valence behaviors (M = 281) were judged to be enacted less oftenthan negative valence behaviors (M = 376) F(1 29) = 1961 p lt 001 Forthe negativendashdiagnostic traits negative valence behaviors (M = 294)were judged to be enacted less often than positive valence behaviors (M= 331) F(1 29) = 7103 p = 012

Two sets of materials both containing the same 20 behavior stemswere created Each behavior stem was matched with a dispositional at-tribution in one set and a situational attribution in the other set For ex-ample in set A the behaviors ldquoI didnrsquot help my dad with yard workbecause I wanted to stay inside to watch TVrdquo [negative behavior in nega-tivendashdiagnostic domain with dispositional attribution] ldquoI sent my sistera present and a birthday bouquet because my grandmother asked me todo itrdquo [positive behavior in negativendashdiagnostic domain with situational

INTERACTION GOALS AND MEMORY 421

attribution] ldquoI missed morning lectures because my roommate playedpranks with my alarm clockrdquo [negative behavior in positivendashdiagnosticdomain with situational attribution] and ldquoI worked as a research assis-tant because I am very interested in the subject matterrdquo [positive behav-ior in positive diagnostic domain with dispositional attribution] wouldbe counterbalanced in set B with different causal attributions These in-cluded ldquoI didnrsquot help my dad with yard work because he couldnrsquot findthe keys to the tool shedrdquo [situational attribution] ldquoI sent my sister apresent and a birthday bouquet because I wanted to give her a pleasantsurpriserdquo [dispositional attribution] ldquoI missed morning lecture becauseI donrsquot feel like waking up for themrdquo [dispositional attribution] and ldquoIworked as a research assistant because my professor said I had to do itrdquo[situational attribution] As will be described below the valence of thebehaviors and attributions (dispositional vs external) were manipu-lated independently to ensure that subsequent differences in recall oftheoryndashconsistent or inconsistent information could not simply be dueto an artifact based on the memorability of the particular behaviors

The two sets of stimuli were presented to 17 participants for pretest-ing One participant misunderstood how to use the response scale andwas dropped from the analysis Participants were first presented withthe behavior stems and were asked to rate how favorable each behaviorwas on a 7ndashpoint scale with higher values indicating greaterfavorability Then they were presented with the same behavior stemslinked to either a dispositional or situational attribution and were askedto judge the locus of causation for the behaviors on a 7ndashpoint scale withhigher values indicating more dispositional causation The responseswere submitted to a 2 (valence)acute 2 (attribution)withinndashsubject ANOVAFor the valence judgments the positive behaviors in both stimulus sets(Mset A = 544 Mset B = 559) were judged as more favorable than the nega-tive behaviors (Mset A = 284Mset B = 215) Fset A (1 8)= 3416 p lt 0004Fset B

(1 8) = 9782 p lt 0001 For the second set of judgments a main effect forattribution was obtained for both sets of stimuli Behaviors linked todispositional attributions (Mset A = 575 Mset B = 646) were judged to bemore dispositionally caused than behaviors linked to situational attribu-tions (M set A = 235 Mset B = 298) F set A (1 8) = 4574 p lt 0001 Fset B (1 8) =15826 p lt 0001 No other main effects or interaction effects were ob-tained The two sets of stimuli were presented under three different ran-domization schemes

422 CHAN AND YBARRA

PROCEDURE

Upon arrival at the lab the participants were provided the cover storyfor the study in which they were told that the study would examine howpersonal information would affect subsequent interactions betweenstrangers They were told that another set of participants was recordingsome personal information on a tape for them to listen to To increase thecredibility of the tape contents participants were told that the partici-pants recording the tape had been asked to provide distinct personal in-formation about ldquothings that they have done in the past few weeksrdquoAdditional information about the experiment including whether theywere in the cooperation competition or control (impression formation)condition was given to each participant in a sealed envelope so that theexperimenter could remain blind to the conditions Experimental groupparticipants were told that after they listened to a tape with personal in-formation recorded by another participant they would get a chance tomeet them in either a cooperative or competitive setting Control partici-pants were informed that they had been randomly assigned to the con-trol group and would listen to some information about a previousparticipant instead of having an interaction with them

Participants in the cooperation condition read that they would engagein a ldquoCommunity Simulation Gamerdquo in which they and their assignedpartners ldquowill work together to try to maximize the number of points foryour team The team with the highest points will receive a prizerdquo Thecooperative nature of the interaction was emphasized with repeated useof the word ldquopartnerrdquo In contrastparticipants in the competition condi-tion read that they would engage in a ldquoWall Street Simulation Gamerdquo inwhich they would ldquocompete against [their] competitor to maximize[their] own profit (points) and minimize their opponentrsquos profit Theperson with the highest points will win a prizerdquo The competitive aspectof the upcoming interaction was emphasized further with repeated useof the words ldquocompetitorrdquo and ldquoopponentrdquo This type of reward struc-ture manipulation has been used to study cooperation and competitionby other researchers (Deutsch 1973)

Participants in the control condition read that they would listen to atape that had been recorded in a previous experimental session Theywere told to form an impression of the person using the informationfrom the tape as part of a ldquostandardizationrdquo procedure for the experi-ment To bolster the believability of the cover story participants then

INTERACTION GOALS AND MEMORY 423

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

suggests that people who are forming impressions focus more on the af-fective and general evaluative qualities of others (Levy amp Dugan 1960Ybarra 2001 Zajonc 1980) and that they tend to make such evaluationsbased mainly on negativendashdiagnostic (moralityndashrelated) information(Wojciszke Bazinska amp Jaworski 1998) Thus according to this per-spective people under a general impression formation goal should onlyattend to information related to the negativendashdiagnostic traits similar topeople with a cooperation goal (ie good recall for positive behaviorscaused by situational factors and negative behaviors caused bydispositional factors) Behaviors from the positivendashdiagnostic trait do-main should not be well processed because this information is of littlerelevance to the goal of forming general evaluative impressions There-fore it was expected that perceivers under an impression goal would notdiscriminate in their memory between theory consistent and inconsis-tent behavioral information related to the positivendashdiagnostic traitsThetheoretical model is summarized in Figure 1

Depending on the experimental group to which they were assignedparticipants were told that they would either cooperate or compete withanother participant Then they learned about the other participantrsquos be-haviors which were related to traits in the negativendashdiagnostic trait do-

INTERACTION GOALS AND MEMORY 419

FIGURE 1 Summary of theoretical model

main (hon esty h elpfu lness and fr iendliness ) and thepositivendashdiagnostic trait domain (hardworking image scoring) Imagescoring refers to people engaging in costly positive behaviors to signal toothers that they are valuable community members (Nowak amp Sigmund1998Wedekind amp Milinski 2000) Controlparticipants learned the samematerial but they were asked to form an impression of the person andwere told that they would not interact with the other participant

METHOD

Design and ParticipantsNinetyndashone students participated in the study for course credit Theywere randomly assigned to the cooperation competition or control (im-pression formation) condition All of the participants were presentedwith behaviors with a positive or negative valence from the negativendashdi-agnostic and positivendashdiagnostic trait domains and these behaviorswere linked to either explicit dispositional or situational attributionsThus the overall design of the study was a 3 (interaction goal coopera-tion competition control) acute 2 (trait domain positivendashdiagnostic nega-tivendashdiagnostic) acute 2 (behavior valence positive negative) acute 2(attribution dispositional situational) mixed design with the first fac-tor varying between participants and the latter three factors varyingwithin participants The participants were randomly assigned to condi-tions and were run in nonndashinteracting groups of two to six

Stimulus MaterialsThe behavioral information that the participants processed was pre-sented on a cassette tape allegedly recorded by another participant Thegender of the speaker on the tape and that of the participant were alwaysthe same Each tape contained 20 behavioral statements half of whichwere negative and half positive in valence Also half of the behaviorswere linked to dispositional attributions and half to situational attribu-tions Twelve of the statements pertained to negativendashdiagnostic traits(eg ldquoI entered a concert through the fire exitrdquo [negative valence] ldquoI sentmy sister a present and a birthday bouquetrdquo [positive valence]) Eight ofthe statements pertained to the positivendashdiagnostic traits (eg ldquoI missedmorning lecturesrdquo [negative valence] ldquoI reported a crimerdquo [positive va-lence])

420 CHAN AND YBARRA

Our choice of the traits hardworking and image scoring was deliberatebecause it allows us to show that perceiverrsquos attention to a targetrsquosstrengths (overestimation) is not confounded with emphasizing infor-mation about the targetrsquos competence For example it is usually as-sumed that there is an inverse relationship between the amount of worka person engages in and the presence of the related ability (Darley ampGoethals 1980 Jones 1989) but being hardworking should still signal apositive quality a strength In addition image scoring refers to peopleengaging in costly positive behaviors such as extreme levels of philan-thropy to enhance their reputation and status (Wedekind amp Milinski2000 McAndrew 2002) It is completely unrelated to competence but itsdiagnosticity is in line with that of a positivendashdiagnostic trait

The behavior stems (no attributions) were pretested to ensure thatthey were indeed regarded as positive or negativendashdiagnostic Thirtyparticipants rated each behavior on how common they are For exampleone item queried them as to ldquoWhat percentage of college students doyou think enter a concert through the fire exitrdquo (6ndashpoint scale with 0ndash1010ndash30 30ndash50 50ndash70 70ndash90 and 90ndash100) For positivendashdiagnos-tic traits behaviors that are positive in valence should be less common(and hence more diagnostic of the actorrsquos dispositions) than behaviorsthat are negative in valence whereas for negativendashdiagnostic traits be-haviors negative in valence should be less common (and hence more di-agnostic of the actorrsquos dispositions) than behaviors positive in valenceResults showed an interaction effect that confirmed our classification ofthe behaviors into the respective positive and negativendashdiagnostic cate-gories F(1 29) = 4913 p lt 001 For the positivendashdiagnostic traits posi-tive valence behaviors (M = 281) were judged to be enacted less oftenthan negative valence behaviors (M = 376) F(1 29) = 1961 p lt 001 Forthe negativendashdiagnostic traits negative valence behaviors (M = 294)were judged to be enacted less often than positive valence behaviors (M= 331) F(1 29) = 7103 p = 012

Two sets of materials both containing the same 20 behavior stemswere created Each behavior stem was matched with a dispositional at-tribution in one set and a situational attribution in the other set For ex-ample in set A the behaviors ldquoI didnrsquot help my dad with yard workbecause I wanted to stay inside to watch TVrdquo [negative behavior in nega-tivendashdiagnostic domain with dispositional attribution] ldquoI sent my sistera present and a birthday bouquet because my grandmother asked me todo itrdquo [positive behavior in negativendashdiagnostic domain with situational

INTERACTION GOALS AND MEMORY 421

attribution] ldquoI missed morning lectures because my roommate playedpranks with my alarm clockrdquo [negative behavior in positivendashdiagnosticdomain with situational attribution] and ldquoI worked as a research assis-tant because I am very interested in the subject matterrdquo [positive behav-ior in positive diagnostic domain with dispositional attribution] wouldbe counterbalanced in set B with different causal attributions These in-cluded ldquoI didnrsquot help my dad with yard work because he couldnrsquot findthe keys to the tool shedrdquo [situational attribution] ldquoI sent my sister apresent and a birthday bouquet because I wanted to give her a pleasantsurpriserdquo [dispositional attribution] ldquoI missed morning lecture becauseI donrsquot feel like waking up for themrdquo [dispositional attribution] and ldquoIworked as a research assistant because my professor said I had to do itrdquo[situational attribution] As will be described below the valence of thebehaviors and attributions (dispositional vs external) were manipu-lated independently to ensure that subsequent differences in recall oftheoryndashconsistent or inconsistent information could not simply be dueto an artifact based on the memorability of the particular behaviors

The two sets of stimuli were presented to 17 participants for pretest-ing One participant misunderstood how to use the response scale andwas dropped from the analysis Participants were first presented withthe behavior stems and were asked to rate how favorable each behaviorwas on a 7ndashpoint scale with higher values indicating greaterfavorability Then they were presented with the same behavior stemslinked to either a dispositional or situational attribution and were askedto judge the locus of causation for the behaviors on a 7ndashpoint scale withhigher values indicating more dispositional causation The responseswere submitted to a 2 (valence)acute 2 (attribution)withinndashsubject ANOVAFor the valence judgments the positive behaviors in both stimulus sets(Mset A = 544 Mset B = 559) were judged as more favorable than the nega-tive behaviors (Mset A = 284Mset B = 215) Fset A (1 8)= 3416 p lt 0004Fset B

(1 8) = 9782 p lt 0001 For the second set of judgments a main effect forattribution was obtained for both sets of stimuli Behaviors linked todispositional attributions (Mset A = 575 Mset B = 646) were judged to bemore dispositionally caused than behaviors linked to situational attribu-tions (M set A = 235 Mset B = 298) F set A (1 8) = 4574 p lt 0001 Fset B (1 8) =15826 p lt 0001 No other main effects or interaction effects were ob-tained The two sets of stimuli were presented under three different ran-domization schemes

422 CHAN AND YBARRA

PROCEDURE

Upon arrival at the lab the participants were provided the cover storyfor the study in which they were told that the study would examine howpersonal information would affect subsequent interactions betweenstrangers They were told that another set of participants was recordingsome personal information on a tape for them to listen to To increase thecredibility of the tape contents participants were told that the partici-pants recording the tape had been asked to provide distinct personal in-formation about ldquothings that they have done in the past few weeksrdquoAdditional information about the experiment including whether theywere in the cooperation competition or control (impression formation)condition was given to each participant in a sealed envelope so that theexperimenter could remain blind to the conditions Experimental groupparticipants were told that after they listened to a tape with personal in-formation recorded by another participant they would get a chance tomeet them in either a cooperative or competitive setting Control partici-pants were informed that they had been randomly assigned to the con-trol group and would listen to some information about a previousparticipant instead of having an interaction with them

Participants in the cooperation condition read that they would engagein a ldquoCommunity Simulation Gamerdquo in which they and their assignedpartners ldquowill work together to try to maximize the number of points foryour team The team with the highest points will receive a prizerdquo Thecooperative nature of the interaction was emphasized with repeated useof the word ldquopartnerrdquo In contrastparticipants in the competition condi-tion read that they would engage in a ldquoWall Street Simulation Gamerdquo inwhich they would ldquocompete against [their] competitor to maximize[their] own profit (points) and minimize their opponentrsquos profit Theperson with the highest points will win a prizerdquo The competitive aspectof the upcoming interaction was emphasized further with repeated useof the words ldquocompetitorrdquo and ldquoopponentrdquo This type of reward struc-ture manipulation has been used to study cooperation and competitionby other researchers (Deutsch 1973)

Participants in the control condition read that they would listen to atape that had been recorded in a previous experimental session Theywere told to form an impression of the person using the informationfrom the tape as part of a ldquostandardizationrdquo procedure for the experi-ment To bolster the believability of the cover story participants then

INTERACTION GOALS AND MEMORY 423

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

main (hon esty h elpfu lness and fr iendliness ) and thepositivendashdiagnostic trait domain (hardworking image scoring) Imagescoring refers to people engaging in costly positive behaviors to signal toothers that they are valuable community members (Nowak amp Sigmund1998Wedekind amp Milinski 2000) Controlparticipants learned the samematerial but they were asked to form an impression of the person andwere told that they would not interact with the other participant

METHOD

Design and ParticipantsNinetyndashone students participated in the study for course credit Theywere randomly assigned to the cooperation competition or control (im-pression formation) condition All of the participants were presentedwith behaviors with a positive or negative valence from the negativendashdi-agnostic and positivendashdiagnostic trait domains and these behaviorswere linked to either explicit dispositional or situational attributionsThus the overall design of the study was a 3 (interaction goal coopera-tion competition control) acute 2 (trait domain positivendashdiagnostic nega-tivendashdiagnostic) acute 2 (behavior valence positive negative) acute 2(attribution dispositional situational) mixed design with the first fac-tor varying between participants and the latter three factors varyingwithin participants The participants were randomly assigned to condi-tions and were run in nonndashinteracting groups of two to six

Stimulus MaterialsThe behavioral information that the participants processed was pre-sented on a cassette tape allegedly recorded by another participant Thegender of the speaker on the tape and that of the participant were alwaysthe same Each tape contained 20 behavioral statements half of whichwere negative and half positive in valence Also half of the behaviorswere linked to dispositional attributions and half to situational attribu-tions Twelve of the statements pertained to negativendashdiagnostic traits(eg ldquoI entered a concert through the fire exitrdquo [negative valence] ldquoI sentmy sister a present and a birthday bouquetrdquo [positive valence]) Eight ofthe statements pertained to the positivendashdiagnostic traits (eg ldquoI missedmorning lecturesrdquo [negative valence] ldquoI reported a crimerdquo [positive va-lence])

420 CHAN AND YBARRA

Our choice of the traits hardworking and image scoring was deliberatebecause it allows us to show that perceiverrsquos attention to a targetrsquosstrengths (overestimation) is not confounded with emphasizing infor-mation about the targetrsquos competence For example it is usually as-sumed that there is an inverse relationship between the amount of worka person engages in and the presence of the related ability (Darley ampGoethals 1980 Jones 1989) but being hardworking should still signal apositive quality a strength In addition image scoring refers to peopleengaging in costly positive behaviors such as extreme levels of philan-thropy to enhance their reputation and status (Wedekind amp Milinski2000 McAndrew 2002) It is completely unrelated to competence but itsdiagnosticity is in line with that of a positivendashdiagnostic trait

The behavior stems (no attributions) were pretested to ensure thatthey were indeed regarded as positive or negativendashdiagnostic Thirtyparticipants rated each behavior on how common they are For exampleone item queried them as to ldquoWhat percentage of college students doyou think enter a concert through the fire exitrdquo (6ndashpoint scale with 0ndash1010ndash30 30ndash50 50ndash70 70ndash90 and 90ndash100) For positivendashdiagnos-tic traits behaviors that are positive in valence should be less common(and hence more diagnostic of the actorrsquos dispositions) than behaviorsthat are negative in valence whereas for negativendashdiagnostic traits be-haviors negative in valence should be less common (and hence more di-agnostic of the actorrsquos dispositions) than behaviors positive in valenceResults showed an interaction effect that confirmed our classification ofthe behaviors into the respective positive and negativendashdiagnostic cate-gories F(1 29) = 4913 p lt 001 For the positivendashdiagnostic traits posi-tive valence behaviors (M = 281) were judged to be enacted less oftenthan negative valence behaviors (M = 376) F(1 29) = 1961 p lt 001 Forthe negativendashdiagnostic traits negative valence behaviors (M = 294)were judged to be enacted less often than positive valence behaviors (M= 331) F(1 29) = 7103 p = 012

Two sets of materials both containing the same 20 behavior stemswere created Each behavior stem was matched with a dispositional at-tribution in one set and a situational attribution in the other set For ex-ample in set A the behaviors ldquoI didnrsquot help my dad with yard workbecause I wanted to stay inside to watch TVrdquo [negative behavior in nega-tivendashdiagnostic domain with dispositional attribution] ldquoI sent my sistera present and a birthday bouquet because my grandmother asked me todo itrdquo [positive behavior in negativendashdiagnostic domain with situational

INTERACTION GOALS AND MEMORY 421

attribution] ldquoI missed morning lectures because my roommate playedpranks with my alarm clockrdquo [negative behavior in positivendashdiagnosticdomain with situational attribution] and ldquoI worked as a research assis-tant because I am very interested in the subject matterrdquo [positive behav-ior in positive diagnostic domain with dispositional attribution] wouldbe counterbalanced in set B with different causal attributions These in-cluded ldquoI didnrsquot help my dad with yard work because he couldnrsquot findthe keys to the tool shedrdquo [situational attribution] ldquoI sent my sister apresent and a birthday bouquet because I wanted to give her a pleasantsurpriserdquo [dispositional attribution] ldquoI missed morning lecture becauseI donrsquot feel like waking up for themrdquo [dispositional attribution] and ldquoIworked as a research assistant because my professor said I had to do itrdquo[situational attribution] As will be described below the valence of thebehaviors and attributions (dispositional vs external) were manipu-lated independently to ensure that subsequent differences in recall oftheoryndashconsistent or inconsistent information could not simply be dueto an artifact based on the memorability of the particular behaviors

The two sets of stimuli were presented to 17 participants for pretest-ing One participant misunderstood how to use the response scale andwas dropped from the analysis Participants were first presented withthe behavior stems and were asked to rate how favorable each behaviorwas on a 7ndashpoint scale with higher values indicating greaterfavorability Then they were presented with the same behavior stemslinked to either a dispositional or situational attribution and were askedto judge the locus of causation for the behaviors on a 7ndashpoint scale withhigher values indicating more dispositional causation The responseswere submitted to a 2 (valence)acute 2 (attribution)withinndashsubject ANOVAFor the valence judgments the positive behaviors in both stimulus sets(Mset A = 544 Mset B = 559) were judged as more favorable than the nega-tive behaviors (Mset A = 284Mset B = 215) Fset A (1 8)= 3416 p lt 0004Fset B

(1 8) = 9782 p lt 0001 For the second set of judgments a main effect forattribution was obtained for both sets of stimuli Behaviors linked todispositional attributions (Mset A = 575 Mset B = 646) were judged to bemore dispositionally caused than behaviors linked to situational attribu-tions (M set A = 235 Mset B = 298) F set A (1 8) = 4574 p lt 0001 Fset B (1 8) =15826 p lt 0001 No other main effects or interaction effects were ob-tained The two sets of stimuli were presented under three different ran-domization schemes

422 CHAN AND YBARRA

PROCEDURE

Upon arrival at the lab the participants were provided the cover storyfor the study in which they were told that the study would examine howpersonal information would affect subsequent interactions betweenstrangers They were told that another set of participants was recordingsome personal information on a tape for them to listen to To increase thecredibility of the tape contents participants were told that the partici-pants recording the tape had been asked to provide distinct personal in-formation about ldquothings that they have done in the past few weeksrdquoAdditional information about the experiment including whether theywere in the cooperation competition or control (impression formation)condition was given to each participant in a sealed envelope so that theexperimenter could remain blind to the conditions Experimental groupparticipants were told that after they listened to a tape with personal in-formation recorded by another participant they would get a chance tomeet them in either a cooperative or competitive setting Control partici-pants were informed that they had been randomly assigned to the con-trol group and would listen to some information about a previousparticipant instead of having an interaction with them

Participants in the cooperation condition read that they would engagein a ldquoCommunity Simulation Gamerdquo in which they and their assignedpartners ldquowill work together to try to maximize the number of points foryour team The team with the highest points will receive a prizerdquo Thecooperative nature of the interaction was emphasized with repeated useof the word ldquopartnerrdquo In contrastparticipants in the competition condi-tion read that they would engage in a ldquoWall Street Simulation Gamerdquo inwhich they would ldquocompete against [their] competitor to maximize[their] own profit (points) and minimize their opponentrsquos profit Theperson with the highest points will win a prizerdquo The competitive aspectof the upcoming interaction was emphasized further with repeated useof the words ldquocompetitorrdquo and ldquoopponentrdquo This type of reward struc-ture manipulation has been used to study cooperation and competitionby other researchers (Deutsch 1973)

Participants in the control condition read that they would listen to atape that had been recorded in a previous experimental session Theywere told to form an impression of the person using the informationfrom the tape as part of a ldquostandardizationrdquo procedure for the experi-ment To bolster the believability of the cover story participants then

INTERACTION GOALS AND MEMORY 423

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

Our choice of the traits hardworking and image scoring was deliberatebecause it allows us to show that perceiverrsquos attention to a targetrsquosstrengths (overestimation) is not confounded with emphasizing infor-mation about the targetrsquos competence For example it is usually as-sumed that there is an inverse relationship between the amount of worka person engages in and the presence of the related ability (Darley ampGoethals 1980 Jones 1989) but being hardworking should still signal apositive quality a strength In addition image scoring refers to peopleengaging in costly positive behaviors such as extreme levels of philan-thropy to enhance their reputation and status (Wedekind amp Milinski2000 McAndrew 2002) It is completely unrelated to competence but itsdiagnosticity is in line with that of a positivendashdiagnostic trait

The behavior stems (no attributions) were pretested to ensure thatthey were indeed regarded as positive or negativendashdiagnostic Thirtyparticipants rated each behavior on how common they are For exampleone item queried them as to ldquoWhat percentage of college students doyou think enter a concert through the fire exitrdquo (6ndashpoint scale with 0ndash1010ndash30 30ndash50 50ndash70 70ndash90 and 90ndash100) For positivendashdiagnos-tic traits behaviors that are positive in valence should be less common(and hence more diagnostic of the actorrsquos dispositions) than behaviorsthat are negative in valence whereas for negativendashdiagnostic traits be-haviors negative in valence should be less common (and hence more di-agnostic of the actorrsquos dispositions) than behaviors positive in valenceResults showed an interaction effect that confirmed our classification ofthe behaviors into the respective positive and negativendashdiagnostic cate-gories F(1 29) = 4913 p lt 001 For the positivendashdiagnostic traits posi-tive valence behaviors (M = 281) were judged to be enacted less oftenthan negative valence behaviors (M = 376) F(1 29) = 1961 p lt 001 Forthe negativendashdiagnostic traits negative valence behaviors (M = 294)were judged to be enacted less often than positive valence behaviors (M= 331) F(1 29) = 7103 p = 012

Two sets of materials both containing the same 20 behavior stemswere created Each behavior stem was matched with a dispositional at-tribution in one set and a situational attribution in the other set For ex-ample in set A the behaviors ldquoI didnrsquot help my dad with yard workbecause I wanted to stay inside to watch TVrdquo [negative behavior in nega-tivendashdiagnostic domain with dispositional attribution] ldquoI sent my sistera present and a birthday bouquet because my grandmother asked me todo itrdquo [positive behavior in negativendashdiagnostic domain with situational

INTERACTION GOALS AND MEMORY 421

attribution] ldquoI missed morning lectures because my roommate playedpranks with my alarm clockrdquo [negative behavior in positivendashdiagnosticdomain with situational attribution] and ldquoI worked as a research assis-tant because I am very interested in the subject matterrdquo [positive behav-ior in positive diagnostic domain with dispositional attribution] wouldbe counterbalanced in set B with different causal attributions These in-cluded ldquoI didnrsquot help my dad with yard work because he couldnrsquot findthe keys to the tool shedrdquo [situational attribution] ldquoI sent my sister apresent and a birthday bouquet because I wanted to give her a pleasantsurpriserdquo [dispositional attribution] ldquoI missed morning lecture becauseI donrsquot feel like waking up for themrdquo [dispositional attribution] and ldquoIworked as a research assistant because my professor said I had to do itrdquo[situational attribution] As will be described below the valence of thebehaviors and attributions (dispositional vs external) were manipu-lated independently to ensure that subsequent differences in recall oftheoryndashconsistent or inconsistent information could not simply be dueto an artifact based on the memorability of the particular behaviors

The two sets of stimuli were presented to 17 participants for pretest-ing One participant misunderstood how to use the response scale andwas dropped from the analysis Participants were first presented withthe behavior stems and were asked to rate how favorable each behaviorwas on a 7ndashpoint scale with higher values indicating greaterfavorability Then they were presented with the same behavior stemslinked to either a dispositional or situational attribution and were askedto judge the locus of causation for the behaviors on a 7ndashpoint scale withhigher values indicating more dispositional causation The responseswere submitted to a 2 (valence)acute 2 (attribution)withinndashsubject ANOVAFor the valence judgments the positive behaviors in both stimulus sets(Mset A = 544 Mset B = 559) were judged as more favorable than the nega-tive behaviors (Mset A = 284Mset B = 215) Fset A (1 8)= 3416 p lt 0004Fset B

(1 8) = 9782 p lt 0001 For the second set of judgments a main effect forattribution was obtained for both sets of stimuli Behaviors linked todispositional attributions (Mset A = 575 Mset B = 646) were judged to bemore dispositionally caused than behaviors linked to situational attribu-tions (M set A = 235 Mset B = 298) F set A (1 8) = 4574 p lt 0001 Fset B (1 8) =15826 p lt 0001 No other main effects or interaction effects were ob-tained The two sets of stimuli were presented under three different ran-domization schemes

422 CHAN AND YBARRA

PROCEDURE

Upon arrival at the lab the participants were provided the cover storyfor the study in which they were told that the study would examine howpersonal information would affect subsequent interactions betweenstrangers They were told that another set of participants was recordingsome personal information on a tape for them to listen to To increase thecredibility of the tape contents participants were told that the partici-pants recording the tape had been asked to provide distinct personal in-formation about ldquothings that they have done in the past few weeksrdquoAdditional information about the experiment including whether theywere in the cooperation competition or control (impression formation)condition was given to each participant in a sealed envelope so that theexperimenter could remain blind to the conditions Experimental groupparticipants were told that after they listened to a tape with personal in-formation recorded by another participant they would get a chance tomeet them in either a cooperative or competitive setting Control partici-pants were informed that they had been randomly assigned to the con-trol group and would listen to some information about a previousparticipant instead of having an interaction with them

Participants in the cooperation condition read that they would engagein a ldquoCommunity Simulation Gamerdquo in which they and their assignedpartners ldquowill work together to try to maximize the number of points foryour team The team with the highest points will receive a prizerdquo Thecooperative nature of the interaction was emphasized with repeated useof the word ldquopartnerrdquo In contrastparticipants in the competition condi-tion read that they would engage in a ldquoWall Street Simulation Gamerdquo inwhich they would ldquocompete against [their] competitor to maximize[their] own profit (points) and minimize their opponentrsquos profit Theperson with the highest points will win a prizerdquo The competitive aspectof the upcoming interaction was emphasized further with repeated useof the words ldquocompetitorrdquo and ldquoopponentrdquo This type of reward struc-ture manipulation has been used to study cooperation and competitionby other researchers (Deutsch 1973)

Participants in the control condition read that they would listen to atape that had been recorded in a previous experimental session Theywere told to form an impression of the person using the informationfrom the tape as part of a ldquostandardizationrdquo procedure for the experi-ment To bolster the believability of the cover story participants then

INTERACTION GOALS AND MEMORY 423

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

attribution] ldquoI missed morning lectures because my roommate playedpranks with my alarm clockrdquo [negative behavior in positivendashdiagnosticdomain with situational attribution] and ldquoI worked as a research assis-tant because I am very interested in the subject matterrdquo [positive behav-ior in positive diagnostic domain with dispositional attribution] wouldbe counterbalanced in set B with different causal attributions These in-cluded ldquoI didnrsquot help my dad with yard work because he couldnrsquot findthe keys to the tool shedrdquo [situational attribution] ldquoI sent my sister apresent and a birthday bouquet because I wanted to give her a pleasantsurpriserdquo [dispositional attribution] ldquoI missed morning lecture becauseI donrsquot feel like waking up for themrdquo [dispositional attribution] and ldquoIworked as a research assistant because my professor said I had to do itrdquo[situational attribution] As will be described below the valence of thebehaviors and attributions (dispositional vs external) were manipu-lated independently to ensure that subsequent differences in recall oftheoryndashconsistent or inconsistent information could not simply be dueto an artifact based on the memorability of the particular behaviors

The two sets of stimuli were presented to 17 participants for pretest-ing One participant misunderstood how to use the response scale andwas dropped from the analysis Participants were first presented withthe behavior stems and were asked to rate how favorable each behaviorwas on a 7ndashpoint scale with higher values indicating greaterfavorability Then they were presented with the same behavior stemslinked to either a dispositional or situational attribution and were askedto judge the locus of causation for the behaviors on a 7ndashpoint scale withhigher values indicating more dispositional causation The responseswere submitted to a 2 (valence)acute 2 (attribution)withinndashsubject ANOVAFor the valence judgments the positive behaviors in both stimulus sets(Mset A = 544 Mset B = 559) were judged as more favorable than the nega-tive behaviors (Mset A = 284Mset B = 215) Fset A (1 8)= 3416 p lt 0004Fset B

(1 8) = 9782 p lt 0001 For the second set of judgments a main effect forattribution was obtained for both sets of stimuli Behaviors linked todispositional attributions (Mset A = 575 Mset B = 646) were judged to bemore dispositionally caused than behaviors linked to situational attribu-tions (M set A = 235 Mset B = 298) F set A (1 8) = 4574 p lt 0001 Fset B (1 8) =15826 p lt 0001 No other main effects or interaction effects were ob-tained The two sets of stimuli were presented under three different ran-domization schemes

422 CHAN AND YBARRA

PROCEDURE

Upon arrival at the lab the participants were provided the cover storyfor the study in which they were told that the study would examine howpersonal information would affect subsequent interactions betweenstrangers They were told that another set of participants was recordingsome personal information on a tape for them to listen to To increase thecredibility of the tape contents participants were told that the partici-pants recording the tape had been asked to provide distinct personal in-formation about ldquothings that they have done in the past few weeksrdquoAdditional information about the experiment including whether theywere in the cooperation competition or control (impression formation)condition was given to each participant in a sealed envelope so that theexperimenter could remain blind to the conditions Experimental groupparticipants were told that after they listened to a tape with personal in-formation recorded by another participant they would get a chance tomeet them in either a cooperative or competitive setting Control partici-pants were informed that they had been randomly assigned to the con-trol group and would listen to some information about a previousparticipant instead of having an interaction with them

Participants in the cooperation condition read that they would engagein a ldquoCommunity Simulation Gamerdquo in which they and their assignedpartners ldquowill work together to try to maximize the number of points foryour team The team with the highest points will receive a prizerdquo Thecooperative nature of the interaction was emphasized with repeated useof the word ldquopartnerrdquo In contrastparticipants in the competition condi-tion read that they would engage in a ldquoWall Street Simulation Gamerdquo inwhich they would ldquocompete against [their] competitor to maximize[their] own profit (points) and minimize their opponentrsquos profit Theperson with the highest points will win a prizerdquo The competitive aspectof the upcoming interaction was emphasized further with repeated useof the words ldquocompetitorrdquo and ldquoopponentrdquo This type of reward struc-ture manipulation has been used to study cooperation and competitionby other researchers (Deutsch 1973)

Participants in the control condition read that they would listen to atape that had been recorded in a previous experimental session Theywere told to form an impression of the person using the informationfrom the tape as part of a ldquostandardizationrdquo procedure for the experi-ment To bolster the believability of the cover story participants then

INTERACTION GOALS AND MEMORY 423

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

PROCEDURE

Upon arrival at the lab the participants were provided the cover storyfor the study in which they were told that the study would examine howpersonal information would affect subsequent interactions betweenstrangers They were told that another set of participants was recordingsome personal information on a tape for them to listen to To increase thecredibility of the tape contents participants were told that the partici-pants recording the tape had been asked to provide distinct personal in-formation about ldquothings that they have done in the past few weeksrdquoAdditional information about the experiment including whether theywere in the cooperation competition or control (impression formation)condition was given to each participant in a sealed envelope so that theexperimenter could remain blind to the conditions Experimental groupparticipants were told that after they listened to a tape with personal in-formation recorded by another participant they would get a chance tomeet them in either a cooperative or competitive setting Control partici-pants were informed that they had been randomly assigned to the con-trol group and would listen to some information about a previousparticipant instead of having an interaction with them

Participants in the cooperation condition read that they would engagein a ldquoCommunity Simulation Gamerdquo in which they and their assignedpartners ldquowill work together to try to maximize the number of points foryour team The team with the highest points will receive a prizerdquo Thecooperative nature of the interaction was emphasized with repeated useof the word ldquopartnerrdquo In contrastparticipants in the competition condi-tion read that they would engage in a ldquoWall Street Simulation Gamerdquo inwhich they would ldquocompete against [their] competitor to maximize[their] own profit (points) and minimize their opponentrsquos profit Theperson with the highest points will win a prizerdquo The competitive aspectof the upcoming interaction was emphasized further with repeated useof the words ldquocompetitorrdquo and ldquoopponentrdquo This type of reward struc-ture manipulation has been used to study cooperation and competitionby other researchers (Deutsch 1973)

Participants in the control condition read that they would listen to atape that had been recorded in a previous experimental session Theywere told to form an impression of the person using the informationfrom the tape as part of a ldquostandardizationrdquo procedure for the experi-ment To bolster the believability of the cover story participants then

INTERACTION GOALS AND MEMORY 423

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

read about the importance of maintaining strict confidentiality of theinformation they were about to listen to on the tape They were asked tosign a confidentiality form to agree not to disclose the personal infor-mation about the other participant once they left the experiment As amanipulation check all participants then completed a ldquolog sheetrdquo inwhich they described briefly whether they were or were not going tomeet the other participant and the tasks in which they were expected toengage

After listening to the tape the participants engaged in a fivendashminutedistraction task [evaluate their introductory psychology class] to re-duce working memory effects A surprise recall task followed the dis-traction task Participants were asked to recall as much of the contentfrom the tape as they could They were given 10 minutes to completethis task and then they were probed for suspicion Participants werethen debriefed given course credit and thanked for their participa-tion

RESULTS

A judge blind to the experimental conditions credited participants withcorrectly recalling a statement if their reproduction captured the gist ofboth the originally presented behavior stem and the associatedcausal at-tribution Similar to Ybarra and Stephan (1996) half credit was givenwhen participants only remembered the behavior stem but no creditwas given if participants failed to remember the behavior stem Recalleditems that only contained the behavioral stem but not the attribution ac-counted for only 6 of the total amount of information recalled The re-call data yielded the same patterns if analyzed without crediting recallthat only involved the behavioral stem

Data from fourteen participants equally distributed between condi-tions (c2(1 14)= 93 p = ns) were excluded from the analysis becausethey either failed the manipulation check (did not believe that therewould be an interaction or did not believe the recording on the tape wasauthentic) The analysis to be reported produced similar results to theone when these participants were included

To aid presentation of the interaction effects the recall data were la-beled according to two categories within each trait domain For both thenegativendashdiagnostic and positivendashdiagnostic trait domains the recalledbehavioral statements were classified as ldquotheory consistentrdquo or ldquotheory

424 CHAN AND YBARRA

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

inconsistentrdquo referring to whether or not they were consistent with thecausal theories of the respective domains2 The effects of interactiongoals on the processing of the different behaviors were analyzed sepa-rately for the negativendashdiagnostic trait domain and the positivendashdiag-nostic trait domain Since there were 12 behavior statements from thenegativendashdiagnostic trait domain and 8 behavior statements from thepositivendashdiagnostic trait domain the number of items recalled in eachcategory was converted to a proportion for the analysis

Recall of Behaviors from the NegativendashDiagnostic Trait DomainIt was expected that participants who anticipated cooperation with theother person (that they listened to on the tape) would focus on the nega-tivendashdiagnostic trait domain (eg honesty helpfulness friendliness)be-cause information from this trait domain should allow them to uncovertheir partnersrsquo negative qualities This consideration should not be ac-tive for the participants in the competition group Participants in the im-press ion format ion g rou p we re expecte d to focus on thenegativendashdiagnostic trait domain because this trait domain should behighly accessible and should provide information for making evaluativejudgments during impression formation (Wojciszke et al 1998) Thusonly participants in the cooperation and impression formation condi-tions should better remember theoryndashconsistent over theoryndashinconsis-tent information in the negativendashdiagnostic trait domain

Participantsrsquo recall scores were submitted to a 3 (interaction goal co-operation competition impression formation) acute 2 (behavior valencepositive vs negative) acute 2 (attribution type dispositional vs situational)mixed design ANOVA with repeated measures on the latter two fac-tors The analysis yielded an interaction of valence and attribution F(1

INTERACTION GOALS AND MEMORY 425

2 The analysis categorizing recalled information as ldquotheoryndashconsistentrdquo and ldquothe-oryndashinconsistentrdquo showed the same pattern of results as when valence (positive negative)and attribution (dispositional situational) were examined as separate factors In all caseswhere theoryndashconsistent recall was greater than theoryndashinconsistent recall participantsrecalled more of both types of theoryndashconsistent information more than both types of the-oryndashinconsistent information (eg in negativendashdiagnostic domain recall of positive be-haviors with situational attributions would be more than positive behaviors withdispositional attributions and recall of negative behaviors with dispositional attributionswould be more than negative behaviors with situational attributions) Presenting them astheoryndashconsistent versus theory inconsistentndashrecall did not conceal other effects but facili-tated presentation of the data in relation to the hypothesis

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

74)= 1338 p lt 0005This interaction effect indicated that theoryndashconsis-tent recall (negative behaviors with dispositional attributions and posi-tive behaviors with situational attributions) (M = 54) was higher thantheoryndashinconsistent recall (positive behaviors with dispositional attri-butions and negative behaviors with situational attributions) (M = 43)This interaction was qualified by a 3ndashway interaction that implicated theinteraction goal factor F(2 74) = 382 p lt 03 The recall patterns for thethree interaction goal conditions are presented in Figure 2 Consistentwith expectations the participants with a cooperation goal showedhigher theoryndashconsistent recall (M = 55) than theoryndashinconsistent recall(M = 45) in this domain (negativendashdiagnostic) F(1 27) = 721 p lt 01 Itwas expected that participants expecting to compete with an opponentwould not discriminate between theory consistent and inconsistent in-formation in memory The analysis confirmed this prediction Theseparticipants showed equivalent levels of theory consistent (M = 45) andtheory inconsistent recall (M = 45) F(1 22) lt 1 Finally also as expectedparticipants in the impression formation group displayed greater theoryconsistent recall (M = 60) than theory inconsistent recall (M = 41)for thenegativendashdiagnostic domain F(1 25) = 1260 p = 001

426 CHAN AND YBARRA

FIGURE 2 Proportion of attributed behavioral information in the negative-di-agnostic trait domain recalled in the three conditions

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

Recall of Behaviors from the PositivendashDiagnostic Trait DomainIt was anticipated that participants who expected to compete with theopponent would focus on the positivendashdiagnostic trait domain (eghardworking image scoring) because information from this domain isdiagnostic of the opponentsrsquo strengths These considerations should notbe active for the cooperation goal and control participants Thus onlyparticipants in the competition condition were expected to show betterrecall for theory consistent over theory inconsistent information in thepositivendashdiagnostic trait domain

Similar to the previous analysis participantsrsquo recall scores for behav-iors and attributions in the positivendashdiagnostic trait domain were sub-mitted to a 3 (interaction goal cooperation competition impressionformation)acute 2 (behavioral valence positive vs negative) acute 2 (attributiontype dispositionalvs situational)mixed design ANOVA with repeatedmeasures on the latter two factors The analysis yielded a main effect forbehavior favorability F(1 74) = 1620 p lt 0001 which indicated thatnegative behaviors (M = 99)were better recalled than positive behaviors(M = 70) The analysis also yielded an effect for attribution type F(1 74)= 1248 p lt 0007 This effect indicated that situationally attributed be-haviors (M = 49) were better recalled than dispositionally attributed be-haviors (M = 36) There was also a goal acute attribution interaction F(2 74)= 339p lt 01 This effect reflected the tendency for participants in the co-operation and competition conditions to recall more situationally attrib-uted (Ms = 53 and 58 respectively) than dispositionally attributedbehaviors (Ms = 37 and 31 respectively) Participants in the impressionformation condition remembered situationally attributed behaviors (M= 37) to the same extent as dispositionally attributed behaviors (M =39)

As we found for the negativendashdiagnostic trait domain there was an in-teraction of behavior favorability and attribution type F(1 74) = 919p lt003 There was higher theory consistent recall (positive behaviors withdispositional attributions and negative behaviors with situational attri-butions) (M = 47) than theory inconsistent recall (M = 40) Of greater in-terest the 3ndashway interaction including interaction goal was marginallysignificant F(2 74)= 272p lt 07 The recall patterns for the three interac-tion goal conditions are summarized in Figure 3 Closer inspection of therecall patterns indicates that participants expecting to compete with anopponent distinguished among the information in the positivendashdiag-nostic trait domain They displayed greater theory consistent (M = 56)

INTERACTION GOALS AND MEMORY 427

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

than inconsistent recall (M = 34) F(1 22) = 1497 p lt 001 Participantswho expected to cooperate with the partner did not distinguish amongthe information from the positivendashdiagnostic trait domain and dis-played an equal amount of theory consistent and inconsistent recall (Ms= 45) Participants in the impression formation group displayed similarlevels of theoryndashconsistent (M = 41) and inconsistent recall (M = 35) aswell F(1 25) = 113 p = ns

DISCUSSION

The findings from Study 2 indicated that participants who had a cooper-ation interaction goal similar to participants in the impression forma-tion condition showed no discernible memory patterns for thepositivendashdiagnostic trait domain but had better theory consistent recallthan theory inconsistent recall for the negativendashdiagnostic trait domainThis latter effect indicated that participants in the cooperation conditionbetter remembered negative dispositionally attributed and positivesituationally attributed behavioral information This memory pattern issimilar to the misanthropic person memory effect found by Ybarra andStephan (1996) This pattern of recall casts people in a negative light bygiving them more blame and less credit than their behaviors would sug-

428 CHAN AND YBARRA

FIGURE 3 Proportion of attributed behavioral information in the positive-di-agnostic trait domain recalled in the three conditions

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

gest In the current study participants in the competition conditionshowed no distinction in memory for behavioral information in the neg-ativendashdiagnostic trait domain but in the positivendashdiagnostic trait do-main the y disp layed bet ter th eory ndashcon si sten t recal l thantheoryndashinconsistent recall This represented a heretofore undocu-mented pattern with participants showing better memory for positivedispositionally attributed and negative situationally attributed behav-ioral information This memory pattern is the reverse of misanthropicperson memory and thus might be called ldquophilanthropic person mem-oryrdquo because the information tends to cast people in a positive light bygiving them credit for doing well and dismissing their shortcomings

In general what is noteworthy about the present findings is that peo-ple in the cooperation condition preferentially processed and recallednegative information about someone with whom they were to play co-operatively The information participants in the competition conditionpreferentially processed and recalled about an opponent was in con-trast positive These counterndashintuitive findings follow directly from thecurrent analysis which posits that when a person is to cooperate with apartner the prevailing strategy is to assess a potential partnerrsquos weak-nesses which are most likely to be signaled in the negativendashdiagnostictrait domain However when a person is to compete with an opponentthe prevailing strategy is to assess the competitorrsquos strengths These cuesare most likely to be discerned in the positivendashdiagnostic trait domain(Reeder amp Brewer 1979 Skowronski amp Carlston 1989)

It might be argued that the present findings do not reflect the positedmechanisms associated with the underestimation of a partnerrsquos weak-nesses and the overestimation of an opponentrsquos strengths Instead thefindings could reflect the tendency in people to find important in a part-ner social warmth qualities such as honesty helpfulness and friendli-ness but competencendashrelated qualities such as hardworking in anopponent Although not implausible there are several aspects of this ex-planation that render it untenable

First without any mention of competence or social warmth the find-ings from Study 1a and Study 1b showed the predicted patterns mdash peo-ple were poised to seek an opponentrsquos strengths but a partnerrsquosweaknesses Second we selected the positivendashdiagnostic traits so thatthey would not be related to competence especially the trait of imagescoring Finding out that a person reported a crime to the police shouldbe of little relevance in assessing a personrsquos competence In addition it is

INTERACTION GOALS AND MEMORY 429

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

usually assumed that there is an inverse relationship between effort andcompetence (Darley amp Goethals 1980 Jones 1989) Thus being hard-working may actually be a cue that a person is not very competent

Finally research indicates that people remember goalndashconsistent in-formation better than goal inconsistent information (Anderson ampPichert 1978) This would suggest that participants in the cooperationcondition for example should have had better memory for socialwarmth information than other types of information but instead theytended to remember information indicative of social coldness and im-morality So for different reasons the alternative account is lacking in ex-plaining the memory patterns for Study 2 and the strategy selectionpatterns from Study 1a and Study 1b

We conducted the first three studies to investigate whether peopletend to underestimate their potential partners and overestimate theirpotential opponents In Studies 1a and 1b participants were asked to in-dicate in the absence of concrete behavioral information how they orsomebody else would be affected by either overestimating or underesti-mating their partners or opponents In Study 2 we assessed participantsrsquomemory after they read a list of behaviors To further validate our con-ceptualization we conducted Study 3 to investigate the type of informa-tion participants would actively seek out regarding their future partnersor opponents

STUDY 3

With limited amounts of time and cognitive resources people oftenneed to seek and process only a subset of all the information that is avail-able to them If interaction goals indeed affect the type of informationpeople elaborate and remember as the findings from Study 2 suggestedthen interaction goals may also affect peoplersquos information seeking pat-terns Various studies lend preliminary support to the idea that goals af-fect peoplersquos information seeking strategies For example research hasshown that people desired more information when evaluating someoneperceived as a date than someone who was not perceived as a date (Le-one amp Leone 1984) Other research has shown that people sought moretraitndashrelated information and less appearancendashrelated informationwhen they expected the other person to be a condashworker or a friend than ifthey expected the other person to be a potential date (Shaw amp Steers1996)

430 CHAN AND YBARRA

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

It has also been shown that depending on the goals of the perceiverpeople who hold an a priori expectation sometimes seek informationthat confirms their initial expectation (eg Snyder amp Swann 1978) butat other times seek information to determine if the initial expectationsare correct (eg Bassok amp Trope 1984 Trope amp Bassok 1983) Peoplewho have a goal that places a value on accuracy or accountability preferto seek information that allows them to determine whether the initial ex-pectations are accurate Other research has shown that people prefer di-agnostic questions (eg ldquoDo you like to work alonerdquo) to leadingquestions (eg ldquoWhat situations have you been in where you wishedthat you could have been more talkativerdquo) when they are given an in-formationndashseeking goal but prefer leading questions to diagnostic oneswhen they are given an impressionndashverification goal (LeyensDardenne amp Fiske 1998) In sum these studies suggest that perceiversrsquogoals can play a significant role in information seeking patterns

How might the interaction goals of cooperationand competition affectinformation seeking In this study we asked participants to imaginethemselves in a scenario in which they would either cooperate or com-pete with another person Participants were given a chance to choosefrom a list of behaviors a subset to determine whether or not those be-haviors had actually been committed by their future partneropponentSince cooperation should lead people to be sensitive to their potentialpartnersrsquo weaknesses it was expected that participants who expected tocooperate would select more behaviors related to the negativendashdiagnos-tic than the positivendashdiagnostic trait domain for verification In contrastbecause competition should lead people to be sensitive to their potentialopponentsrsquo strengths it was expected that participants expecting com-petition would select more behaviors related to the positivendashdiagnostictrait domain than the negativendashdiagnostic trait domain for verificationThe last experiment tested this hypothesis

METHOD

Design and ParticipantsForty students participated in the study They were approached on auniversity campus and asked if they would volunteer to fill out ashort questionnaire They were randomly assigned to the cooperationor competition condition All of the participants were presented withpositive and negative behaviors from the negativendashdiagnostic and

INTERACTION GOALS AND MEMORY 431

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

positivendashdiagnostic trait domains Thus the overall design of thestudy was a 2 (interaction goal cooperation competition) acute 2 (traitdomain positivendashdiagnostic negativendashdiagnostic) acute 2 (behavior va-lence positive negative) mixed design with the first factor varyingbetween participants and the latter two factors varying within partic-ipants

Stimulus MaterialsParticipants read 24 behaviors (no attributions) half of which were neg-ative and half positive in valence Twelve of the statements pertained tonegativendashdiagnostic traits and 12 of the statements pertained to the posi-tivendashdiagnostic traits For this study we added behaviors to the stimu-lus set used in Study 2 to create a set with an equal number of behaviorsfrom the two trait domains Thus we prendashtested the behaviors again toensure that they were indeed regarded as positive or negativendashdiagnos-tic Twenty participants rated the behaviors on how common they wereFor example one item queried them as to ldquoWhat percentage of collegestudents do you think have an essay published in a collegiate Economics mag-azinerdquo (6ndashpoint scale with 0ndash10 10ndash30 30ndash50 50ndash70 70ndash90and 90ndash100) For positivendashdiagnostic traits behaviors with positivevalence should be less common (and hence more diagnostic of the ac-torrsquos dispositions) than behaviors with negative valence whereas fornegativendashdiagnostic traits behaviors with negative valence should beless common (and hence more diagnostic of the actorrsquos dispositions)than behaviors with positive valence Results showed an interaction ef-fect that confirmed our classification of behaviors F(1 19) = 48537 p lt001 For the positivendashdiagnostic traits positive behaviors were judgedto be enacted less often (M =171) than negative behaviors (M = 279) F(119) = 6247 p lt 001 For the negativendashdiagnostic traits negative behav-iors were considered less common (M = 224) than positive behaviors (M= 290) F(1 19) = 1813 p lt 001

ProcedureParticipants were each given a questionnaire that asked them to eitherimagine that they were to cooperate or compete with another personThe cooperation scenario read

You will be cooperating with Person A in an upcoming task The goal is tomaximize the number of points the two of you will get on the task You and

432 CHAN AND YBARRA

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

your partner (Person A) will have to work together to succeed on the task Ifyou and your partner were to earn more than 300 points both of you will re-ceive a prize

The competition scenario read

You will be competing with Person A in an upcoming task The goal is tomaximize the number of points you get and minimize the number of pointsyour opponent (Person A) gets You will have to compete against your op-ponent (Person A) to succeed on the task The person who has more pointswill receive a prize

Participants in both conditions were then asked to imagine that ldquobeforebeginning the cooperationcompetition you have a chance to learnabout your partneropponent (Person A) so that you can decide whatstrategy to take during the cooperationcompetition in order to achievethe most successful outcomerdquo They were then given the list of 24 behav-iors and were told that some of these behaviors had actually been en-acted by the targets whereas others had not They were asked to choose 9of the behaviors to find out if their partneropponent had actually en-gaged in those behaviors After they made their selections participantswere debriefed and thanked for their participation

INTERACTION GOALS AND MEMORY 433

FIGURE 4 The number of behaviors participants who expected to cooperate or competeselected for verification about their future partneropponent

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

RESULTS

Participantsrsquo selections of behaviors were submitted to a 2 (interactiongoal cooperation competition) acute 2 (trait domain positivendashdiagnosticnegativendashdiagnostic) acute 2 (behavior valence positive vs negative) mixeddesign ANOVA with repeated measures on the latter two factorsConsis-tent with the hypothesis the analysis yielded an interaction of trait do-main and interaction goal F(1 38) = 8241 p = 007 (Figure 4) Thisinteraction effect indicated that participants who expected to cooperateselected more behaviors from the negativendashdiagnostic trait domain (M =2625) than the positivendashdiagnostic trait domain (M = 1900) F(1 19) =3937 p = 062 In contrast participants who expected to compete selectedmore behaviors from the positivendashdiagnostic domain (M = 2525)than thenegativendashdiagnostic domain (M = 1850) F(1 19) = 4368 p = 050

The only other effect was a marginal interaction of interaction goal acutevalence F(1 38) = 3459 p = 071 This effect was based on two trendsshowing that cooperation participants wished to verify fewer negative(M = 210) than positive behaviors (M = 243) In contrast competitionparticipants wished to verify fewer positive (M = 193) than negative be-haviors (M = 245)

DISCUSSION

The findings from Study 3 indicated that people who had a cooperationinteraction goal were more interested in their potential partnerrsquos behav-iors related to the negativendashdiagnostic trait domain whereas peoplewho had a competition interaction goal were more interested in their po-tential opponentsrsquo behaviors related to the positivendashdiagnostic trait do-main These findings are congruent with the notion that people tend tounderestimate potential partners and overestimate potential oppo-nents consistent with the findings of the previous two studies

GENERAL DISCUSSION

Study 1a and Study 1b showed that people anticipating cooperation pre-ferred to underestimate rather than overestimate how good their part-ners were and people who anticipated competition preferred tooverestimate rather than underestimate how good their opponentswere Study 2 demonstrated the important role of interaction goals in so-cial information processing by showing how cooperation and competi-

434 CHAN AND YBARRA

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

tion goals can determine the type of information (positivendashdiagnostic ornegativendashdiagnostic) people focus on and give additional processing toPeople who expected to cooperate with the other person focused on andtended to remember information from the negativendashdiagnostic trait do-main In contrast people who expected to compete focused on informa-tion from the positivendashdiagnostic trait domain When people onlyintended to form a general impression of others they focused on nega-tivendashdiagnostic information Study 3 further validated our conceptual-ization by showing that people who anticipated cooperation withanother person sought more information from the negativendashdiagnostictrait domain whereas people who anticipated competition sought moreinformation from the positivendashdiagnostic trait domain

The current framework is applicable not only to interpersonal inter-actions but also possibly to intergroup interactions as well Althoughintergroup judgments are not always negative (Tajfel amp Turner 1979)in general outgroup members tend to be treated unfairly (TurnerBrown amp Tajfel 1979) are more likely to be perceived as homogenous(Linville Fischer amp Salovey 1989Ostrom amp Sedikides 1992) and theirnegative behaviors tend to be perceived as representative of thegrouprsquos negative ldquocollective dispositionrdquo (Pettigrew 1979) Consistentwith the research on ingroupndashoutgroup bias Ybarra Stephan andSchaberg (2000) found evidence of misanthropic memory when peoplewere perceiving outgroup members but not when they were perceiv-ing ingroup members

It may be useful to apply the current research on interaction goals tothe processing of outgroup informationbecause intergroup contact is of-ten competitive or perceived to be competitive (Rabbie amp Wilkens 1971)The current research suggests that people who expect to compete withoutgroup members might focus on the positivendashdiagnostic traits of theiroutgroup opponents instead of focusing on their negativendashdiagnosticqualities By focusing on the positivendashdiagnostic traits people should bemore likely to remember positive behaviors with dispositional attribu-tions and negative behaviors with situational attributions This leads toa counterndashintuitive prediction that people may also be more likely tomake positive than negative dispositional inferences about their oppo-nents

This prediction runs counter to previous findings that have shownthat competition tends to increase the salience of group boundaries andaccentuates intergroup bias (eg Bettencourt Brewer Croak amp Miller

INTERACTION GOALS AND MEMORY 435

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

1992 Johnson amp Johnson 1985 Sherif 1958) This difference could po-tentially be resolved by noting that philanthropic information process-ing under the competition goal in the current studies was elicited whenthe perceivers were sizingndashup the opponent before actual competitionDuring this initial period of evaluation it should be more important todiscover the opponentsrsquo strengths so that one can avoid underestimat-ing the opponent However after a competition people may no longerneed to size up the opponent Instead since competition should have ledto an increase in the salience of group boundaries there might be an in-crease in the need to enhance the ingroup and derogate the outgroupThis should revert peoplersquos focus to negativendashdiagnostic informationabout the outgroup which would lead to misanthropic memory andother negativity effects Therefore the current analysis would suggestthat philanthropic processing would be obtained before competitivecontact whereas misanthropic processing would occur after competi-tive contact

In sum the present research suggests that interaction goals affectwhether people adopt a general strategy to either underestimate theirpartners or to overestimate their opponents The adopted strategy thenoperates in conjunction with the implicit causal theories held byperceivers to determine how information is processed The current find-ings indicate that the concerns and needs of people expecting to cooper-ate counterndashintuitively may lead them to seek negative informationabout others while the concerns and needs of people who expect to com-pete may lead them to seek positive information about others These re-sults point to the importance of incorporating perceiversrsquo goals andinformationalcontent into our understanding of person memory and so-cial information processing

REFERENCESAnderson R C amp Pichert J W (1978) Recall of previously unrecallable information fol-

lowing a shift in perspective Journal of Verbal Learning and Verbal Behavior 17 1ndash12Bassok M amp Trope Y (1984) Peoplersquos strategies for testing hypotheses about anotherrsquos

personality Confirmatory or diagnostic Social Cognition 2 199ndash216Bettencourt B A Brewer M B Croak M R amp Miller N (1992)Cooperation and the re-

duction of intergroup bias The role of reward structure and social orientation Jour-nal of Experimental Social Psychology 28 301ndash319

Bornstein G amp Rapoport A (1988) Intergroup competition for the provision ofstepndashlevel public goods Effects of preplay communication European Journal of So-cial Psychology 18 125ndash142

436 CHAN AND YBARRA

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

Brewer M B (1979) Inndashgroup bias in the minimal intergroup situation A cognitivendashmoti-vational analysis Psychological Bulletin 86 2 307ndash324

Darley J M amp Goethals G R (1980) Peoplersquos analyses of the causes of ability linked per-formances In L Berkowitz (Ed) Advances in Experimental Social Psychology 131ndash37

Deutsch M (1973) The resolution of conflict New Haven YaleDevine P G Sedikides C amp Fuhrman R W (1989) Goals in social information process-

ing The case of anticipated interaction Journal of Personality and Social Psychology56 680ndash690

Hertel G amp Fiedler K (1994) Affective and cognitive influences in a social dilemmagame European Journal of Social PsychologySpecial Issue Affect in social judgments andcognition 24 131ndash145

Insko C A Schopler J Hoyle R H Dardis G J amp Graetz K A (1990) Individ-ualndashgroup discontinuity as a function of fear and greed Journal of Personality and So-cial Psychology 58 68ndash79

Johnson D W amp Johnson R T (1985)Relationships between Black and White students inintergroup cooperation and competition Journal of Social Psychology125 421ndash428

Jones E E (1989)The framing of competence Personality and Social PsychologyBulletin 15477ndash492

Jones E E amp Nisbett R E (1972)The actor and the observer Divergent perceptions of thecauses of behavior In E E Jones D Kanouse H H Kelley R E Nisbett S Valinsamp B Weiner (Eds) Attribution Perceiving the causes of behavior (pp 79ndash94)Morristown NJ General Learning Press

Kahneman D amp Tversky A (1979) Prospect theory An analysis of decision under riskEconometrica 47 263ndash91

Kerr N L (1999) Anonymity and social control in social dilemmas In M Foddy MSmithson S Schneider M Hogg (Eds) Resolving Social Dilemmas Dynamic Struc-tural and Intergroup Aspects (pp 103ndash119) Philadelphia Psychology Press TaylorFrancis

Lassiter G D Geers A L Apple K J amp Beers M J (2000) Observational goals and be-havior unitization A reexamination Journal of Experimental Social Psychology 36649ndash659

Leone C amp Leone CD (1984) Some effects of outcome dependency temporal proximityof interaction and sex differences on social information seeking Representative Re-search in Social Psychology 13 66ndash76

Levy L H amp Dugan R D (1960)A constant error approach to the study of dimensions ofsocial perception Journal of Abnormal and Social Psychology 61 21ndash24

Leyens J Dardenne B amp Fiske S T (1998) Why and under what circumstances is a hy-pothesisndashconsistent testing strategy preferred in interviews British Journal of SocialPsychology 37 259ndash274

Linville P W Fischer G W amp Salovey P (1989)Perceived distributions of the character-istics of inndashgroup and outndashgroup members Empirical evidence and a computersimulation Journal of Personality and Social Psychology 57 165ndash188

Malle B F amp Pearce G E (2001) Attention to behavioral events during interaction Twoactorndashobserver gaps and three attempts to close them Journal of Personality and So-cial Psychology 81 278ndash294

McAndrew F T (2002) New evolutionary perspectives on altruism Multilevelndashselectionand costly signaling theories Current Directions in Psychological Science 11 79ndash82

McGuire W J amp McGuire C V (1986)Differences in conceptualizing self versus concep-tualizing other people as manifested in contrasting verb types used in naturalspeech Journal of Personality and Social Psychology 51 1135ndash1143

INTERACTION GOALS AND MEMORY 437

Neuberg S L amp Fiske S T (1987)Motivational influences on impression formation Out-come dependency accuracyndashdriven attention and individuation processes Journalof Personality and Social Psychology 53 431ndash444

Nowak M A amp Sigmund K (1998) Evolution of indirect reciprocity by image scoringNature 393 573ndash577

Ostrom T M amp Sedikides C (1992)Outndashgroup homogeneity effects in natural and mini-mal groups Psychological Bulletin 112 536ndash552

Parks C D Henager R F amp Scamahorn S D (1996) Trust and reactions to messages ofintent in social dilemmas Journal of Conflict Resolution 40 134ndash151

Pettigrew T F (1979)The ultimate attribution error Extending Allportrsquos cognitive analy-sis of prejudice Personality and Social Psychology Bulletin 5 461ndash476

Prentice D A (1990)Familiarity and differences in selfndash and otherndash representations Jour-nal of Personality and Social Psychology 59 369ndash383

Rabbie J M amp Wilkens G (1971) Intergroup competition and its effects on intragroupand intergroup relations European Journal of Social Psychology 1 215ndash234

Reeder G D (1985) Implicit relations between dispositions and behaviors Effect ondispositional attribution In J H Harvey amp G Weary (Eds) Attribution Basic issuesand applications (pp 87ndash116) New York Academic

Reeder G D amp Brewer M B (1979) A schematic model of dispositional attribution in in-terpersonal perception Psychological Review 86 61ndash79

Reeder G D amp Fulks J L (1980) When actions speak louder than words Implicationalschemata and the attribution of ability Journal of Experimental Social Psychology 1633ndash46

Rosen J amp HaagenD A F (1998)Facilitating cooperation in a social dilemma A persua-sion approach Journal of Psychology 132 143ndash153

Ruscher J B amp Fiske S T (1990) Interpersonal competition can cause individuating pro-cesses Journal of Personality and Social Psychology 58 832ndash843

Shaw J I amp Steers W N (1996)Effects of perceiver sex search goal and target person at-tributes on information search in impression formation Journal of Social Behavior ampPersonality Special Issue Handbook of gender research 11 209ndash227

Sherif M (1958) Superordinate goals in the reduction of intergroup conflict AmericanJournal of Sociology 63 349ndash356

Skowronski J J amp Carlston D E (1987) Social judgment and social memory The role ofcue diagnosticity in negativity positivity and extremity biases Journal of Personal-ity and Social Psychology 52 689ndash699

Skowronski J J amp Carlston D E (1989) Negativity and extremity biases in impressionformation A review of explanations Psychological Bulletin 105 131ndash142

Snyder M amp Swann W B (1978) Hypothesisndashtesting processes in social interaction Jour-nal of Personality amp Social Psychology 36 1202ndash1212

Srull T K amp Brand J F (1983) Memory for information about persons The effect of en-coding operations upon subsequent retrieval Journal of Verbal Learning and VerbalBehavior 22 219ndash230

Tajfel H amp Turner J C (1979)An integrative theory of intergroup conflict In W G Aus-tin amp S Worchel (Eds) The social psychology of intergroup relations (pp 33ndash47)Monterey CA BrooksCole

Tesser A amp Danheiser P (1978) Anticipated relationship salience of partner and atti-tude change Personality and Social Psychology Bulletin 4 33ndash38

Trope Y amp Bassok M (1983) Informationndashgathering strategies in hypothesisndashtestingJournal of Experimental Social Psychology 19 560ndash576

Turner J C Brown R J amp Tajfel H (1979) Social comparison and group interest iningroup favouritism European Journal of Social Psychology 9 187ndash204

438 CHAN AND YBARRA

Wedekind K amp Milinski M (2000) Cooperation through image scoring in humans Sci-ence 288 561ndash574

Williams K D amp Karau S J (1991)Social loafing and social compensation The effects ofexpectations of condashworker performance Journal of Personality amp Social Psychology61 570ndash581

Wittenbaum G M Vaughan S I amp Stasser G (1998) Coordination in taskndashperforminggroups Social psychological applications to social issues 4 177ndash204

Wojciszke B BazinskaR amp Jaworski M (1998)On the dominance of moral categories inimpression formation Personality and Social Psychology Bulletin 24 1251ndash1263

Ybarra O (2001) When first impressions donrsquot last The role of isolation and adaptationprocesses in the revision of evaluative impressions Social Cognition 19 491ndash520

Ybarra O (2002) Peoplersquos naiumlve causal understanding of valenced behaviors and its im-plications for social information processing Psychological Bulletin 128 421ndash441

Ybarra O amp Stephan W G (1996)Misanthropic person memory Journal of Personality andSocial Psychology 70 691ndash700

Ybarra O amp Stephan W G (1999)Attributional orientations and the prediction of behav-ior the attributionndashprediction bias Journal of Personality and Social Psychology 76718ndash727

Ybarra O Stephan WG amp Schaberg LA (2000)Misanthropic memory for the behaviorof group members Personality and Social Psychology Bulletin 26 1515ndash1525

Zajonc R B (1980) Feeling and thinking Preferences need no inferences American Psy-chologist 35 151ndash175

INTERACTION GOALS AND MEMORY 439